Processing of LiDAR for Traffic Scene Perception of Autonomous Vehicles

The future of automated driving system deals with the emergence of autonomous vehicles popularly referred as self-driving cars or driver-less cars. The ability of sensing its environment with or without human interaction is the redeeming feature of autonomous vehicles. The traditional segmentation algorithms mainly uses camera data,whose accuracy is degraded by the effect of shadows, bright sunlight or headlight of other cars. This paper focuses on Lidar sensors which provides 3-D geometry information of the vehicle surroundings with high accuracy. Apart from the previous works which deals only with segmentation using data from Lidar sensor, it deals with predicting the probabilities of detecting things such as human, vehicles, traffic signals, stationary objects etc. Moreover with the use of Ford campus vision and KITTI vision detection bench marks it achieves a high accuracy of about 90 percentage.