Multi-technology Fusion Indoor Monitoring Application Research Based On The Pedestrian Detection And Target Tracking

In the field of computer vision, the pedestrian detection is a very common application technology, and detect the pedestrian is not only a research hot spot in the field, but also a research difficulties.Pedestrian detection and tracking technology is small in volume, high in accuracy and low in cost. What's more important, it can save manpower, financial and material resources. That's why the pedestrian detection has an extremely widespread application, such as intelligent auxiliary driving, intelligent monitoring, pedestrian analysis and intelligent robot, etc.In the video monitoring system,it will realize the intelligent and distant observation and carry out procession and analysis on the video image shooting by monitors.It will enable the detection,location and track of moving targets in various scenes, send out warnings when unsafe factors and suspicious behaviors are detected,and prompt managers to prevent and handle in time.It will ensure the routine monitoring under normal circumstances. What's more important is to give timely alarm information under the emergency. Although at present the pedestrian detection and tracking technology becomes much maturer, the shade occurred in the complex situation is also the emphasis as well as difficulty in current pedestrian detection research. The method proposed in this article receives some effect in pedestrian detection and tracking, but still it needs further improvement.The main content of this article is as follows:At the very beginning, an introduction is given on research background and significance of the pedestrian detection and tracking of indoor surveillance camera,and the present situation of the related researches at home and abroad. Next, some methods on how to detect and extract moving targets are listed and illustrated. What's followed then is a brief description on how to combine HOG and SVM, which are both very classic pedestrian detection methods in the history. In order to improve the detection accuracy of human body and reduce the feature dimension so as to speed up the operating time, this article uses the multi-scale HOG features, and extractsfeatures of pedestrian strong selection based on Fisher criterion. The experimental results show that the algorithm, relatively to the global scanning, reduces the number of detection window, speeds up the detection rate, and ensure the pedestrian detection accuracy to reach 90% and above.However, the pedestrian tracking is a process which is to study based on the previous frame and then forecast the information of current frame. This article selects Mean shift algorithm for pedestrian target tracking. In the case of shade, it chooses to use kalman filter to forecast the pedestrian targets. The experimental results show that the proposed algorithm in target tracking not only can accurately track pedestrians,but also when targets are under the condition of partial sheltering or serious barriers, it can track the pedestrian targets stably and effectively, guaranteeing the integrity of the target motion.The author put forward the strong resolution that the combine pedestrian detection method of pedestrian feature extraction and pedestrian tracking method of the Mean shift algorithm and kalman filtering, before the pedestrian tracking to strengthen detect pedestrians, in the speed and precision of the pedestrian detection at the same time, and it can improve the accuracy of pedestrian tracking.