Research On Moving Object Tracking Algorithm With Template Updating Strategy In Complex Background

Artificial intelligence has developed rapidly in recent years,and computer vision,as an important area of artificial intelligence,has received close attention from researchers.Object tracking,as an important part of the field of computer vision,is also widely used in many fields.However,the object tracking algorithm in the actual industrial production environment faces the interference of challenging complex environments such as fast object movement and high product similarity.Therefore,effective tracking of moving objects in complex backgrounds is a current research problem.Based on the current problems,we propose a visual object tracking algorithm based on a template update strategy.The use of this strategy can effectively solve the problem of object tracking failure under complex backgrounds and effectively improve the overall success rate and accuracy of the algorithm.The implementation of our strategy has three main processes:First,we judge the current environment of the target and update the strategy using the proposed template.Environmental judgment uses a comprehensive confidence domain,gray-scale histogram,and APCEm value method.If the current environment is complex,the proposed template update strategy is used.The fast LK optical flow method is used to predict where the target might appear in the next frame.Then,the target detection is performed in parallel at the predicted position and the target current position,and the more responsive is selected as the final update result.Conversely,if the current environment is normal,the original template is used for tracking.Subsequently,we apply the proposed strategy to the particle filter Diagnose algorithm and the correlation filter KCF algorithm and BACF algorithm.Experimental verification on the OTB2015 data set,the experimental results show that the DiagnoseM,KCFM and BACF_M algorithms using the proposed strategy have significantly improved the overall success rate and accuracy compared to the original algorithms.And it has strong tracking robustness under challenging rotation changes,illumination changes and fast movements.Especially in the complex background,our improved tracking algorithm has obvious advantages.Finally,we use the optimized BACF_M algorithm to track objects in the actual industrial production environment.We independently collect complex industrial production videos and perform corresponding preprocessing,and finally use the proposed BACF_M algorithm to track industrial video sequences.Experiments show that the BACF_M algorithm using template update strategy has better tracking effect in the actual industrial environment.It shows that the strategy we mentioned has a positive impact on actual industrial production.