Research And Implementation Of Face Detection In High-speed Video

With the development of the traffic, more and more people have their own cars, so that high-speed traffic problem got more and more concern. At present, many major highways have been installed monitoring equipments to improve supervision and eliminate traffic events. However, 24 hours a day of traffic video volume is huge, traditionally, workers often analysis test video recording by their eyes and experience, which is one inefficient method. More and more research focused on the quickly and efficiently method of processing a huge amount of monitoring video. Based on this background, this paper is to focus on surveillance video in the face detection. It is an important prerequisite for the content-based video retrieval, and has great application value.As the complexity of human face patterns, the existing face detection algorithms have some short comings such as the prevalence of large, computing slow and poor robustness. Skin-based detection algorithm is simple, fast, but not a good way to overcome the light, background color, interference and other factors.Firstly, this paper described the development of high-speed video surveillance system, leaded to the need of the application of face detection technology in high-speed video monitoring system. Based of these, research on the current status of face detection technology, as well as some existing research difficult. Then, because of the good of Adaboost-based face detection, this paper analyzed the characteristics, the basic ideas and mainly content of face detection. After then, this paper designed and implemented a Adaboost-based rapid face detection of video system, the system presented one strategy of using random sampling video frame, and one pre-filtering mechanism of skin-based face detection algorithms, by these, the system can reduce the processing burden, which greatly improved the efficiency of video processing, and presented detail of the system design and implementation processes. Finally, this paper introduced the OpenCV relevant theories and techniques, and used the VC+ +6.0 and OpenCV technologies to realize of the system, combined with experimental results, analyzed some deficiencies exist in the system and put forward some corresponding solutions.