ABSTRACT:
Biometric system is the system that needs the identification of a pattern in order to indicate the difference in the features from one person to another. This study focuses on palmprint classification which is considered as an effective method of obtaining accurate results in the wide development. The main problem of palmprint classification is the accuracy rate which needs to be improved, especially when these features are affected by orientation, illumination, and scaling changes from one person to another. In this research, a prototype for a biometric system was developed, which is a palmprint for personal verification using the bag of words (BoW) method to improve the accuracy of palmprint verification. This can be achieved by extracting interesting points from palmprint by determining the distance of patches from the centroids of codebook built by clustering scale-invariant character transform features utilising k-means. Hence, the clustering of features is indicated as the most significant tool for the task of image classification, where this task assists in minimising the testing features in the palmprint images. The proposed system was tested on two types of databases, namely the Indian Institute of Technology Kanpur (IITK) and Chinese Academy of Sciences Institute of Automation (CASIA) databases of 900 and 900 images, respectively. The method for classifying palmprints was examined in a case study to assess and determine the efficiency of the proposed method utilising BoW clustering. The experiment gave an accuracy rate of 99.88% when using the IITK database and 96.87% with the CASIA database. The approach is effective for a highlevel biometric security.

Keywords: Biometric, BoW, features, K-means clustering, SIFT.