Minimum obstacle distance weighted tracking method

Abstract

The invention discloses a minimum obstacle distance weighted tracking method. Sampling in a search window is carried out to generate a set of candidate samples; the boundary box of the current frame is divided to non-overlapped image blocks and is expanded to an expanded boundary box, and features of each image block are extracted; with a box image as a seed node, the minimum obstacle distance between other nodes and the seed node is calculated, and a distance conversion graph is then obtained; the minimum obstacle distance is added to the corresponding image block feature as a weight, a spatially-ordered weighted image block feature descriptor is obtained, and combination weights with image features are merged into a structured support vector machine to perform tracking. In combination ofimage color histogram features and directional gradient histogram features, the distance conversion graph based on a background seed node set is obtained through calculation, and according to the minimum obstacle distance between the image blocks and the background node, the problems of occlusion and deformation are solved, drift can be reduced and the tracking stability and the robustness are enhanced.

Description

technical field [0001] The invention relates to a computer vision technology, in particular to a minimum obstacle distance weighted tracking method. Background technique [0002] The widespread popularization of intelligent video surveillance systems has promoted the continuous deepening of computer vision theory research, and target tracking is an important issue in video analysis in the field of computer vision, and has a wide range of applications in surveillance, human-computer interaction and medical imaging. A typical target tracking scenario is that the position and size of the target to be tracked are given in the first frame of the video sequence, and the tracker is used to track the target in the next consecutive frames. Although the emergence of many tracking algorithms in recent years has promoted the development of object tracking, many factors faced in the tracking process, such as illumination changes, occlusion and rapid deformation, have not been well resolv...

AI Technical Summary

Problems solved by technology

However, it defines the static structure graph of the eight-neighborhood, and the edge weight is calculated based on the low-level feature distance. It performs poorly in dealing with severe occlusion, rapid deformation, and illumination changes in the tracking sequence.

[0004] The disadvantage of the existing technology is: the traditional target tracking method generally uses the Euclidean distance to calculate the similarity of two regions

The weight of the edges used to construct the static structure graph is based on the low-level feature distance calculation, and the Euclidean distance cannot accurately distinguish the target from the background

Images