基于Swin Transformer和注意力机制的红外无人机检测算法OA

Infrared drone target detection has broad application prospects in both military and civilian fields,and it is a hot research topic in the field of computer vision.Due to the small scale of drone targets and the complex and ever-changing aerial environment,existing detection algorithms generally have low detection rates and high false alarm rates.Aiming at issues such as poor detection of infrared drone targets in complex scenarios,this article proposes a ST-YOLOA infrared unmanned aerial vehicle target detection model.Firstly,in order to improve model performance and effectively capture global information,an STCNet backbone feature extraction network is constructed using the Swin Transformer network architecture and coordinated attention(CA)mechanism;Secondly,in the feature fusion section,a PANet path aggregation network with residual structure is used to construct a feature pyramid to enhance the overall feature extraction ability,while improving the up and down sampling method to enhance detection ability;Finally,the decoupled detection head is used to predict the position of the drone target.The proposed model has a detection accuracy of 92.8%and a detection speed of 22frames/s,which is verified by experiments on an infrared drone dataset.This indicates that the model has better detection performance compared to other models,especially in complex environments,and basically meets the real-time detection requirements.It has practical significance for detection in multi drone target scenarios.