Target tracking method based on deep and shallow feature adaptive fusion and context information

Abstract

The invention discloses a target tracking method based on deep and shallow feature adaptive fusion and context information, and the method comprises the steps: firstly obtaining a first frame image ofa video image sequence, and building a deep feature model and a shallow feature model based on a context sensing framework; acquiring a plurality of second frame images of the video image sequence, and calculating a deep feature response and a shallow feature response of a corresponding tracking target by using the deep feature model and the shallow feature model; obtaining the position of the tracking target in the corresponding second frame image according to the response sum after the deep feature response and the shallow feature response are adaptively fused; judging the average peak related energy based on the threshold, and updating the deep feature model and the shallow feature model until the video image sequence is finished, so that the target can be effectively tracked, and theaccuracy is relatively high.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an object tracking method based on adaptive fusion of deep and shallow features and context information. Background technique [0002] Visual tracking has always been a key issue in the field of computer vision. In recent years, visual tracking algorithms based on correlation filtering have developed rapidly, and have shown certain advantages in tracking speed and accuracy. However, there are still some difficulties in the research of target tracking. External interference factors such as target occlusion, fast motion, and illumination changes directly affect the performance of tracking algorithms. [0003] Correlation filtering and deep learning-based methods are the current mainstream target tracking algorithms. Correlation filtering has become one of the current research hotspots in the field of target tracking due to its speed advantage brought by its fast calculati...

AI Technical Summary

Problems solved by technology

[0004] The above method only uses a single feature to describe the appearance of the target. In complex scenes, the recognition of the target is poor and it is easy to be disturbed.

The traditional correlation filter tracking algorithm has serious boundary effects, the artificial features used cannot describe the appearance of the tracking target well, and there is a lack of a suitable feature fusion strategy. These factors will seriously affect the accuracy of the algorithm and cannot be effective. tracking target of

Images