Motion Matching Based On HECS Structure Feature

The extraction and expression of image features can be used for the research and application of computer vision in various fields,and the research on the characteristic extraction,description,matching and tracking of video flow data is also an important part of machine vision research and research hotspot.However,there are still many defects in the field of feature matching and tracking in video stream,which is difficult to achieve better practical application.Common problems are: lack of correlation between features,lack of local characteristics of the ability to resist background interference,the ability to accurately express the image structure,etc.,the need to use triangulation and other man-made follow-up treatment can be used for practical research and application.Therefore,it is obvious to use the image itself information to associative and establish feature extraction and description method with anti-background interference ability as a prerequisite for video stream feature matching and tracking.In addition,edge information as an important part of image information,although in recent years the edge detection algorithm has been close to the accuracy of human edge recognition,but the edge attribution problem has not been solved.Further study on the problem of image edge attribution is still of great significance.This paper studies the above-mentioned problems,the main work is:(1)The problem of image edge attribution is studied,inspired by graphics to describe the level of objects,and a method of separating half-edge image edges is proposed to solve the problem of lack of attribution of edges produced by different objects and the inability to directly use its grayscale and angle information for structural description.The completeness and correctness of extracting half is verified by experiment.(2)The image features against background interference are studied,and a method of extraction and matching based on half-edge local combination features(HECS)is proposed,so that the image features can be independent of the local region changes caused by the interlaced objects in the video stream,and the stability of the features in the video stream is strengthened.Experiments on public and own data sets,especially in the edge area of the image,a good feature matching result was obtained.By comparing the experiment,it is proved that the matching and tracking of the proposed features in the video stream has a better experimental effect.(3)The correlation of image features is studied,and a combination and matching method based on HECS graphic structure features is proposed,which can establish the image feature according to the image information itself,and more complete lying the face information of the image.The correspondence of the picture structure is established by means of incomplete matching,which strengthens the stability of the image features.Its effectiveness has been proved by experiments.(4)The visual residual phenomenon is studied,a video flow feature matching method based on the characteristics of the graph structure is proposed,which can be used to reason and supplement the image features of the current frame according to the matching results of the historical graph structure,so that the image features in the video stream gradually stabilize with the passage of time,and strengthen the matching effect and tracking time of the image features in the video stream.Compared with the traditional image feature extraction method,a better matching tracking effect and stronger stability are obtained.