3D Point Cloud Matching Technology Based on Depth Image Based Rendering

摘要

According to the rapid improvement of technology, the Internet of Things (IoT), big data, automated systems and artificial intelligence technology have also gradually developed. The study of self-driving vehicles is the closest to life, the biggest benefit is to make transportation more convenient. In addition, reducing traffic accident is also a major purpose, so the system that the autonomous vehicle is equipped with playing a very important role, and the autonomous vehicle will be equipped with many sensor elements. Its main purpose is to collect and build environmental information, which is equivalent to the eyes of autonomous vehicles. Through the obtained environmental information, the system can determine whether there are obstacles around the vehicle, and then perform its corresponding action to avoid danger. Three-dimensional LiDAR (Light Detection and Ranging) plays a very important role in autonomous vehicles. It uses optical laser to project to the surrounding environment of the vehicle and receives different reflectance generated by different objects to obtain data. Therefore, the paper proposes to combine 3D LiDAR point cloud information with depth maps. The two information are matched to obtain high-precision point cloud information, and different objects are distinguished to observe the results. The depth map is also used to generate two-dimensional point cloud information to repair the problems of missing point cloud information due to environmental factors. In addition, this paper uses the digital chip design flow to implement the decoding of 3D LiDAR packet information.