Design Of Embedded Target Tracking System Based On Improved Camshift Algorithm

Vision was an important communication channel between human beings and the outside world.Computer vision understood objective things by acquiring,recognizing and understanding objective information as well as simulating human vision.Moving target detection and tracking technology,widely used in intelligent monitoring,robot,intelligent transportation and other fields,was one of the most important research directions.In order to solve such problems as poor tracking performance in Camshift algorithm with tracking targets similar to interference and sheltered,an improved CamShift tracking algorithm combined with KNN detection technology of moving target and Kalman filtering was proposed.In this paper,common motion detection methods were introduced: the background difference method was unstable,and the mixed Gauss model was influenced by the shadow.Therefore,the tradition methods combined the motion detection algorithm based on the KNN background segmentation could achieve the self-adaptive learning according to the changes of the background,and the automatic segmentation of the moving target is carried out effectively.Kalman algorithm was used to predict and correct the position of moving objects,reducing the tracking range and increasing the speed of CamShift algorithm.As for multi-target tracking problem,we set up target class and tracked each target separately.The embedded raspberry 3B+ platform was built to track moving objects in embedded system.On the embedded raspberry platform,image processing,moving object detection and improved algorithm tracking were carried out,with the test video collected.According to the experiment analysis,the algorithm could track the target accurately in the case of similar interference and shelter.And the target class was set up in the multi target tracking,whose independence was based on the improved algorithm tracking.Compared with the traditional tracking algorithm,the improved algorithm raised the robustness and accuracy of the tracking.