| The face recognition systems which the public security department used are mostly desktop equipments or professional equipments currently, and take photos under relatively good imaging conditions and getting the cooperation of the target person.The algorithm of software aims at the face image which is acquired under constrained condition. However, police works also involve a lot of outdoor works, such as security patrols, checkpoint interrogates and so on. If the policeman want a instant face acquisition and comparison toward the suspect, the existing facilities could not meet the need of portability and instantaneity, furthermore its algorithm of software for face acquisition under random environment and unconstrained conditions has been lacked a certain adaptability. Therefore, this thesis designed a kind of mobile terminal face acquisition system based on face detection and image processing, which can be better to accomplish the face acquisition and feature extraction under the above situations, and then with the help of wireless-internet, it connects the back-end server platform and accomplishes the work of face recognition. Academically, aiming at those situations of large-scale face database, non-ideal imaging conditions, target person is not cooperative, etc. How to improve the robustness of face recognition system is currently a hot research topic in the field of Computer Vision.In this thesis, the research work mainly includes the following points,(1) In order to solve a series of problems that acquiring face images under unconstrained conditions, which will affect face feature extraction and recognition greatly, such as illumination changes, complex background, diversity of target person’s position and posture. This thesis uses a series of targeted digital image processing technologies on some key links, for example, face detection combined eye detection, face alignment based on eye detection, histogram equalization of block faces, elliptical mask and so on.(2) Took PCA, Fisher face, LBP method for further study,analyzed and compared the advantages and disadvantages of these methods, do some work to pave the way for choosing a more suitable algorithm for face feature extraction under unconstrained conditions.(3) Using the NDK+JNI technology, transplant C++module of face acquisition and image processing to Android platform, which is programmed and debugged on Windows platform, solved the problem of the Android application invokes the OpenCV library functions and related C/C++image processing programs.Based on the above researches, this thesis realized a mobile face acquisition prototype system on Android platform aiming at the police application.The prototype system is fast, flexible, accurate and well-adapted to the face acquisition under unconstrained conditions, and has a certain values of reference and application in the police work, this system is a beneficial supplement to the existing application equipments and technologies for the public security department. |
