The Complex Movement Of Multi Moving Target Detection And Tracking

Currently, machine vision technology has been widely used in intelligent life, industrial,medical and military fields. Extraction of target detection, identification and tracking technology are its main study contents. Because of its great value and wide application prospect, these research areas have become one of the hot topics of scholars.This paper sets a fixed scene,chooses complex background multiple moving target detection and tracking as the main study contents. For the current mainstream target detection and tracking algorithms, on the basis of their theoretical research, the advantages and disadvantages of each algorithm are analyzed. And then, some important areas of detection and tracking algorithms are optimized. Through theoretical analysis and experimental simulations, the paper demonstrates the performance advantages of the improved algorithm in moving target detection and tracking. The main contents of this paper are organized as follows:Firstly, we proceed the research of moving target detection technology, inter-frame difference and background subtraction are the two most typical moving target detection algorithm, this paper adequately analyzes the full scope of its applications and shortcomings,and makes the theoretical research and experimental simulation on the reason which causing the shadow of a moving object and its removal method. For the moving target tracking algorithm, kalman filtering algorithm and Mean-Shift algorithm are the two excellent target tracking algorithm, this paper carries on the thorough study of these two algorithms, and laid a good foundation for its optimized algorithm.Secondly, three differential detection method is improved in this paper, and we propose a three-frame difference method with a combination of Gaussian mixture model algorithm for moving target detection, and combines a new multiple moving target detection and tracking system of kalman filtering prediction tracking, then the performance benefits of the new method is proved by two simulations.Finally, the paper proposes a new specific target tracking system, it combines multi-feature fusion Mean-shift and kalman filtering algorithm. The two innovations of this tracking system is: Firstly we use texture and color features to for the Mean-shift target modeling; Secondly, to further ensure the accuracy immunity, we use the similarity judgments combined with filtering prediction method for real-time updates on the target template, this method can effectively balance the robustness and instantaneity. And under the circumstances when the tracking target is shield on a large scale, the accurate tracking is only through kalman filtering prediction, thus the numbers of iteration of the algorithm is reduced. Finally,after three sets of simulation experiments show that the improved specific target tracking system has a good performance for the situation which the target is seriously covered or foreground and background interference of a wide range of similar colors,or the targets are moving too fast. The improved target detection and tracking system in this paper can provide theoretical support and reference for the study in the field of intelligent video surveillance.