Based On The Complexity Of The Context Of Face Detection And Recognition

The research on face detection and recognition aims at enabling machines to possess the ability of face recognition just as that of human beings. It is essential applied in the domain of personal identification, human-computer interface, image retrieval, visual surveillance, etc. Although the topic has received much attention in pattern recognition and computer vision community in 30's years , but the recognition result is affected by expression,pose and illumination. In order to run in real-time, many problems remain to be resolved.This paper made a further research on face detection and face recognition based on complex background. AdaBoost makes a high real-time performance and has good detected results. It is well known that the Foley-Sammon optimal set of discriminant vectors based on the Fisher discriminant criterion is an efficient linear feature extraction method. So this paper presents that the Face Detection and Recognition Demo platform to implement detector and recognized result use in the AdaBoost and the algebraic method.The main contributions of this thesis include following issues:1. Using Haar-like features, which are computed by integral image rapidly, and applying AdaBoost algorithm to select important detecting features. Then, a cascaded classifier combining many classifiers is trained and allows background regions of the image to be quickly discarded. Not only does the detector ensure an enough intriguing detection rate, but also it persist a substantial small false positive. The human face detection system is realized by the OpenCV development kit with VC++ development environment.2. The human face detail drawing obtains which to the examination carries on the pretreatment likely, including the geometry, the greyhound, the size normalization, obtains the unification size training sample and the test sample data.3. We propose a PCA plus LDA method to solve the small sample size problem. As we known that the Fisher discriminant criterion is an effective method of extracting the information used for classifying. The Foley-Sammon optimal discriminant vectors obtained based on the Fisher discriminant crmterion is calculated one by one, i. e., every discriminant direction has the best separability respectively. After we can obtain the optimal discriminant vectors set calculating from the preprocessed training sample set, the test sample is projected on the Foley-Sammon discriminant vectors set. Then feature vector can be obtained from the test image in the feature subspace. We extract feature from training set and testing set with vc++ and matlab development environment.4. The system carries on the classifier with the most close neighbor classification, the test images are classified according to the discriminant vectors obtained from the training sample.5. This paper introducing the entire face detection and recognition flow including reading the image the data, locating the human eye position, preprocessing the test image, recognizing the human face image by the classifier.