资源描述
[人脸识别Step]在opencv下作人脸检测
———————————————————————————————— 作者:
———————————————————————————————— 日期:
18
个人收集整理 勿做商业用途
[人脸识别Step1]在opencv下作人脸检测
本帖被 admin 设置为精华(2007-08-03)
人脸识别的第一步,就是人脸检测。把人的脸部从一张照片中用计算机自动识别出来,作为下一步人脸识别的基础。
在opencv 中,库中自带了一个利用harr特征的人脸检测训练及检测函数:cvHaarDetectObjects.它利用训练好的检测器,在图片中间检测你想要的 物体,如人脸。opencv自带了很多检测器,在%opencv%data/haarcascades目录下,你可以随意取用。或者你也可以自己用图片训 练自己的检测器,之后拿来使用.
下面是检测人脸的源代码:
// DetectFaces.c
//
// Example code showing how to detect faces using
// OpenCV's CvHaarClassifierCascade
//
// See also, facedetect。c, in the samples directory。
//
// Usage: DetectFaces 〈imagefilename〉
#include <stdio.h>
#include "cv.h"
#include "highgui.h"
// *** Change this to your install location! ***
// *********************************************
#define OPENCV_ROOT ”C:/Program Files/OpenCV”
// *********************************************
void displayDetections(IplImage * pInpImg, CvSeq * pFaceRectSeq, char* FileName);
int main(int argc, char** argv)
{
// variables
IplImage * pInpImg = 0;
CvHaarClassifierCascade * pCascade = 0; // the face detector
CvMemStorage * pStorage = 0; // memory for detector to use
CvSeq * pFaceRectSeq; // memory-access interface
// usage check
if(argc < 2)
{
printf(”Missing name of image file!\n”
”Usage: %s <imagefilename〉\n”, argv[0]);
exit(-1);
}
// initializations
pInpImg = (argc > 1) ? cvLoadImage(argv[1], CV_LOAD_IMAGE_COLOR) : 0;
pStorage = cvCreateMemStorage(0);
pCascade = (CvHaarClassifierCascade *)cvLoad
((OPENCV_ROOT”/data/haarcascades/haarcascade_frontalface_default。xml”),
0, 0, 0 );
// validate that everything initialized properly
if( !pInpImg || !pStorage || !pCascade )
{
printf("Initialization failed: %s\n",
(!pInpImg)? "can't load image file" :
(!pCascade)? "can’t load haar—cascade —- ”
"make sure path is correct" :
"unable to allocate memory for data storage", argv[1]);
exit(-1);
}
// detect faces in image
pFaceRectSeq = cvHaarDetectObjects
(pInpImg, pCascade, pStorage,
1。1, // increase search scale by 10% each pass
3, // merge groups of three detections
CV_HAAR_DO_CANNY_PRUNING, // skip regions unlikely to contain a face
cvSize(40,40)); // smallest size face to detect = 40x40
// display detected faces
displayDetections(pInpImg, pFaceRectSeq, argv[1]);
// clean up and release resources
cvReleaseImage(&pInpImg);
if(pCascade) cvReleaseHaarClassifierCascade(&pCascade);
if(pStorage) cvReleaseMemStorage(&pStorage);
return 0;
}
void displayDetections(IplImage * pInpImg, CvSeq * pFaceRectSeq, char* FileName)
{
const char * DISPLAY_WINDOW = "Haar Window";
int i;
// create a window to display detected faces
cvNamedWindow(DISPLAY_WINDOW, CV_WINDOW_AUTOSIZE);
// draw a rectangular outline around each detection
for(i=0;i〈(pFaceRectSeq? pFaceRectSeq->total:0); i++ )
{
CvRect* r = (CvRect*)cvGetSeqElem(pFaceRectSeq, i);
CvPoint pt1 = { r->x, r->y };
CvPoint pt2 = { r—〉x + r-〉width, r—>y + r—>height };
cvRectangle(pInpImg, pt1, pt2, CV_RGB(0,255,0), 3, 4, 0);
cvSetImageROI(pInpImg, *r);
//char* FileName = argv[1];
IplImage* dst = cvCreateImage( cvSize(92,112), pInpImg—>depth, pInpImg—〉nChannels);
cvResize(pInpImg, dst, CV_INTER_LINEAR);
strcat(FileName,”.pgm");
cvSaveImage(FileName, dst);
}
// display face detections
cvShowImage(DISPLAY_WINDOW, pInpImg);
cvWaitKey(0);
cvDestroyWindow(DISPLAY_WINDOW);
}
程序会将人脸检测出来,显示在屏幕上,并存回原来的文件将其覆盖。
在window xp, devcpp 4。9。9.2 下编译通过。本文为互联网收集,请勿用作商业用途文档为个人收集整理,来源于网络
人脸识别Step2]对人脸图片作PCA降维
本帖被 admin 设置为精华(2007—08—03)
在上一篇文章中,我们将人脸从图片中检测了出来,并存回了同名的文件。下一步,就要进行识别了。
对 一幅图片而言,我们可以将它看成是一个二位数组,每个元素表达一个像素的特征,如颜色等。那么,对一幅我们处理好的大小为92*112的黑白单通道图片, 就需要一个长度为10304的浮点箱量来表示.这么高维度的数据,处理起来计算代价很大,因为它将所有像素一视同仁,效果也不一定好.
