A Study Of Face Recognition System On Android Platform

With the development of internet technology, the mobile devices has gained popularity. At the same time, due to the gradual improvement of the intelligent mobile devices and hardware level, many applications based on mobile devices are directly related to people’s personal privacy and property. Therefore, how to realize the information security protection on mobile devices has become a urgent social problem. As an effective and convenient identification technology, Face recognition has already been mature after years of development. However, the previous research on the face recognition technology is mainly in the PC terminal, and face recognition based on the mobile device also has not obtained the very good realization. Meanwhile, as the most popular operating system for mobile devices, Android development is convenient for its open source,and it makes the study of face recognition system based on Android platform received wide attention of scholars both at home and abroad.In this thesis, based on the study of the relevant theory of face detection and recognition, a face recognition system based on Android platform has been constructed, the following works are completed.(1) In order to solve the problem that skin color segmentation can not divide correctly when influenced by the light and shadow due to its threshold is fixed, an adaptive skin color segmentation method is put forward. The between-cluster variance method(OTSU) can adjust threshold adaptively, and make the skin segmentation more accurately. Then, only the segmentation region using Adaboost algorithm for face detection. The method combining the adaptive skin color segmentation and adaboost can forecast face more accurately. It has a certain anti-interference ability to resistance to the impact of the light and shadow existing Android platform. At the same time, because of the scope that Adaboost need to traverse is narrowed, makes the detection time decreased and detection precision is improved. Therefore, this method can meet the real-time and accuracy requirements for Android platform.(2) The Shearlet multi-orientation features fusion and weighted histogram are proposed to overcome the disadvantage of Shearlet transform, which has data redundance in extracting features and can not sparse represent the global characters. By coding fuse the features from different directions of the same scale of the Shearlet transform, and then blocking weighted fusion the fused scale image, can significantly improve the recognition rate.(3) For the reason of the traditional sparse representation classifier is not sensitive to the changes of characteristics only has enough training samples, a face recognition method based on Shearlet multi-orientation adaptive weighted fusion and sparse representation is proposed. Shearlet multi-orientation feature fusion is applied to construct sparse representation classifiers for sparse representation of coefficient vectors, which makes its robustness against illumination, expression, pose and local deformation have a better effect.(4) Constructing the face recognition system based on Android platform gives the selection of system building, the environment of system development and the implement function of system, as well as providing a basic for further investigation of face detection and recognition algorithm.