Supervised Deep Statistical Analysis Model Using Improved Canny Edge Detection Based on Optimal Double-Threshold Estimation for Palmprint Recognition System

In this research, an efficient Supervised Deep Statistical Analysis Model based on Improved Canny Edge Detection (SDSA-ICED) system is proposed for the Palmprint Recognition System (PRS). To impart the novelty, Statistical Analysis Model on Improved Canny Edge Detection (SA-ICED) and Supervised Deep Learning (SDL) approaches are designed for feature extraction and classifier processes. Initially, acquired 2D-Palmprint Region of Interest (2D-PROI) images are normalized, smoothened, and enhanced to detect the correct edges for extending the determination of authenticity. To detect the correct edges of the 2D-PROI images, Improved Canny edge detection (ICED) based on Optimal Double-Threshold Estimation algorithm (ODTE) can be employed and produce Thin Edge Detection of Smoothened and Enhanced Image (TESEI). To extract a set of statistical unique features, find the inclination value of the line (m) between the endpoints of all edges of TESEI. Feed a set of features into the Supervised Deep Learning (SDL) classifier to authenticate the person’s palmprint at a higher accuracy rate. The proposed ICED algorithm’s edge detection quality has been observed by edge detection quality checking benchmarks such as Peak Signal Noise Ratio, Mean Square Error, and entropy. The proposed SDSA-ICED system has been scrutinized and evaluated with various benchmarks and scored 97.25% of authentication accuracy.