Research On Infrared Target Tracking In Sky Background

Infrared target tracking is widely used,and it is the core technology of photoelectric measurement and control,aerospace,security monitoring and so on.However,infrared object tracking is a complex and difficult task,in the actual tracking process,the target may be in a variety of complex situation,including low contrast,size changes,occlusion,interference and illumination changes,etc.In order to realize stable tracking,the tracking algorithm must be adapted to these situation.As the target moving in the sky background most of the time,they mostly appear as area target or small target,and often show dynamic changes.What’s more,the infrared detector has high noise,also there is local florist,which makes it difficult to distinguish when the target is weak.In the sky background,clouds are the main interference factors to target tracking,the target may be moving close to clouds,occluded by clouds,changing weak under clouds and leaving clouds,etc.,and this should be solved to achieve stable tracking.This paper focus on appropriate tracking algorithms for target tracking under sky background,which is based on the characteristics of the infrared target in sky background.Struck is an target tracking algorithm based on online learning,by training classifier online to classify the candidate samples and achieve the goal of tracking.In this paper,the principle of Struck algorithm is deeply studied,and the performance of the algorithm is analyzed comprehensively.The stable tracking of infrared area target and dim target is achieved based on Struck,the main research contents are as follows:1)In-depth study on characteristics of background,target and noise of infrared image in sky background,their impact on tracking is analyzed and the main difficulties of infrared target tracking under sky background are summarized.2)Infrared area target tracking methods under sky background are studied.Through comprehensive experimental analysis of several different trackers,a tracking scheme is worked out based on Struck and an adaptive tracking method based on Struck is proposed which is robust against occlusion and target scale change.Experiments demonstrated that the framework is of high robustness for infrared area object tracking under sky background.3)Infrared dim target tracking methods under sky background are studied.Firstly,III infrared image preprocessing algorithm is studied.Then the stable tracking of dim target with low SCR under sky background is achieved through the combination of preprocessing algorithm and Struck.4)The efficiency optimization methods for Struck are studied,and two accelerated Struck frameworks are proposed.Experiments demonstrated that the framework based on not-all sampling can significantly improve the processing speed without effect the precision of Struck,and the method based on Median Flow can remarkably improve the processing speed of Struck when tracking the targets with a certain edge and texture feature.