Pine Seedling Detection and Registration Except where reference is made to the work of others, the work described in this thesis is my own or was done in collaboration with my advisory committee. This thesis does not include proprietary or classi ed information. Je rey K. Hunt Certi cate of Approval: John Y. Hung, Co-Chair Professor Electrical and Computer Engineering Timothy P. McDonald, Co-Chair Associate Professor Biosystems Engineering Thomas Baginski Professor Electrical and Computer Engineering John P. Fulton Assistant Professor Biosystems Engineering George T. Flowers Dean Graduate School Pine Seedling Detection and Registration Je rey K. Hunt A Thesis Submitted to the Graduate Faculty of Auburn University in Partial Ful llment of the Requirements for the Degree of Master of Science Auburn, Alabama December 18, 2009 Pine Seedling Detection and Registration Je rey K. Hunt Permission is granted to Auburn University to make copies of this thesis at its discretion, upon the request of individuals or institutions and at their expense. The author reserves all publication rights. Signature of Author Date of Graduation iii Vita Je rey Hunt, son of Raymond Hunt and Teresa Surrey, was born February 26, 1985, in Mobile, Alabama. He graduated from Robertsdale High School in 2003. He attended Auburn University in September 2003 and graduated with a Bachelor of Electrical Engineer- ing degree in May 2007. He enrolled in Graduate School at Auburn University, June 2007. He worked as a graduate teaching assistant for the Department of Electrical Engineering and as a research assistant for the Department of Biosystems Engineering. He graduated December 2009 with a Master?s of Science in Electrical Engineering. iv Thesis Abstract Pine Seedling Detection and Registration Je rey K. Hunt (B.S., Auburn University, 2007) 91 Typed Pages Directed by John Y. Hung and Timothy P. McDonald Pine seedling nurseries across the United States depend on antiquated techniques and equipment to manage the country?s supply of pine seedling stock. This research is directed towards the development of a method to autonomously inventory and manage pine seedlings while in the nursery bed. Several electronic technologies are analyzed and compared to de- termine their performance as a pine seedling detection and registration sensor for nurseries. The technologies of interests are: photo-interrupt sensors, capacitive sensors, microwave sensors, computer vision, and radar. A key component in determining which technology to pursue was its ability to obtain an accurate inventory with low processing cost. The ability to determine the overall health of the pine seedlings was also an important consider- ation. The nal goal of this research is to catalog each technologies proformance as a pine seedling detection and registration sensor for the future development of an auto-indexing pine seedling counter. v Master of Science, December 18, 2009 Acknowledgments This thesis would not have been possible without the support and dedicated faculty found at both the Department of Electrical Engineering and the Department of Biosystems Engineering. A special thanks goes to each of my advisers. Dr. Hung, thank you for providing focus during the extent of the research process. Dr. McDonald, thank you for your direction and constant support. Dr. Baginski, thank you for your willingness to think outside the box and try new ideas. Dr. Fulton, thank you for editing and your words of encouragement. I would also like to recognize Dr. Riggs for his extensive assistance. His knowledge in the area of electromagnetics was true north for this research. vi Style manual or journal used Journal of Approximation Theory (together with the style known as \aums"). Bibliograpy follows van Leunen?s A Handbook for Scholars. Computer software used The document preparation package TEX (speci cally LATEX) together with the departmental style- le aums.sty. vii Contents List of Figures x 1 Introduction 1 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Current Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.3 Research Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.4 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.5 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.5.1 Method Speci cations . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.6 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2 Current Inventory Management Technologies 7 2.1 Photo-Interrupt Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Capacitive Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Microwave Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4 Object Registration via Computer Vision . . . . . . . . . . . . . . . . . . . 10 2.5 Object Registration via Radar . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3 Research Procedures 12 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.2 Photo-Interrupt Detection Experiment . . . . . . . . . . . . . . . . . . . . . 12 3.3 Capacitive Sensing Experiment . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.4 Computer Vision Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.5 Radar Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.5.1 Radar Return Signal Measurement . . . . . . . . . . . . . . . . . . . 17 3.5.2 Radar Backscatter Measurement . . . . . . . . . . . . . . . . . . . . 17 4 Results and Discussion 23 4.1 Photo-Interrupt Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.2 Capacitance Sensor Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.3 Computer Vision Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.4 Radar Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 4.4.1 Monopole Tree Model . . . . . . . . . . . . . . . . . . . . . . . . . . 30 4.4.2 Ideal Loblolly Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.4.3 Sample Seedling Model . . . . . . . . . . . . . . . . . . . . . . . . . 35 viii 5 Summary and Conclusions 37 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Bibliography 41 Appendices 43 A Photo-Interrupt Microcontroller Schematic 44 B Photo-Interrupt Microcontroller Code 45 C Capacitive Sensor Schematic 58 D Capacitive Sensor MATLAB Code 59 E Capacitive Sensor Collected Data 61 F MATLAB Code for Image Processing 74 G NEC Win-Pro Code for Pine Seedling Models 76 H Simulated Radar Return Data 78 ix List of Figures 3.1 Typical Pine Seedling Nursery Bed [2] . . . . . . . . . . . . . . . . . . . . . 12 3.2 Arti cial Pine Seedling Test Bed . . . . . . . . . . . . . . . . . . . . . . . . 13 3.3 Photo-Interrupt Experiment Setup . . . . . . . . . . . . . . . . . . . . . . . 14 3.4 Capacitive Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.5 Incident Plane Wave Referenced to the Model . . . . . . . . . . . . . . . . . 18 3.6 Ideal Loblolly Pine & Sample Pine Seedling [21] . . . . . . . . . . . . . . . . 19 3.7 NEC Win-Pro Wire Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.8 Sampled Dielectric Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 4.1 Seedling Position and Measured and Modeled Capacitance . . . . . . . . . 24 4.2 Seedling Harvester Infra-Red Guidance Camera . . . . . . . . . . . . . . . . 25 4.3 Infra-Red Photographs of Arti cially Spaced Pine Seedlings . . . . . . . . . 26 4.4 Ultra-Violet Photographs of Arti cially Spaced Pine Seedlings . . . . . . . 26 4.5 Plant Absorption of Light in the Visible Spectrum [14] . . . . . . . . . . . . 27 4.7 Series of Images with Data Vector . . . . . . . . . . . . . . . . . . . . . . . 29 4.6 Visible Spectrum Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.8 Voltage Waveform Induced from the Current Flow through a Pine Seedling due to the presence of an Electromagnetic Field . . . . . . . . . . . . . . . . 31 4.9 Radiation Pattern of the Monopole Tree Model . . . . . . . . . . . . . . . . 32 4.10 Monopole Antenna Reradiated Power . . . . . . . . . . . . . . . . . . . . . 33 4.11 Radiation Pattern of Ideal Loblolly Model . . . . . . . . . . . . . . . . . . . 34 x 4.12 Loblolly Antenna Re-radiated Power . . . . . . . . . . . . . . . . . . . . . . 34 4.13 Radiation Pattern of the Sample Seedling Model . . . . . . . . . . . . . . . 35 4.14 Sample Seedling Antenna Re-radiated Power . . . . . . . . . . . . . . . . . 36 A.1 Photo-Interrupt Microcontroller Schematic . . . . . . . . . . . . . . . . . . . 44 C.1 Capacitive Sensor (Cx represents electrodes) . . . . . . . . . . . . . . . . . . 58 xi Chapter 1 Introduction 1.1 Motivation The production of pine seedlings for the purpose of reforestation is a crucial part of sustaining the earth?s forests. Pine seedlings and other tree seedlings are grown and purchased every year to reforest millions of acres of timberlands. The timber industry is the largest producer and purchaser of pine seedlings for reforestation. The timber company Weyerhaeuser produces the most annually in the United States at over 20 million seedlings a year just for their own timberlands [4]. Seedlings grown for reforestation in the United States are mainly a variety of pine seedlings. Pine seedling nurseries across the United States depend on antiquated techniques and equipment to manage the countries supply of pine seedling stock. The high volume of pine seedlings produced each year combined with the limited amount of nursery land has pushed nurseries to become more e cient in production and inventory management. Nursery man- agers are already pushing the limits of production by increasing the number of seedlings planted per square foot to the maximum viable limits [20]. Interest in creating a more e - cient process has generated research into the development of a better inventory management system. 