| Face recognition has a wide development prospects such as secure identification because of its advantages of easy to obtain, without direct contact, reliable in wide fields. Face recognition system consists of face detection and orientation, face feature points extraction and recognition, Facial feature points extraction as the key steps of face recognition, directly affect the accuracy of face recognition, which is a research hotspot in recent years. But in real life, the changes of factors such as illumination, posture, facial expression will affect the feature point positioning, for different face database, experimental results of the algorithm will also be different.Face database has great significance for face recognition system, establish proper database has an important foundation for the establishment、research and evaluation of face recognition system, the recognition rate of face recognition system to a certain extent also depends on if the face database established well. This paper establishes a new face database——SSDUT face database, which collected 581 volunteers’ images of the 15 different point of view in the same place, in accordance with the unified standards in the eyebrows, eyes, nose, lips, face marked 72 orderly feature points. Taking LFW database and SSDUT database as experimental objects, this paper do the research about face alignment by explicit shape regression, changing the number of training pictures and adopting the combinations of different angle pictures as training pictures and testing pictures to contrast test result and analysis the performance of algorithm, In the implementation process of algorithm, parsing the main ideas and defects of algorithm, researching in in which use condition the algorithm has the highest efficiency, to improve the efficiency and robustness of the algorithm. And this paper selected the classical texture-based, shape-based, regression-based and invariant-based algorithm to implement and achieve the compare results. |
PDF Full Text Request