所以,我们对于人脸数据,采用了主成分分析 ( Principal Component Analysis , PCA ) 方法来做数据降维。那么,什么是PCA呢?
主 成分分析 ( Principal Component Analysis , PCA ) 是一种掌握事物主要矛盾的统计分析方法,它可以从多元事物中解析出主要影响因素,揭示事物的本质,简化复杂的问题。计算主成分的目的是将高维数据投影到较 低维空间。给定 n 个变量的 m 个观察值,形成一个 n ′ m 的数据矩阵, n 通常比较大。对于一个由多个变量描述的复杂事物,人们难以认识,那么是否可以抓住事物主要方面进行重点分析呢?如果事物的主要方面刚好体现在几个主要变量 上,我们只需要将这几个变量分离出来,进行详细分析.但是,在一般情况下,并不能直接找出这样的关键变量。这时我们可以用原有变量的线性组合来表示事物的 主要方面, PCA 就是这样一种分析方法。
PCA 的目标是寻找 r ( r<n )个新变量,使它们反映事物的主要特征,压缩原有数据矩阵的规模.每个新变量是原有变量的线性组合,体现原有变量的综合效果,具有一定的实际含义。这 r 个新变量称为“主成分",它们可以在很大程度上反映原来 n 个变量的影响,并且这些新变量是互不相关的,也是正交的.通过主成分分析,压缩数据空间,将多元数据的特征在低维空间里直观地表示出来。
Opencv 中,事先预置了PCA方法的函数,我们去调用即可。在下面给出的程序中,我将训练集人脸图片的PCA结果输出到TrainFace。txt,测试人脸图片 的PCA结果输出到TestFace.txt中,以备以后调用别的学习算法(如SVM等)使用。在程序中,直接使用欧式距离或余弦距离(需要在源代码中手 动更改编译)度量测试人脸和各训练人脸的相似度并输出,以供作简单的观察。
以下是源程序:
// eigenface.c, by Robin Hewitt, 2007
// edited by Xin QIao, AUG 2007
//
// Example program showing how to implement eigenface with OpenCV
// Usage:
//
// First, you need some face images。 I used the ORL face database.
// You can download it for free at
// www.cl.cam。ac.uk/research/dtg/attarchive/facedatabase。html
//
// List the training and test face images you want to use in the
// input files train.txt and test。txt。 (Example input files are provided
// in the download.) To use these input files exactly as provided, unzip
// the ORL face database, and place train。txt, test。txt, and eigenface.exe
// at the root of the unzipped database.
