Research And Implementation Of Object Tracking Algorithm

The Object Tracking technology which is one of the research in the filed of computer vision is been widely used in recent years.The using of the technology eonsist of the camera focus in tracking.automatic object tracking of drones,face tracking,gesture tracking and so on.And this technology also be used in the intelligent surveillance video to track specific people or targets.Object tracking technology as an import.ant research topic has also been rapidly developed and researched in recent years.In this paper,theoretical research and experimental verification of object dectection and the object tracking algorithms in the filed of computer vision are performed,which mainly include two major aspects:Moving target detection and Object tracking technology.The innovation of the essay is based on the study of traditional typical Moving target detection and Object tracking.It also conducts detailed research and experimental verification on Tracking-by-Leaming algorithm in recent years,including Gaussian Mixture Model and Bayesian inference background segmen-tation model,The TLD?MIL tracking algorithm.The main difficulty of essay is how to achieve stable tracking of multiple targets in an actual dynamic environment.Finally,this paper applies the object-oriented idea and combines the single-target tracking algorithm and Kalman filter to achieve stable tracking of multiple targets based on the background segmentation model.The mainly work and research content of the essay is described as follows:1.Improving the effect of Objects Tracking with Kalman filter algorithm because of the reason that Meanshift,Camshift algorithm is susceptible to the envirnmental factors that may lost targets and exist the partial or incomplete occlusion.Combining the Kalman filter algorithm with background segmentation model algorithm(Mixed Gaussian Model,Probability the Foreground Bayesian Estimation,etc.),propose a multi-target tracking algorithm that can detect and track the target in static background.2.Intensive studies of various object tracking algorithms.Including of the study and the experimental verification of TLD algorithm,MIL algorithm,and improved KCF algorithm based on correlation filtering,etc.It has been studied and experimentally verified through the advantages and disadvantages of the actual video sequence detection algorithm.Finally,a combining Kalman filter for multi-target detection and short-time tracking based on background segmentation is firstly proposed.And a background segmentation algorithm for detecting moving targets combined with Meanshift and Kalman filtering algorithm is secondly proposed that designed to detect and track multi-objects in the simple environment stably.The effect of the algorithm is verified by the actual video sequence in the final part of the essay.