Parallel matching system for digital non text information (fingerprint image)

摘要

Ever since we were born, there are few parts of our body have been developed uniquely to represent our identities. Parts like retina patterns, veins, DNA, fingerprints are all unique for every human being. In order to recognize these unique patterns, scientists have spent billions of dollars and years of research and development to produce tools and techniques to identify humans based on their unique body parts. Today, the most highly utilized parts of our body for the identification purposes are fingerprints. Although there are so many readers and applications have been made for the job, there are still weaknesses that need to be addressed especially in the area of fingerprint matching. Therefore, the aim of this research is to find ways to increase the performance of fingerprint matching in response to the increased number of fingerprint records available in the system. This research focuses on developing a parallel matching process in a computer cluster to increase the performance of fingerprint matching, without taking into account additional features such as high throughput and others. The main idea is to minimize processing time with multiple processes executing parallels in several computers. As a proof of concept, a prototype consists of client and server processes has been developed. An interface which been integrated with a fingerprint device on client side enables a request (a fingerprint image) been send to a server process. Upon receiving a fingerprint image, the server process will place the request into a queue before distributing to a cluster of matching server throughout the network. The processing time for fingerprint matching has been improved especially when the amount of records increased. The detailed results of identification, duration of matching and similarity are discussed.