Feature Extraction Approach Based On Enhanced Antcolonyoptimization Algorithm For Higher Accuracy In Human Iris Identification

Zainal Abidin, Zaheera and Ahmad, Rabiah and Abal Abas, Zuraida and Abdul Latip, Sheikh Faisal (2019) Feature Extraction Approach Based On Enhanced Antcolonyoptimization Algorithm For Higher Accuracy In Human Iris Identification. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (24 Pages) Feature Extraction Approach Based On Enhanced Antcolonyoptimization Algorithm For Higher Accuracy In Human Iris Identification.pdf - Submitted Version Restricted to Registered users only Download (536kB)

Abstract

Iris recognition is an autonomous approach to measure unique feature using False Reject Rate (FRR).Due to high rate in FRR,caused by internal (i.e. aging and health condition) and external factors (i.e. technical fault and source of lighting) produce noisy iris image.Current ways of reducing FRR is by decreasing noise in iris image, multiple biometric modality and selection of unique iris features.However,the elimination of noise is not at the expected level,multiple modalities demand higher cost of maintenance and feature selection still shows low accuracy performance,around 50%.Therefore,to mimic human behavior,computer learns through the use of natural computing languages such as particle swarm optimization (PSO),Bee Optimization (BO) and ant colony optimization (ACO).ACO algorithm is chosen because it produces good performance and able to reduce noise,which convince us toimplement this in iris recognition.Moreover,it consumes less memory consumption and scans the images pixel by pixel although in high noise images.The research is conducted in two phases.In phase 1,ACO searches unique features pixels in random manner,meanwhile in phase 2; ACO algorithm is enhanced in feature extraction based on pheromone weight age which is at various angles (0o, 45o, 90o,135o and 180o). Later,in matching phase,the iris features are compared using Euclidean Distance with other iris templates in iris databases (i.e. CASIA and UBIRIS) to evaluate the accuracy performance.To evaluate the new approach,the genuine acceptance rate (GAR) is measured based on the acceptance of accuracy performance,(1 – FRR%) compared to the existing one.This proposes approach adapts unique iris features in robust and use small amount of information in unique iris features to determine the genuine.

Actions (login required)

View Item