opencv 手势识别
我使用OpenCV2.4.4的windows版本+Qt4.8.3+VS2010的编译器做了一个手势识别的小程序。
本程序主要使到了Opencv的特征训练库和最基本的图像处理的知识，包括肤色检测等等。
废话不多，先看一下基本的界面设计，以及主要功能：
相信对于Qt有一些了解的人都不会对这个界面的设计感到陌生吧！（该死，该死！）我们向下走：
紧接着是Qt导入OPenCV2.4.4的库文件：（先看一下Qt的工程文件吧）
#-------------------------------------------------
#
# Project created by QtCreator 2013-05-25T11:16:11
#
#------------------------------------------------- QT += core gui CONFIG += warn_off greaterThan(QT_MAJOR_VERSION, 4): QT += widgets TARGET = HandGesture
TEMPLATE = app INCLUDEPATH += E:/MyQtCreator/MyOpenCV/opencv/build/include SOURCES += main.cpp\
handgesturedialog.cpp \
SRC/GestrueInfo.cpp \
SRC/AIGesture.cpp HEADERS += handgesturedialog.h \
SRC/GestureStruct.h \
SRC/GestrueInfo.h \
SRC/AIGesture.h FORMS += handgesturedialog.ui #Load OpenCV runtime libs
win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_core244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_core244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_features2d244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_features2d244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_haartraining_engine
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_haartraining_engined INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_highgui244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_highgui244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_objdetect244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_objdetect244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_video244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_video244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_calib3d244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_calib3d244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_contrib244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_contrib244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_imgproc244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_imgproc244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_legacy244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_legacy244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_ml244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_ml244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_photo244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_photo244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10 win32:CONFIG(release, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_nonfree244
else:win32:CONFIG(debug, debug|release): LIBS += -L$$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10/lib/ -lopencv_nonfree244d INCLUDEPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
DEPENDPATH += $$PWD/../../../MyQtCreator/MyOpenCV/opencv/build/x86/vc10
当做好以上的基本配置之后，我们进行手势识别的开发：
第一：要采集到原始的图片
采集好原始图片后进行修正，包括尺寸大小，那时我还使用到了matlab这个强大的工具，
紧接着进行图像的样本特征提取，到网上把，CSDN中有大量的关于对图像特征训练库的识别与训练，按照他们一步一步的操作模式不会有问题的饿
下面是要通过摄像头进行图像的采集，直接贴代码：
void HandGestureDialog::on_pushButton_OpenCamera_clicked()
{
cam = cvCreateCameraCapture(0);
timer->start(time_intervals);
frame = cvQueryFrame(cam); ui->pushButton_OpenCamera->setDisabled (true);
ui->pushButton_CloseCamera->setEnabled (true);
ui->pushButton_ShowPause->setEnabled (true);
ui->pushButton_SnapImage->setEnabled (true);
afterSkin = cvCreateImage (cvSize(frame->width,frame->height),IPL_DEPTH_8U,1);
}
void HandGestureDialog::readFarme()
{
frame = cvQueryFrame(cam);
QImage image((const uchar*)frame->imageData,
frame->width,
frame->height,
QImage::Format_RGB888);
image = image.rgbSwapped();
image = image.scaled(320,240);
ui->label_CameraShow->setPixmap(QPixmap::fromImage(image));
gesture.SkinDetect (frame,afterSkin); /*next to opencv*/ if(status_switch == Recongnise)
{
