Face Detection And Recognition Algorithm Based On Skin And Pca

Today's society is a highly developed information society, and technology applications infiltrates various industries. Face detection and recognition technology is just a product of this era. This technology has been applied to various security areas, such as building access security controls, customs security checks and smart card authentication and so on. It is one major development direction of identifications in the future.This paper uses the combination of the color space based on YCgCb and face template matching algorithm for face detection. YCgCb-based face detection method is a knowledge-based face detection method; this algorithm is intuitive, and not limited to the shape and the size of the face. This algorithm is easy to understand, and the clustering of YCgCb color space is better than the clustering of other color space. Face template matching algorithm is based on the statistics algorithm for face detection. This algorithm avoids that the human visual observation brings error as result of incomplete and inaccurate visual observation. The integration of these two algorithms make uses of the advantages of two algorithms, using less time for face detection and having a better detection rate for face detection. The experiment validates the time of face detection costs 3.5s to 24s, and has a better test result for different size, expression face gesture, many people and the complex background, etc. A single face detection rate is 83.7%, many face detection rate is 76.2%.This paper uses a weighted 2DPCA for the gray face recognition. This method which is different from the traditional 2DPCA is not only to use the between-class divergence matrix but also to use within-class divergence matrix. This method not only effectively expands the differences between the between-class samples, but also effectively reduces the difference between within-class samples, thus enhances the face recognition rate. The experiment is based on ORL face pool. After anglicizing the test result, this algorithm greatly improves the accuracy of face recognition, comparing the traditional principal component analysis.