Face Fatigue And Identity Recognition Based On Landmarks Analysis

Automatic driver fatigue detection and face recognition based on facial landmarks location have important influence to scientific research, which belong to the interdisciplinary of pattern recognition and computer vision. With the emergence of low-cost image acquisition device and the rapid developed computer technology, as well as the increasing demand of commercial and military applications, driver fatigue detection and face recognition has become a hot research topic. Many factors lead to the diversity of natural face images, which makes face analysis challenging. Based on the study of facial landmarks location, this paper propose new methods about fatigue detection and face recognition, the main achievements are given as follows.A non-intrusive fatigue detection method based on fast facial feature analysis is proposed. Firstly, the facial landmarks are obtained by the supervised descent method, which automatically tracks the faces and fits the facial appearance very fast and accurately. Secondly, the aspect ratios of eyes and mouth are computed with the coordinates of the detected facial feature points. We interpolate and smooth those aspect ratios by a forgetting factor to deal with the occasionally missing detection of facial features. Thirdly, the degrees of eye closure and mouth opening are evaluated with two Gaussian based membership functions. Finally, the driver fatigue state is inferred by several IF-THEN logical relationships by evaluating the duration of eye closure and mouth opening.A live face recognition algorithm based on multi-scale high-dimensional features is proposed. We also build a platform for face recognition. The facial landmarks are localized by supervised descent method. Whether the target is living or not is determined by the extracted facial features. The high-dimensional global and local features based on local binary pattern are extracted around every facial landmark from each face image, which are processed by PCA and LDA to reduce the dimensionality of the features. We can recognize face using k-nearest neighbor and voting algorithm based on the extracted feature. For those faces, which do not exist in the database, we set rules to reject them automatically. The experiments show satisfying results.