Research On Detection And Tracking Of Weak And Small Targets In Sky Background

Targets such as missiles and drones have small radar cross-section area,which makes them difficult to detect quickly by radar during warfare.It is significant to use complementary means to quickly counter the weak moving targets in the air.This paper systematically studies the key technical aspects of image preprocessing,target detection and target tracking for weak target detection and tracking problem.A preprocessing method combining bottom-hat and corners detection is proposed for the image preprocessor,which realizes a fast but rough extraction of weak targets in complex images,and outputs the extraction results as candidate targets to detector.Aiming at the problem of a large number of false alarms in the candidate targets,a logistic regression classifier with high recall is proposed to filter the candidate targets,which reduces the time consumed in subsequent confirmation which need multiple frames.Aiming at the problems of small imaging area and low signal intensity of weak targets,an idea of optical flow trajectory is proposed and realizes the detection of weak moving targets in high-dimensional feature space.Aiming at the problem that traditional anomaly detection algorithms are too complex,the two schemes of fast-LOF and isolation-forest are introduced to achieve more efficient target detection in complex environment.Aiming at the problem that the appearance model of weak target is unreliable,guided filter is introduced for KCF(Kernelized Correlation Filter)to keep edges while reducing noise,which improves the anti-noise ability of the tracker.Aiming at the tracking drift which caused by rapid motion of weak target or background while shrinking the tracking window,the particle filter is introduced to estimate the target position roughly for KCF,which realizes stable,accurate and fast tracking of fast moving targets even if the camera is shaking.Multiple sets of experiments are carried out for the detector and the tracker.The efficiency and accuracy of detector are verified.The accuracy and efficiency of the tracker in various scenes are analyzed,which proves the value of the tracker in practical engineering.