1.2 Current Method Generating an accurate count of the number of seedlings in a eld is the main concern presented by nursery managers. Currently there is no standard method or system in place to inventory seedlings in the eld or during harvest. The method commonly used for obtaining a seedling count prior to harvesting is to count the number of seedlings, manually, in a 1 predetermined section of a row. The number of seedlings counted is divided by the row length yielding a count-length. The count-length is considered the average number of trees found in a unit length of row. Nursery managers then use the count-length to estimate the overall inventory for the eld and as a guide to determine how much of a row must be harvested to ll an order. Nursery managers would also like to maintain a count of seedlings during the harvesting process. Seedling orders arrive at the nursery and an estimation is made out in the eld as to how much of the seedling bed to harvest to ll the order using the count-length method. The seedlings are pulled from the nursery bed with a harvester and taken to a shed for processing. The current method for counting seedlings post harvest is to count, by hand, a sub-sample. The sub-sample is then weighed and a count-weight is established. A count- weight is the weight of a pre-established number of trees, in most cases the number is 50 or 100. Groups of seedlings are weighed and bundled according to their weight and grade. A seedling?s weight is mainly a function of its moisture content, which changes dynamically from one batch to the next. The count-weight is updated constantly to average dynamic changes in moisture for a more accurate count. The count-weight method is the current inventory process for most pine seedling nurseries in the Southeastern part of the United States. Both methods are inaccurate, so the nursery managers must ensure generous portions of tree stock for the consumer, which cuts into the e ciency of the nursery operation. Although automated inventory systems have been developed in the past, none are currently in use due to their poor performance. 1.3 Research Value Managers are primarily in need of a automated system to obtain an accurate count of pine seedlings in the eld. The development of a pine seedling detection and registration system would be bene cial to the entire nursery operation. The value of this research is to provide nursery managers with a modern tool to inventory more e ciently. Nursery 2 managers could maintain better quality control by making proper irrigation and fertiliza- tion adjustments according to seedling density plots. A system capable of grading seedlings would increase the e ciency of the operation by identifying seedling rows ready for harvest. Grading the seedlings in the eld would reduce the time to package. The eld inventory would also ensure that the proper quantity of seedlings are harvested and packaged. De- veloping an automated system will provide nursery manager?s with a tool to better control the e ciency and the quality of the entire inventory process. 1.4 Previous Work A system for auto-indexing seedlings was patented in 1977 by an inventor, Rath [3]. Rath designed a harvester he claimed was capable of counting, grading, and bundling seedlings as they were lifted from the nursery bed. His invention used a photo-interrupt sensor to count and grade seedlings, but it was never adopted by the seedling industry. The United States government has funded research for a device to count seedlings. In 1991 a seedling counter was developed by the USDA Forest Service?s Missoula Technology and Development Center (MTDC) in cooperation with Dr. Glenn Kranzler of Oklahoma State University. The device was built to count a single row of conifer seedlings in a nursery bed. Their objective was to automate the inventory process. The outcome was a custom- designed photo-interrupt system controlled by a portable computer which could count a single row of seedlings with rows spaced at least 6 inches apart. The seedlings were required to be at least .08 inches (2mm) in diameter and 2 inches (5.1 cm) in height to be counted. Due to the large number of seedlings in a nursery it was suggested that the gross inventory could be estimated based on the count obtained from one row1. The results of this study were disappointing, often recording between 70-90% of the total number of seedlings in a row. The study recommended that a precision seeder would improve the accuracy by evenly spacing the seedlings [1]. Several other explanations were given to account for the counter?s 1This method is similar to the count-length method previously described in the current method section. 3 inaccuracy, such as mounds that occur at the root collar of the seedlings due to the e ects of irrigation, false interrupts from branches, and false interrupts from foliage density. Other types of tree registration technologies have been developed from radar and Li- DAR2 for applications in forestry management [5]. Researchers have been interested in developing a way to monitor and map forests on a global scale to ensure sustainable forestry. Several techniques have developed from this interest and most of those techniques make use of two technologies: Li-DAR3 and InSAR4. Li-DAR is typically used to obtain stand heights from an airborne system. The system?s airborne Li-DAR sensor scans the ground to obtain the aircraft to ground distance and the aircraft to tree crown distance. The di erential between the two measurements provides the stand height at the scanned location. InSAR is used to obtain characteristics that cannot be detected optically such as canopy volume and biomass. InSAR is currently being researched as a possible tool to make indi- vidual tree measurements. Researchers are suggesting that airborne InSAR data of forest stands can be used to build a two dimensional map of a forest stand. The same InSAR system could tag each tree on the map with speci c information pertaining to that tree such as height and stocking volume. These technologies can potentially be applied to pine seedling detection and are examined in this thesis.[5] 1.5 Approach An evaluation of the best applicable sensing technology is needed based on the limited amount of literature and research performed on pine seedling detection and registration. The development of a pine seedling detection and registration method will require studying several detection methods. The approach of this research is to take a broad view of current inventory management technologies and analyze each technology?s strengths and weaknesses for its application in a seedling nursery. The evaluation of each technology?s performance 2Light Detection and Ranging 3Light Detection and Ranging 4Interferometric Synthetic Aperture Radar, aka IFSAR 4 is based on the method speci cations. The method speci cations presented represent the needs of nursery managers and the limitations that exist in the nursery environment. 1.5.1 Method Speci cations During the November 7, 2007 Auburn University-Southern Forest Nursery Manage- ment Cooperative (SFNMC) Advisory Meeting, nursery management from the south east- ern United States were asked to provide speci cations for a proposed pine seedling sensor. The sensor would be incorporated into an inventory management and control system. The desired speci cations are presented in the following subsections. Inventory The sensor should have the ability to inventory seedling populations during various stages of the seedling growth cycle, primarily the mid to late stages. This counting includes seedlings at variable heights and stem diameters. The sensor would need to operate in the nursery?s outdoor environment, which includes but is not limited to precipitation, dust, dirt, and wind. Real-time Harvest Count The sensor will need to count seedlings while harvesting to prevent inventory loss due to the perishability of the product. This system requirement would demand the sensor to operate on-the- y5 through a row of seedlings. A sensor mounted to the front of a tractor or possibly mounted on the harvester was proposed by one nursery manager. He also proposed that a two dimensional plot of seedling?s location be established prior to harvesting, giving the harvester an accurate guide to determine how much product should be harvested. 5typically 1 mile per hour 5 Non-Invasive The sensor system will need to remain non-invasive to protect the quality of the seedlings while in their early stages of development. A non-invasive system is also de- sirable during harvesting to ensure no damage or disorientation is caused to the trees on their way to and through the harvester. Grading Capability The nal system requirement provides for an automatic grading capability giving the user an option to identify trees based on root collar diameter. An option to identify diseased trees was also proposed by Dr. Timothy McDonald. 1.6 Objectives The need for automated pine seedling detection and registration for the better in- ventory management of pine seedling nurseries is the reason for conducting this research. Inventory acquisition techniques currently used by seedling nurseries are in need of mod- ernization. A more e cient inventory process is needed to improve the productivity of pine seedling nurseries. The goals of this research will be reached by accomplishing the following objectives. 1. Review fundamental technologies for modern inventory control and management as they apply to the problem of pine seedling detection and registration. 2. Model and test each technology using a selected detection and registration method. 3. Catalog the strengths and weaknesses of each detection and registration method as it pertains to the future development of automated pine seedling detection and registra- tion. 6 Chapter 2 Current Inventory Management Technologies Thousands of sensors have been developed for general inventory management and are available commercially. Most inventory sensors were developed for use in controlled envi- ronments such as libraries, factories, and warehouses. Each of those inventory sensors are based on central principles from a particular detection technology. For instance, light de- tection has evolved from a single detection diode to Li-DAR and computer vision which are completely di erent types of detection. Five basic detection principles exist for a majority of the inventory management sensors: photo-interrupt, capacitance, microwave, computer vision, and radar. These fundamental sensing techniques will be analyzed for their potential application to the problem of pine seedling detection and registration. 2.1 Photo-Interrupt Sensing A photo-interrupt sensor consists of two components, an emitter diode and a detection diode. The emitter diode sends a beam of light directed to the detector diode or the beam is re ected to the detector diode. The detector diode receives the light-beam and sends an electrical current to a control unit. The current is non-linearly proportional to the amount of light gathered by the detector. The sensor control unit sets a detection threshold which limits the output of the sensor?s controller to two states, detection or no detection of the light beam. The output of the controller is e ectively a binary signal, representing the interruption of the light beam. Photo-interrupt is commonly used in industry due to its simplicity and dependabil- ity. The sensor provides an easy way to count, maintain proper spacing, or ensure size 7 parameters on an assembly line. Gasvoda?s paper [1], describes the photo-interrupt sys- tem developed by MTDC. It uses multiple sensors arranged in a vertical column to detect conifer seedlings. The vertical geometry of the photo-interrupt network helps to distinguish between horizontal limbs and vertical stems. A pine seedlings stem is a good way to count a pine seedling because all pine seedlings have only one central stem. The paper states the main weakness of a photo-interrupt system is false interrupts contributed by the nursery?s environment. 