Research And Application On Moving Object Tracking Algorithm Based On Video Sequence

Moving object tracking is a processing of detection, recognition and tracking on video sequence. The aim of moving object tracking is to find the location of the object on the video sequences and process it. After decades of development, moving object tracking algorithm is applied in many domains.With the rapid development of these fields, people increasing high requirement for the moving object tracking Algorithm. Only by constantly improve the real-time, accuracy and robustness on the algorithm design of moving object tracking can meet the needs of real life.Compared with static images, video sequence has a wealth of raw data, strong correlation between adjacent video images and dynamic change of information over the time. In this paper, analysis and research on the moving object tracking at home and abroad of the existing algorithm based on video sequence, presented a drift correction algorithm according to the template drift. Firstly, according the knowledge of visual image processing to transform color space,the gray value and preprocess the quality of image on the obtained video image sequence, reduce the influence of exterior condition in the process of object tracking effectively.In the video sequence using background subtraction method for moving objects detecting and correlation processing.At last, the characteristics of accumulative histogram of the image template is used, implementing the tracking processing by the Mean shift algorithm. In the following tracking then set drift coefficient of the initial template and the current template by the Bhatthacharyya to correct the template.The experimental using the standard video library show that the algorithm meet the requirements of real-time and robustness and effectively solve the problem of object tracking template drift.At the end of this paper using the template drift correction algorithm and combining with the LBP feature which is suitable for face detection and identification to improve the particle filter algorithm and apply in the field of human face tracking. Experiments of the video collect by spherical camera achieves good results, prove the feasibility of the algorithm.