Twin network target tracking method and system based on convolutional self-attention module

Abstract

The invention provides a twin network target tracking method and system based on a convolutional self-attention module. The method comprises the following steps of constructing a feature fusion network model; learning the target image features on the template branches and local regions in the target image features of a search region through a convolutional neural network model to obtain the corresponding local semantic information, and aggregating the local semantic information to obtain the global context related information; pre-training the feature fusion network model; using the pre-trained feature fusion network model to extract the target image features and the search area target image features in the template branches, and introducing the features into the classification branches and regression branches of an area suggestion network with the anchor points; carrying out depth cross-correlation convolution calculation to obtain similarity scores; and performing target tracking on the target candidate block with the maximum similarity score. According to the invention, the global matching accuracy of the target image and the target image of the search area can be improved, and the more accurate tracking is realized.

Description

technical field [0001] The present invention relates to the technical field of computer vision and image processing, in particular to a twin network target tracking method and system based on a convolutional self-attention module. Background technique [0002] In the field of computer vision and image processing, visual tracking has many fundamental applications in computer vision. For example, autonomous driving, video surveillance, traffic vehicle monitoring, and human-computer interaction. Therefore, as tracking becomes more practical and real-time, more and more tracking-related applications become more and more common in real life, which also becomes more and more valuable for the research of video tracking technology. [0003] In general, visual tracking remains a challenging task due to many reasons such as appearance change, deformation, fast motion, and occlusion. In recent years, Siamese network trackers based on convolutional neural networks (CNN) have been wide...

AI Technical Summary

Problems solved by technology

[0005] In view of the above situation, it is necessary to solve the problem that some visual tracking algorithms in the prior art ignore the context-related information generated in the time dimension of continuous frames, resulting in the loss of a large amount of foreground information and the generation of redundant background information, which leads to It is difficult to deal with the impact of complex appearance changes

Images