Research On Target Tracking Algorithm Against Cloud Occlusion Based On KCF

In recent years,cruise missiles,air-to-ground missiles and other infrared imaging precision guided weapons have developed rapidly and played an increasingly important role in modern warfare.When the seeker based on infrared imaging is tracking the ground target,the cloud cover often occurs,resulting in the loss of the tracking target.Based on the KCF algorithm,the paper makes an in-depth study on the long-term tracking of ground targets under the condition of random cloud occlusion.The main work of this paper is as follows:In order to solve the problem that the KCF algorithm is not strong in feature description,cannot adapt to the target scale change,and the search area is fixed,a scale adaptive kernel correlation filter tracking algorithm which integrates the color statistical features is proposed.The integration of color statistical features can find the location of the target more accurately;The scale filter can solve the scale change of the target.Searching the feature of the edge of the area and replacing it with the inner feature extension can speed up the algorithm.Experimental results show that the algorithm can track the target more accurately and quickly.In order to solve the problem that KCF algorithm tracking drift leads to tracking failure in the case of occlusion,an anti-cloud occlusion template adaptive update kernel correlation filter tracking algorithm is proposed.According to the peak value of the response graph and the average energy of related peaks,the target can be judged to be blocked or not.If the target is blocked,stop updating the template parameters.When the target is completely blocked,the detection module is used to find the target.When the target scale changes too much,the template of large scale was retrained.Experimental results show that the algorithm can track the target stably for a long time in the case of cloud occlusion.In order to solve the problem of tracking drift caused by the lag of the occlusion discrimination basis when the target is partly blocked by the cloud,considering that the block algorithm will affect the tracking stability,this paper proposes a tracking algorithm of anti-cloud occlusion kernel correlation filtering based on the mask strategy.The targetis processed with different masks,and each mask module trains the corresponding filter template respectively.When the target is partially blocked by the cloud,the mask module with better output response is used to continue tracking the target.Experimental results show that the algorithm can track the target more accurately and stably when the target is partly blocked by the cloud.