Research And Implementation Of Face Recognition System Based On LBP

Face recognition technology is an important biometric technology. It has been widely used in many fields because of its special advantages. The essence of face recognition technology is making your computer has the ability to authentication. It relates to image processing, artificial intelligence, pattern recognition, and it has important significance. Now the difficulties consist in the improvement and enhance of recognition methods under non-ideal conditions.Based on carefully studying the face recognition technology, this paper designs to detect real-time faces based on skin color and capture images. The paper uses Local Binary Patterns operator to extract texture feature and nearest neighbor classifier combined with Adaboost classifier to match the facial feature. Ultimately the system achieves the face recognition function.After determining the overall program, the paper simulates in MATLAB and chooses histogram equalization to pre-selection. Then the paper simulates variety of face detection algorithms based on skin color and chooses the method of skin color region to capture images. The next step is to establish face database. This paper uses4×4scales to block the face image and uses circular rotation invariant LBP operator to extract the feature for each part of image. Then it needs to converse the feature to their equivalent mode. On this basis, the amount of computation is reduced by improving feature reduction method. Then the paper connects each part of feature to form a complete eigenvector. Finally, the paper uses sample images to train classifiers and matches the faces. In the end, this paper tests the feasibility of the matching algorithm.The system is achieved on the DSP hardware platform. In the process of implementation, firstly the paper uses illumination compensation to improve image quality. Then it uses a method based on skin color region to detect faces and finds the face position. In the next step, the system capture the180×272image to extract LBP feature and bring the eigenvector to achieve the function of face match. Finally the screen displays the identifiable information.This article simulates the overall design of the system in the MATLAB software platform. Then it debugs and modifies the program in Microsoft Visual C++6.0and CCS3.3. By setting the DSP/BIOS environmental parameters, configuring files, and loading works, the paper completes the system on ICETEK-DM6437-B Evaluation Module hardware platform. The core processor of the hardware platform is TM320DM6437. The experiment showed that the detection rate of the algorithm in this paper is beyond90%and the system has a certain degree of stability and a strong practicality. It lays a good foundation for the further study of face recognition.