Research On Autonomous Recognition And Tracking In Surveillance Mission Of Cataloged Space Targets

The surveillance mission of catalogued space targets is crucial to space-based situational awareness and other surveillance services.However,most existing recognition and tracking methods for catalogued space targets face issues such as slow recognition speed,poor recognition effect,and easy loss of tracked targets.These challenges significantly limit the effectiveness of catalogued spatial target surveillance in complex space missions.Therefore,this thesis aims to provide a detailed study of autonomous recognition and tracking in the surveillance mission of catalogued space targets.Given the large number of catalogued space targets in the sensing field of the space-based space target surveillance system,existing recognition algorithms for space targets face the problems of slow recognition speed and low recognition accuracy.This thesis proposes a recognition method of catalogued space targets based on improved YOLOv5 s to address these challenges.The proposed method includes a pretreatment method based on Discrete Wavelet Transform,which sharpens and enhances recognition images.Additionally,a Global Attention Module is added to reduce the false detection of black objects.This recognition method improves the mission’s speed and accuracy by adjusting Anchor’s number and size.Simulation experiment results show that the proposed catalogued space target recognition method achieves an improved accuracy of 95.30% and a speed of 81frames/second.Regarding the existing tracking algorithms for space targets,the problems of easy target loss when the tracking target is occluded need to be addressed.This thesis proposes a tracking method of catalogued dynamic space targets based on a Convolutional Neural Network to address this challenge.This method includes a dynamic space target detection fusion to separate all dynamic space targets from the space background.Further separation of catalogued dynamic or static and suspicious dynamic spatial targets is achieved through threshold matching of 2D location information.Finally,an online tracking method of catalogued dynamic space targets based on Deepsort is developed,which can accurately and stably accomplish the tracking task in the surveillance mission of catalogued space targets.Simulation experiment results show that the developed online tracking method achieves stable tracking of all catalogued dynamic space targets under occlusion,improving tracking accuracy to 93.88% and tracking precision to 71%.Based on the above-researched autonomous recognition and tracking of catalogued dynamic space targets,a ground simulation experiment platform for space grasping is built in the laboratory.The vision-sensing device is integrated with the proposed autonomous recognition and tracking method.Verification experiments of the function of space target recognition and tracking tasks are carried out on the ground simulation experiment platform for space grasping.An upper computer visualization interface is designed for intuitive and convenient access to space target surveillance results.The functional verification experiments show that the developed autonomous recognition and tracking of catalogued dynamic space targets has good feasibility in the surveillance mission of catalogued space targets.