Research On Robust Motion Object Tracking Algorithm Based On Cascade Detector-Discriminative Correlation Filter

Object tracking in sequence images or videos is a classic problem in the field of computer vision,it is also the basis for applications such as intelligent monitoring and visual navigation which has important significance in both civil and military fields.However,how to perform long-term and stable tracking on the basis of ensuring real-time performance is still a problem in the field of tracking.In recent years,algorithms based on correlation filtering have good performance in the field of target tracking.It has great potential in both speed and performance.Therefore,this paper mainly studies the tracking algorithm based on correlation filtering,but the classic correlation filter tracking still has certain limitations,and it is difficult to solve the problem of long-term tracking.This paper first expounds the theoretical basis and related problems of target tracking,then introduces the correlation filter tracking algorithm and analyzes its advantages and disadvantages.In order to introduce the advantages of correlation filtering into the field of long-term tracking,this paper studies and proposes a kind of robust tracking algorithm for moving targets based on cascaded detector.This paper analyzes the advantages and disadvantages of trackers based on correlation filters,and proposes a long-term tracking algorithm based on cascaded classifier.The method introduces a cascade classifier as a detection module which performs feature modeling around the target and trains different classifiers in the initial frame.In the normal tracking process,the classifier performs cascade detection to determine the exact position of the target and update the classifier.The result of the detection is then combined with the tracking result to reduce the tracking drift caused by the accumulation of errors.For the case of interference,the peak response and the average peak correlation energy are used to judge in time to ensure high confidence update of the model.When the interference disappears,we design a recapture algorithm to retrieve the target in time so that we can continue to track.Finally the performance of the proposed algorithm is tested on different data sets.Experiments show that the proposed method can effectively overcome the drift and loss of long-term tracking process,and has broad prospects in practical application.