基于LM算法的三维点云与二维图像标定方法

打开文本图片集

中图分类号：TN911.73-34；TP391 文献标识码：A 文章编号：1004-373X（2026）01-0059-07

DOI：10.16652/j.issn.1004-373x.2026.01.010引用格式：，等.基于LM算法的三维点云与二维图像标定方法[J].现代电子技术，2026，49（1）：59-65.

3D point cloudsand 2D image calibrationmethod based on LM iterative algorithm

WULong'，TAOYifan，YANG Xu1 ，XU Lu1 ，CHENShuyu² （1.SchoolofComputerScienceand Technology，ZhejiangSci-TechUniversityHangzhou31oo18，China; 2.KeyiCollegeofZhejiangSci-TechUniversity，Shaoxing312369，China）

Abstract：Theinsuffcientcalibrationaccuracy between LiDARandcameradetectionleads toerrorsinthespatial alignentofLiDARpointcloudsandcameraimages，affctingsubsequentfeaturematching，objectdetection，and 3D reconstructionaccuracy.Therefore，acalibrationmethodbasedonLiDAR3Dpointcloudsand monocularcamera2Dimages is proposed.Themethodaimstoachieve precisedetectionoflarge-scaleobjectsand 3Denvironmentreconstructio.Inthemethod， multi-framepointclouddataacumulationisusedtoobtainrelativelydensepointcloudmeasurementsandacornerdetection algorithmisappliedtodetectfeaturecornersintheimages.Subsequentlytheparametersaresolvedwithpartialleastsquares （PLS）.Finally，theLevenberg-Marquardt（LM）iterativealgorithmisemployedtominimizethereprojectioneror，thereby improvingcalibrationacuracy.ThecalibrationresultsshowthatthereprojectionerroroftheSPAAMalgorithmisreduced by 8.6%incomparisonwiththatof theclasical methods，whiletheproposedmethodachievesanearly 38.2%reductionin reprojection error incomparison with that of the classical methods，validating its accuracyand efectiveness.

Keywords：LiDAR; monocular camera; calibration method;point cloud data; PLS; LM iterativealgorithm

0 引言

激光雷达在检测大型物体形变（如火车站顶棚）中发挥着重要作用，但单一的激光雷达无法实现全方位检测，因此通常需要进行多传感器数据融合，在各种融合算法中，激光雷达与相机的组合最为常见[1-3]。（剩余10730字）