Research And Implementation Of Face Recognition Algorithm Aiming At Head Swung

The face recognition technology is the technology that uses computer to analysis face image and recognizes identification using the effective information extracted from face image. Due to many impacts such as abundance of face expression, variation of aging face, different lighting conditions, imaging angle or imaging distance, using computers to detect faces becomes very difficult.Face detection under complex conditions has been the difficult point of face recognition areas for a long term. Face recognition method, in simple conditions (such as common face expression, normal lighting, front view, etc.), has obtained very good results, even up to the standards of the commerce. But complexity of the image acquisition conditions, such as different lighting conditions, different poses, different shooting angles, different backgrounds, often can not get the satisfying recognition effect. Therefore, the face recognition method under complex conditions is worthy of being further researched.At present, the amount of researches aiming at multi pose is little. The recognition rate of face recognition algorithm will decrease, under the condition of big amplitude of head swung. In addition, emotions can be directly reflected on face, the variation of multi face pose is extremely complex. It’s difficult to recognize multi face pose images for face recognizing system. Therefore, finding a reasonable algorithm to make the face recognition system well adapt to posture changes becomes the main purpose of this project.In practice, many images contain human faces in variant rotations, but most of previous algorithms are effective only for upright faces. To locate faces rotated in different orientations, we use a Radial Template (RT) to detect face-like areas in edge map. The model is designed to describe face feature configuration and its orientation.In this paper, we propose a radial template face detection algorithm to locate faces rotated in any orientation. Detecting rotated faces is important for a face detection system. First we present a novel model named Radial Template (RT) to detect rotated faces. This template is designed to find stable features of center-rotated objects in edge maps. Based on skin detection, our method searches for face-like areas and get face features (brow, eye, nose and mouth) and geometric center of gravity. We use radial template to get face orientation. A system integrating these techniques is presented. Experimental results show that our algorithm is effective to detect human faces under head swung conditions with different sizes, lighting conditions and backgrounds.