Video target tracking method based on local weighted sparse feature selection

Abstract

The embodiment of the invention discloses a video target tracking method based on local weighted sparse feature selection, and the method comprises the following steps: carrying out the local appearance modeling of a target, and constructing a local discrimination dictionary for representation; sampling the search area of the current frame, and selecting a target candidate sample by taking the sparse decomposition coefficient of the sample as a feature; adding local noise energy to the reconstruction error through a weighting function, enhancing the representation capability and discriminationstability of the reconstruction error under noise interference, and selecting an optimal tracking result; utilizing an updating method combining long time and short time, using a self-adaptive noiseenergy threshold value as a condition for executing updating, and selecting two local dictionaries randomly for updating. According to the embodiment of the invention, the performance of sparse feature selection is improved through the methods of local discrimination dictionary learning, noise energy analysis and weighted decision, the purpose of improving the tracking accuracy is achieved, the interference of background noise information on target detection is suppressed, and the stability of model discrimination is improved.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a video target tracking method based on local weighted sparse feature selection. Background technique [0002] In the past two decades, many classic algorithms for specific scenes have emerged in the field of visual tracking research, such as tracking methods based on sparse representation, deep learning and correlation filtering. The sparse tracking method based on the particle filter framework belongs to the generative tracking method, and its core is to regard the target tracking as a matching optimization or similarity measurement problem in the feature space. For example, methods such as L1 tracking and prototype sparse tracking use sparse representation dictionaries and orthogonal PCA basis vectors to establish static target appearance models, which are very effective for occlusions. However, the problem of target template mismatch caused by target appearance changes is i...

AI Technical Summary

Problems solved by technology

However, the multi-layer structure of the deep network will increase the computational complexity and reduce the real-time performance of the tracking process.

In the follow-up, methods combining deep learning and correlation filtering appeared, but these methods have a large amount of calculation and are easily limited by CF boundary effects.

Images