Spatial-semantic perception attention tracking method based on adaptive weight template updating

The invention discloses a space-semantic perception attention tracking method based on adaptive weight template updating, belongs to the technical field of target tracking, and is used for solving the problem that a traditional tracking algorithm based on classification-regression is difficult to process continuous appearance change of a target. According to the method, firstly, a space-semantic perception attention model is established, and the importance of different regions and channels on a target feature map to target representation is identified online by using a single convolutional attention network with a loss function; then, enhancing a specific region and semantic information on the feature map by adding corresponding weights in space and channel dimensions; secondly, in order to better adapt to the appearance change of the target, a new template updating strategy is provided to adaptively adjust the contribution weight of the tracking result to a new template, and the reliability of the template i