Identification of vagina and pelvis from iris region using artificial neural network

Iris recognition not only can be used in biometrics technology but also in medical application by identifying the region that relates to the body part. This paper describes a technique for identification of vagina and pelvis regions from iris region using Artificial Neural Network (ANN) based on iridology chart whereby the ANN process utilized Feed Forward Neural Network (FFNN). The localization of the iris is carried out using two methods namely Circular Boundary Detector (CBD) and Circular Hough Transform (CHT). The iris is segmented based on the iridology chart and unwrapped into polar form using Daugman’s Rubber Sheet Model. The vagina and pelvis regions are cropped into pixel size of 40×7 for feature extraction using Principal component Analysis (PCA) and classified using FFNN. In the experiments, 15 pelvis and 20 vagina regions are used for classification. The best result obtained gives overall correct identification from localization using CBD and CHT of about 67% and 81% respectively. From the experiments, it is observed that vagina and pelvis regions are able to be identified even though the results obtained are not 100% accurate.