الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 141 |  |  |
| | | from | 141 | | |
Abstract
The demand on devices and systems empowered by biometric verification and identification mechanisms has been increasing in recent years as they have become a significant part of our lives. Palm vein biometric is an emerging technology that has drawn considerable attention from researchers and scientists over the last decade. Palm vein authentication has taken significant attention because of its uniqueness, stability and non-intrusiveness. In the present thesis, two palm vein authentication models for personal identification and verification were proposed. The first model is based on convolutional neural network (CNN), which is the most popular deep learning architecture and Bayesian optimization. To verify and evaluate the proposed approach, an extensive set of experimented is carried on two publicly available databases comprising images in the Near-infrared (NIR) spectral range. The first set of experiments is conducted on contactless palm vein images collected from PolyU MS palm vein database. The Same experiments are conducted on CASIA multispectral palmprint database. from the first and second database, 6000 and 1200 images, respectively were assigned to training, validation and testing the model. First and foremost, region of interest (ROI) of the palm vein is extracted as an image and filtered by Jerman enhancement filter to enhance the gray levels of the vein patterns. The proposed CNN model allows different numbers of convolutional layers to be added to optimize the network structure. Furthermore, the model is trained with training data to extract the highly representative features of the different classes. The training process is performed at every objective function evaluation, each with a different network structure and training options using a Bayesian optimization algorithm to find the optimal network structure and training options in a search space of possible solutions. The CNN model serves as the palm vein template creator or feature extractor for the proposed identification and verification experiments. Receiver operating characteristic (ROC) curve and equal error rate (EER) were plotted for evaluating the performance of the proposed model. Another palm vein method for the same purpose is proposed using different feature extraction and recognition techniques. In this method, a novel feature extraction methodology named Gabor positional Weber’s law descriptor (GPWLD) combing the Gabor features with positional Weber’s local descriptor (PWLD) features is proposed. Weber’s law descriptor (WLD) is a highly representative micro-pattern descriptor that performs well against noise and illumination changes in images. However, it lacks the ability capture the vein pattern features at different orientations. Moreover, its descriptor packs the local information content into a single histogram that does not take the spatial locality of micro-patterns into consideration. To solve these two issues, firstly, the palm vein image is passed through Gabor filter with different orientations to capture the salient rotational features found in the output feature maps. Secondly, the spatiality is achieved by uniformly dividing each feature map into several blocks. Next, Weber’s feature descriptor (WLD) histogram is computed for each block in every feature map. Finally, these histograms are concatenated to compose the final feature vector. Due to the high dimensionality of the final feature vector, principal component analysis (PCA) algorithm is utilized for feature size reduction. In the classification stage, a deep neural network comprising an optimized stacked autoencoder (SAE) with Bayesian optimization and a softmax layer is used. Optimization of the SAE is carried out using Bayesian optimization to find the optimal SAE structure and training options. The Experimental results verify the discriminative power of the extracted features and the accuracy of the proposed deep neural network. For both identification and verification experiments, the proposed method attains higher identification rate and lower EER than state-of-the-art methods.