2.2 Capacitive Sensing Capacitive sensors typically consist of two electrodes and a voltage source to generate an electric eld between the two electrodes. The amplitude of the source and geometry of the electrodes determines the size and shape of the generated electric eld. Equation 2.1 describes the relationship between capacitance, dielectric medium, and the electrode?s geometry for a parallel plate capacitor. The capacitance C is proportional to the area of the electrodes A divided by the distance between the electrodes d multiplied by the permittivity of the dielectric medium between the plates. The sensor detects the change in the electric eld when an insulator, also known as a dielectric, passes between the electrodes. The change in the electric eld is dependent on the materials relative static permittivity "r. The measured characteristic of the material is also referred to as its relative dielectric constant.[6] C = "r"0Ad (2.1) Capacitive sensors are typically selected for their robustness and low cost. They are used in industry as position sensors, determining the distance between objects. This prop- erty allows for their use in pressure sensors by determining the distance between a xed position and an elastic diaphragm. Capacitance sensors are also used as liquid level ll sensors and as material detection sensors. Capacitive sensors have previously been used in 8 agriculture as a moisture sensor for grain and other solid ow applications [14]. A capaci- tance sensor could possibly be con gured as a material detection sensor and used to register pine seedlings by detecting their presence between two electrodes. 2.3 Microwave Sensing Microwave sensors typically consist of a controller, a transmitting antenna, and a re- ceiving antenna. The transmitter produces microwaves that are directed through a material of interest and into the receiver. The receiver detects the microwave signal and the controller records the phase shift and attenuation observed between the sent and received signals. The controller uses the phase shift and attenuation as parameters to measure and determine the materials dielectric permittivity. Phase shift and attenuation are dependent upon the per- mittivity of the medium between the transmitting and receiving antennas. Equation 2.21 de nes phase shift and attenuation as the electric displacement eld vector ^D. The electric displacement eld is shown as function of the electric eld intensity ^E produced by the transmitting antenna multiplied by a material?s permittivity ^".[8] ^D = ^"^E (2.2) Microwave sensors are commonly used in industry as moisture detection sensors. Mi- crowave moisture sensors take advantage of water?s molecular dipole structure and the e ect it has on attenuating microwaves. Microwave technology has been applied to inventory man- agement with the development of RFID2 tags. The tags are uniquely coded for each type of inventoried item and are physically attached to that item. The inventory count is modi ed if the item, with attached RFID, passes between a transmitting and detecting antenna. Placing RFID tags on each seedling is not feasible and could possibly be invasive causing damage to the seedlings. A microwave sensor, like the capacitive sensor, could be setup for 1Equation 2.2 is for a linear, homogeneous, and isotropic material. 2Radio Frequency Identi cation 9 material detection of pine seedlings. However the transmitting and receiving antennas must be generously spaced apart for far eld operation. 2.4 Object Registration via Computer Vision Computer vision is the youngest sensing technique and employs the use of a digital camera and light source. The light source is dependent on which spectrum is of interest, because cameras are capable of processing infrared, ultraviolet, and the visual spectrum. The sensor works by capturing the re ected light from the camera?s eld of view and pro- ducing a digital image of the scene. The image must then be processed by a controller to determine if any valuable data is present. Computer vision is capable of object detection, recognition, and tracking. Object de- tection is of interest for its possible applications in inventory management and pine seedling registration. Object detection uses a controller to process an image based on an object iden- ti cation algorithm. The algorithm?s output is processed to detect if the object of interest is present. If the controller detects the object is present, then it may be able to reprocess the image and determine other optical characteristics of the object such as size. Several algorithms for object detection have been developed using computer vision such as edge detection, corner detection, cross correlation, and statistical analysis. The algorithms have possible applications to the problem of pine seedling detection and registration. [12] 2.5 Object Registration via Radar Traditional CW3 radar consists of a controller, a transmitting antenna, and a receiv- ing antenna. The transmitting and receiving antenna are sometimes combined into one structure. The transmitting antenna sends out a signal of a sine wave carrier modulated with narrow rectangular pulses. The signal is directed at the object of interest also known as the target. The distance from the target to the radar is determined by measuring the 3Continuous-wave 10 time it takes for a radar signal to travel to the target and return to the receiver, TR. Since electomagnetic energy propagates at the speed of light c, the range R is R = cTR2 (2.3) Radar is best known for its military defense applications, however it has been used in forestry to map the earth?s topography and also to monitor the condition of our planet?s forests. Researchers have used radar in combination with Li-DAR to develop a more accu- rate map of a forest?s biomass and stand heights. There is also an interest in using radar to spatially map forest stands as mentioned above in the previous work section. Radar has the ability to collect data over a large area very quickly and potentially retrieve spatial information as demonstrated in [5]. 2.6 Summary Five principle sensing technologies have been reviewed for sensing abilities. Of the ve reviewed, four were selected for further research and testing. The four technologies are photo-interrupt detection, capacitive sensing, computer vision, and radar detection. Microwave sensing was omitted due to its similarity to capacitive sensing in its ability to determine the dielectric constant and its limitation to far eld operation. The remain- ing technologies were evaluated through experimentation for their application to the pine seedling detection and registration problem. 11 Chapter 3 Research Procedures 3.1 Introduction A sample set of mature pine seedlings were donated for experimentation by a seedling nursery. The seedlings were transplanted into a large container and spaced to arti cially resemble the row density found in a typical nursery bed. Figure 3.1 depicts pine seedlings in a nursery bed (left) and the density that pine seedlings are grown (right). The arti cial test bed shown in Figure 3.1 was subjected to the selected sensing methods to test for their capabilities and limitations. The following experiments were conducted to provide insight into sensing technologies for a method of pine seedling detection and registration. 3.2 Photo-Interrupt Detection Experiment A Photo-interrupt system is the logical place to start and is the only technology pre- viously tested for pine seedling detection. An experiment was set up to resemble the Figure 3.1: Typical Pine Seedling Nursery Bed [2] 12 Figure 3.2: Arti cial Pine Seedling Test Bed photo-interrupt seedling counter developed by MTDC?s previous research [1]. The Ban- ner QS18VP6LLP retro-re ective laser sensor was selected as the photo-interrupt sensor for this experiment because of its ability to operate with the emitter and detector in close proximity [18]. Another bene t of this retro-re ective sensor is the beam must be re ected back to the sensor?s detector at an angle. The re ection angle provides an increase in real- ized beam height. Like MTDC?s device, the sensor?s emitter, detector, and re ector were mounted to a frame for traveling down a single row of seedlings. The beam width was selected to be 3mm, half the average diameter of the sample set of seedlings. The beam width was chosen to ensure the count of seedlings that may be under grown while screening out small vertical limbs and foliage. A sensor controller was developed from a 16F747 PIC microcontroller. The sensor was debounced with the controller permitting a maximum count rate of 100 seedlings per second. The assembly code for the controller can be found in Appendix A. Unlike MTDC, this experiment does not use a controller to monitor a vertical array of photo-interrupts. 13 The vertical array allowed the system to di erentiate between horizontal obstructions and vertical stems. Instead, the sensor relies on its retro-re ective structure to provide a narrow and tall beam. The nal con guration gives the sensor a realized beam height of one centimeter and is shown in Figure 3.2. Figure 3.3: Photo-Interrupt Experiment Setup 3.3 Capacitive Sensing Experiment Capacitive sensors e ectively measure the dielectric properties of any material that passes into it?s electric eld. Since pine seedlings are non-conductive they are considered insulators or more commonly called dielectrics. According to equation 2.1, a materials dielectric constant is proportional to its capacitance. The dielectric constant of a pine seedling can be determined by measuring the capacitance of two parallel plates with a seedling as the dielectric medium. Placing more dielectric medium between the electrodes 14 will increase the capacitance. A pine seedling count could be determined if capacitance is related to the number of seedlings between two electrodes. An experiment was designed to determine if capacitance is related to the number of seedlings between two electrodes. The size of the electrodes was designed to maximize the seedling stem?s e ect on the electric eld while maintaining a measurable capacitance. The electrodes were mounted to a frame and spaced 10 cm apart. The frame allowed sample seedlings to pass through the electric eld between the electrodes. A custom capacitance measuring circuit was constructed to convert capacitance into a measurable voltage. Care was taken to shield the electrodes, sensing cables, and the custom capacitance measuring circuit. The sensor setup is shown in Figure 3.3. As the electrodes passed down the row of seedlings the sensor measured the capacitance between the electrodes and converted the capacitance into a representative voltage. The sensor?s output was then measured by a digital multimeter. 