Research and Implementation of Intelligent Vehicle Localization Technology Based on Segmentation and Matching of High-Precision Campus Maps
Year: 2026
Type: conference-paper
Abstract: Intelligent vehicle positioning technology is the most fundamental and critical technology in the field of autonomous driving. Only by knowing the specific position, orientation, and attitude in the corresponding coordinate system can an intelligent vehicle achieve reasonable path planning and reach its destination. However, due to their inherent limitations, single sensors cannot meet all the positioning requirements of intelligent vehicles. This paper focuses on a campus shuttle bus project and proposes an intelligent vehicle positioning technology based on segmentation and matching of a high-precision campus point cloud map. It achieves integrated positioning using three sensors: the BeiDou Navigation Satellite System (BDS), Velodyne LiDAR, and the Strap down Inertial Navigation System (SINS). Fusion algorithms are employed to process the corresponding data, obtaining precise position and attitude information of the intelligent vehicle with centimeter-level accuracy. The main research contents of this paper are as follows: (1) The definitions and concepts of SINS are introduced, various coordinate systems are studied, and the coordinate transformation formulas between commonly used coordinate systems are derived and summarized. After completing the above work, an error analysis of the SINS used in this paper is conducted, dividing errors into random errors and systematic errors, and a calibration model is established for error compensation. Finally, filtering algorithms are used to obtain the vehicle's attitude and to calculate its velocity and position. (2) The positioning principle and error sources of the BeiDou Navigation Satellite System are analyzed. The Kalman filter algorithm is used to fuse the positioning information of BDS with the position and attitude information of the integrated navigation system, establishing a loosely coupled BDS/INS positioning and navigation system to obtain real-time position and attitude information of the intelligent vehicle. (3) The integrated navigation system and multi-line LiDAR are used to construct a high-precision campus point cloud map, which is then divided according to corresponding rules to obtain a dynamically loaded, relatively uniform point cloud map. Based on the BDS/INS integrated navigation system, onboard LiDAR is used to perform position matching with the point cloud map, obtaining a displacement matrix to calibrate the positioning information derived from the BDS/INS system. This constitutes a three-sensor integrated positioning system, ensuring the safety of the intelligent vehicle in autonomous driving mode. The intelligent vehicle positioning system based on segmentation and matching of a high-precision campus point cloud map, as developed in this paper, has been successfully applied to an intelligent vehicle platform operating on campus. Through extensive real-vehicle tests and experimental simulation data analysis, it has been verified that the system meets the positioning requirements of intelligent vehicles in a campus environment.
Authors: Baoqing Du, Mingxi Cai
Institutions: Chengdu Qinggong Polytechnic University, MediaTek (China)
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7214218485