Object fine contour tracking method based on low-order sparse expression

Abstract

The present invention provides an object fine contour tracking method based on low-order sparse expression. The essence features of an object in video are analyzed to obtain the relation of the object and the background between two front and back frames, so that the object fine contour tracking method based on the low-order sparse expression can be applied to most scenes. The method is characterized in that: images including a tracking object are subjected to blocking by using super pixels in the problem of object fine tracking, each super pixel is taken as a point so as to calculate the complexity; the modeling for the object tracking problem is configured to solve the problem of the low-order sparse expression of the matrix; and after the sparse expression coefficient is obtained, the energy minimizing method is employed to cut the object and the background, the decision function of an energy function model is provided, and the decision function is used in the energy minimizing method to take as the basis of the object and the background cutting result.

Description

technical field [0001] The invention belongs to the field of computer vision, in particular to the field of intelligent monitoring. Background technique [0002] Video-based target fine contour tracking technology is a field that has attracted much attention and is developing rapidly in the field of computer vision. This technology is one of the most basic technologies in the field of computer vision and can obtain the tracking results of target contours. The upper-level algorithm further analyzes and processes according to the target contour tracking results to realize the application of scene understanding, target action recognition, and human behavior recognition. The wide application prospect and high research value of this technology have aroused the strong interest of researchers at home and abroad. [0003] The fine contour tracking of the target in the video is considered to be a binary classification problem. At present, many related algorithms have appeared at hom...

AI Technical Summary

Problems solved by technology

[0003] The fine contour tracking of the target in the video is considered to be a binary classification problem. At present, many related algorithms have appeared at home and abroad, such as the method based on level set, which divides the motion estimation and target segmentation into two separate stage, but in a video with many camera movements, it is difficult to estimate the motion to get a good tracking effect

In order to solve the situation of camera motion, people propose a method based on graph cut, which fuses multiple clue functions together. The motion information of the target is usually one of the important clue functions, but the background motion field usually interferes. The target's motion information makes the tracked target outline inaccurate

There are also some semi-automatic segmentation methods, which need to manually mark some target and background areas, which greatly limits their application fields.

[0004] There are still many defects in the various tracking algorithms for the fine outline of the target that are currently known. These algorithms are all for a specific scene, and there is no general algorithm that can be applied in most scenes.

Images