3D Face Recognition Using Fuzzy ARTMAP Based On Multi-Features

How to identify a person’s identity accurately and protect information security is a critical issue which must be resolved in today’s society. Compared with the traditional identity recognition technology, biometric identification technology has the following advantages: ever-lasting or lost, difficult to counterfeit or be stolen, portable and available. It is widely used in the national public security area. Biometric identification technology has been listed as one of the top ten technologies which has a revolutionary influence on the human society in the 21st century. In numerous biometrics technologies, face recognition technology has become a frontier research direction in biometric identification field for its advantages of initiative, noninvasive, friendly, and convenient collection, etc.In recent years, the research on 2D face recognition has been quite mature. However, the problems like the illumination change, posture and obstacle interference, which hinder the application of face recognition technology greatly. As the technology research on three-dimensional data acquisition equipment continuing to mature, the recognition research based on 3D face data has become another important research direction.3D face data has the characteristics like information richness, illumination and posture invariance, which has great significance on the research of improving face recognition rate.This paper research 3D face recognition algorithm based on global features and local contour feature fusion, which focus on 3D face feature extraction, also the design and application of intelligent classifier. Mainly include the following contents:Firstly, according to the characteristics of the structure of 3D face data, designing preprocessing algorithm of face depth data mainly solve the problems about positioning in the face critical areas and interference removal, etc. The experimental results show that the preprocessing algorithm is effective and practical.Secondly, aiming at the problem of face recognition which is easily affected by expression changes, this paper design an algorithm to extract the face contour curve characteristics based on facial bone structure characteristics. The experimental results show that these features have preferable robustness.Thirdly, after analyzing and comparing the improved LBP algorithm, Gabor algorithm and Log-Gabor algorithm, this paper do research on single feature extraction and features fusion respectively. The experimental results show that the double filter features based on LBP and Log-Gabor in classification test have higher recognition rate.Fourthly, research on single training sample identification. In some practical applications, everyone could be collected only one face samples, it is required that system could recognise face images in different gestures and expressions, which is based on single sample model. Currently the recognition rate of mainstream classification algorithm has been greatly reduced under the condition of single sample input. This paper introduces fuzzy adaptive resonance mapping network (Fuzzy ARTMAP) as the 3D face recognition classifier. The experimental results show that the Fuzzy ARTMAP classifier for single sample 3D face recognition has real-time performance, incremental nature and higher recognition rate.