openCV的人脸识别主要通过Haar分类器实现，当然，这是在已有训练数据的基础上。openCV安装在 opencv/opencv/sources/data/haarcascades_cuda（或haarcascades）中存在预先训练好的物体检测器（xml格式），包括正脸、侧脸、眼睛、微笑、上半身、下半身、全身等。
openCV的的Haar分类器是一个监督分类器，首先对图像进行直方图均衡化并归一化到同样大小，然后标记里面是否包含要监测的物体。它首先由Paul Viola和Michael Jones设计，称为Viola Jones检测器。Viola Jones分类器在级联的每个节点中使用AdaBoost来学习一个高检测率低拒绝率的多层树分类器。它使用了以下一些新的特征：
1. 使用类Haar输入特征：对矩形图像区域的和或者差进行阈值化。
2. 积分图像技术加速了矩形区域的45°旋转的值的计算，用来加速类Haar输入特征的计算。
3. 使用统计boosting来创建两类问题（人脸和非人脸）的分类器节点（高通过率，低拒绝率）
4. 把弱分类器节点组成筛选式级联。即，第一组分类器最优，能通过包含物体的图像区域，同时允许一些不包含物体通过的图像通过；第二组分
类器次优分类器，也是有较低的拒绝率；以此类推。也就是说，对于每个boosting分类器，只要有人脸都能检测到，同时拒绝一小部分非人脸，并将其传给下一个分类器，是为低拒绝率。以此类推，最后一个分类器将几乎所有的非人脸都拒绝掉，只剩下人脸区域。只要图像区域通过了整个级联，则认为里面有物体。
此技术虽然适用于人脸检测，但不限于人脸检测，还可用于其他物体的检测，如汽车、飞机等的正面、侧面、后面检测。在检测时，先导入训练好的参数文件，其中haarcascade_frontalface_alt2.xml对正面脸的识别效果较好haarcascade_profileface.xml对侧脸的检测效果较好。当然，如果要达到更高的分类精度，可以收集更多的数据进行训练，这是后话。
// faceDetector.h
// This is just the face, eye, smile, profile detector from OpenCV's samples/c directory
//
/* *************** License:**************************
Jul. 18, 2016
Author: Liuph
Right to use this code in any way you want without warranty, support or any guarantee of it working.
OTHER OPENCV SITES:
* The source code is on sourceforge at:
http://sourceforge.net/projects/opencvlibrary/
* The OpenCV wiki page (As of Oct 1, 2008 this is down for changing over servers, but should come back):
http://opencvlibrary.sourceforge.net/
* An active user group is at:
http://tech.groups.yahoo.com/group/OpenCV/
* The minutes of weekly OpenCV development meetings are at:
http://pr.willowgarage.com/wiki/OpenCV
************************************************** */
#include "cv.h"
#include "highgui.h"
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
using namespace std;
static CvMemStorage* storage = 0;
static CvHaarClassifierCascade* cascade = 0;
static CvHaarClassifierCascade* nested_cascade = 0;
static CvHaarClassifierCascade* smile_cascade = 0;
static CvHaarClassifierCascade* profile = 0;
int use_nested_cascade = 0;
void detect_and_draw( IplImage* image );
/* The path that stores the trained parameter files.
After openCv is installed, the file path is
"opencv/opencv/sources/data/haarcascades_cuda" or "opencv/opencv/sources/data/haarcascades" */
const char* cascade_name =
"../faceDetect/haarcascade_frontalface_alt2.xml";
const char* nested_cascade_name =
"../faceDetect/haarcascade_eye_tree_eyeglasses.xml";
const char* smile_cascade_name =
"../faceDetect/haarcascade_smile.xml";
const char* profile_name =
"../faceDetect/haarcascade_profileface.xml";
double scale = 1;
int faceDetector(const char* imageName, int nNested, int nSmile, int nProfile)
{
CvCapture* capture = 0;
IplImage *frame, *frame_copy = 0;
IplImage *image = 0;
const char* scale_opt = "--scale=";
int scale_opt_len = (int)strlen(scale_opt);
const char* cascade_opt = "--cascade=";
int cascade_opt_len = (int)strlen(cascade_opt);
const char* nested_cascade_opt = "--nested-cascade";
int nested_cascade_opt_len = (int)strlen(nested_cascade_opt);
const char* smile_cascade_opt = "--smile-cascade";
int smile_cascade_opt_len = (int)strlen(smile_cascade_opt);
const char* profile_opt = "--profile";
int profile_opt_len = (int)strlen(profile_opt);
int i;
const char* input_name = 0;
int opt_num = 7;
char** opts = new char*[7];
opts[0] = "compile_opencv.exe";
opts[1] = "--scale=1";
opts[2] = "--cascade=1";
if (nNested == 1)
opts[3] = "--nested-cascade=1";
else
opts[3] = "--nested-cascade=0";
if (nSmile == 1)
opts[4] = "--smile-cascade=1";
else
opts[4] = "--smile-cascade=0";
