Development of face recognition system based on raspberry pi card

In recent years, methods based on the human face biometry have become widespread due to their importance in different applications. In the thesis, face detection and recognition methods based on facial localization, feature extraction and data classification algorithms have been proposed. At first, the face of human has been extracted from the background in order to increase accuracy of the method and reduce computational time that is required for the processing the image. Two-dimensional discrete wavelet transforms that represent high level feature extraction have been used as a robust method to explore the edges. In addition, the wavelet method reduces the size of the view and provides important data. The application of the discrete wavelet transform to the face images yields four sub bands such as low-low, low-high, high-low and high-high. Only the low-low sub band was used for forward processing as the feature vector. Finally, artificial neural network has been exploited as the classification method for the face images. Back-propagation training algorithm has been utilized in learning stage where the images of the faces are considered as input data for the neural network. A minicomputer raspberry pi was used for the implementation of the proposed method. The method is based on the python programming language for generating the software codes. To assess performance of proposed method, its results were compared with the results of the methods based on discrete cosine transform (DCT) and histogram of oriented gradients (HOG). The comparison results are satisfactory and demonstrate the superiority of the proposed method.