Research On Visual Object Tracking Algorithm Based On Correlation Filter

The advent of the industrial revolution has freed people from heavy manual labor.In the21 st century,the advancement of computer technology has enabled some human brains to complete tasks that can be done by computers.The purpose of computer vision is to allow image acquisition equipment to be as capable as humans.Many excellent target tracking algorithms have emerged in the field of target tracking.Among them,correlation filtering has received extensive attention from scholars,and many target tracking algorithms based on correlation filtering have been proposed.However,there are still many challenges in practical applications that affect the effect of the tracking algorithm.This paper studies related filtering algorithms,and makes corresponding improvements to their shortcomings in the target tracking process,and conducts experiments to verify the effectiveness of the algorithm.The main work is as follows:The application of color histogram in STAPLE algorithm ignores the importance of element position.At the same time,when the target and the background are similar,the distribution of the background histogram appears similar to the target histogram,which will affect the effect of the histogram classifier.This paper proposes a tracking algorithm based on the histogram of background weights.According to the Euclidean distance to the center of the target and background elements,the weights are assigned,and the weights are used as the statistical weighted values of the histogram.In addition,the inhibitor is introduced,and the inhibitor is applied to the target histogram,thereby suppressing the part where the target and the background are similar.Finally,an adaptive update strategy is introduced to choose not to update the template with unstable tracking,so as to avoid the accumulation of tracking errors caused by model deviation.Experiments show that the algorithm in this chapter further improves the tracking accuracy of the algorithm.The use of a single feature will cause the tracker to reduce tracking accuracy in environments where some features are not robust to it.This paper proposes a multi-feature adaptive fusion tracking algorithm,which combines HOG features and CN features with adaptive weighting.HOG features are robust to lighting changing environments,and CN features are robust to some target deformations and background clutter environments,so the target can still be tracked stably in scenes of illumination changes and target deformation.At the same time,the two features are fused in the response layer,and the peak sidelobe ratios of the two features and their respective filter responses are used as the judgment of tracking accuracy.Finally,we can get the weight of the fusion.