if (nProfile == 1)
opts[5] = "--profile=1";
else
opts[5] = "--profile=0";
opts[6] = (char*)imageName;
for( i = 1; i < opt_num; i++ )
{
if( strncmp( opts[i], cascade_opt, cascade_opt_len) == 0)
{
cout<<"cascade: "<\"]\n"
" [--nested-cascade[=\"nested_cascade_path\"]]\n"
" [--scale[=\n"
" [filename|camera_index]\n" );
return -1;
}
storage = cvCreateMemStorage(0);
if( !input_name || (isdigit(input_name[0]) && input_name[1] == '\0') )
capture = cvCaptureFromCAM( !input_name ? 0 : input_name[0] - '0' );
else if( input_name )
{
image = cvLoadImage( input_name, 1 );
if( !image )
capture = cvCaptureFromAVI( input_name );
}
else
image = cvLoadImage( "../lena.jpg", 1 );
cvNamedWindow( "result", 1 );
if( capture )
{
for(;;)
{
if( !cvGrabFrame( capture ))
break;
frame = cvRetrieveFrame( capture );
if( !frame )
break;
if( !frame_copy )
frame_copy = cvCreateImage( cvSize(frame->width,frame->height),
IPL_DEPTH_8U, frame->nChannels );
if( frame->origin == IPL_ORIGIN_TL )
cvCopy( frame, frame_copy, 0 );
else
cvFlip( frame, frame_copy, 0 );
detect_and_draw( frame_copy );
if( cvWaitKey( 10 ) >= 0 )
goto _cleanup_;
}
cvWaitKey(0);
_cleanup_:
cvReleaseImage( &frame_copy );
cvReleaseCapture( &capture );
}
else
{
if( image )
{
detect_and_draw( image );
cvWaitKey(0);
cvReleaseImage( &image );
}
else if( input_name )
{
/* assume it is a text file containing the
list of the image filenames to be processed - one per line */
FILE* f = fopen( input_name, "rt" );
if( f )
{
char buf[1000+1];
while( fgets( buf, 1000, f ) )
{
int len = (int)strlen(buf), c;
while( len > 0 && isspace(buf[len-1]) )
len--;
buf[len] = '\0';
printf( "file %s\n", buf );
image = cvLoadImage( buf, 1 );
if( image )
{
detect_and_draw( image );
c = cvWaitKey(0);
if( c == 27 || c == 'q' || c == 'Q' )
break;
cvReleaseImage( &image );
}
}
fclose(f);
}
}
}
cvDestroyWindow("result");
return 0;
}
void detect_and_draw( IplImage* img )
{
static CvScalar colors[] =
{
{{0,0,255}},
{{0,128,255}},
{{0,255,255}},
{{0,255,0}},
{{255,128,0}},
{{255,255,0}},
{{255,0,0}},
{{255,0,255}}
};
IplImage *gray, *small_img;
int i, j;
gray = cvCreateImage( cvSize(img->width,img->height), 8, 1 );
small_img = cvCreateImage( cvSize( cvRound (img->width/scale),
cvRound (img->height/scale)), 8, 1 );
cvCvtColor( img, gray, CV_BGR2GRAY );
cvResize( gray, small_img, CV_INTER_LINEAR );
cvEqualizeHist( small_img, small_img );
cvClearMemStorage( storage );
if( cascade )
{
double t = (double)cvGetTickCount();
CvSeq* faces = cvHaarDetectObjects( small_img, cascade, storage,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
|CV_HAAR_DO_CANNY_PRUNING
//|CV_HAAR_SCALE_IMAGE
,
cvSize(30, 30) );
t = (double)cvGetTickCount() - t;
printf( "faces detection time = %gms\n", t/((double)cvGetTickFrequency()*1000.) );
for( i = 0; i < (faces ? faces->total : 0); i++ )
{
CvRect* r = (CvRect*)cvGetSeqElem( faces, i );
CvMat small_img_roi;
CvSeq* nested_objects;
CvSeq* smile_objects;
CvPoint center;
CvScalar color = colors[i%8];
int radius;
center.x = cvRound((r->x + r->width*0.5)*scale);
center.y = cvRound((r->y + r->height*0.5)*scale);
radius = cvRound((r->width + r->height)*0.25*scale);
cvCircle( img, center, radius, color, 3, 8, 0 );
//eye
if( nested_cascade != 0)
{
cvGetSubRect( small_img, &small_img_roi, *r );
nested_objects = cvHaarDetectObjects( &small_img_roi, nested_cascade, storage,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
//|CV_HAAR_DO_CANNY_PRUNING
//|CV_HAAR_SCALE_IMAGE
,
cvSize(0, 0) );
for( j = 0; j < (nested_objects ? nested_objects->total : 0); j++ )
{
