Research On Face Detection Technology And Implementation

As a classic one of pattern recognition problems, Face Detection has been concerned a lot for a long time. As computer technology matures ,Face detection have been extensively applied to the new human-machine interfaces, content-based retrieval, object-based video compression, digital video processing, visual monitoring and many other fields.In this paper, we use Color information of the fast, accurate and the accuracy of AdaBoost algorithm to realize the composite application. Where the more popular algorithm AdaBoost algorithm is chosen as the core detection algorithm, and Color information as the treatment of early detection.In this paper, AdaBoost algorithm, through the establishment of the classifier trained as a core element detection, color detection as a pre-processing, Cascade structure of the late detection of the composite structure, a face detection algorithm is studied. It detailedly introduces the three commonly used color space and color model, and different models of color clustering situation are compared. YCbCr space with better effects of color segmentation is btained. AdaBoost algorithm for classification theory and implementation steps is detailedly introduced. According to classification error rate ceiling of Strong classifier, a strong classifier training closing conditions is obtained. The superposition of Weak classifier threshold is used to reduce the repeated computation amount of weak classifier threshold selection. And also effective shorten the training time by reducing the characteristics number. Cascade constructure is applied to construct a multi-layer classifier , the number of sub-window of the advanced stage strong classifier is effectively reduced. Finally, clustering of face samples in the YCbCr color space is used to exclude sub-window of non-face color background before testing image, So that the target area which is going into multi-layer classifier become more clear. Thereby the speed of face detection is accelerated. Experiments show that, the method both has high accuracy rate and the low rate, which can be effectively applied in the context of people face and the situation of complex background, with better detection results.Composite applications followed is somehow explored.