Improved Arithmetic Optimization with Deep learning Driven Contactless Biometric Verification on Iris Images

Biometrics makes use of detection methods based on unique physical attributes, namely an iris, fingerprint, electrocardiogram, gait, voice, or image patterns to prevent property and privacy loss. Recently, contactless biometric detection has received significant interest because of the COVID-19 outbreak. The most popular detection technique is iris identification. The four major phases of the iris detection systems are feature matching, feature extraction, segmentation, and normalization. Even though precise iris segmentation is essential, abstracting discriminatory factors of iris images is the most difficult feature of this system. Convolution Neural Network (CNN) learns and extracts features from extremely complicated datasets like iris images. Therefore, this research develops an Improved Arithmetic Optimization with Deep learning Driven Contactless Biometric Verification on Iris Images (IAODL-CBVII) technique. The presented IAODL-CBVII technique initially performs iris localization process to identify the iris region from the background. Next, the iris region normalization process takes place to enable proper comparison of two iris images with varying sizes. For feature extraction, the IAODL-CBVII technique uses improved arithmetic optimization algorithm (IAOA) with Faster SqueezeNet model. Finally, biometric verification takes place by means of hybrid cascade forward neural network with Elman Neural Network (HCFNN-ENN) prototype. The experimental authentication of the IAODL-CBVII method is verified on iris databases and the comparative research portrayed the improvement results of the IAODL-CBVII technique through other current approaches.