// Flips the frame into mirror image
cvFlip(frame,frame,1); // Call the function to detect and draw the hand positions
StartRecongizeHand(frame);
}
}
查看一下样例图片：
开始训练的核心代码：
void HandGestureDialog::on_pushButton_StartTrain_clicked()
{
QProgressDialog* process = new QProgressDialog(this);
process->setWindowTitle ("Traning Model");
process->setLabelText("Processing...");
process->setModal(true);
process->show ();
gesture.setMainUIPointer (this);
gesture.Train(process);
QMessageBox::about (this,tr("完成"),tr("手势训练模型完成"));
}
void CAIGesture::Train(QProgressDialog *pBar)//对指定训练文件夹里面的所有手势进行训练
{
QString curStr = QDir::currentPath ();
QString fp1 = "InfoDoc/gestureFeatureFile.yml";
fp1 = curStr + "/" + fp1;
CvFileStorage *GestureFeature=cvOpenFileStorage(fp1.toStdString ().c_str (),0,CV_STORAGE_WRITE);
FILE* fp;
QString fp2 = "InfoDoc/gestureFile.txt";
fp2 = curStr + "/" + fp2;
fp=fopen(fp2.toStdString ().c_str (),"w");
int FolderCount=0; /*获取当前的目录，然后得到当前的子目录*/
QString trainStr = curStr;
trainStr += "/TraningSample/";
QDir trainDir(trainStr);
GestureStruct gesture;
QFileInfoList list = trainDir.entryInfoList(); pBar->setRange(0,list.size ()-2); for(int i=2;i<list.size ();i++)
{
pBar->setValue(i-1); QFileInfo fileInfo = list.at (i);
if(fileInfo.isDir () == true)
{
FolderCount++; QString tempStr = fileInfo.fileName ();
fprintf(fp,"%s\n",tempStr.toStdString ().c_str ());
gesture.angleName = tempStr.toStdString ()+"angleName";
gesture.anglechaName = tempStr.toStdString ()+"anglechaName";
gesture.countName = tempStr.toStdString ()+"anglecountName"; tempStr = trainStr + tempStr + "/";
QDir subDir(tempStr);
OneGestureTrain(subDir,GestureFeature,gesture);
}
}
pBar->autoClose ();
delete pBar;
pBar = NULL;
fprintf(fp,"%s%d","Hand Gesture Number: ",FolderCount);
fclose(fp);
cvReleaseFileStorage(&GestureFeature);
}
void CAIGesture::OneGestureTrain(QDir GestureDir,CvFileStorage *fs,GestureStruct gesture)//对单张图片进行训练
{
IplImage* TrainImage=0;
IplImage* dst=0;
CvSeq* contour=NULL;
CvMemStorage* storage;
storage = cvCreateMemStorage(0);
CvPoint center=cvPoint(0,0);
float radius=0.0;
float angle[FeatureNum][10]={0},anglecha[FeatureNum][10]={0},anglesum[FeatureNum][10]={0},anglechasum[FeatureNum][10]={0};
float count[FeatureNum]={0},countsum[FeatureNum]={0}; int FileCount=0;
/*读取该目录下的所有jpg文件*/
QFileInfoList list = GestureDir.entryInfoList();
QString currentDirPath = GestureDir.absolutePath ();
currentDirPath += "/";
for(int k=2;k<list.size ();k++)
{
QFileInfo tempInfo = list.at (k);
if(tempInfo.isFile () == true)
{
QString fileNamePath = currentDirPath + tempInfo.fileName ();
TrainImage=cvLoadImage(fileNamePath.toStdString ().c_str(),1);
if(TrainImage==NULL)
{
cout << "can't load image" << endl;
cvReleaseMemStorage(&storage);
cvReleaseImage(&dst);
cvReleaseImage(&TrainImage);
return;
}
if(dst==NULL&&TrainImage!=NULL)
dst=cvCreateImage(cvGetSize(TrainImage),8,1);
SkinDetect(TrainImage,dst);
FindBigContour(dst,contour,storage);
cvZero(dst);
cvDrawContours( dst, contour, CV_RGB(255,255,255),CV_RGB(255,255,255), -1, -1, 8 );
ComputeCenter(contour,center,radius); GetFeature(dst,center,radius,angle,anglecha,count);
for(int j=0;j<FeatureNum;j++)
{
countsum[j]+=count[j];
for(int k=0;k<10;k++)
{
anglesum[j][k]+=angle[j][k];
anglechasum[j][k]+=anglecha[j][k];
}
}
FileCount++;
cvReleaseImage(&TrainImage);
}
}
for(int i=0;i<FeatureNum;i++)
{
gesture.count[i]=countsum[i]/FileCount;
for(int j=0;j<10;j++)
{
gesture.angle[i][j]=anglesum[i][j]/FileCount;
gesture.anglecha[i][j]=anglechasum[i][j]/FileCount;
}
}
cvStartWriteStruct(fs,gesture.angleName.c_str (),CV_NODE_SEQ,NULL);//开始写入yml文件 int i=0;
for(i=0;i<FeatureNum;i++)
cvWriteRawData(fs,&gesture.angle[i][0],10,"f");//写入肤色角度的值 cvEndWriteStruct(fs);
