Face Image Recognition System With Pose Variations

Since the1960s, face recognition research has become an important re-search topic in the field of machine vision. With the gradually developmentdemand for security in every field, as a kind of biological recognition, facerecognition technology has the advantages of non-contact, easy and difficultto disguise, and has already become the important identity authenticationtechnology in the field of security. However, the external environment changehas great influence on the face recognition effect. The change of the externalenvironment or camera shooting angle, often can produce larger interference,and make face recognition effect poorer. How to make use of different poseface samples for effective face recognition has become a one of the key re-search content.Throughout the full text of the research work, the results and conclu-sions obtained mainly embodied in the following aspects:1) This paper proposes a fusion method based on face images with posevariations.This paper proposes face images fusion method to create a front faceimage with pose varied face images. First we create the shape informationthrough the position of the feature points, and create texture informationthrough the pixels of face images, and then we get the final fusion frontal face image. Finally, we verified the effectiveness of the proposed fusionmethod.2) This paper proposes improved feature extraction method based onface symmetryAccording to the symmetry of the face image, we introduce a measureindex to measure the face symmetry. And then, a reasonable treatment on theeven-odd symmetry images is taken to reconstruct face image. Experimentproved that this method can overcome some of the confounding factors, andeffectively improve the recognition rate. Finally, according to the measureindex of symmetry, we divide the test gallery into collections and test sepa-rately to analysis of the effect of symmetry in face recognition.3) This paper proposes a face images database with pose variationsThis database can be used for experimental research of the video sur-veillance and network retrieval applications. This database contains25per-sons, and each person contains25varied pose face images as a test set,3frontal face images as gallery set. Database has been gain at the laboratoryweb site.4) This paper proposes practical face recognition system with pose vari-ationsThe face recognition system can be applied to the Internet retrieval andvideo monitor applications. Human face image can gain from the network orfrom the camera. The system using the proposed method can confirm theidentity of the persons.