Unmanned aerial vehicle ground target tracking method based on multi-layer feature self-attention transformation network

Abstract

The invention relates to the technical field of computer vision, and particularly discloses an unmanned aerial vehicle ground target tracking method based on a multi-layer feature self-attention transformation network. According to the method, an Alexnet network and a self-attention transformation network are fused, and the specific steps are as follows: firstly, the Alexnet network is adopted to extract third, fourth and fifth layers of features of a template image and search the image to obtain a high-resolution feature map and a low-resolution feature map of the image, and then the high-resolution feature map and the low-resolution feature map are input into the self-attention transformation network to realize self-attention transformation of the multi-layer feature map; target features from different hierarchies are aggregated, the dependency relationship of the features between the hierarchies is increased, and the ability of an unmanned aerial vehicle platform to track complex scene targets is adapted; and finally, performing related convolution operation on the obtained target feature maps of the template branch and the search branch to obtain a similarity score map of the target, performing classification and regression on the target object, and determining the optimal position of the target tracked by the unmanned aerial vehicle. According to the invention, accurate tracking of the ground target by the unmanned aerial vehicle platform is realized.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method for tracking ground targets of unmanned aerial vehicles based on a multi-layer feature self-attention transformation network. Background technique [0002] UAV-to-ground target tracking is an important research topic in the field of computer vision. In recent years, with the rapid development of the UAV industry, UAVs have been widely used in the fields of urban safety prevention and control, water conservancy survey, forestry survey, forestry pest monitoring, road maintenance and inspection. [0003] Among the many application tasks of UAVs, UAV-to-ground target tracking technology has important research significance and has become a research hotspot in the field of UAVs. UAV-to-ground target tracking is to use computer vision methods to track ground targets in aerial video, obtain their motion trajectory information (such as position, speed, acceleration, etc....

AI Technical Summary

Problems solved by technology

[0007] Although the tracking effect of the twin network method on some videos is better, because this type of method only uses the network model to extract the target features and does not further process the features, when the ground target acquired by the UAV encounters such as severe occlusion, large When there are technical problems such as deformation or target disappearance, the tracking effect of the traditional twin network method is significantly reduced, so it is difficult to meet the robustness requirements of the actual application scenario.

Images