Research on SLAM of Indoor Environment Based on Dynamic Feature Point Matching
This paper mainly focuses on the problem of effectively removing dynamic objects in dynamic environ-ments to construct more accurate indoor semantic maps.Based on ORB-SLAM3,a method is proposed to achieve se-mantic map construction in dynamic scenes by adding a video image keyframe selection mechanism and calculating the probability of feature point switching in image frames.Firstly,the original architecture of ORB-SLAM3 was sim-plified and only monocular vision was used as the image data source;Secondly,in order to reduce the amount of cal-culation,a key frame was selected for processing through an evaluation selection mechanism in a group of continuous image frames;Then,in order to eliminate dynamic objects,a switching probability method was used to calculate the dynamic changes of feature points,which was segmented by Mask R-CNN;Finally the construction of real-time se-mantic map was realized.The experimental results show that the proposed method is accurate and real-time.The ex-perimental results show that this method can eliminate dynamic objects accurately to solve the problem of tracking loss,and the absolute trajectory error and relative attitude error are greatly improved in the indoor dynamic scene,and can meet the real-time requirements of the system operation.
NETL
NSTL
万方数据
