A Study of Accuracy Enhancement of Vehicle Self-localization Using Pole-like Object Maps

• Main Authors: CHIANG, YEN-HUNG, 江彥宏
• Other Authors: HSU, CHIH-MING
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/fuxuf2

碩士 === 國立臺北科技大學 === 機械工程系機電整合碩士班 === 107 === The self-driving system has become the mainstream in the development of international research. At present, road detection and mapping is a project that the research institutes and many company try to study it. The localization system also plays a pivotal role in the self-driving system. If you can localization very well, the trajectory can be improved to prevent the offset, and a good trajectory path can be used to establish a map with higher precision. In this paper, we propose a method which use pole-like object map to enhance vehicle positioning and correction previous road trajectory at the same time. We have a pole-like object map, which only contains the pole-like object point cloud around the road, and these cloud are marked with space coordinate information to enhance the vehicle positioning accuracy. We use 3D LiDAR to be the main sensor to scan an environmental. Then uses the dynamic segment threshold and projection point cloud to image to extract the pole-like object, to avoid the segmentation error caused by scattering characteristics of 3D LiDAR. The extracted pole-like map will use the occupied grid to filter out the noise points to improve the accuracy of the pole-like map. In terms of positioning, we continue to detect the feature of the pole-like while the vehicle is moving, and compare the found pole-like with the pole-like map. When the matching is successful, we will estimation and correction of the current positioning. Use this method can effectively improve the performance of vehicle positioning accuracy.