Study On Multi-face Detection Method

With the rapid development of information technology, face detection is the important research field of computer vision, video security monitoring system usually need to detect multi-face images. So the development of society production and living inseparable from multi-face detection technology. The multi-face detection technology in the biometric identification technology rapid development today, is widely applied. Multi-face detection refers to contain more than one face image within the input face image, based on the facial feature points location and geometric eigenvectors of face image further verification recognition and matching. Facial feature information and geometrical characteristics can be used for face detection, is a stable facial features. The geometrical characteristics of the human face for multi-face detection technology has the practical value and significance.This paper makes a deep research for multi-face detection, involved in face detection and facial feature points localization and the recognition of the facial geometric eigenvectors based on video sequence. First, this paper presents the face detection based on AdaBoost algorithm to detect many faces, include the size and location information. Through lots of face images in video sequence,it can find that AdaBoost face detection algorithm is effective to detect the facial images with the resolution,rotation angle in left and right side, pitching angle and so on.As well as analysis of face detection algorithm the most suitable occasion. Secondly,the feature points positioning based on gray statistic and the distribution of face, realize the facial eye and nose and mouth feature location. Facial feature points location results directly verify the accuracy of face detection. Finally, the geometric eigenvectors composed of the distance between the facial feature points, it can comprehensive reflect the facial proportion relationship between organs and properties. By calculating weighted Euclidean distance between the facial geometric eigenvectors, it is used to identify the multi-face images of video sequence.To experiment on video image with multi-face detection method discussed above,achieved a good multi-face detection. Extensive experiences and analysis illustrates that these methods can be applied different expressions and rotation angle and size of face detection,realize the dynamic analysis of successive frames multi-face image. Multi-face detection canbe effectively realized face detection with high detection rate, speed, strong robustness.