Fast Feature Point Matching Algorithm in VSLAM System
Feature point matching between consecutive image frames is a key technology in visual simultaneous localization and mapping(VSLAM).To address the issues of time-consuming and low accuracy in feature point matching between consecutive image frames in VSLAM systems,a fast feature point matching algorithm based on local pixel motion model(LPMM)is proposed.Based on the assumption of motion smoothness constraint,this algorithm adopts the consistency of pixel motion within local regions in consecu-tive image frames to divide the image into local grid regions,and estimate the motion vector of each grid by using some feature points within the grid.On this basis,the center point of the search range for matching feature points in the next frame is calculated.Finally,the feature points that match the current frame's feature points are searched for within the local neighborhood of the calculated center point of the next frame.Experimental results show that the proposed matching algorithm,compared with the widely used projection matching algorithm in ORB-SLAM2,has an average matching speed increase of over 50％and an accuracy improvement of approxi-mately 4％.
VSLAMfeature point matchinglocal pixel motion modellocal region grid
万方数据
维普
