The Research Of Face Recognition Based On Linear Discriminant Analysis

With the progress of science and technology,new technologies such as computer vision,pattern recognition and artificial intelligence have been developed rapidly.Among them,face recognition technology is the main research hotspot in the field of computer vision.The face recognition process mainly includes three parts: Face detection,feature extraction and face recognition.In the present study,under the circumstance of facial expression changes greatly as well as the light,part of the occlusion and other natural environment,the recognition effect remains to be improved.On the basis of the existing technology,the face recognition technology in this complex situation is studied in this thesis.The main contents are as follows:1.In the aspect of face detection technology: First of all,studies the selection of features in AdaBoost face detection algorithm,the training of weak classifier,the composition and detection process of cascade classifiers,Besides,for the detection of complex environments,the introduction of NPD differential features,A depth binary tree is proposed to train the classifier,and the skin color algorithm is used for pre-detection.The experimental results show that the algorithm improves the detection rate by 8.7%,the false positive rate is 4.1% and the detection speed is 27.7% higher than that of the original algorithm in multiple faces and multiview detection background environment.2.In the aspect of face recognition technology: the traditional linear discriminant analysis method is analyzed and found that the LDA algorithm is not ideal for the extraction of nonlinear features in the complex case.The kernel function And RBF neural networks are introducted.The improved algorithm maps the input space to the high-dimensional feature space through the kernel function,and more non-linear feature information is preserved.Then,the null space linear discriminant analysis is used to extract the features.Finally,the RBF neural network is used to realize the classification and recognition.The experimental results on the AR face database show that the recognition rate of this method is 8.5% higher in complex natural environment.3.Implementation of face recognition system: A real-time face recognition system is designed and implemented based on the improved detection and recognition algorithm.Based on the experimental results,the improved algorithm can still be accurately identified in the environment with large change of face,which further proves that theimprovement the validity and practicability of the algorithm.