The Research On Moving Human Detection And Tracking Of Intelligent Video Surveillance

The detection and tracking of human motions is always one of the most important fields to computer visions research, involving pattern recognition, artificial intelligence and image processing knowledge, which has an extensive application prospect on the fields of intelligent monitoring system and security control. Owing to the non-rigid, clothing and color of human target, it is still quite difficult to research the detection and tracking of human body. This paper mainly studies the moving human detection and tracking technology under static background. After analyzing the characteristics and methods of the detection and tracking of human target, it developed a way to detect and track moving human detection suitable for this environment by researching and improving the defects of the existing methods.In the aspect of image denoise, Aiming at the non-ideal denoise effect of the traditional median filtering algorithm for noise larger image, this paper uses a threshold median filtering algorithm.The algorithm inserts a threshold decision based on the traditional median filtering algorithm, of which the denoise effect is more ideal than traditional median filtering algorithm, and which also has a great effect on desoising the salt and papper noise that contains larger area in images.In the moving object detection, this paper studies a three frame difference based on the combination of points multiplied by the edge detection algorithm for moving target detection. Comparing with the traditional frame difference method, the algorithm can reduce the influence of light and ambient noise on the moving target detection result, and also makes the target contour extracting more accurate. For the identification method in the human target, with the head shoulder characteristics of the human body as the feature recognition of thde human boy, it used a human motion recognition method based on Hu matrix shoulder characteristics. AS the characteristics of human head and shoulder part of the nature in the process of movement are stable, using the algorithm has higher recognition rate and is more reliable than the one that using body’s "aspect ratio" to recognize characteristics of the human body.On the tracking of human target, this paper adjusted the Calman of traditional forecasting Calman gain coefficient algorithm, which greatly reduced the amount of calculation of tracking and improved the efficiency of the tracking algorithm. At the same time, using the Calman filter algorithm that automatically adjusts window position and size, effectively resolved the problem that tracking position and size is not accurate owing to the estimation error accumulation with the traditional Calman filter algorithm.This paper deeply studied the detection, identification and tracking of moving human, and Simulated the experiment, which proved that this algorithm has certain feasibility and superiority.