Road boundary detection based on information entropy

Detecting the road boundary is an important step for the applications of Intelligent Vehicle (IV), such as generating the region of interest (ROI) as a prior information for object detection or path planning. Hereafter we propose a new method for detecting road boundaries using Laser Interferometry Detection and Ranging (LIDAR). Ground points are firstly removed from LIDAR point cloud through a preprocessing procedure. Then information entropy is applied here for estimating the steering angle of the host vehicle. The estimated steering angle is later employed to rectify the point cloud. Road boundaries are finally detected based on the maximal value of the corresponding histogram. We compare this approach to the traditional histogram based road boundary detection method. Experiments showed that the proposed method can effectively detect road boundaries even in steering situations and outperform the traditional method.