基于分散注意力机制的无人机目标跟踪算法OACSTPCD

A UAV target tracking algorithm based on the distracting attention mechanism is proposed for the target tracking scenarios of UAV platforms with small targets,large deformations,complex backgrounds,and easy target occlusion.Firstly,for the problems of complex background and target occlusion,a distraction mechanism is introduced in the feature extraction residual net-work to increase the interactions between convolution channels and give self-attention weights between channels.Secondly,for small targets and large scale variations,the feature pyramid structure is added to increase the proportion of high semantic informa-tion in the feature map to improve the tracking ability of multi-scale targets in UAV scenes and improve the tracking accuracy.The proposed algorithm is tested on the UAV public dataset UAV123 by using one-pass evaluation mode,and the accuracy of the pro-posed algorithm reaches 82.7%and the success rate reaches 61.6%.The experimental results demonstrate that the test results of the proposed algorithm outperform the current mainstream target tracking algorithms and effectively improve the accuracy and robust-ness of UAV tracking for specified targets in complex scenarios.