3.4 Computer Vision Experiment Computer vision uses an object?s color, intensity, and shape for identi cation and reg- istration. These characteristics are known about pine seedlings only in the visible spectrum. An experiment was designed to examine the set of arti cially planted pine seedlings in the infrared, ultraviolet, and visible spectrum?s. The goal is to reveal any potential color or intensity qualities that can be exploited to distinguish pine seedlings in an image. A Sony Handycam camera was used to take digital images of the pine seedling for image processing and analysis. The camera used the sun as a source of light and collected images from the side of one arti cial pine seedling row. Sunlight was the logical source of illumination for the camera because sunlight emits infrared, visible, and ultraviolet light. Also, nurseries grow pine seedlings outdoors and nursery operations are performed during the day. 15 Figure 3.4: Capacitive Experiment 16 3.5 Radar Experiments 3.5.1 Radar Return Signal Measurement Two experiments were developed to test radar?s ability to detect pine seedlings. The rst experiment tested a pine seedling?s measurable radar return signal. The experiment uses a discone antenna as a signal transmitter to propagate a plane wave. The discone antenna design allows it to transmit a vertically polarized wideband signal omnidirectional. The vertical polarization of this antenna design is expected to help couple the propagated wave to the vertical stem of the pine seedling. A signal generator was used to produce a wideband signal and propagate it from the discone antenna to a sample pine seedling in the far eld. The sample seedling will generate a return if it has internal current ow. A Pearson coil1 is used to detect any internal current ow in the seedling and convert the current ow into a readable voltage on an oscilloscope. During the experiment the Pearson coil will only be coupled to the pine seedling allowing the seedling?s return to be isolated from the return of other objects in the room. 3.5.2 Radar Backscatter Measurement The second experiment tests a pine seedling?s signal backscatter characteristics. The concern with using radar for pine seedling detection is radar?s ability to distinguish a pine seedling from foliage. The seedlings measured backscatter will determine if the return is a function of only the seedling or if foliage contributes. The portion of backscatter that returns in the direction from which the transmitted plane wave originates is called the radar return. A radar system will need to detect a return based on each seedling?s stem. The test for radar return requires a software simulation tool to model a pine seedling?s backscatter response to an electromagnetic eld. Backscatter measurements are used to identify a target with a radar system. Several methods of computing radar backscatter have been developed for a variety of target geometries. The geometry of a pine seedling 1A Pearson coil measures the induced voltage from the magnetic eld generated by a change in current through the coil 17 Figure 3.5: Incident Plane Wave Referenced to the Model resembles a monopole antenna with branches and foliage. This analogous geometry of a pine seedling to a wire antenna is exercised when choosing the analysis tool. For this research the antenna analysis software NEC Win-Pro2 is used to simulate a pine seedling model?s backscatter. NEC is typically used to simulate wire antennas but will be used for this research to simulate a wire structure resembling a pine seedling. The transmitted power from a radar is modeled in NEC as an incident plane wave [16]. Although a radar?s signal originates from a localized antenna, it can be modeled as a plane wave in the far eld. NEC uses a polar coordinate system to relate the orientation of the models, the incident plane wave, and the resulting backscatter. The incident plane wave or the transmitted radar signal will originate from a phi angle of zero degrees and a theta angle of ninety degrees as shown in Figure 3.5. The seedling models are suspended in free space and ground parameters are ignored for this experiment. 2Numerical Electromagnetics Code (NEC) uses Method of Moments Analysis (MoM) 18 Model Geometry The model?s geometry is important since a pine seedling?s structure di ers for each species of pine. The Loblolly pine seedling structure was selected for one of the geometries for this experiment. The Loblolly pine seedling ideally consist of a central stem with two branches that project from the body at an acute angle. The branches commonly grow out from each row to maximize the amount of sun light absorbed by the foliage. The optimum structure for a Loblolly pine seedling as pictured left in Figure 3.6. Most of the pine seedlings examined in nursery beds resemble the seedling on the right side of the image in Figure 3.6. The shape of a pine seedling is determined from its genetics and the environment. All pine seedlings have a central stem that runs vertically from the root base to the top of the seedling. This commonality among seedlings will be exploited by vertically polarizing the simulated radar signal to couple with a pine seedling?s stem [17]. Figure 3.6: Ideal Loblolly Pine & Sample Pine Seedling [21] Three pine seedling models were developed in the NEC Win-Pro software (Appendix C). Two of the models mimic the geometry of the optimum Loblolly and the sample seedling shown in Figure 3.6. A third branch free NEC model was also included as a structure similar to a monopole antenna for comparison. The pine seedling?s needles will not be modeled 19 because of their mainly horizontal orientation and their small size. The pine needles are not viewed as major contributors to the vertically polarized backscatter. The branches will need to be modeled because of their similarity to both the stem structure and orientation. All three model geometries are shown in Figure 3.7. Figure 3.7: NEC Win-Pro Wire Models Frequency According to Ikrath [19], trees can be used as monopole antennas for radios. It is known that monopole antennas have a resonant frequency equal to the half wavelength of the antenna. This type of half wavelength resonance is a common principle of operation for monopole antennas. The stem of the pine seedling will then conceivably experience peak resonance when the half wavelength of the radar?s carrier frequency is equal to the height of the pine seedling?s stem. It is proposed that any additional return produced by branches will be negligible due to the peak resonance of the central stem. This experiment will test the concept of using a half wavelength carrier frequency to generate stem dependent radar return. Loading The wire models were loaded under the hypothesis that a loaded antenna is a more accurate model of a pine seedling. They are, by nature, insulators and have been shown to have dynamic resistance and reactance with respect to frequency [7]. Sample pine seedlings have not been measured for their dielectric characteristics. Instead, a broad grouping of 20 loads common to living pine trees is selected based on Ranson [9] and Sugimoto [10]. Ranson measured Siberian and White pine?s dielectric properties at C-band which is approximately 5 GHz. The work from Sugimoto [10] determined dielectric properties of Hinoki wood from 20 Hz to 10 GHz. Both studies used the complex form of electrical permittivity to measure the wood?s dielectric constant. Both studies show the real permittivity "0 to be 2 to 4 times larger than the complex loading "". Equation 3.1 gives the notation for a complex dielectric constant. ^"r = "0r j"r" (3.1) Figure 3.8 graphically illustrates the complex permittivity range of White and Siberian pines recorded in [9]. Five sample dielectric values were selected from the graph to provide a broad range of loadings for each of the pine seedling models. The NEC software tool allows for the seedling models to be loaded as homogeneous structures with each of the ve dielectric sample values for each simulation. 21 Figure 3.8: Sampled Dielectric Loads 22 Chapter 4 Results and Discussion 4.1 Photo-Interrupt Results One of the bene ts of the photo-interrupt sensor?s con guration was its narrow and tall aperture to help distinguish between stems and horizontal branches. However, exper- imentation revealed the photo-interrupt sensor to be unable to distinguish between stems and large vertically positioned branches. Lowering the sensor?s height to the base of the seedling?s stem where the probability of dense foliage was lower caused false interrupts to occur with low branches. Four runs were conducted with the photo-interrupt sensor at dif- ferent heights from the ground. The results of the experiments are shown in Table 4.1 and indicate false interrupts occur at all heights. The false interrupts caused at 8 centimeters are generated mainly by dense foliage. False interrupts occurring below 8 centimeters were generated primarily by thick vertically oriented branches. The sensor performed well when it was positioned at a height of 6 centimeters, because the foliage was not too dense and the lower branches were not too large to cause an interrupt. The conclusions from this experiment coincided with the study performed in [1]. Height (cm) Count Percent Error 2 25 47 4 23 35 6 20 18 8 40 135 Table 4.1: Photo Interrupt Test Results 23 Figure 4.1: Seedling Position and Measured and Modeled Capacitance 4.2 Capacitance Sensor Results The expectations of the capacitive sensor experiment were that a correlation could be detected between capacitance of the two electrodes and the number of seedlings in their electric eld. During experimentation the capacitance of the electrodes changed as indi- vidual seedlings passed between the electrodes verifying that pine seedlings are detectable. The capacitance measurement of an individual seedling was recorded and used to generate a model of the sensors output across the arti cial test-bed. The model was coded with the assumption that the capacitance of a seedling is based on the amount of stem between the electrodes. The sensor?s model and output from the test bed are graphed in Figure 4.2. The model code and the sensor?s output data are located in appendices D and E respectively. The capacitive sensor?s output is at across the arti cial test-bed revealing that two seedlings between the electrodes do not add linearly. Furthermore, the sensor o ered no distinction between seedlings as they consecutively passed through the electrodes. A second experiment was constructed to test the contribution of foliage to the measured capacitance. The results, Table E, revealed the capacitance of a seedling to be a function of foliage not stem thickness. 