Research On Face Detection Based On Adaboost Algorithm

The definition of face detection is that determine whether or not there are any faces in the image and,if present,return the image location and extent of each face. Face detection originated in face recognition on the subject, Since the 1990s,face detection began as an independent subject of attention by researchers.With the fast rapid development of biological technology, face detection technology in face recognition, expression recognition, video conferencing, content-based indexing, image and video search, intelligent human-computer interaction, obtain the widespread application.Viola proposed the method of face detection based on Adaboost algorithm which has high detecting speed and hit rate,but it also has high miss rate,so it is necessary to verify the face regions after detecting the face candidate regions and removes the none face regions quickly.Since the algorithm has not taken the skin color information into account, this paper proposes that it increases a skin color detection to verify the face candidate regions which was detected by using the mehtod of based on Adaboost algorithm. Namely after detecting the face candidate regions, it uses the skin color model which establishes in the YCbCr color space to make the further skin color confirmation and removes in the candidate regions which has not the skin color infomation,namely rules out the none face region which has detected by mistake.The work of this paper mainly includes the following several aspects:1. Studied a method of face detection based on Adabost algorithm.First it calculated a large number of Haar-Like feathers by integral image,the Adaboost algorithm selects a small number of critical features to construct a set of weak classifiers.Then yields a strong classifiers using some weak classifiers.At last combine increasingly more complex classifiers in a "cascade" which allows background regions of the image to be quickly discarded.This paper improves the weak classifier and the experimental results show that it can reduce the training time.2. Studied a method of face detection based on skin color.After comparing with several color space,this paper establishes the Gaussian color model using YCbCr color space in which chrominance and luminance are separated. Then skin color segmentation is carried out with the method of dynamic threshold optimizing and two-value processed.In order to eliminate the impact of noise,this paper use the filter based on mathematical morphology.3. Proposed a method of face detection based on Adaboost algorithm and skin color Verification. A classifier is trained by the Adaboost algorithm at first,which work for color image detection.And the detection results form candidate face regions.Then these candidate regions are further verified by a skin color mode in YCbCr chrominnce space.Thus the none face regions which contain no skin color pixels are ruled out. The algorithm improves detection accuracy rate and reduces false positives rate under the premise of guaranting not to reduce the speed of face detection.