火力与指挥控制2025，Vol.50Issue(6)：71-77,88,8.DOI:10.3969/j.issn.1002-0640.2025.06.010
基于SuperGlue的相机和激光雷达位姿变换的自动检测
Automatic Detection of Camera And Lidar Pose Transformation Based on SuperGlue
刘全超 1何舒文 1张涛 1王飞 2郝强国2
作者信息
- 1. 北方自动控制技术研究所,太原 030006
- 2. 陆军装备部驻北京地区军事代表局,北京 100166
- 折叠
摘要
Abstract
A method for automatic detection of camera and LiDAR sensor pose transformation changes during the driving process of unmanned vehicles due to collisions and bumps is proposed.Firstly,the multi frame LiDAR point cloud is accumulated,and the superglue convolutional neural network calculates the feature matching points between the radar images and the camera images.The RANSAC and reprojection error are minimized to calculate the pose transformation matrix,and the pose transformation relationship between the camera and LiDAR is obtained.The experimental results show that the rotation error of this method is better than 2°,and the translation error is not greater than 1 cm,which can effectively detect the pose transformation of the camera.关键词
自动标定/点云密集化/相机模型/SuperGlue神经匹配网络/位姿变换检测Key words
automatic calibration/point cloud densification/camera model/SuperGlue matching neural network/pose transformation detection分类
信息技术与安全科学引用本文复制引用
刘全超,何舒文,张涛,王飞,郝强国..基于SuperGlue的相机和激光雷达位姿变换的自动检测[J].火力与指挥控制,2025,50(6):71-77,88,8.
