Face orientation detection in video stream based on harr-like feature and LQV classifier for civil video surveillance
Zhiguo Yan, Jian Wang, Mingxia Sun, Yongjie Shi, Fang Yang, Chao Li · 2013
Most face recognition and tracking techniques employed in surveillance and human-computer interaction (HCI) systems rely on the assumption of a frontal view of the human face. In alternative approaches, knowledge of the orientation angle of the face in captured images can improve the performance of techniques based on non-frontal face views. Face orientation detection plays important role in city surveillance video for the successive specific application, such as the face identification, face recognition and screening face snapshot image for saving the storage volume. In this paper-we propose a kind of method on face orientation by combining Haar-feature and LVQ technique. First, we execute the eye location based on the haar-like feature. Then, we divide the face image into several sub-images and statistical information of the binary sub-image at the eye location. After acquiring the statistical pixel distribution, we exploit the LVQ classifier to execute the classification on face orientation. According to the result, the algorithm we proposed can detect correctly up to 95%. By executing the face orientation classification, we can get the upright fontal face image with the best recognizable and distinctive quality for the further application.