Using Genetic Algorithms For Implementing Human Face Recognition

Human face recognition technology is a hot and difficult research area incomputer vision and computer graphics,and it has been developed rapidly in the passeddecades. Because human face is nonrigid and its expression is changeful, human facerecognition has been facing tremendous difficulties in actual applications, which makeshuman face recognition become a challenging issue.In this article, GAs are applied in the principal steps of face recognition,including image segmentation, location of face and curing the angle, and a mathematicalmodel is built.The concrete means is to use GAs in image segmentation of themaximum entropy. The 1-D and 2-D maxmum entropy thresholding segmentationmethods are discussed and an improved maximum entropy thresholding segmentationmethod with GAs is presented. In the detection and location of face, a mathematicalmodel is established to measure the existence of two eyes, nose and mouse in a rectanglearea, and weight-based sum of these measurements is adoptedas genetic algorithms'sfitness. In the design of GA's operators and performing strategies, selection operatoris designed to be 'elite selection'; multi-point crossover can perform between fathergeneration and son generation; Mutation requires two chromosomes, thus premature isavoided by the greatest extent. Besides, an inversion operator is given to generate a newbinary string by choosing randomly two gene positions and inversing the substringbetween the two positions in a chromosome.For the GAs given in the article, a mathematical description is presented and theGAs' convergence properties are analyzed based on GA probability convergence theoryand Markov chain theory, and the proposed algorithms are proved to be globalconvergent.The emulation illustrates that the convergence velocity, optimal solution of the pre-sented genetic algorithms and the rate of face recognition, compared with the formerlyimproved genetic algorithm, are enhanced significantly.