24 4.3 Computer Vision Results Infrared Spectrum Pine seedlings were shown to be re ectors of infrared light during a nursery visit in the fall of 2008. An image was retrieved from a guidance camera mounted on the rear of a seedling harvester shown in Figure 4.2. The image is rotated clockwise and the pine seedling stems, bright broad lines, appear slanted left. Figure 4.2: Seedling Harvester Infra-Red Guidance Camera The images taken from the experiment are shown in Figure 4.3. They were taken with the Sony Handycam set in IR mode. The images taken with the Handycam match the intensity of the nursery guidance camera image and rea rm that pine seedlings are good re ectors of infrared light. However, all color information was immediately lost when processing images in infra-red. The remaining image was gray scale measure of the intensity of the re ected infrared light. The pine seedling?s intensity was uniform over the entire seedling preventing any distinction between the seedling?s stem and the surrounding foliage. 25 Figure 4.3: Infra-Red Photographs of Arti cially Spaced Pine Seedlings Ultraviolet Spectrum The ultraviolet spectrum has been known to show uorescence in certain plants. The ultraviolet spectrum was analyzed using Wood?s glass as a lens to lter out the visible spectrum from entering the camera?s aperture. The remaining re ected light entering the camera is ultraviolet. Figure 4.4 shows two images taken by the Sony Handycam with a Wood?s glass lens. The images reveal no distinction between the pine seedling stems and their foliage. Figure 4.4: Ultra-Violet Photographs of Arti cially Spaced Pine Seedlings 26 Visible Spectrum In the visible spectrum images are divided into three channels: red, green and blue. Each of these channels is represented by a two-dimensional pixel matrix. Each value in the matrix corresponds to the intensity of the channel?s color at that particular pixel location. The visible spectrum has obvious color advantages for distinguishing between stems and foliage. The stems re ect brown light while the foliage typically re ects green light. Figure 4.5 shows light absorption for chlorophyll, a key color component of seedling foliage. Figure 4.5: Plant Absorption of Light in the Visible Spectrum [14] The visible spectrum experiment was developed to capitalize on the color distinction between stem and foliage in the visible spectrum. A simple statistical algorithm was devel- oped using the light absorption information from Figure 4.5. MATLAB code was developed to run the algorithm and process the image (Appendix F). The algorithm determines pixel columns that contain a higher concentration of the color brown. These columns should correlate with vertically oriented brown seedling stems. Following equation 4.1, each of the channel?s columns, n, were summed and normalized creating a vector for each channel ^Cn. 27 The vectors ^R, ^G, and ^B represent the average color intensity across the image. The three resulting vectors were weighted, based on the light absorption chart, and added together to form equation 4.2. The remaining vector ^D shows where high concentrations of brown pixels existed in the image. ^Cn = Pn 1 Cm;n m (4.1) Figure 4.6 is an image of a pine seedling row from a nursery visit. The red, green, and blue lines plotted across the image represent the vectors ^R, ^G, and ^B respectively. The white line represents the calculated vector ^D. The calculated vector equation is based on the color properties of a seedling stem. The results of the algorithm show the vector peaking at locations where seedlings were present. The peaks were obscured by a dynamic threshold across the entire image. Some of the peaks were also obscured by noise in the image. The noise is introduced by the presence of a white shirt sleeve and foliage in the background. The color white especially a ects the noise oor because white shares equal parts of high intensity color from each channel. The experiment for the visible spectrum considered that the accuracy of the system could be improved by placing a black background behind the seedling row. Figure 4.7 contains four images from the test-bed with the data vector plotted across each image. The simple algorithm indicates that individual seedlings can be detected in the arti cial test-bed. ^D = 2 ^R ^G+ 2 ^B (4.2) 28 Figure 4.7: Series of Images with Data Vector Figure 4.6: Visible Spectrum Analysis 29 4.4 Radar Results Radar Return Results Figure 4.8 shows the results of the rst radar experiment from a Tektronics oscilloscope display. The top graph is a control and shows the noise oor. The control was measured from the Pearson coil when it was de-coupled from the seedling. The bottom graph is the signal generated through the seedling by the propagating wideband signal. The signal was measured from the Pearson coil coupled around the base of the seedling. Both measurements were made during a transmission of a wideband pulse. The signal generated by the seedling is shown to be well above the noise oor and is considered detectable by radar. Radar Backscatter Measurement The pine seedling models are simulated with a carrier frequency wavelength of 0:14m< f < 3m, where the pine seedling stem height corresponds to a wavelength P = 0:3m and is the average height of a pine seedling stem. Sweeping the simulation frequency will make any resonance obvious when analyzing the radar?s return. The wavelength is normalized = P f , also making the pine seedling wavelength with respect to the frequency wavelength more easily observable. The stem is expected to resonate at the half wavelength frequency and generate a similar return for all three models. A similar radar return for each model will enable the branch geometry to be neglected, making individual pine seedlings realizable by a radar system. 4.4.1 Monopole Tree Model Figure 4.9 shows the magnitude of the backscatter spreading away from the monopole seedling model at the half wavelength frequency. The plot veri es the expected uniform backscatter for a monopole antenna. Figure 4.10 is a magnitude plot of the returned backscatter with respect to frequency and loading of the monopole seedling model. The plot represents the radar return from each 30 Figure 4.8: Voltage Waveform Induced from the Current Flow through a Pine Seedling due to the presence of an Electromagnetic Field 31 Figure 4.9: Radiation Pattern of the Monopole Tree Model of the ve dielectric loadings. The results show an exponentially decaying increase in the re-radiated power. The expected spike or increase at the resonant frequency is not present for any of the model?s dielectric loads. 32 Figure 4.10: Monopole Antenna Reradiated Power 4.4.2 Ideal Loblolly Model The Loblolly model adds two branches to the previously simulated monopole model. The branches are oriented almost completely on the horizontal axis and are expected to contribute little to the return. Figure 4.11 is a polar plot of the Loblolly model?s backscatter at the half wavelength frequency. The backscatter plot shows a slight manipulation to the uniform return observed by the monopole model. Figure 4.12 plots the returned backscatter with respect to frequency and dielectric loading of the Loblolly model. The results do not show resonance at the half wavelength frequency, but the magnitudes almost perfectly match the monopole seedling model. 33 Figure 4.11: Radiation Pattern of Ideal Loblolly Model Figure 4.12: Loblolly Antenna Re-radiated Power 34 4.4.3 Sample Seedling Model The sample seedling model adds several branches to the monopole model. The branches are positioned more vertically than the previous Loblolly model. The return of the seedling model?s stem is expected to be the main contributor to the backscatter and the branches are expected to be negligible. Figure 4.13 is the radar backscatter of the sample seedling model at the half wavelength frequency. The magnitude of the backscatter is almost uniform like the Loblolly and monopoles models, but the magnitude is greater than both models by about 4:5dB. Figure 4.13: Radiation Pattern of the Sample Seedling Model Figure 4.14 con rms that there is an increase in the amplitude across all measured frequencies. The sample seedling model?s return averages to about 3-4 decibels higher when compared to the monopole and Loblolly models over the same frequency span. 35 Figure 4.14: Sample Seedling Antenna Re-radiated Power The results of the models demonstrate that vertically positioned limbs will signi cantly contribute to radar return. The seedling models did not show any resonance at the half wavelength frequency. The results demonstrate that individual seedlings cannot be deter- mined from radar return at the half wavelength frequency of the seedling?s stem. 36 Chapter 5 Summary and Conclusions 5.1 Introduction The goal of this research was to propose a pine seedling detection and registration method for the development of an auto-indexing pine seedling counter. Previous research and literature on the problem was initially reviewed. Two prior studies were found, a patent and a study preformed by the USDA forestry service. The patent and study both used photo- interrupts as the method for detection. This thesis identi ed four sensing technologies and determined their strengths and weaknesses