Research Of Face Recognition Based PCA And Artificial Neural Network

Face detection and recognition is a subject of multi-discipline, including computer science, automate controlling, model recognition, image processing. It has immense studying value and application prospective. Owing to its three dimensional pattern, varied features and affect of outside surroundings, make face detection and recognition highly difficult. This thesis studies deeply about face detection and recognition. The main works are as follows:(1) Image Preprocessing, including lighting compensation, contrast enhancement and fore-back-ground separation. To every aspect of preprocessing, some technique's feats and defeats are compared by experiments and the most suitable is select to be applied in this thesis.(2) A gradient fusion face detection method is proposed. This method integrates the virtues of color difference, the size of eyes and outline of eyes, and finally gets the precise, non-fake eyes location of real eyes.(3) Principal Component Analysis is studied here. PCA is used to extract one-dimensional characteristics from two-dimensional images and collects those characteristics whose contribution sum are below 80 percentage.(4) Through many series of experiments, the affect of various recognition methods, the amount of collected features, neural network hide-layer and neurons in the hide-layer to the consuming time and recognition accuracy are repeatedly compared and analyzed and then the best suitable one is found.