Video Tracking Based On Camshift Algorithm And Multi-feature Fusion

Video tracking has always been one of the research hotspots in the field of computer vision.In addition,it has important application value in military,public security,human-computer interaction,intelligent transportation and many other fields,so this technology has a very broad application prospect.However,because video tracking is different from the recognition of non-moving objects,the scene changes a lot,such as the tracking target color is similar to the background color,the target scale is scaled,and the rotation,shielding,lighting changes and so on,all of which will lead to the target tracking process becoming more complex and unstable.Although many researchers have proposed various tracking algorithms or improved existing algorithms over the years,no algorithm has been found that can achieve real-time and robust performance.Therefore,tracking targets accurately under complex conditions is still a challenge and still has great research significance.Camshift algorithm is an algorithm to track the target by using the color characteristics of the target,with good real-time performance and good tracking effect for the target with a single background color.However,in the case of target occlusion or similar background color interference and other responsible scenes,the limitation is very great,which may easily cause the problem of target tracking loss.In order to improve the efficiency of target tracking and enhance the anti-interference and stability of the algorithm in complex scenes,this paper first makes some improvements to the shortcomings of the Camshift algorithm:introducing texture features for feature fusion to enhance the reliability of trackingresults;A Kalman prediction mechanism is proposed,which predicts the location of the target in the next frame,and uses the prediction result as the initial candidate window of the next frame to reduce the number of iterations of the algorithm.Finally,the template is updated in a timely manner.Then based on the feature-shifted Camshift algorithm,a new target tracking algorithm is proposed.First,the Gaussian kernel function is used to weight the target area so that the center weight of the target is different from other positions,which can prevent interference from other positions and improve the anti-noise ability of the algorithm.Finally,in order to accurately track moving targets and prevent occlusion,this article uses kalman filter to predict moving target position.