图像角点与点云曲率渲染边界特征下的雷视一体标定  

LiDAR and camera integrated calibration based on image corner points and point clouds curvature rendering boundary

作　　者：张代聪 李倩[1] 余继龙 陈文博 罗宝琪 ZHANG Daicong;LI Qian;YU Jilong;CHEN Wenbo;LUO Baoqi(School of Mechanical and Electronic Engineering,Xi’an Polytechnic University,Xi’an 710048,China)

机构地区：[1]西安工程大学机电工程学院,陕西西安710048

出　　处：《西安工程大学学报》2025年第2期109-117,共9页Journal of Xi’an Polytechnic University

基　　金：国家自然科学基金项目(52005380)。

摘　　要：激光雷达和相机组成的多传感器感知系统首先要经过标定才能进行建模、识别等工作。随着激光雷达的扫描范围逐步扩大及应用场景的多样化,为确保雷视标定在室内外及远近距离等场景中均能灵活使用,提出了一种无需标定板,通过对建筑物或者其他自然场景特征点选取的方法,对激光雷达与相机的外参进行标定。该方法结合了点云的曲率渲染边界特征辅助特征点的选取以及二维图像角点提取算法,能够方便快捷地实现具有自然特征的场景下的雷视一体外参标定。实验结果表明:标定的最大平移向量均值误差为34 mm,最大旋转向量均值误差为0.493°,能够满足常用的厘米级激光雷达和相机雷视一体标定的需求。A multi-sensor perception system composed of light detection and ranging(LiDAR)and cameras must first undergo calibration before it can perform tasks such as modeling and recognition.As the scanning range of LiDAR gradually expanded and its application scenarios diversified,to ensure the flexibility of LiDAR-camera calibration in various environments,including indoor,outdoor,and long-or short-distance scenarios,the paper proposed a method that does not require calibration boards.Instead,it leverages feature points from buildings or natural scenes to calibrate the extrinsic parameters between LiDAR and cameras.This method combined curvature-rendered boundary features of point clouds to assist in feature point selection and a 2D image corner extraction algorithm,enabling convenient and efficient extrinsic parameter calibration in scenes with natural features.Experimental results show that the maximum mean error of the translation vector is 34 mm,and the maximum mean error of the rotation vector is 0.493°,which meets the requirements for common centimeter-level LiDAR-camera calibration.

关 键 词：多传感器 雷视标定 图像角点 曲率渲染 激光雷达 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]