//
// To run the learning phase of eigenface, enter
// eigenface train
// at the command prompt. To run the recognition phase, enter
// eigenface test
#include 〈stdio.h〉
#include 〈string.h〉
#include <math。h>
#include ”cv.h”
#include ”cvaux.h”
#include ”highgui。h"
//// Global variables
IplImage ** faceImgArr = 0; // array of face images
CvMat * personNumTruthMat = 0; // array of person numbers
int nTrainFaces = 0; // the number of training images
int nEigens = 0; // the number of eigenvalues
IplImage * pAvgTrainImg = 0; // the average image
IplImage ** eigenVectArr = 0; // eigenvectors
CvMat * eigenValMat = 0; // eigenvalues
CvMat * projectedTrainFaceMat = 0; // projected training faces
//// Function prototypes
void learn();
void recognize();
void doPCA();
void storeTrainingData();
int loadTrainingData(CvMat ** pTrainPersonNumMat);
double * findNearestNeighbor(float * projectedTestFace);
int loadFaceImgArray(char * filename);
void printUsage();
//////////////////////////////////
// main()
//
int main( int argc, char** argv )
{
// validate that an input was specified
if( argc != 2 )
{
printUsage();
return -1;
}
if( !strcmp(argv[1], "train”) ) learn();
else if( !strcmp(argv[1], ”test") ) recognize();
else
{
printf(”Unknown command: %s\n", argv[1]);
printUsage();
}
return 0;
}
//////////////////////////////////
// learn()
//
void learn()
{
int i, j, k, offset;
// load training data
nTrainFaces = loadFaceImgArray(”train.txt");
if( nTrainFaces 〈 2 )
{
fprintf(stderr,
"Need 2 or more training faces\n"
”Input file contains only %d\n", nTrainFaces);
return;
}
// do PCA on the training faces
doPCA();
// project the training images onto the PCA subspace
projectedTrainFaceMat = cvCreateMat( nTrainFaces, nEigens, CV_32FC1 );
offset = projectedTrainFaceMat-〉step / sizeof(float);
for(i=0; i〈nTrainFaces; i++)
{
//int offset = i * nEigens;
cvEigenDecomposite(
faceImgArr,
nEigens,
eigenVectArr,
0, 0,
pAvgTrainImg,
//projectedTrainFaceMat—〉data.fl + i*nEigens);
projectedTrainFaceMat->data.fl + i*offset);
}
// store the projectedTrainFaceMat as TrainFace。txt
FILE * TrainfaceFile = 0;
if( TrainfaceFile = fopen(”TrainFace。txt”, ”w”) )
{
for(j = 0 ; j 〈 nTrainFaces ; j++){
fprintf(TrainfaceFile,"%d ", j);
for(k = 0; k 〈 nEigens ; k++){
fprintf(TrainfaceFile, ” %d : %f ", k, (projectedTrainFaceMat->data.fl + j*offset)[k] );
}
fprintf(TrainfaceFile,” -1 : ? \n”);
}
}
// store the recognition data as an xml file
storeTrainingData();
}
//////////////////////////////////
// recognize()
//
void recognize()
{
int i, j, nTestFaces = 0; // the number of test images
CvMat * trainPersonNumMat = 0; // the person numbers during training
float * projectedTestFace = 0;
// load test images and ground truth for person number
nTestFaces = loadFaceImgArray(”test.txt”);
printf(”%d test faces loaded\n", nTestFaces);
// load the saved training data
if( !loadTrainingData( &trainPersonNumMat ) ) return;
// project the test images onto the PCA subspace
projectedTestFace = (float *)cvAlloc( nEigens*sizeof(float) );
double * sim = (double *)cvAlloc( nEigens*sizeof(double) );;
for(i=0; i〈nTestFaces; i++)
{
int iNearest, k;
//int star[nTrainFaces];
// project the test image onto the PCA subspace
cvEigenDecomposite(
faceImgArr,
nEigens,
eigenVectArr,
0, 0,
pAvgTrainImg,
projectedTestFace);
// store the projectedTestFace as TestFace。txt
FILE * TestfaceFile = 0;
if( TestfaceFile = fopen(”TestFace。txt", ”w”) )
{
fprintf(TestfaceFile,"%d ", 0);
for(k = 0; k < nEigens ; k++){
