Research On Moving Target Tracking Algorithm Based On Kernel Correlation Filtering

Moving object tracking,as a key research content in the field of machine vision,is widely used in various fields such as intelligent monitoring,intelligent transportation,and humancomputer interaction.However,because the tracking scene is complex and changeable,and the tracking effect is easily affected by interference factors such as occlusion,size change,and fast movement,there are still many challenges and problems in designing a robust,versatile,and accurate object tracking algorithm.Therefore,this paper studies Kernel Correlation Filtering tracking algorithm and improves it from four aspects: position prediction,feature fusion,scale update and model update,which significantly improves the accuracy and robustness of the algorithm.The main work of this paper is as follows:(1)KCF and particle filter tracking algorithm are studied.Among them,a particle filtering algorithm based on weighted random resampling is proposed to solve the "particle degradation" problem in the particle filter algorithm.Combining the characteristics of the KCF algorithm and improved particle filter algorithm,KCF tracking algorithm is improved from three aspects of position prediction,interference detection and model update,making the algorithm more robust under occlusion interference.(2)The characteristics of different features such as gray features,color features,and gradient features are studied.In order to solve the problem of insufficient object representation ability of a single feature,this paper proposes a multi-feature fusion strategy,which makes full use of the multi-channel feature advantage of the kernel correlation filtering algorithm.The principle and advantages of the Vibe object detection algorithm and the multi-block detection mechanism are studied,and a multi-block detection mechanism based on the Vibe algorithm is established to accurately analyze the changes in the scale during tracking.Based on this,a scale adaptive update strategy is proposed to realize adaptive adjustment of the object scale.(3)In order to test and verify the improved tracking algorithm of this paper,this paper designed a moving target tracking test system on Visual Studio 2013 platform using Open CV library and Qt cross-platform development tool.The improved tracking algorithm in this paper can accurately and stably track moving targets,and feedback the coordinates,frame rate,and size of the tracking frame of the tracking object.