Method for extracting video object contour based on centroid tracking and improved GVF Snake

Abstract

The invention discloses a method for extracting video object contour based on centroid tracking and improved GVF Snake. The method is characterized by comprising the following steps of according to the condition that the movement trend of adjacent frames is similar in short time, dividing a video sequence into a plurality of sections, wherein each section has k frames of video, taking the first two frames in the section as key frames, detecting the effect of eliminating the background border based on the change of t obvious test, and acquiring an initial movement area; extracting a critical quadrangle as an initial contour of the key frames, performing intra-frame GVF Snake evolution, searching the exact contour, then forecasting initial contours of the subsequent frames through the movement vector between centroids of the intra-frame movement object contours of the key frames, and then performing the exact intra-frame GVF Snake contour positioning of the subsequent frames; and by the analogy, extracting the object contour of all frames. The improved model adopts four-direction anisotropic diffusion, and adopts fidelity term coefficients with fast descending speed to enhance the performance to enter into concave; and the convergence to a weak edge is kept. By the method, the shortcoming of manually acquiring the initial contours is conquered.

Description

technical field [0001] The invention relates to a processing method for video object contour extraction, in particular to a centroid tracking and improved GVF Snake video object contour extraction method. This method overcomes the shortcomings of manually drawing the initial contour, improves the traditional GVF Snake method, and greatly improves the accuracy. Experiments prove that this method successfully extracts the concave part of the target and has a better convergence effect on weak boundaries, which improves the accuracy of contour extraction and makes the method more practical and popular. Background technique [0002] Moving target contour extraction is a popular research direction in the field of computer vision, and is widely used in many fields such as security monitoring and behavior recognition. Moving target contour extraction is to accurately find the contour of a moving target of interest in each frame of a video sequence . Objects with complex shapes are...

AI Technical Summary

Problems solved by technology

[0033] 1. When the target has a weak boundary, excessive smoothing will lead to the loss of image edge information and cause the force field at the weak boundary to have a wrong direction;

[0034] 2. Excessive restoration of the smoothed original image gradient will produce edge enhancement and displacement. When the image has a deep concave edge, the ability of GVF Snake to enter the deep concave part will be weakened due to the enhancement of the gradient value of the normal direction of the edge at the depression.

Images