Research On The Anomaly Face Detection Method For ATM Monitoring

With the development of society, as well as the increase of the ATM, the higher securityperformance of the ATM is put forward. In order to enhance the intelligence of themonitoring system, an abnormal face detection system which can be applied to ATM hasbeen designed.In the first part of the paper, the main content is the video image preprocessing. Here, thepretreatment means is to make preparations for the next step of the research work. It mainlyincludes enhanced image quality and separating foreground image by background modeling.In the process of image enhancement, this paper put forward an improved algorithm of imageenhancement towards the surveillance video images on the analysis of existing methods. Andthe algorithm has reached some experimental conclusion. In the process of backgroundmodeling, the paper eliminated interference by connecting prospects domain filter for imageprocessing.The second part is the core part of this paper, the mainly content is the abnormal facedetection algorithm. This paper trained the classifier based on the haar-like feature. During theanomaly detection in the eye area, we use the classifier to search the eye area. And during theanomaly detection in the mouth area, the method used is classifier search combined horizontalline detection based on Hough transform.In the third part, this paper designed the abnormal face detection system based on theVisual C++6.0. The paper established the interface based on MFC and achieve all functionsof this anomaly face detection system.