Siamese network video single target tracking method based on feature fusion

The invention relates to a Siamese network video single-target tracking method based on feature fusion, in particular to a Siamese network monitoring video single-target tracking method based on feature fusion, and aims to solve the problems that a Siamese network single-target tracking algorithm is relatively low in tracking capability when facing a complex environment and obvious background interference exists near a tracked target; the problems that the target cannot be accurately tracked, and the output tracking area is not accurate enough when some specific targets are tracked are solved. The constructed model is trained through the template area image set and the search area image set, feature maps of template images and feature maps of search images are output respectively, and the model sequentially comprises a ResNet-50 network based on a mixed attention mechanism and a twin feature fusion network; and inputting the feature map of the template image and the feature map of the search image into an RPN