WCSE

System Design of Road and Vehicle Detection Based on LiDAR Point

Abstract— Autonomous vehicles have been hot topics in recent years. It is very important to sense the environment for autonomous vehicles. There are many sensors can be used to sense the environment including camera, sonar, radar, LiDAR and so on. Especially LiDAR [1-3] is a device with high accuracy, high precision and fast characteristics that can capture the distance and contour of object precisely and quickly. It can get 2.2 million points per second and the scanning range is wide. LiDAR can scan the environment from 0.3m to 100m, the scanning horizontal angle is 0° to 360°, and the horizontal resolution is 0.08°. The vertical angle is -15° to 15°, and the vertical resolution is 0.4°. Therefore, LiDAR is a good choice for autonomous vehicles application. Besides, the technology of neural network is maturing day by day. It is good in object recognition and semantic segmentation. Therefore, we propose a method to use LiDAR, color image, and neural network to detect the environment. The scheme pre-processes point cloud data and color image at first. Then we use neural network to ext ract drivable roads and cars. Finally, we combine point clouds and neural network outputs to mark the road and car on point clouds by mapping algorithm. Besides, we also propose the hardware design with multiple neural network functions in our system.

Index Terms— Autonomous vehicles, Environment detection, LiDAR, neural network

Chun-Ju Huang, Bo-Tai Wu, Yu-Cheng Fan
Department of Electronic Engineering, National Taipei University of Technology, TAIWAN

[Download]

Cite: Chun-Ju Huang, Bo-Tai Wu, Yu-Cheng Fan, "31-34," Proceedings of 2019 the 9th International Workshop on Computer Science and Engineering WCSE_2019_SPRING, pp. 31-34, Yangon, Myanmar, February 27-March 1, 2019.