Implementation Of Infrared UAV Detection And Tracking System Based On Deep Learning

Low-altitude UAVs are developing rapidly in the field of military and civilian applications,and the security risks it brings have become a hot topic of social concern.How to effectively identify and track UAV targets is a prerequisite for implementing countermeasures.Therefore,UAV identification and tracking technology has become one of the key research directions in the field of anti-drone.In the field of visual monitoring,infrared image has the characteristics of long imaging distance and can work day and night,which can better accomplish the task of UAV identification and tracking.However,due to the imaging characteristics of infrared images(the target image is blurred,low contrast and easily blends into the complex background,which makes the target recognition accuracy low),the detection and tracking are much more difficult.To address the above problems,this paper investigates the problem of infrared UAV target recognition and tracking by means of deep learning,and the main research work is as follows:(1)Aiming at the problems of less feature information,serious feature loss and low recognition accuracy in the process of infrared UAV target recognition,an infrared UAV target detection method based on YOLOv7 is proposed.By introducing attention mechanism,the feature expression ability of target region is enhanced and spatial information content of the image is improved.The improved serial connection mode is used to connect the channel attention module to spatial attention module.In combination with the target channel feature information and spatial feature information,the improved structure reduces negative impact of the channel attention on infrared image recognition,and can better realize strengthening effect on the infrared target feature.The SIo U loss function based on angle vector regression is selected as the frame loss function,which further improves the convergence and detection accuracy of the model.Experimental results show that reasoning speed of the improved algorithm model reaches 43 frames/s,accuracy is 95.4%,recall rate is 87.3%,and m AP is 96.1%.Better results are obtained in the infrared UAV detection task.(2)Aiming at tracking failures caused by irregular movements such as acceleration,sudden stop,and vertical climb during the movement of unmanned aerial vehicles,which lead to large positional deviations between the prediction frame and the detection frame,an improved Strong SORT tracking algorithm is proposed,which uses the BIo U edge matching method to increase the search area during edge matching and mitigate the impact of motion estimation bias in the matching space on the tracking process.(3)A set of gimbal active tracking control algorithm is designed to solve the problem brought by the limitation of fixed monitoring field of view in the process of UAV identification and tracking to the UAV monitoring work.By delineating a stable observation area to realize the lens following of the target,the system obtains a larger monitoring field of view and improves the flexibility of UAV target tracking.(4)Combining the above proposed algorithms,an infrared UAV detection and tracking system is designed,which can realize video file detection and tracking,real-time gimbal monitoring,and gimbal active tracking by displaying the recognition tracking results and completing function switching through the visual interactive interface.The minimum detectable IR UAV target of the system is 10×9 pixels,and the processing speed is 25 frames/s,which meets the demand of real-time tracking.