Research And Implementation On Moving Object Detection For Video Surveillance System

In recent years, with the rapid growth of the national economy, many public areas, including banking, transportation, security and other needs of intelligent security monitoring systems growing increasingly demanding. Although the security monitoring system already widely used in the above-mentioned public places, but in fact monitoring tasks and did not reflect the "intelligent" and did not fully reflect the real-time and proactive monitoring. Intelligent monitoring should be able to do real-time video moving objects analysis and recognition, judgment and can give tips when an abnormal event occurs, the auxiliary humans provide meaningful information and save effort.Moving target detection technology and one of the key underlying technologies as computer vision technology, the rapid development in recent years, and is widely used, many scholars have proposed a variety of algorithms, but the actual scene diverse and complex, so that the moving object detection technology is still challenging of the subject.The main work is as follows: to build a monitoring system framework, the current real-time video capture, real-time monitoring of moving objects, for the number of Gaussian distribution Gaussian mixture model is fixed and the computational complexity of the problem is difficult to real-time detection is proposed based on Gaussian mixture model and Three frame difference algorithm improved adaptive background model fusion target detection algorithm, and the algorithm used in the monitoring system. Initially for each pixel to create the same number of Gaussian models, in which the learning process, the fusion of three difference method in complex regional surveillance video into the foreground region, the background noise region and exposed region, updated dynamically adjusted for each type of region update rate, and according to the state of each pixel of a plurality of Gaussian models, dynamic adjustment of the number of Gaussian model, namely, adaptive adjustment of the Gaussian distribution and the number of different areas of the model with different refresh rate, making the background model to adapt to changes in the scene.In this paper, a method used in the monitoring system compared with the traditional Gaussian mixture algorithm, adaptive Gaussian model number, different areas of adaptive update rate, and target detection algorithm meet the accuracy requirements; compared with the inter-frame difference method, the motion good target detection extraction effect; compared with the optical flow method to calculate the amount is relatively small amount of calculation and speed fast.Experimental results show that real-time video surveillance This article describes the use of MFC framework of human-computer interaction interface display monitor, display and monitor the scene moving object detection results, the ability to pause, save frames, playback and other functions, experiments show that the proposed target detection algorithm for many scenarios have a stronger ability to adapt in real time to achieve complete extraction of moving objects, and it has practical value.