as a pine seedling detection and registration method. Each technology has been tested for its ability to detect a pine seedling. The results indicated photo-interrupt and capacitive sensing to be poor choices for accu- rate pine seedling detection. The photo-interrupt system cannot account for false interrupts that can occur in a nursery?s environment. Thick foliage and vertically positioned branches were a couple of the nursery conditions identi ed as a primary causes for false interrupts. The capacitive sensor measures a pine seedling?s dielectric constant which is a function of foliage. The foliage of a seedling varies signi cantly making detection of individual seedlings in a row impossible. Computer vision is a better choice for pine seedling identi cation. The results of the computer vision experiment demonstrated an accurate indication of a pine seedling?s location. However, a computer vision system would experience problems in the tight spaces between rows in a nursery bed. In addition most of the natural sunlight will be blocked by foliage between rows making color distinction very di cult. The nursery?s envi- ronment would also be harsh on a camera sensor?s lens along with the high computational cost that is associated with the large amounts of data a camera collects. A camera system has to detect and register each pine seeding individually in a two dimensional data set. This 37 system could create an immense processing cost in a nursery consisting of millions of pine seedlings. CW radar was shown to generate a radar return, however simulation did not reveal a resonant frequency for the modeled seedling. CW radar has the potential ability to map a large area of seedlings very quickly. CW radar can also potentially retrieve seedling properties that are not optically detectable such as moisture content which is proportional to the seedling?s permittivity and mass. Strengths Weaknesses Photo-Interrupt Simple and dependable Easy to count and ensure spacing Maintain size parameters Subject to nursery environment Cannot distinguish stems from thick foliage or vertical limbs Capacitive Robust and low cost Poor resolution between seedlings Foliage dependent instead of stem Computer Vision Object recognition and tracking Color di erentiation between stem and foliage High computational cost Tracking in foliage and nursery environment Needs arti cial lighting Radar Collect data over a large area High computational cost High frequency operation for good resolution Table 5.1: Sensor Strengths and Weaknesses Objective 1 This research identi ed ve major fundamental sensing methods used in modern in- ventory control and management. The sensing methods were analyzed and adapted for the detection of pine seedlings. The methods were applied to the problem of pine seedling de- tection and registration based on their detection capabilities. The method for pine seedling detection was then suggested as experiment parameters. 38 Photo-Interrupt Capacitive Computer Vision CW Radar Impervious to nursery environment no yes no yes Inventory during various stages of seedling growth cycle yes no yes no On-the- y operation yes yes yes yes Harvester mountable yes yes yes no Determine seedlings location (2-D plot) yes no yes - None Invasive yes yes yes yes Grading capability (RCD) - no yes - Identify diseased seedlings no no yes - Table 5.2: System Requirements vs. Sensor Capabilities Objective 2 Four of the ve fundamental sensing methods were selected for further testing. The technologies were developed into an experiment to test their ability to detect pine seedlings. The experiments used an arti cial pine seedling nursery row as a test bed. The technologies were modeled and analyzed in software when physical experiments were not feasible to test. Objective 3 Each detection and registration method was analyzed as it pertained to the future de- velopment of automated pine seedling detection and registration. The sensing technology?s strengths and weaknesses were cataloged in Table 5.1. The sensing methods were then compared to the ideal sensing system in Table 5.2. 5.2 Future Work The ultimate goal for pine seedling detection and registration is to develop a high res- olution detection and mapping system for pine seedlings in the nursery bed. The resolution of the system would need to be accurate down to one centimeter for mapping purposes and 39 down to one millimeter for sizing and grading. Computer vision meets the most require- ments for pine seedling detection and registration system and is logical choice for further development. Ultra-wideband (UWB) radar is a relatively new technology and was not tested in this thesis. However UWB could conceivably be used to image the stem and branch structure of a group of seedlings using a computed tomography approach. The system would be a millimeter wave version of the InSAR system developed in [5]. The development of an ultra-wideband system would require a more realistic model. For high frequency analysis the foliage would need to be included along with accurate dielectric measurements for di erent species of pine seedlings. The use of L-system and OpenGL is suggested to grow pine seedlings of a particular species virtually and model them three-dimensionally. The ability to accurately grow a species of pine seedling in software will enable the radar simulation of an entire nursery bed. The radar could then be extensively tested prior to prototyping. 40 Bibliography [1] Gasvoda, Dave and Herzberg, Diane. \Seedling Counter Field Tests". Tree Planters? Notes 44(1):8-12; 1993. [2] Nix, Steve. \Nursery Bed of Pine Seedlings". About.com. June 17, 2009. [3] Rath, Karl Friedrich. \Apparatus for Harvesting and Bundling Plants". United States Patent 4037666. July 26, 1977 [4] Weyerhaeuser.com. April 23, 2009. [5] D. W. Liu, Y. Du, G. Q. Sun, W. Z. Yan, and B.-I. Wu. "Analysis of InSAR Sensitivity to Forest Structure Based on Radar Scattering Model." Progress In Electromagnetics Research, PIER 84, 149{171, 2008 [6] Baxtor, Larry. Capacitive Sensors: Design and Applications. John Wiley and Sons. 1996 [7] Yamamoto, Y.; Harada, H.; Yasuhara, K.; Nakamura, T.; Instrumentation and Mea- surement, IEEE Transactions on Volume 44, Issue 3, June 1995 Page(s):729 - 732 [8] Trabelsi, S.; Nelson, S.O.; Ramahi, O.; Microwave Conference, 2006. 36th European 10-15 Sept. 2006 Page(s):447 - 450 [9] Ranson, K.J.; Rock, B.N.; Salas, W.A.; Smith, K.; Williams, D.L.; Geoscience and Remote Sensing Symposium, 1992. IGARSS ?92. International Volume 2, 1992 Page(s):1283 - 1285 [10] Sugimoto, Hiroyuki.; Takazawa, Ryosuke.; Norimoto, Misato.; Dielectric Relaxation Due to Heterogeneous Structure in Moist Wood. Journal of Wood Science, Volume 51, Number 6 / December, 2005 [12] Rafael C. Gonzalez and Richard E. Woods. Digital Image Processing Third Edition. Upper Saddle River, New Jersey: Pearson Prentice Hall, 2008 [14] "absorption-spectrum". wikispaces.com. June 20, 2009. [14] Kumhala, Frantisek.; Prosek, Vaclav.; Kroulik, Milan.; Kviz, Zdenek.; Parallel Plate Mass Flow Sensor for Forage Crops and Sugar Beet. ASABE Annual International Meeting. 2008 Providence, Rhode Island, June 29 - July 2, 2008 084700. 41 [15] Currie, Nicholas C. Techniques of Radar Re ectivity Measurement. Dedham, MA. Artech House, Inc. 1984. [16] Nittany Scienti c, Inc. NEC-Win Pro User?s Manual. Riverton, UT. 1997 [17] Skolnik, Merrill I. Introduction to Radar Systems. New York. 1980 [18] Banner Sensors website. July 3, 2009. [19] Ikrath, K.; Kennebeck, W.; Hoverter, R.; Antennas and Propagation, IEEE Transac- tions on. Volume 23, Issue 1, Jan 1975 Page(s):137-140 [20] Boyer, James N. and South, David B.. \Loblolly Pine Seedling Morphology and Pro- duction at 53 Southern Forest Nurseries". Tree Planters? Notes 39(3):13-16; 1988. [21] South, David B.;VanderSchaaf, Curtis L.; \RCDlob". 42 Appendices 43 Appendix A Photo-Interrupt Microcontroller Schematic Figure A.1: Photo-Interrupt Microcontroller Schematic 44 Appendix B Photo-Interrupt Microcontroller Code ;********************************************************************** ; This le is assembly code for eight binary sensors. Each I/O pin ; contribute to the over all count. There is approximately 0.5 ms ; between the checking of each sensors status. Each sensor must ; must signal a detection and a non-detection to be counted. This ; limits the number of detections made by one to sensor to 100 per ; minute. ;********************************************************************** ; Filename: TreeCounter.asm ; Date: December 2007 ; File Version: 1.0 ; Author: Je Hunt ; Company: Auburn University ;********************************************************************** ; Files required: p16F747.inc ;********************************************************************** ; Notes: This software was developed for the detection of an eight ; sensor network that could count eight rows of seedlings at once. ;********************************************************************** #include ; Processor Include le, for standard names CBLOCK 0x20 ASCII TEMP FLAG W TEMP STATUS TEMP SENSORS SENSORSF SENSOR 0 SENSOR 1 SENSOR 2 45 SENSOR 3 SENSOR 4 SENSOR 5 SENSOR 6 SENSOR 7 ONES TENS HUND THOU TTHO HTHO MILL TMIL CARR ENDC ;Start Code ORG 0 ; Start of code (location 0) GOTO SETUP ORG 0X04 GOTO ISR SETUP BANKSEL OSCCON MOVLW B?01101000? MOVWF OSCCON BANKSEL ADCON1 MOVLW 0x0F ;set all A/D pins to Digital I/O MOVWF ADCON1 BANKSEL TRISC ; BANK 1 MOVLW B?11111111? ; RC7-RC0 are inputs for PORTC MOVWF TRISC CLRF TRISB ; all PORTB pins con gured for output mode CLRF TRISA ; all output ;*********************************MAIN********************************* MAIN 46 CALL LCD SETUP CLRF ASCII CLRF TEMP CLRF FLAG CLRF MSD CLRF LSD CLRF SENSORSF CLRF CARR MOVLW 0X30 MOVWF ONES MOVWF TENS MOVWF HUND MOVWF THOU MOVWF TTHO MOVWF HTHO MOVWF MILL MOVWF TMIL MAIN LOOP ; CALL CHECK SENSORS ; CALL UPDATE DISPLAY GOTO MAIN LOOP CHECK SENSORS BANKSEL PORTC MOVF PORTC,W MOVWF SENSORS BTFSS SENSORS,0 GOTO $+3 BCF SENSORSF,0 CLRF SENSOR 0 BTFSC SENSORSF,0 GOTO $+6 INCF SENSOR 0,f BTFSS SENSOR 0,3 GOTO $+3 47 BSF SENSORSF,0 GOTO UPDATE COUNT BTFSS SENSORS,1 GOTO $+3 BCF SENSORSF,1 CLRF SENSOR 1 BTFSC SENSORSF,1 GOTO $+6 INCF SENSOR 1,f BTFSS SENSOR 1,3 GOTO $+3 BSF SENSORSF,1 GOTO UPDATE COUNT BTFSS SENSORS,2 GOTO $+3 BCF SENSORSF,2 CLRF SENSOR 2 BTFSC SENSORSF,2 GOTO $+6 INCF SENSOR 2,f BTFSS SENSOR 2,3 GOTO $+3 BSF SENSORSF,2 GOTO UPDATE COUNT BTFSS SENSORS,3 GOTO $+3 BCF SENSORSF,3 CLRF