Design And Implementation Of Aircraft Target Detection And Tracking Syste Based On Deep Learning

In recent years,with the popularization of artificial intelligence technology,technologies such as face recognition and autonomous driving have begun to emerge and become more mature,and people’s lives have become more and more intelligent.The application of target detection and tracking to aircraft identification will bring new changes to both civil aviation and military aviation.In the field of civil aviation,the flight scheduling of airports has a long-term application prospect,and in the field of military aviation,the reconnaissance of enemy aircraft can play a special role in combat.This paper proposes a method based on deep learning for target detection and tracking of aircraft,designs a real-time detection system for input video,and implements it.This paper mainly does the following work:1.Combining the research background and research significance of the subject,starting from the research status at home and abroad,analyzing the current cutting-edge research results in this research field and comparing the latest excellent algorithms,clarifying the ultimate goal of the system,the desired effect,and the purpose of this paper.2.This paper starts from the demand analysis,designs the aircraft target detection and tracking system,and divides the system into eight functional modules,namely:parameter management module,real-time image acquisition and processing module,training management module,user login module,user management module module,data set management module,model management module,real-time detection module.This paper analyzes and designs the functions and functions of each module one by one.They are interrelated and communicate with each other to ensure the normal operation and functional integrity of the entire system.3.Implement it on the basis of system design,achieve the purpose of training model and real-time detection,use related technologies such as front-end framework VUE and back-end framework Django to carry out system coding development,and complete a set of visualization based on web technology.The tool is used for users to upload self-made data sets,customize parameters to train models,and integrate functions such as model,data set,and training and detection task management in system design.Finally,the effect of real-time detection of the aircraft target in the picture can be achieved during the playback of the input aircraft video.4.This system implements the aircraft target detection algorithm based on YOLOv5,adopts the moasic data enhancement method at the input end to show a good effect in the detection of small target aircraft,and adopts the adaptive anchor frame calculation to make the network get Optimal anchor box location.The focus structure is combined in the backbone network to improve the training speed.The traditional FPN+PAN structure is adopted in neck to strengthen the fusion ability of network features.The bounding box loss function is introduced in the output layer.In the model training process,the parameters are continuously adjusted,and the final algorithm can meet the requirements of the system requirement analysis in the aircraft target detection and tracking of this system.5.The function test is carried out for each module of the system,and on this basis,the performance test of some indicators of the system and the accuracy test of the accuracy of the algorithm to identify the aircraft arc carried out.This paper finally completes an aircraft target detection system.Users can use this system and deploy the connected cameras in airports and other areas,so as to achieve target detection of aircraft at the airport and then schedule flights.Or used in reconnaissance in the military field,it can also effectively identify enemy aircraft.