Research about Face Recognition in the Access Control System
|School||Wuhan University of Technology|
|Course||Control Science and Engineering|
|Keywords||Face recognition Face preprocess elliptical skin color model quotientimage method|
As one kind of pattern recognition, face recognition has been widely applied in door control system and information security. The face recognition include face detection and face determination, there are many factors can affect the recognition rate, the three main factors are Lighting conditions, face position and facial expressions. Light condition is subject to change during the door control system, it has an obvious impact on the face recognition. So we carry out the theory and applied research about illumination compensation for face recognition in door control system. The identify matching rate will not achieve the desired value if we just use face image collected by the camera in door control system, we can through preprocess the collected by the camera to eliminate noise, reduce the interference of light.In the paper we first introduce basic method about face image preprocessing, such as gray level transformation, histogram correction and image filter. Through the Face color model experiments, we know that we can easier detect face contour if the image has been pretreated. Comparing several basic methods in face detection, the Face detection algorithm based on the elliptical color model has good affection. The advantage of this algorithm is split fast as well as segmentation accuracy can achieve the test requirements. So it is used widely. We can implement expansion and corrosion treatment on skin color model, if we would like to extract better human contours. Finally, one of the bottlenecks in face recognition is:light conditions change significantly can affect the accuracy of face recognition. In the paper we expound the principle of light affect the accuracy of face recognition. In quotient image we can find one facial feature that do not vary with illumination change and can largely eliminate the adverse effects of illumination changes. At the same time, the quotient image has some poor performances, the improvements are as follows:(1) Increasing the number of the light source from three to five.(2) By specifying the condition for illumination compensation, this paper estimates the unnecessary for illumination compensation so as to reduce the calculation. The results of the experiments in Extended Yale B database show that quotient image method can widely reduce the affect of face recognition rate when the light condition changes significantly. The improved method can not only further improve the recognition rate but also reduce the computational complexity.