fprintf(TestfaceFile, ” %d : %f ”, k, projectedTestFace[k] );
}
fprintf(TestfaceFile,” —1 : ? \n");
}
sim = findNearestNeighbor(projectedTestFace);
//truth = personNumTruthMat-〉data.i;
//nearest = trainPersonNumMat->data。i[iNearest];
/*
//sort
int order[nTrainFaces];
order[0] = iNearest;
double temp[nTrainFaces];
for ( j = 0; j < nTrainFaces; j++){
temp[j] = sim[j];
}
for ( k = 0; k 〈 nTrainFaces; k++){
//记录temp中最小元素的index
int result=0;
double leastSim = DBL_MAX;
//找到temp中最小元素的index,记录入result
for ( j = 0; j < nTrainFaces; j++){
if( temp[j] 〈 leastSim ){
leastSim = temp[j];
result = j;
}
}
order = result;
//将temp中找到的最小元素置为无穷大
temp[result] = DBL_MAX;
//}
*/
printf("%d test image : \n”,i+1);
for ( k = 0; k < nTrainFaces; k++)
printf(”star : %d, simility : %f\n”, k, sim[ k ]);
//printf("nearest = %d, Truth = %d\n\n", nearest, truth);
}
}
//////////////////////////////////
// loadTrainingData()
//
int loadTrainingData(CvMat ** pTrainPersonNumMat)
{
CvFileStorage * fileStorage;
int i;
// create a file-storage interface
fileStorage = cvOpenFileStorage( "。/data/facedata。xml", 0, CV_STORAGE_READ );
if( !fileStorage )
{
fprintf(stderr, ”Can’t open facedata.xml\n”);
return 0;
}
nEigens = cvReadIntByName(fileStorage, 0, "nEigens", 0);
nTrainFaces = cvReadIntByName(fileStorage, 0, ”nTrainFaces", 0);
*pTrainPersonNumMat = (CvMat *)cvReadByName(fileStorage, 0, ”trainPersonNumMat", 0);
eigenValMat = (CvMat *)cvReadByName(fileStorage, 0, ”eigenValMat”, 0);
projectedTrainFaceMat = (CvMat *)cvReadByName(fileStorage, 0, "projectedTrainFaceMat", 0);
pAvgTrainImg = (IplImage *)cvReadByName(fileStorage, 0, ”avgTrainImg”, 0);
eigenVectArr = (IplImage **)cvAlloc(nTrainFaces*sizeof(IplImage *));
for(i=0; i〈nEigens; i++)
{
char varname[200];
sprintf( varname, "eigenVect_%d”, i );
eigenVectArr = (IplImage *)cvReadByName(fileStorage, 0, varname, 0);
}
// release the file-storage interface
cvReleaseFileStorage( &fileStorage );
return 1;
}
//////////////////////////////////
// storeTrainingData()
//
void storeTrainingData()
{
CvFileStorage * fileStorage;
int i;
// create a file—storage interface
fileStorage = cvOpenFileStorage( ”./data/facedata.xml", 0, CV_STORAGE_WRITE );
// store all the data
cvWriteInt( fileStorage, "nEigens", nEigens );
cvWriteInt( fileStorage, ”nTrainFaces", nTrainFaces );
cvWrite(fileStorage, "trainPersonNumMat", personNumTruthMat, cvAttrList(0,0));
cvWrite(fileStorage, "eigenValMat”, eigenValMat, cvAttrList(0,0));
cvWrite(fileStorage, "projectedTrainFaceMat”, projectedTrainFaceMat, cvAttrList(0,0));
cvWrite(fileStorage, "avgTrainImg", pAvgTrainImg, cvAttrList(0,0));
for(i=0; i〈nEigens; i++)
{
char varname[200];
sprintf( varname, "eigenVect_%d", i );
cvWrite(fileStorage, varname, eigenVectArr, cvAttrList(0,0));
}
// release the file-storage interface
cvReleaseFileStorage( &fileStorage );
}
//////////////////////////////////
// findNearestNeighbor()
//
double * findNearestNeighbor(float * projectedTestFace)
{
//double leastDistSq = 1e12;
double leastDistSq = DBL_MAX;
int i, iTrain, iNearest = 0;
double * sim = (double *)malloc(nTrainFaces * sizeof(double));
for(iTrain=0; iTrain〈nTrainFaces; iTrain++)
{
double distSq=0;
long double distSx=0;
long double distSy=0;
for(i=0; i〈nEigens; i++)
{
float d_i =
projectedTestFace -
projectedTrainFaceMat—>data.fl[iTrain*nEigens + i];
//distSq += d_i*d_i / eigenValMat—>data.fl; // Mahalanobis
distSq += d_i*d_i; // Euclidean
/*
/////////////////////////////////////////////////////////////////
double d_i =
projectedTestFace *
projectedTr
展开阅读全文