Multi-scale anti-occlusion target tracking method based on manual feature fusion

Abstract

The invention provides a multi-scale anti-occlusion target tracking method based on manual feature fusion, which is high in robustness and small in calculation amount, and can be operated on most tracking platforms. According to the invention, the scale pool is designed to carry out multi-resolution sampling, and the scale change of the target in the movement process can be quickly processed; a method for fusing response results of two characteristic filters according to a correlation peak average energy index is provided, so that the integration of the advantages of HOG characteristics and CN characteristics is realized, and the distinguishing capability of a tracker is improved; by performing shielding judgment and evaluation on the tracking result of each frame and additionally designing an SVM re-detector, re-detection is performed on a nearby area before a target is lost after the target is lost, so that the anti-shielding capability of the tracker is improved. Tracking results of the two trackers are independently evaluated, and the learning rate is adjusted according to an APCE index, so that respective self-adaptive updating is realized.

Description

technical field [0001] The invention relates to the technical field of digital image processing, in particular to a multi-scale anti-occlusion target tracking method based on manual feature fusion. Background technique [0002] In the daily production and life applications of computer vision technology, it is often necessary to stably track specific targets in video images, and obtain information such as the position and size of specific targets in the picture in real time. [0003] The current mainstream target tracking algorithms can be divided into two categories: the first category is the target tracking algorithm based on correlation filtering, which has high calculation speed and good tracking effect. The basic principle is to train the filter template in the initial frame, calculate the similarity between the candidate sample and the filter template in the subsequent frame, output the response map, and take the position of the maximum response value as the target posi...

AI Technical Summary

Problems solved by technology

The second category is the target tracking algorithm based on deep learning, which appeared relatively late, but developed very rapidly

Although the traditional correlation filter tracking algorithm has a great advantage in speed, its performance in common complex scenes still needs to be improved

In order to improve the robustness of the algorithm, in recent years people have proposed many improved methods based on it, but they are often at the expense of the real-time performance of the algorithm, so that the correlation filter tracking method loses its most basic advantages, which also limits The scope of application of the algorithm

Images