一种融合点线特征的视觉SLAM方法

打开文本图片集

中图分类号：TN911.73-34；TP242 文献标识码：A 文章编号：1004-373X（2025）15-0144-08

Avisual SLAM method that integrates point and line features

SUNHaidi，YUXiayi，ZHENGEnhui

5choolofMechanicalandElectricalEngineering，China Jiliang University，Hangzhou31oo18，China

Abstract：Avisual inertial SLAM（simultaneous localizationandmapping）methodintegrating pointandline featuresis proposedbecauseof thechalengesofpurevisual SLAMalgorithmsuchassharpilluminationfluctuation，lowtextureareaand complextextureenvironment，frequentfeaturetrackingfailureandsignificantdeclineinpositioningaccuracyduetohigh-speed cameramovement.ThegradientthresholdofLSDalgorithmisimprovedadaptivelyandsomefixedparametersofLSDalgorithm areadjustedtoimprovetheaccuracyoflinesegment extraction.Forsomeextractedshortlinefeatures，themergingconditions areusedtojudgewhethertheshortlinecanbemergedinto highqualitylinesegments.Thesliding windowoptimizationmethod isused toestimatethepose，andthereprojectionerorsofIMUmeasurementresiduals，pointsandlinesareoptimizedglobaly. Theproposed method is testedontheEurocdataset，andthetestresultsshowitcanimprovethepositioning acuracyand robustness oftheoverallSLAM system effectively.

Keywords：SLAM； line featureextraction;line feature matching；pointand line featurefusion;adaptive improvement; nonlinear optimization

0 引言

当前，在地面机器人与自动驾驶等领域中，同步定位与地图构建（SimultaneousLocalization and Mapping，SLAM）占据着至关重要的地位，它作为核心定位导航策略，被广泛采纳与应用。（剩余11502字）