外文科技资料翻译
英文原文
Research on a face recognition system by the genetic algorithm
Computer vision and recognition is playing anincreasingly important role in modern intelligent control.Object detection is the first and most important step
inobjectrecognition.Traditionally,a special object can berecognized by the template matching method,but the recognition speed has always been a problem.In this article,animproved general genetic algorithm-based face recognitionsystem is proposed.The genetic algorithm(GA)has beenconsidered to be a robust and global searching method.Here,the chromosomes generated by GA contain the information needed to recognize the object.The purpose of thisarticle is to propose a practical method of face detection andrecognition.Finally,the experimental results,and a comparison with the traditional template matching method,andsome otherconsiderations,are also given.
1 Introduction
If we search on the web or in a conference proceedingsabout intelligent control,lots of papers and applications arepresented.Among them,image processing and
recognitionoccupy a very large percentage.The higher the degree ofintelligence,the more important the image detection andrecognition technology.
For controlling an intelligent system(autonomous mobile vehicle,robot,etc.),the most important element is thecontrol strategy,but before automatically making it move,image recognition is needed.For an intelligent control system,it is necessary to acquire information about the external world automatically by sensors,in order to recognize itsposition and the surrounding situation.A camera is one ofthe most important sensors for computer vision.That is tosay,the system endeavors to find out what is in an image(the environment of the robot)taken by the camera:trafficsigns,obstacles,guidelines,etc. The reliability and time-response of object detection andrecognition have a major influence on the performance andusability of the whole object recognition
system.1Thetemplatematching method is a practicable and reasonablemethod for object detection.2This article gives an improvement in the general template matching method. In addition,in order to search for the object of interestin an image,lots of data need to be processed.The geneticalgorithm(GA)has been considered to be a robust andglobal searching method(although it is sometimes said thatGA can not be used for finding the globaloptimization3).Here,the chromosomes generated by GA contain information about the image data,and the genetic and evolutionoperations are used to obtain the best match to the template:searching for the best match is the goal of this article.This thought emerged from the features of the GA,andthe need to recognize the faces of people easily and quicklyby an intelligent system.The single concept and features ofimage processing and the GA will not be introduced here,because there is already extensive literature on this subject.
In this article,Sect.2 gives the encoding method of theGA and the experimental setting that is used.In Sect.3,theexperiment and the analysis are addressed.Some conclusions are given in Sect.4.
Theory and experimental setting
For an image recognition system,the most interesting partthat has special features has first to be detected in the original image.This is called object detection.After that,thispart will be compared to a template to see if it is similar ornot.This is called object recognition.For example,if wewant to find a special person in an image,we first have todetect people in the image,and then recognize which one isthe person of
interest(sometimes these two steps will beexecuted simultaneously).The whole procedure is shown inFig.1.
Fig.1. Object recognition system
Statistical object recognition involves locating and isolating the targets in an image,and then identifying them bystatistical decision theory.One of the oldest techniques ofpattern recognition is matching filtering,4which allowsthe computation of a measure of the similarity between theoriginal image f(x,y)and a template h(x,y).Define themean-squared distance
2d2fh???{f(x,y?h(x,y))}dxdy
(1)
Rfh???f(x,y)h(x,y)dxdy,if the image and template are normalized by
??f(2)
2(x,y)dxdy???h2(x,y)dxdy
And then
d2fh???{f(x,y)?h(x,y)}dxdy
2 =??{f2(x,y)?2f(x,y)h(x,y)?h2(x,y)}dxdy
=2??f2(x,y)dxdy?2Rfh (3)
For the right-hand side of Eq.3,the first term is constant,and thusRfh can measure as the least-squared similaritybetween the original image and the template.5If Rfh has alarge value(which means that d2fhis small enough),then theimage is judged to match the template.If Rfh is less than apreselected threshold,the recognition process will eitherreject
the match or create a new pattern,which means thatthe similarity between the object in the original image andthe template is not satisfied. 2.1 Genetic encoding
As introduced above,the chromosomes generated by theGA contain information about the image data,so the firststep is to encode the image data into a binary
string.6Theparameters of the center of a face(x,y)in the originalimage,the rate of scale to satisfy eq.2,and the rotating
angleθare encoded into the elements of a gene.Some important parameters of the GA used here are given inTable 1,and the search field and region are given in Table2.As shown in Table2,one chromosome contains 4 bytes:the coordinate(x,y)in the original image,the rate of scale,
and the rotation angleθ.
2.2 Experimental setting
The experiment is done by first loading the original and thetemplate images.The GA is used to find whether or notthere is the object of a template in the original image.If theanswer isYES,then in the original image the result givesthe coordinates of the center of the object,the scale,and therotation angle from the template.For comparison,thegeneral template matching method is also presented.7The execution time shows the effectiveness of the GA-
based recognition method.Figures 2 and 3 are the original images and the templatesfor the experiment.The values are the width×height inpixels of the image.In Fig.2,three images are presented,the content and size of which are different.Figure 2a hastwo faces(the faces of a person and a toy),Fig.2b shows aface tipped to one side,and the person in Fig.2c wears a hat
and the background is more complicated than in Figs.2a and b.
The two templates in Fig.3 are not extracted from thesame image.For normal use,the template should be extracted as the average of several feature images.In Fig.4,the template(a)-0 is generated from(a)-1,(a)-2,and(a)-3,and takes the average value of the gray levels from the threemodels.The same is also true for(b)-0.
相关推荐：
loading