Research On Mean Shift Video Target Tracking Algorithm Based On Multi Feature Fusion

Video object tracking is a core topic in the field of computer vision.It has been widely concerned in the fields of video surveillance,intelligent transportation,robot navigation,human-machine interaction and so on.With the development of information technology and the interdisciplinary integration of machine learning,video and image processing,pattern recognition and computer vision,video target tracking technology and tracking performance have been continuously improved.Although the domestic and foreign researchers have proposed many effective tracking algorithm,but due to the actual scene in the presence of noise,illumination changes,its deformation and occlusion and so o n,the design of a robust and real-time video target tracking algorithm is still a challenging task at presentThe main contents of this dissertation are as follows:(1)For the traditional MeanShift tracking algorithm,which uses only color features to achieve accurate tracking,this paper proposes an improved local three-value texture color fusion histogram method.The algorithm first analyzes the original local three-valued texture features,and finds that the key texture pattern of the target can accurately describe the target object and can reduce the computational complexity.Therefore,this paper extracts key textures and color features for effective adaptive weights fusion.,making the tracking results more accurate and reliable.(2)When the size of the target in the tracking changes,the existing fixed frame of most algorithms tends to have a large deviation when the target object and the background size do not match,resulting in the problem of tracking failure.In this paper,the ratio of the target model and the candidate model determined by the texture color features is used to estimate the area,and the method of constructing the Mahalanobis distance is used to determine the candidate position of the next frame target.(3)Taking into account the occlusion inevitably often exist in object tracking,this paper introduces the Kalman filter to estimate the center position of the current frame,to provide more accurate and reliable prediction information,determine the severity of occlusion,with the parameters of adaptive filter,the final results show that this algorithm is effective and can better handle the target under occlusion.