视觉SLAM深度相机无重叠视域下自标定方法研究

打开文本图片集

中图分类号：TP242.6；TP391.4 文献标识码：A 文章编号：2096-4706（2025）20-0035-05

Abstract： Visual sensors are widelyused in SLAM（Simultaneous Localization and Mapping）technology.Multi-sensor fusionhas becomeanimportantresearch direction inrobotic SLAM.Accuratedataof visual extrinsic parameters directlyfects the successateofSAM.However，inpractice，extrinsicpaametercalirationoftencotaiserors，andsnsorsandeuipmet mayshift or deform duringlong-termuse，leading tochanges inthe extrinsic parameters.Meanwhile，calibration methods using calibration plates becomedificult when non-overlapping feldsofviewexist betweenmultiplecameras.This paper studies the calibrationofdual-Kinectdepthcameras，andutilzespointfeature matchingrelationshipstocalculatetherelativetransforation betweenthe two sensors，therebyobtaining the extrinsic parametercalibrationresults forboth cameras.Experimentalresults showthat inenvironments rich intexture information，theproposedcamera self-calibration method yieldsresults consistent with Zhang'scabrationmethod.Moreover，itenablesutonomouscalibationfersonvenientapplicationndcaneetthe demands of practical use.

Keywords：SLAM; calibration; depth camera; camera extrinsic parameter

0 引言

机器人同时定位与制图SLAM1]（SimultaneousLocalizationandMapping，SLAM）是指机器人在没有先应信息的未知环境中，依靠自身传感器与运动来估计环境地图与自身位姿的技术。（剩余5680字）