Core-related Filtering Tracking Algorithm Optimization And Hardware Testing In Occlusion Scenarios

With the development of computer technology,computer vision technology has now become a hot issue of current research and has been applied in many fields.Although the target tracking technology has made great progress after years of development.However,due to the unpredictability of tracking object motion and the complexity of the environment,there are still many difficulties to be solved in target tracking.For example,when the target is moving,there are problems such as occlusion,deformation,and rapid movement due to target tracking.These factors will cause the target model to drift and even cause the failure of target tracking.Aiming at the problem of target loss when moving targets are occluded in a complex environment,this paper improves the related filtering algorithm,adding confidence update strategy and SVM redetection strategy on the basis of KCF algorithm to improve the robustness of the algorithm.Experiments show that compared with the KCF algorithm,the proposed algorithm has improved target tracking accuracy and success rate by 13.8% and 17.4%,respectively.When the target is blocked or the target field of view is lost,the algorithm can still retrieve the target and achieve stable tracking.The main work of this paper is as follows:1.In the process of target tracking,many algorithms do not judge the reliability of the tracking results.Using unreliable data to update the model will only make the tracker increasingly unable to identify the target and cause model drift.In response to the above problems,we have improved the algorithm and added a part to determine the confidence of the tracking result.When the tracking target result is in a high confidence state,the tracking result can be considered valid,and the template is updated at this time.When the tracking target result is in a low-confidence state,it can be considered that the target may be blocked at this time,and the template is no longer updated to prevent model drift.2.When the tracking target is occluded,severely deformed or even loses the field of view,the traditional target tracking algorithm cannot perform effective re-detection,resulting in the loss of the tracked target and the tracking failure.To solve the above problems,we improve the tracking algorithm and add a re-detection mechanism.When the target is occluded until it reappears,we will find the target again.3.Carry out simulation experiments based on the above theory and build a hardware experiment platform.Use the UAV monocular camera to select the tracking target of interest and transmit the dynamic video to the lower computer for simulation processing.Track the effect when it arrives.