| As one of biometrics, face recognition has gotten much interest and become a research focus in applied mathematics and information technology.Face recognition algorithm is the critical part of face recognition system, and it can directly affect the performance of the system. In this paper face recognition algorithm based on local binary pattern (LBP) and adaboost is researched with modern applied mathematics. The main contribution of this paper is as following:First, a new way of jumping block of LBP is proposed to form features. Face characteristics of different regions and size are reflected at the same time by the features. And the information has relatively lower repeatability, which can make the machine learning algorithm select the best features more easily and efficiently.Second, a fast method to make training samples and calculate threshold is designed to greatly improve the speed of training. Random sort of the training samples is proposed to improve the robustness of the training result.Third, the local binary pattern is extended (ELBP) to make up some deficiency of LBP. ELBP can not only reflect the trend of texture difference like LBP, but also can reflect the precise variation of gray. Uniform pattern for ELBP is proposed through a lot of statistics, and this can greatly reduce the computation.Last, the algorithm designed in this paper is tested on CASIA_NIR near infrared and CAS-PEAL-R1visible lighting face database. The system recognition rate reached99.0%and99.2%separately. |
