Motion Matching Based On Half-edge Combined Structural Features

Computer vision is a research hotspot in image processing.Image matching is one of the important technologies,which provides an important basis for subsequent image analysis.Feature detection and description are important links in image matching.It is a very difficult task to detect which features and how to describe features.Image matching technology is mainly used in target tracking,image stitching,3D reconstruction,medical diagnosis and other fields.In the motion scene,due to the spatial displacement of the moving target,the background information near the target contour changes randomly,and the local feature matching algorithm has a poor matching effect in the target contour region,especially when the target longitudinal moving image scale changes cause image deformation,which may result in Image matching failed.Aiming at this problem,this paper proposes a matching algorithm based on half-edge combined structure features to achieve image matching in motion scenes.The main work of this paper is as follows:(1)The characteristics of local feature matching in moving scenes are analyzed,and the mechanism of motion matching against background interference and longitudinal motion image distortion is studied.(2)A half-edge structure feature is proposed,which combines the half-edge structure adjacent to the region to realize corner detection.(3)The feature model of half-edge-angle-half-edge composite structure is established,and the combined feature vectors are calculated by fusing the attributes of half-edge and corner to realize the local feature description of anti-background interference and anti-longitudinal motion deformation.(4)Establish matching rules based on half-edge composite structure features to realize local feature matching in motion scenes.The experimental results show that the proposed algorithm is feasible in motion scenarios,has stronger correlation between structural features,more comprehensive matching on moving objects,richer information contained in structural features,and better matching performance in edge contour neighborhoods,and the continuity of matching is better.