A Target Tracking Method Based on Posterior Template Dictionary Learning

Abstract

The invention discloses a target tracking method based on posterior template dictionary learning, which mainly includes the following steps: local recombination based on target template sub-blocks, local sub-block search based on current observation, half-image search based on sparse reconstruction, based on Modules such as template incremental learning for sparse principal component analysis. The present invention proposes a posteriori template dictionary learning strategy based on current frame observation data for the first time. On the one hand, this strategy utilizes the information of target candidate images to guide the template dictionary construction process, so that the constructed dictionary template is as consistent as possible with the current target candidate images, which can effectively represent the apparent changes of targets. On the other hand, target candidate images are only used in template dictionary construction, so over-fitting can be effectively avoided. Experimental results show that, compared with traditional dictionary construction and update methods, the method proposed by the present invention can track the apparent changes of the target more effectively. This algorithm is a general algorithm and has wide application prospects.

Description

technical field [0001] The invention mainly relates to the field of target tracking in computer vision, in particular to a target tracking method based on posterior template dictionary learning. Background technique [0002] Object tracking is a very active research topic in the field of computer vision. Since it was proposed at the end of the last century, many scholars have joined the ranks of research. It is one of the frontier issues of current research at home and abroad. Object tracking is a middle-level part in the field of visual analysis, which lays the foundation for subsequent high-level visual analysis, so it has very important research value. [0003] The most critical component in object tracking algorithms is the appearance model. The appearance model, simply put, how to express the appearance of the tracking target object concisely and effectively. Generally speaking, the appearance model of the target can be mainly divided into two categories: generative a...

AI Technical Summary

Problems solved by technology

[0006] Although the appearance model based on sparse representation has achieved great success in dealing with occlusion and noise, the model still has the following problems: under the tracking framework of sparse representation, the candidate samples of the current frame image pass the linear reconstruction error of the template dictionary to evaluate

Therefore, when the target's appearance changes greatly, the template dictionary determined by the previous tracking results will not be able to represent the target's apparent change, resulting in tracking failure

In addition, the traditional sparse representation tracking algorithm simply replaces the old target template with the latest tracking result, and it is easy to introduce errors in the tracking result (such as caused by noise and occlusion) into the template dictionary. When the error accumulates to To a certain extent, it will lead to the problem of model drift (model drifting)

Images