Research On Infrared Small Target Detection And Tracking Algorithm Under Sky Background

As the key techniques of infrared warning and infrared guidance,infrared small target detection and tracking are widely used in military defense.An infrared small target detection and tracking system consists of four parts:Image data collection,image preprocessing,target detection and target tracking.This paper focuses on the researches of false alarm suppression and target occlusion problems in the infrared small target detection and tracking process,and the details are listed as follows:To cope with the false alarm problem,this paper proposed an infrared small target detection method based on guided filtering and data correlation.Firstly,the background of the image was suppressed via guided filtering;the target was enhanced based on a set of adaptive gain adjustment coefficients,and the spatial locations of candidate targets were obtained through an appropriate threshold segmentation.Due to the fact that the spatial locations of false alarms are different from the ones in the frequency domain,this paper utilized the minimum Euclidean distance to associate these two sets of test results and suppressed the false alarms.Finally,the target trajectory was drawn by means of inter-frame data association,and the detection and localization of infrared small targets in the infrared sequence were realized.For the target occlusion problem,this paper proposed an infrared small target tracking method based on sparse representation and transfer constrained particle filter.Firstly,a sparse representation model of infrared small targets under occlusion constraint was established.Secondly,the saliency information was taken as a priori to implement the target state prediction and particle sampling processes.Lastly,a particle filtering model based on the differences of the sparse reconstruction errors of the target samples was established,and the target state was estimated with an online template updating strategy.The theoretical analysis and experimental results show that the proposed algorithm has a good suppressing effect on the false alarms appearing in the flocculent clouds and cloud-sky boundaries.During the tracking process,this model has a good characterization of the areas where infrared small targets are actually obscured,and can achieve a robust and accurate tracking of infrared small targets in complex sky backgrounds.