计算机应用研究2023，Vol.40Issue(12)：3821-3827,3833,8.DOI:10.19734/j.issn.1001-3695.2023.03.0147
基于动态物体跟踪的语义SLAM
Semantic SLAM based on dynamic object tracking
摘要
Abstract
This paper proposed a semantic SLAM algorithm based on dynamic object tracking to address the issue of decreased localization accuracy in traditional visual SLAM methods due to feature matching errors in dynamic scenes.Based on the clas-sic visual SLAM framework,The algorithm extracted dynamic objects for inter-frame tracking and utilized their pose informa-tion to assist the camera's own localization.Firstly,it employed YOLACT,RAFT,and SC-Depth networks in the data prepro-cessing stage to extract semantic masks,optical flow vectors,and pixel depths from the images.Subsequently,the visual fron-tend module utilized the extracted information to compute probability maps,employing semantic segmentation masks,motion consistency checks,and occlusion point verification algorithms.These probability maps aided in effectively distinguishing be-tween dynamic and static features in the scene.Then,the bundle adjustment module in the back-end integrated multiple fea-ture constraints derived from object motion to enhance the algorithm's pose estimation performance in dynamic scenes.Finally,comprehensive comparisons and validations were conducted on the dynamic scenes of the KITTI and OMD datasets.The expe-rimental results demonstrate that the proposed algorithm accurately tracks dynamic objects and exhibits robust and accurate lo-calization performance in both indoor and outdoor dynamic scenes.关键词
视觉SLAM/语义信息/动态物体跟踪/捆集调整Key words
visual SLAM/semantic information/dynamic object tracking/bundle adjustment分类
信息技术与安全科学引用本文复制引用
刘家麒,高永彬,姜晓燕,方志军..基于动态物体跟踪的语义SLAM[J].计算机应用研究,2023,40(12):3821-3827,3833,8.基金项目
国家自然科学基金—民航联合重点项目(U2033218) （U2033218）
