Offline Extrinsic Calibration of GPS and LiDAR Sensors Using Static Object Matching and Size-Constrained Optimization

This paper presents an offline extrinsic calibration method for autonomous vehicles, which combines integrated navigation and LiDAR (light detection and range) systems. The method utilizes generic, static roadside objects, which are detected during regular vehicle operation, thus eliminating the need for predefined calibration targets. We propose a multi-object matching algorithm and design a two-stage optimization framework: A coarse calibration based on object size constraints and a fine calibration using Improved Chamfer Distance. Experimental results in the KITTI dataset and Gazebo simulation show that the proposed method achieves an average translation error of 3.7 cm and a rotation error of 0.67 \(^{\circ }\) , demonstrating higher precision and flexibility than the existing methods.