الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 131 |  |  |
| | | from | 131 | | |
Abstract
Biometrics based personal authentication is regarded as an effective method for automatically recognizing, with a high confidence, a person’s identity. As a relatively new branch of biometrics, palmprint authentication has attracted increasing amount of attention due to its ease of acquisition, high user acceptance and reliability. In the present thesis, an accurate palmprint segmentation technique was applied to 480 images taken from Chinese Academy of Sciences Institute of Automation (CASIA) database. Results have shown perfect segmentation for the required region of interest (ROI) for all images. Palmprint identification systems try to identify the person providing the biometric data. This data is compared against a number of users’ biometric data. Therefore; identification is generally referred as 1: N (one-to-N or one-to-many). The system then recognizes a query palmprint image by searching for its nearest neighbor from all of the stored templates. A set of feature extraction techniques based on principal component analysis (PCA), discrete cosine transform (DCT), Radon transform, and Gabor filter, have been implemented and utilized in identification process. A neural network classifier and an Euclidean distance classifier have been used for matching purpose. A proposed fusion technique on the feature level between PCA and discrete Wavelet transform (DWT) has achieved an identification ratio of 99.16%. Palmprint verification systems verify or reject users’ identity. The user should first claim an identity then he/she provides his palmprint to be compared against his/her enrolled biometric data. The biometric system returns one of two possible answers, verified or not verified. Verification is usually referred to as 1:1 (one-to-one). A modified segmentation technique has been applied to CASIA palmprint images for preparation of low-resolution-based palmprint identity verification. Features have been obtained using principal lines of the palm and the log Gabor filter. The investigated approach was tested using CASIA database with 60 genuine users and 60 imposters’ images. Experimental results have shown that the equal error rate (EER) was 0.06% at threshold of 0.45. For high-security applications, such as forensic ones, high-resolution palmprints are required. A Radon transform-based technique was applied to THUPALMLAB database with 500 dpi. Tests were done with 60 genuine users and 60 imposters’ images. Results have shown that the EER was 0.05% at threshold of 0.55. Thesis Objectives: 1. A proposed method for palmprint preprocessing and segmentation is introduced. 2. Subspace-based method using principal component analysis (PCA) and another set of transform-based methods including DCT, Gabor transform, wavelet transform, and Radon transform, were used for feature extraction phase. Artificial neural networks and Euclidean distance are used for classification purpose. Fusion on the feature level between PCA and wavelet transform features is also proposed and implemented to obtain better identification results. A comparison between different identification methods is introduced. 3.Another segmentation method is introduced as a precursor for verification of low resolution palmprint images. Another verification technique based on Radon transform and Gabor filter is implemented using high resolution database. Both techniques are evaluated and compared using performance curves of FAR and FRR.