Using Simple Genetic Algorithm for a Hand Contour Classification: An Experimental Study

Abstract

The area of biometric systems has passed through considerable advancement in the past two decades. Supporting of security provision plays a key role in many branches. There are large amount of the biometrical markers which can be utilized in the person identification process. One of the possible ways is a method which uses a hand shape contour classification. The presented paper solves the problem of hand contours classification with use of a Simple Genetic Algorithm (SGA). The foundations of the SGA were established in 1950’s, but an improvement process of the SGA continues. The hand contour for the classification purposes is obtained from a color image from a biometric scanner. The biometric scanner has fixed pegs to hold the hand, or the hand can be freely placed on the scanning area. A core of the proposed estimator is an Iterative Closes Point algorithm which enables matching of the two point-clouds and expressing their dissimilarity regarding the elected metrics. In the experimental section, a large number of experiments were performed with a different setting of the SGA working parameters. Beside the capability to correctly align/match the hand contours, selected standard benchmark tests were performed with a corresponding number of dimensions. The presented estimator solves the thee-dimensional optimization task. Based on experimental results, it was proven that in the case of identical contours the proposed method, which utilizes the SGA optimizer, provides very accurate results.