Target Tracking System Based on Dynamic Scene

Abstract Under the scenario of moving target tracking is very important in fields computer vision and research direction, its purpose to give vision, double eye for simulation human visual information processing feedback process, process can obtain trajectory provide related data tracking, The relevant results have been successfully developed many such as human-computer interaction, security monitoring, intelligent monitoring face recognition. However scene vulnerable disorder interference, light, shade, scale complex situations, change apparent dynamic movement target, achieve different stable method has significance engineering events field value, existing popular nuclear-related filter algorithm was improved, To solve problem occlusion multi-scale a new based on multi-feature anti-occlusion proposed by using kernel correlation filtering method.