SENSOR 3 BTFSC SENSORSF,3 GOTO $+6 INCF SENSOR 3,f BTFSS SENSOR 3,3 GOTO $+3 BSF SENSORSF,3 48 GOTO UPDATE COUNT BTFSS SENSORS,4 GOTO $+3 BCF SENSORSF,4 CLRF SENSOR 4 BTFSC SENSORSF,4 GOTO $+6 INCF SENSOR 4,f BTFSS SENSOR 4,3 GOTO $+3 BSF SENSORSF,4 GOTO UPDATE COUNT BTFSS SENSORS,5 GOTO $+3 BCF SENSORSF,5 CLRF SENSOR 5 BTFSC SENSORSF,5 GOTO $+6 INCF SENSOR 5,f BTFSS SENSOR 5,3 GOTO $+3 BSF SENSORSF,5 GOTO UPDATE COUNT BTFSS SENSORS,6 GOTO $+3 BCF SENSORSF,6 CLRF SENSOR 6 BTFSC SENSORSF,6 GOTO $+6 INCF SENSOR 6,f BTFSS SENSOR 6,3 GOTO $+3 BSF SENSORSF,6 GOTO UPDATE COUNT 49 BTFSS SENSORS,7 GOTO $+3 BCF SENSORSF,7 CLRF SENSOR 7 BTFSC SENSORSF,7 GOTO $+6 INCF SENSOR 7,f BTFSS SENSOR 7,3 GOTO $+3 BSF SENSORSF,7 GOTO UPDATE COUNT RETURN UPDATE COUNT MOVLW 0x06 MOVWF CARR INCF ONES,f MOVF ONES,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF ONES MOVLW 0x06 MOVWF CARR INCF TENS,f MOVF TENS,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF TENS MOVLW 0x06 MOVWF CARR INCF HUND,f 50 MOVF HUND,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF HUND MOVLW 0x06 MOVWF CARR INCF THOU,f MOVF THOU,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF THOU MOVLW 0x06 MOVWF CARR INCF TTHO,f MOVF TTHO,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF TTHO MOVLW 0x06 MOVWF CARR INCF HTHO,f MOVF HTHO,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF HTHO MOVLW 0x06 51 MOVWF CARR INCF MILL,f MOVF MILL,0 ADDWF CARR,1 BTFSS CARR,6 RETURN MOVLW 0X30 MOVWF MILL MOVLW 0x06 MOVWF CARR INCF TMIL,f MOVF TMIL,0 ADDWF CARR,1 BTFSS CARR,6 RETURN GOTO MESSAGE 1 UPDATE DISPLAY MOVLW 0X40 CALL LCD COMMAND MOVF TMIL,0 CALL DISPLAY MOVF MILL,0 CALL DISPLAY MOVF HTHO,0 CALL DISPLAY MOVF TTHO,0 CALL DISPLAY MOVF THOU,0 CALL DISPLAY MOVF HUND,0 CALL DISPLAY MOVF TENS,0 CALL DISPLAY MOVF ONES,0 52 CALL DISPLAY RETURN MESSAGE 1 RETURN ;******************************END MAIN***************************** ;********************************************************************** ;*********************LCD DISPLAY SUBROUTINES******************** ;********************************************************************** DISPLAY CALL LCDWAIT MOVWF PORTB BSF PORTA,0 BCF PORTA,1 ;R/W is cleared. BSF PORTA,2 NOP NOP BCF PORTA,2 NOP NOP BCF PORTA,0 RETURN ;LCD WAIT ROUTINE IT POLES THE MSB OF THE LCD LCDWAIT BANKSEL TRISB BSF TRISB,7 BANKSEL PORTA BSF PORTA,1 BCF PORTA,0 NOP BSF PORTA,2 BSF PORTA,3 CLRWDT BTFSC PORTB,7 53 GOTO $-2 BCF PORTA,3 BCF PORTA,2 BCF PORTA,2 BCF PORTA,1 BANKSEL TRISB BCF TRISB,7 BANKSEL PORTA RETURN ;END OF WAIT ROUTINE ;*************************LCD SETUP ROUTINES*********************** ;********************************************************************** LCD SETUP BANKSEL PORTA BCF PORTA,0 ;CLEARS OUTPUTS TO LCD BCF PORTA,1 BCF PORTA,2 SLEEP MOVLW 0X38 MOVWF PORTB BSF PORTA,2 NOP NOP NOP NOP BCF PORTA,2 NOP NOP NOP NOP MOVLW 0X38 MOVWF PORTB BSF PORTA,2 NOP 54 NOP NOP NOP BCF PORTA,2 CALL WAIT MOVLW 0X38 MOVWF PORTB BSF PORTA,2 NOP NOP NOP NOP BCF PORTA,2 ;INTIAL SETUP IS COMPLETE THE LCD IS SET FOR 8 BITS AND MAY BE POLLED MOVLW 0X0C ;DISPLAY IS ON CURSOR IS OFF CALL WAIT CALL LCD COMMAND MOVLW 0X01 ;CLEARS DISPLAY CALL LCD COMMAND MOVLW 0X06 ;ENTRY MODE CURSOR MOVES LEFT TO RIGHT NO SCREEN SHIFT CALL LCD COMMAND MOVLW 0X80 ;RETURNS CURSOR TO HOME (TOP LEFT) CALL LCD COMMAND RETURN ;SETUP OF THE LCD PANNEL IS COMPLETE ;GENERIC SUBROUTINE THAT SENDS A COMMAND BYTE TO THE LCD LCD COMMAND ;ACTUALLY SENDS COMMAND LINE TO LCD CALL LCDWAIT ;SETUP BYTE IS IN W BANKSEL PORTB MOVWF PORTB ;OUTPUT TO B BCF PORTA,0 NOP 55 BSF PORTA,2 ;SEND INFO NOP NOP ;WAIT NOP NOP BCF PORTA,2 ;STOP BYTE RETURN ;*************************END OF COMMAND************************* ;********************************************************************** ;Displays WAR EAGLE WAR EAGLE BANKSEL PORTA BSF PORTA,3 BANKSEL 0 MOVLW 0X57 CALL DISPLAY MOVLW 0X41 CALL DISPLAY MOVLW 0X52 CALL DISPLAY MOVLW 0x20 CALL DISPLAY MOVLW 0X45 CALL DISPLAY MOVLW 0X41 CALL DISPLAY MOVLW 0X47 CALL DISPLAY MOVLW 0X4C CALL DISPLAY MOVLW 0X45 CALL DISPLAY RETURN WAIT 56 BANKSEL 0 BSF T1CON,5 BCF T1CON,4 CLRF TMR1L CLRF TMR1H BCF PIR1,0 BSF T1CON,0 BTFSS PIR1,0 GOTO $-2 BCF T1CON,0 RETURN ISR RETFIE END 57 Appendix C Capacitive Sensor Schematic Figure C.1: Capacitive Sensor (Cx represents electrodes) 58 Appendix D Capacitive Sensor MATLAB Code capsens.m % Seedling stem location (1) Seedling stem present (0) Seedling stem absent T=[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0,0... ,0,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1... ,1,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,0,0,0,0... ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0,0... ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0... ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1... ,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,0,0,0,0... ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1... ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0... ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0... ,0,0]; % mask represents electrodes with a width of 50 mm mask=[ones(1,50),zeros(1,410)]; mask=circshift(mask,[0 -25]); % capacitance of electrodes were measured for 20 Seedlings seedling C=[35,35.1,34.8,35.4,35.1,35.4,35.6,35.3,35.1,35,35.2,35.4,35.7,35.3,35.1,35,35.2,35... ,35.1,34.8]; % Base capacitance of electrodes baseC=32.1; % Root Collar Diameter of 17 Seedlings Diameter=[3,6,5,7,5,3,4,4,7,3,5,4,5,3,4,7,5]; % Average capacitance per mm of pine seedling stem avgC=mean(seedlingC-baseC)/mean(Diameter); % model generated by shifting the mask (electrodes) over the seedling positions for i=1:460; model(1,i)=sum(mask.*T); mask=circshift(mask,[0 1]); 59 end % model of over lapping electrode multipled by the average capacitance per % mm of seedling plus the base capacitance model=(model*avgC)+baseC; C=[32.1,32.1,32.1,32.1,32.1,32.1,32.1,32.1,32.1,32.1,32.3,32.3,32.4,32.6,32.8,32.8... ,32.9,33.1,33.2,33.2,33.3,33.4,33.4,33.5,33.5,33.7,33.8,33.9,33.9,34,34.3,34.3,34.4... ,34.4,34.5,34.5,34.7,34.6,34.7,34.8,35,35.1,35.2,35.4,35.6,35.7,35.8,36,36.2,36.3... ,36.4,36.5,36.5,36.6,36.7,36.8,36.8,36.9,36.9,36.9,37,37.1,37.1,37.3,37.3,37.4,37.5... ,37.5,37.5,37.6,37.6,37.6,37.6,37.7,37.7,37.8,37.8,37.9,37.9,38,38,38.1,38.1,38.1... ,38,38,38.1,38,38,37.9,37.9,37.9,38,38,38,38.1,38.1,38.1,38.2,38.1,38.1,38.1,38.1... ,38,38,38,38.1,38.1,38.1,38,38,37.9,37.9,37.8,37.9,37.9,37.9,38.1,38.1,38,38,38,... 38,38.1,38.1,38.1,38.1,38.1,38.1,38,38,38,38,37.9,37.9,37.9,38,37.9,37.9,37.9... ,37.9,38.1,38.1,38,38,38,38,38.1,38.1,38,38,38,38.1,38.1,38.1,38,38,38,38,37.9... ,37.9,37.9,37.9,37.9,37.8,37.8,37.9,37.8,37.9,37.9,37.9,37.8,37.8,37.9,37.8,37.8... ,37.9,37.9,37.9,38,38,38.1,38.1,38.1,38,38,38.1,38.1,38.1,38,38,38.1,38.1,38.1... ,38.2,38.1,38.1,38.1,38.1,38,38,38,38.1,38,37.9,38,38,38,38,37.9,37.9,38,38,38... ,38,38,37.9,38,38,38,38,37.9,37.9,37.9,38,38,38,37.9,38,37.9,38,38,38,38,38,38... ,38,38,38,38.1,38.1,38,38,38,38,37.9,37.9,37.9,37.8,37.9,37.9,37.9,38,38,38,38... ,38,38,38,37.9,37.9,37.9,37.9,38,38,37.9,37.9,38,37.9,37.9,37.8,37.8,37.8,37.9... ,37.8,37.9,37.8,37.9,37.9,37.9,37.9,37.9,38,38,37.9,37.9,38,38,38,38,37.9,37.9... ,37.9,38,38,38,38,38.1,38.1,38.1,37.9,37.9,38,38,37.9,38,38,38,38,38,38,38,38.1... ,38,38,37.9,37.9,37.9,38,37.9,37.9,38.1,38.1,38.1,38,38.1,38.1,38.1,38,38.1,38.1... ,38,38,37.9,38,38,38.1,38.1,38.1,38.1,38.2,38.2,38.2,38.1,38,38,38,37.9,37.9... ,37.9,37.9,38,38,38,38.1,38.1,38,38,38,38,37.9,37.9,37.9,38,38,37.9,37.8,37.8... ,37.8,37.8,37.7,37.6,37.7,37.8,37.7,37.7,37.6,37.6,37.4,37.6,37.6,37.7,37.6,37.6... ,37.6,37.7,37.7,37.7,37.8,37.7,37.6,37.4,37.7,37.6,37.6,37.6,37.5,37.4,37.3,37.3... ,37.1,37.1,37,36.9,36.9,36.8,36.7,36.6,36.6,36.5,36.4,36.3,36.3,36.2,36.1,36.1... ,36,35.7,35.7,35.4,35.2,35,34.9,34.8,34.6,34.5,34.4,34.2,34.2,34.1,34.1,34,33.9... ,33.9,33.8,33.8,33.7,33.7,33.6,33.5,33.3,33.2,33.1,33,33,32.9,32.8,32.7,32.6... ,32.6,32.5,32.4,32.4,32.3,32.2,32.2,32.1,32.2,32.2,32.2]; % Error calculation SSerr=sum((C-model).^2); SStot=sum((C-mean(C)).^2); R=1-(SSerr/SStot); 60 Appendix E Capacitive Sensor Collected Data 61 Seedlings with Foliage (pF) Seedlings without Foliage (pF) 35.0 32.3 35.1 32.1 34.8 31.8 35.4 31.9 35.1 32.3 35.4 32.3 35.6 32.9 35.3 32.6 35.1 32.5 35.0 32.0 35.2 32.5 35.4 32.3 35.7 32.1 35.3 31.8 35.1 31.8 35.0 32.3 35.2 32.3 35.0 32.4 35.1 32.6 34.8 32.5 Table E.1: Capacitance of Seedlings with and without Foliage Position (mm) Root Collar Diameter (mm) 33 3 49 6 81 5 104 7 130 5 154 3 175 4 202 4 217 7 248 3 273 5 292 4 314 5 340 3 365 4 400 7 432 5 Table E.2: Seedling Position and Root Collar Diameter 62 Distance (mm) Capacitance (pF) Model (pF) 1 32.1 32.1 2 32.1 32.1 3 32.1 32.1 4 32.1 32.1 5 32.1 32.1 6 32.1 32.1 7 32.1 32.1 8 32.1 32.8 9 32.1 33.4 10 32.1 34.1 11 32.3 34.1 12 32.3 34.1 13 32.4 34.1 14 32.6 34.1 15 32.8 34.1 16 32.8 34.1 17 32.9 34.1 18 33.1 34.1 19 33.2 34.1 20 33.2 34.1 21 33.3 34.1 22 33.4 34.7 23 33.4 35.4 24 33.5 36.0 25 33.5 36.7 26 33.7 37.3 27 33.8 38.0 28 33.9 38.0 29 33.9 38.0 30 34.0 38.0 31 34.3 38.0 32 34.3 38.0 33 34.4 38.0 34 34.4 38.0 35 34.5 38.0 36 34.5 38.0 37 34.7 38.0 38 34.6 38.0 39 34.7 38.0 40 34.8 38.0 41 35.0 38.0 63 42 35.1 38.0 43 35.2 38.0 44 35.4 38.0 45 35.6 38.0 46 35.7 38.0 47 35.8 38.0 48 36 .0 38.0 49 36.2 38.0 50 36.3 38.0 51 36.4 38.0 52 36.5 38.0 53 36.5 38.0 54 36.6 38.0 55 36.7 38.0 56 36.8 38.0 57 36.8 38.0 58 36.9 37.3 59 36.9 36.7 60 36.9 36.0 61 37 .0 36.0 62 37.1 36.0 63 37.1 36.7 64 37.3 37.3 65 37.3 38.0 66 37.4 38.6 67 37.5 39.3 68 37.5 39.3 69 37.5 39.3 70 37.6 39.3 71 37.6 39.3 72 37.6 38.6 73 37.6 38.0 74 37.7 37.3 75 37.7 36.7 76 37.8 36.0 77 37.8 36.0 78 37.9 36.7 79 37.9 37.3 80 38 .0 38.0 81 38 .0 38.6 82 38.1 39.3 83 38.1 40.0 84 38.1 40.0 64 85 38 .0 40.0 86 38 .0 40.0 87 38.1 40.0 88 38 .0 40.0 89 38 .0 40.0 90 37.9 40.0 91 37.9 40.0 92 37.9 40.0 93 38.0 40.0 94 38.0 40.0 95 38.0 40.0 96 38.1 40.0 97 38.1 40.0 98 38.1 40.0 99 38.2 40.0 100 38.1 40.0 101 38.1 40.0 102 38.1 40.0 103 38.1 40.0 104 38.0 40.6 105 38.0 41.3 106 38.0 41.9 107 38.1 42.6 108 38.1 43.2 109 38.1 43.2 110 38.0 43.2 111 38.0 43.2 112 37.9 43.2 113 37.9 42.6 114 37.8 41.9 115 37.9 41.3 116 37.9 40.6 117 37.9 40.0 118 38.1 40.0 119 38.1 40.0 120 38.0 40.0 121 38.0 40.0 122 38.0 40.0 123 38.0 40.0 124 38.1 40.0 125 38.1 40.0 126 38.1 40.0 127 38.1 39.3 65 128 38.1 38.6 129 38.1 38.6 130 38.0 38.6 131 38.0 38.6 132 38.0 38.0 133 38.0 37.3 134 37.9 37.3 135 37.9 37.3 136 37.9 37.3 137 38.0 37.3 138 37.9 37.3 139 37.9 37.3 140 