The Research On Detection And Tracking Method Of Ground Moving Targets On Moving Platform

Target detection and tracking technology in image processing are widely used in the field of guidance,which improves the precision strike capability of missiles.In recent years,the integration of miniature missiles for detection and positioning as well as strike has gotten more and more attention in the military.As the traditional Shoulder-fired missiles have low accuracy for striking the target with low mobility and small diversification,it is urgent to develop a miniature missile for automatically detect and track ground moving targets.This thesis mainly focus on image registration,target detection,and target tracking in the miniature missile,which is used to improve the accuracy and adaptability of miniature missiles in small-scale wars.The main works include the following four parts:Firstly,the overall flowchart of miniature missile attack targets has been introduce,which includes target aiming phase,target detection and correlation numbering phase,target tracking and striking phase,and has been in-depth analyzed.Then,an improved SURF-LK registration method is proposed by employing image sequence registration and the experimental analysis of the image point registration method in order to deal with image jitter and background change in target aiming phase.The experimental results verify its effectiveness,validating that the proposed method can effectively improve the registration speed while ensuring the registration accuracy.Next,a V-SORT algorithm combining the improved Vibe with SORT multi-objective information is proposed in the target detection and association numbering phase.The proposed V-SORT method optimizes the Vibe+ method based on SURF-LK registration,and adds the part to eliminate foreground noise and long-term pixels.The experimental results obtained by applying the proposed V-SORT method to the two real-time sequences,which respectively includes 5 targets and 9 targets,validate that the V-SORT algorithm can accomplish the tasks of multi-target detection,data association and ID assignment.Finally,four scale-updated tracking strategies including STC,DSST,FDSST,and ECO-HC,are experimented on four real-time video sequences in the target tracking phase.With comparisons of the target center position distance between the tracking results and the true value,the target area overlap rate,and the algorithm’s cost time the ECO-HC algorithm can robustly track the target from small to large in the image when the miniature missile approaching the target.Moreover,The ECO-HC algorithm satisfies the technology requirements including the calculation speed and accuracy in the target tracking phase.