Lane Detection, Calibration, and Attitude Determination with a Multi-Layer Lidar for Vehicle Safety Systems

Abstract

This thesis develops a lane detection algorithm which extracts lane markings from the road's surface in an effort to ascertain the lateral position of the vehicle in the lane using a lidar, with the goal of being incorporated into a larger lane departure warning system. The presented algorithm underwent numerous highway tests. These tests include both a comparison of the reported position of the algorithm using the lidar to a high accuracy global positioning system (GPS) as well tests more commonly found in literature. The results of which show that the presented algorithm is capable of determining its position in the lane to a high degree of accuracy and consistency.

Abstract Additionally, a novel lidar calibration method is developed which is capable of calibrating a 3D forward facing lidar mounted on a vehicle to a plane, through simply causing the vehicle to pitch. The presented algorithm will demonstrate the ability to determine vehicle attitude with sub degree accuracy. The algorithm is tested under a number of conditions ranging from in lab tests to on vehicle tests. During these test both attitude accuracy as well as processing time are analyzed. Finally, it is shown that this algorithm is capable of making highly accurate attitude calculations at a rate that is capable of running real-time.