37.9 37.3 141 37.9 37.3 142 38.1 37.3 143 38.1 37.3 144 38.0 37.3 145 38.0 37.3 146 38.0 37.3 147 38.0 37.3 148 38.1 37.3 149 38.1 38.0 150 38.0 38.6 151 38.0 39.3 152 38.0 40.0 153 38.1 40.0 154 38.1 39.3 155 38.1 38.6 156 38.0 38.0 157 38.0 37.3 158 38.0 36.7 159 38.0 36.7 160 37.9 36.7 161 37.9 36.7 162 37.9 36.7 163 37.9 36.7 164 37.9 36.7 165 37.8 36.7 166 37.8 36.7 167 37.9 36.7 168 37.9 36.7 169 37.9 36.7 170 37.9 36.7 66 171 37.9 36.7 172 37.8 36.7 173 37.8 36.7 174 37.9 36.7 175 37.8 36.7 176 37.8 37.3 177 37.9 38.0 178 37.9 38.6 179 37.9 38.6 180 38.0 38.0 181 38.0 37.3 182 38.1 37.3 183 38.1 37.3 184 38.1 37.3 185 38.0 37.3 186 38.0 37.3 187 38.1 37.3 188 38.1 37.3 189 38.1 37.3 190 38.0 38.0 191 38.0 38.6 192 38.1 39.3 193 38.1 40.0 194 38.1 40.6 195 38.2 41.3 196 38.1 41.9 197 38.1 41.9 198 38.1 41.9 199 38.1 41.3 200 38.0 40.6 201 38.0 40.0 202 38.0 39.3 203 38.1 39.3 204 38.0 39.3 205 37.9 39.3 206 38.0 39.3 207 38.0 39.3 208 38.0 39.3 209 38.0 39.3 210 37.9 39.3 211 37.9 39.3 212 38.0 39.3 213 38.0 39.3 67 214 38.0 39.3 215 38.0 39.3 216 38.0 39.3 217 37.9 39.3 218 38.0 39.3 219 38.0 39.3 220 38.0 39.3 221 38.0 39.3 222 37.9 39.3 223 37.9 40.0 224 37.9 40.6 225 38.0 41.3 226 38.0 40.6 227 38.0 40.0 228 37.9 39.3 229 38.0 38.6 230 37.9 38.6 231 38.0 38.6 232 38.0 38.6 233 38.0 38.6 234 38.0 38.6 235 38.0 38.6 236 38.0 38.6 237 38.0 38.6 238 38.0 38.6 239 38.0 38.6 240 38.1 38.0 241 38.1 37.3 242 38.0 36.7 243 38.0 36.0 244 38.0 35.4 245 38.0 34.7 246 37.9 34.1 247 37.9 34.7 248 37.9 35.4 249 37.8 36.0 250 37.9 36.7 251 37.9 37.3 252 37.9 37.3 253 38.0 37.3 254 38.0 37.3 255 38.0 37.3 256 38.0 37.3 68 257 38.0 37.3 258 38.0 37.3 259 38.0 37.3 260 37.9 37.3 261 37.9 37.3 262 37.9 37.3 263 37.9 37.3 264 38.0 37.3 265 38.0 37.3 266 37.9 38.0 267 37.9 38.6 268 38.0 39.3 269 37.9 40.0 270 37.9 40.0 271 37.8 40.0 272 37.8 40.0 273 37.8 39.3 274 37.9 38.6 275 37.8 38.0 276 37.9 38.0 277 37.8 38.0 278 37.9 38.0 279 37.9 38.0 280 37.9 38.0 281 37.9 38.0 282 37.9 38.0 283 38.0 38.0 284 38.0 38.0 285 37.9 38.0 286 37.9 38.0 287 38.0 38.0 288 38.0 38.6 289 38.0 39.3 290 38.0 40.0 291 37.9 40.6 292 37.9 41.3 293 37.9 41.3 294 38.0 41.3 295 38.0 41.3 296 38.0 41.3 297 38.0 40.6 298 38.1 40.0 299 38.1 39.3 69 300 38.1 38.6 301 37.9 38.0 302 37.9 38.0 303 38.0 38.0 304 38.0 38.0 305 37.9 38.0 306 38.0 38.0 307 38.0 38.0 308 38.0 38.0 309 38.0 38.0 310 38.0 38.0 311 38.0 38.0 312 38.0 38.0 313 38.1 38.0 314 38.0 38.0 315 38.0 38.6 316 37.9 38.6 317 37.9 38.6 318 37.9 38.0 319 38.0 38.0 320 37.9 37.3 321 37.9 37.3 322 38.1 37.3 323 38.1 37.3 324 38.1 37.3 325 38.0 37.3 326 38.1 37.3 327 38.1 37.3 328 38.1 37.3 329 38.0 37.3 330 38.1 37.3 331 38.1 37.3 332 38.0 37.3 333 38.0 37.3 334 37.9 37.3 335 38.0 37.3 336 38.0 37.3 337 38.1 36.7 338 38.1 36.7 339 38.1 36.7 340 38.1 36.7 341 38.2 36.7 342 38.2 36.7 70 343 38.2 36.7 344 38.1 36.7 345 38.0 36.7 346 38.0 36.7 347 38.0 36.7 348 37.9 36.7 349 37.9 36.7 350 37.9 36.7 351 37.9 36.7 352 38.0 36.7 353 38.0 36.7 354 38.0 36.7 355 38.1 36.7 356 38.1 36.7 357 38.0 36.7 358 38.0 36.7 359 38.0 36.7 360 38.0 36.7 361 37.9 36.7 362 37.9 36.7 363 37.9 36.7 364 38.0 36.7 365 38.0 36.0 366 37.9 35.4 367 37.8 34.7 368 37.8 34.7 369 37.8 34.7 370 37.8 34.7 371 37.7 34.7 372 37.6 34.7 373 37.7 35.4 374 37.8 36.0 375 37.7 36.7 376 37.7 37.3 377 37.6 38.0 378 37.6 38.6 379 37.4 39.3 380 37.6 39.3 381 37.6 39.3 382 37.7 39.3 383 37.6 39.3 384 37.6 39.3 385 37.6 39.3 71 386 37.7 39.3 387 37.7 39.3 388 37.7 39.3 389 37.8 38.6 390 37.7 38.0 391 37.6 37.3 392 37.4 36.7 393 37.7 36.7 394 37.6 36.7 395 37.6 36.7 396 37.6 36.7 397 37.5 36.7 398 37.4 36.7 399 37.3 36.7 400 37.3 36.7 401 37.1 36.7 402 37.1 36.7 403 37.0 36.7 404 36.9 36.7 405 36.9 36.7 406 36.8 37.3 407 36.7 38.0 408 36.6 38.6 409 36.6 39.3 410 36.5 40.0 411 36.4 40.0 412 36.3 40.0 413 36.3 40.0 414 36.2 40.0 415 36.1 40.0 416 36.1 40.0 417 36.0 40.0 418 35.7 40.0 419 35.7 40.0 420 35.4 40.0 421 35.2 40.0 422 35.0 40.0 423 34.9 39.3 424 34.8 38.6 425 34.6 38.0 426 34.5 37.3 427 34.4 36.7 428 34.2 36.0 72 429 34.2 35.4 430 34.1 35.4 431 34.1 35.4 432 34.0 35.4 433 33.9 35.4 434 33.9 35.4 435 33.8 35.4 436 33.8 35.4 437 33.7 35.4 438 33.7 35.4 439 33.6 35.4 440 33.5 35.4 441 33.3 35.4 442 33.2 35.4 443 33.1 35.4 444 33.0 35.4 445 33.0 35.4 446 32.9 35.4 447 32.8 35.4 448 32.7 35.4 449 32.6 35.4 450 32.6 35.4 451 32.5 35.4 452 32.4 35.4 453 32.4 35.4 454 32.3 35.4 455 32.2 35.4 456 32.2 34.7 457 32.1 34.1 458 32.2 33.4 459 32.2 32.8 460 32.2 32.1 Table E.4: Measured and Modeled Capacitance as a Function of Position in the Test Bed 73 Appendix F MATLAB Code for Image Processing pic2data.m function [data,R,G,B] = pic2data(pic) pic = getsnapshot(obj); [m,n]=size(pic); R=pic(:,:,1); G=pic(:,:,2); B=pic(:,:,3); R=sum(R)./m; G=sum(G)./m; B=sum(B)./m; data=(2*R-G+2*B); treecount.m function treecount %%%%%%%%%%%%%%%%%% %initilize webcam % %%%%%%%%%%%%%%%%%% % Construct a video input object associated % with a Matrox device at ID 1. obj = videoinput(?matrox?, 1); % Select the source to use for acquisition. set(obj, ?SelectedSourceName?, ?input1?) % View the properties for the selected video source object. src obj = getselectedsource(obj); get(src obj) % Preview a stream of image frames. Important because % webcam shows dark images unless preview is running. preview(obj); %%%%%%%%%%%%%%%%%%%%%% %initialize variables% %%%%%%%%%%%%%%%%%%%%%% master=0; data=0; count=0; %%%%%%%%% 74 %program% %%%%%%%%% data=pic2data; master=treealign(master,data); [count,master]=counter(master); treealign.m function [data] = treealign(data1,data2) %assumed that the data is always moving forward! [row1,column1]=size(data1); [row2,column2]=size(data2); shft=0; di =10; for n=1:column2; sub=abs(sum((data1(column1-n+1:column1)-data2(1:n))?)/n); if sub < di shft=n; di =sub; end end if max(data1) > max(data2) data=[data1,data2(shft:column2)]; else data=[data1(1:column1-shft),data2]; end 75 Appendix G NEC Win-Pro Code for Pine Seedling Models Monopole Antenna CM NEC Input File CM Monopole radius 0.00275m, length 0.3m CM excitation by incident plane wave CE GW 1 21 0 0 0 0 0 0.3 0.00275 GS 0 0 1.000000 GE 0 LD 4 1 1 13 871.5 188.7 EX 1 1 1 0 90 0 0 0 0 0 FR 0 21 0 0 99.930819 99.930819 RP 0 1 360 1000 90 0 1.00000 1.00000 EN Loblolly Antenna Model CM NEC Input File CM Monopole radius 0.01m, length .3 meters CM Also wires that ressemble seedling CM limbs will ajoin the monopole (0.002m diameter) CE GW 1 21 0 0 0 0 0 .30 .00275 !Stem GW 2 5 0 0 .1 0.05 0 .15 .001 !Branch Right GW 3 5 0 0 .1 -.05 0 .15 .001 !Branch Left GS 0 0 1.000000 GE 0 LD 4 1 1 13 871.5 188.7 LD 4 2 1 3 201.1 43.5 LD 4 3 1 5 201.1 43.5 FR 0 21 0 0 99.930819 99.930819 EX 1 1 1 0 90 0 0 0 0 0 RP 0 1 360 1000 90 0 1.00000 1.00000 EN 76 Sample Seedling Antenna Model CM NEC Input File CM Monopole radius 0.01m, length .3 meters CM Also, wires that ressemble seedling CM limbs will ajoin the monopole (0.002m diameter) CE GW 1 21 0 0 0 0 0 .30 .00275 !Stem GW 2 5 0 0 .1 .05 0 .15 .001 !Branch Right GW 3 9 0 0 .15 .02 .01 .28 .001 !Second Stem GW 4 10 0 0 .12 -.02 -.01 .26 .001 !Third Stem GS 0 0 1.000000 GE 0 LD 4 1 1 13 871.5 188.7 LD 4 2 1 5 335.1 72.6 LD 4 3 1 9 603.3 130.6 LD 4 4 1 10 670.3 145.2 FR 0 21 0 0 99.930819 99.930819 EX 1 1 1 0 90 0 0 0 0 0 RP 0 1 360 1000 90 0 1.00000 1.00000 EN 77 Appendix H Simulated Radar Return Data Monopole Antenna Power (dB) Dielectric Constant ^" Frequency (MHz) 15-j5 15-j20 40-j5 40-j20 27.5-j12.5 99.93 -68.48 -66.11 -64.06 -64.84 -65.73 199.86 -56.16 -54.02 -51.81 -52.39 -53.45 299.79 -49.01 -46.95 -44.67 -45.19 -46.30 399.72 -43.97 -41.96 -39.64 -40.12 -41.27 499.65 -40.08 -38.12 -35.78 -36.22 -37.39 599.58 -36.92 -34.99 -32.64 -33.06 -34.24 699.52 -34.25 -32.36 -30.01 -30.4 -31.59 799.45 -31.95 -30.09 -27.73 -28.11 -29.31 899.38 -29.93 -28.09 -25.74 -26.10 -27.30 999.31 -28.13 -26.31 -23.95 -24.30 -25.51 1099.20 -26.50 -24.70 -22.35 -22.68 -23.89 1199.20 -25.02 -23.23 -20.89 -21.21 -22.42 1299.10 -23.66 -21.89 -19.55 -19.86 -21.07 1399.00 -22.40 -20.65 -18.31 -18.62 -19.83 1499.00 -21.24 -19.5 -17.17 -17.46 -18.68 1598.90 -20.15 -18.43 -16.10 -16.39 -17.60 1698.80 -19.13 -17.42 -15.10 -15.38 -16.59 1798.80 -18.18 -16.40 -14.16 -14.44 -15.65 1898.70 -17.26 -15.59 -13.28 -13.54 -14.75 1998.60 -16.40 -14.74 -12.44 -12.70 -13.91 2098.50 -15.59 -13.94 -11.65 -11.90 -13.10 Table H.1: Monopole Antenna Reradiated Power (dB) 78 Loblolly Antenna Model Power (dB) Dielectric Constant ^" Frequency (MHz) 15-j5 15-j20 40-j5 40-j20 27.5-j12.5 99.93 -69.39 -67.57 -65.37 -65.93 -66.92 199.86 -56.54 -54.75 -52.63 -53.05 -54.10 299.79 -49.10 -47.52 -45.19 -45.55 -46.68 399.72 -43.81 -42.51 -40.02 -40.30 -41.50 499.65 -39.69 -38.65 -36.10 -36.30 -37.53 599.58 -36.30 -35.44 -32.96 -33.08 -34.31 699.52 -33.43 -32.65 -30.34 -30.40 -31.58 799.45 -30.97 -30.15 -28.08 -28.11 -29.22 899.38 -28.85 -27.89 -26.06 -26.11 -27.12 999.31 -27.02 -25.85 -24.22 -24.32 -25.26 1099.20 -25.45 -24.03 -22.55 -22.72 -23.60 1199.20 -24.12 -22.45 -21.04 -21.29 -22.13 1299.10 -23.00 -21.12 -19.67 -20.00 -20.84 1399.00 -22.07 -20.01 -18.47 -18.87 -19.73 1499.00 -21.31 -19.13 -17.42 -17.87 -18.80 1598.90 -20.69 -18.43 -16.53 -17.01 -18.02 1698.80 -20.18 -17.90 -15.79 -16.28 -17.37 1798.80 -19.75 -17.50 -15.16 -15.64 -16.84 1898.70 -19.37 -17.20 -14.64 -15.08 -16.38 1998.60 -18.99 -16.96 -14.19 -14.58 -15.98 2098.50 -18.56 -16.74 -13.77 -14.09 -15.58 Table H.2: Loblolly Antenna Model Reradiated Power (dB) 79 Sample Seedling Antenna Model Power (dB) Dielectric Constant ^" Frequency (MHz) 15-j5 15-j20 40-j5 40-j20 27.5-j12.5 99.93 -66.16 -64.99 -63.66 -63.79 -64.57 199.86 -52.12 -50.12 -48.95 -49.40 -50.05 299.79 -44.47 -42.37 -40.81 -41.35 -42.12 399.72 -39.25 -37.24 -35.41 -35.93 -36.82 499.65 -35.30 -33.36 -31.41 -31.89 -32.85 599.58 -32.17 -30.28 -28.28 -28.70 -29.71 699.52 -29.58 -27.74 -25.71 -26.09 -27.14 799.45 -27.40 -25.60 -23.56 -23.90 -24.97 899.38 -25.53 -23.76 -21.71 -22.03 -22.03 999.31 -23.80 -22.16 -20.11 -20.40 -21.49 1099.20 -22.44 -20.75 -18.70 -18.97 -20.06 1199.20 -21.16 -19.49 -17.44 -17.70 -18.79 1299.10 -20.00 -18.36 -16.31 -16.56 -17.65 1399.00 -18.95 -17.33 -15.30 -15.54 -16.62 1499.00 -17.99 -16.39 -14.37 -14.61 -15.68 1598.90 -17.11 -15.54 -13.53 -13.75 -14.83 1698.80 -16.31 -14.75 -12.75 -12.97 -14.04 1798.80 -15.56 -14.01 -12.03 -12.25 -13.31 1898.70 -14.87 -13.33 -11.36 -11.58 -12.63 1998.60 -14.22 -12.69 -10.74 -10.96 -12.00 2098.50 -13.61 -12.10 -10.15 -10.38 -11.41 Table H.3: Sample Seedling Antenna Model Re-Radiated Power (dB) 80