Moving Detection And Ground Target Tracking Based On The Airborne Platform

This paper is devoted to the research of moving target detection and tracking technology on the airborne platform.In the airborne platform,the image background will move with the movement of the imaging platform.The pixel model method will generate pixel-level errors in the registration process,which will lead to the model not being able to exactly match the pixel to be detected and the detection performance is not good.In the aspect of tracking algorithm,the control system of the airborne platform is more complex and requires higher real-time performance.In addition,the target needs to be stably tracked for a long time during the tracking task.The research content of this paper starts from moving target detection and moving target tracking.The detection part of this paper is based on the Two-channel Single Gaussian Moving Background Model technology.Compared with the Temporal Difference method,this algorithm has better detection effect and is faster than the Optical-Flow method.On the basis of the traditional static background modeling,this method reduces the moving background by affine transformation through the region SIFT feature matching,and reconstructs the region model to adapt to the moving target detection under the moving platform.In order to deal with the "aging" of the model which is caused by redundant background information,this method introduces the model self-reconstruction technology.On this basis,combining the detection results in different time scales,this paper proposes an adaptive time domain multiscale detection method based on affine transform similarity measurement,which can improve the detection rate of the original algorithm and effectively reduce the number of false alarms in the detection results.In the part of target tracking algorithm,an adaptive learning rate algorithm based on KCF is proposed.Aiming at the traditional KCF algorithm is not good at resisting similar interference,this paper proposes an adaptive learning rate to improve the model updating process of KCF algorithm,which has better interference ability.The target recapture module is added to make the algorithm applicable to the situation of long time target tracking.Finally,the algorithm is implemented on TMS320C6455 single-core fixed-point processor through feature reduction and fixed-point transformation.The detection algorithm designed in this paper has a detection rate of more than 95% in simple scenes and more than 80% in complex scenes for targets with a target pixel larger than 30×30.The tracking algorithm runs on the TMS320C6455 single-core fixed-point processor,with high real-time performance.