CvRect* nr = (CvRect*)cvGetSeqElem( nested_objects, j );
center.x = cvRound((r->x + nr->x + nr->width*0.5)*scale);
center.y = cvRound((r->y + nr->y + nr->height*0.5)*scale);
radius = cvRound((nr->width + nr->height)*0.25*scale);
cvCircle( img, center, radius, color, 3, 8, 0 );
}
}
//smile
if (smile_cascade != 0)
{
cvGetSubRect( small_img, &small_img_roi, *r );
smile_objects = cvHaarDetectObjects( &small_img_roi, smile_cascade, storage,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
//|CV_HAAR_DO_CANNY_PRUNING
//|CV_HAAR_SCALE_IMAGE
,
cvSize(0, 0) );
for( j = 0; j < (smile_objects ? smile_objects->total : 0); j++ )
{
CvRect* nr = (CvRect*)cvGetSeqElem( smile_objects, j );
center.x = cvRound((r->x + nr->x + nr->width*0.5)*scale);
center.y = cvRound((r->y + nr->y + nr->height*0.5)*scale);
radius = cvRound((nr->width + nr->height)*0.25*scale);
cvCircle( img, center, radius, color, 3, 8, 0 );
}
}
}
}
if( profile )
{
double t = (double)cvGetTickCount();
CvSeq* faces = cvHaarDetectObjects( small_img, profile, storage,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
|CV_HAAR_DO_CANNY_PRUNING
//|CV_HAAR_SCALE_IMAGE
,
cvSize(30, 30) );
t = (double)cvGetTickCount() - t;
printf( "profile faces detection time = %gms\n", t/((double)cvGetTickFrequency()*1000.) );
for( i = 0; i < (faces ? faces->total : 0); i++ )
{
CvRect* r = (CvRect*)cvGetSeqElem( faces, i );
CvMat small_img_roi;
CvSeq* nested_objects;
CvSeq* smile_objects;
CvPoint center;
CvScalar color = colors[(7-i)%8];
int radius;
center.x = cvRound((r->x + r->width*0.5)*scale);
center.y = cvRound((r->y + r->height*0.5)*scale);
radius = cvRound((r->width + r->height)*0.25*scale);
cvCircle( img, center, radius, color, 3, 8, 0 );
//eye
if( nested_cascade != 0)
{
cvGetSubRect( small_img, &small_img_roi, *r );
nested_objects = cvHaarDetectObjects( &small_img_roi, nested_cascade, storage,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
//|CV_HAAR_DO_CANNY_PRUNING
//|CV_HAAR_SCALE_IMAGE
,
cvSize(0, 0) );
for( j = 0; j < (nested_objects ? nested_objects->total : 0); j++ )
{
CvRect* nr = (CvRect*)cvGetSeqElem( nested_objects, j );
center.x = cvRound((r->x + nr->x + nr->width*0.5)*scale);
center.y = cvRound((r->y + nr->y + nr->height*0.5)*scale);
radius = cvRound((nr->width + nr->height)*0.25*scale);
cvCircle( img, center, radius, color, 3, 8, 0 );
}
}
//smile
if (smile_cascade != 0)
{
cvGetSubRect( small_img, &small_img_roi, *r );
smile_objects = cvHaarDetectObjects( &small_img_roi, smile_cascade, storage,
1.1, 2, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
//|CV_HAAR_DO_ROUGH_SEARCH
//|CV_HAAR_DO_CANNY_PRUNING
//|CV_HAAR_SCALE_IMAGE
,
cvSize(0, 0) );
for( j = 0; j < (smile_objects ? smile_objects->total : 0); j++ )
{
CvRect* nr = (CvRect*)cvGetSeqElem( smile_objects, j );
center.x = cvRound((r->x + nr->x + nr->width*0.5)*scale);
center.y = cvRound((r->y + nr->y + nr->height*0.5)*scale);
radius = cvRound((nr->width + nr->height)*0.25*scale);
cvCircle( img, center, radius, color, 3, 8, 0 );
}
}
}
}
cvShowImage( "result", img );
cvReleaseImage( &gray );
cvReleaseImage( &small_img );
}
调整主函数中faceDetect(const char* imageName, int nNested, int nSmile, int nProfile)函数中的参数，分别表示图像文件名，是否检测眼睛，是否检测微笑，是否检测侧脸。以检测正脸、眼睛为例：
如果对分类器的参数不满意，或者说想识别其他的物体例如车、人、飞机、苹果等等等等，只需要选择适当的样本训练，获取该物体的各个方面的参数，训练过程可以通过openCV的haartraining实现（参考haartraining参考文档，opencv/apps/traincascade），主要包括个步骤：
1. 收集打算学习的物体数据集（如正面人脸图，侧面汽车图等， 1000~10000个正样本为宜），把它们存储在一个或多个目录下面。
2. 使用createsamples来建立正样本的向量输出文件，通过这个文件可以重复训练过程，使用同一个向量输出文件尝试各种参数。
3. 获取负样本，即不包含该物体的图像。
4. 训练。命令行实现。