cvStartWriteStruct(fs,gesture.anglechaName.c_str (),CV_NODE_SEQ,NULL); for(i=0;i<FeatureNum;i++)
cvWriteRawData(fs,&gesture.anglecha[i][0],10,"f");//写入非肤色角度的值 cvEndWriteStruct(fs);
cvStartWriteStruct(fs,gesture.countName.c_str (),CV_NODE_SEQ,NULL);
cvWriteRawData(fs,&gesture.count[0],FeatureNum,"f");//写入肤色角度的个数
cvEndWriteStruct(fs); cvReleaseMemStorage(&storage);
cvReleaseImage(&dst);
}
void CAIGesture::SkinDetect(IplImage* src,IplImage* dst)
{
IplImage* hsv = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 3);//use to split to HSV
IplImage* tmpH1 = cvCreateImage( cvGetSize(src), IPL_DEPTH_8U, 1);//Use To Skin Detect
IplImage* tmpS1 = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* tmpH2 = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* tmpS3 = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* tmpH3 = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* tmpS2 = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* H = cvCreateImage( cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* S = cvCreateImage( cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* V = cvCreateImage( cvGetSize(src), IPL_DEPTH_8U, 1);
IplImage* src_tmp1=cvCreateImage(cvGetSize(src),8,3); cvSmooth(src,src_tmp1,CV_GAUSSIAN,3,3); //Gaussian Blur
cvCvtColor(src_tmp1, hsv, CV_BGR2HSV );//Color Space to Convert
cvCvtPixToPlane(hsv,H,S,V,0);//To Split 3 channel /*********************Skin Detect**************/
cvInRangeS(H,cvScalar(0.0,0.0,0,0),cvScalar(20.0,0.0,0,0),tmpH1);
cvInRangeS(S,cvScalar(75.0,0.0,0,0),cvScalar(200.0,0.0,0,0),tmpS1);
cvAnd(tmpH1,tmpS1,tmpH1,0); // Red Hue with Low Saturation
// Hue 0 to 26 degree and Sat 20 to 90
cvInRangeS(H,cvScalar(0.0,0.0,0,0),cvScalar(13.0,0.0,0,0),tmpH2);
cvInRangeS(S,cvScalar(20.0,0.0,0,0),cvScalar(90.0,0.0,0,0),tmpS2);
cvAnd(tmpH2,tmpS2,tmpH2,0); // Red Hue to Pink with Low Saturation
// Hue 340 to 360 degree and Sat 15 to 90
cvInRangeS(H,cvScalar(170.0,0.0,0,0),cvScalar(180.0,0.0,0,0),tmpH3);
cvInRangeS(S,cvScalar(15.0,0.0,0,0),cvScalar(90.,0.0,0,0),tmpS3);
cvAnd(tmpH3,tmpS3,tmpH3,0); // Combine the Hue and Sat detections
cvOr(tmpH3,tmpH2,tmpH2,0);
cvOr(tmpH1,tmpH2,tmpH1,0); cvCopy(tmpH1,dst); cvReleaseImage(&hsv);
cvReleaseImage(&tmpH1);
cvReleaseImage(&tmpS1);
cvReleaseImage(&tmpH2);
cvReleaseImage(&tmpS2);
cvReleaseImage(&tmpH3);
cvReleaseImage(&tmpS3);
cvReleaseImage(&H);
cvReleaseImage(&S);
cvReleaseImage(&V);
cvReleaseImage(&src_tmp1);
}
//To Find The biggest Countour
void CAIGesture::FindBigContour(IplImage* src,CvSeq* (&contour),CvMemStorage* storage)
{
CvSeq* contour_tmp,*contourPos;
int contourcount=cvFindContours(src, storage, &contour_tmp, sizeof(CvContour), CV_RETR_LIST, CV_CHAIN_APPROX_NONE );
if(contourcount==0)
return;
CvRect bndRect = cvRect(0,0,0,0);
double contourArea,maxcontArea=0;
for( ; contour_tmp != 0; contour_tmp = contour_tmp->h_next )
{
bndRect = cvBoundingRect( contour_tmp, 0 );
contourArea=bndRect.width*bndRect.height;
if(contourArea>=maxcontArea)//find Biggest Countour
{
maxcontArea=contourArea;
contourPos=contour_tmp;
}
}
contour=contourPos;
}
//Calculate The Center
void CAIGesture::ComputeCenter(CvSeq* (&contour),CvPoint& center,float& radius)
{
CvMoments m;
double M00,X,Y;
cvMoments(contour,&m,0);
M00=cvGetSpatialMoment(&m,0,0);
X=cvGetSpatialMoment(&m,1,0)/M00;
Y=cvGetSpatialMoment(&m,0,1)/M00; center.x=(int)X;
center.y=(int)Y; /*******************tO find radius**********************/
int hullcount;
CvSeq* hull;
CvPoint pt;
double tmpr1,r=0;
hull=cvConvexHull2(contour,0,CV_COUNTER_CLOCKWISE,0);
hullcount=hull->total;
for(int i=1;i<hullcount;i++)
{
pt=**CV_GET_SEQ_ELEM(CvPoint*,hull,i);//get each point
tmpr1=sqrt((double)((center.x-pt.x)*(center.x-pt.x))+(double)((center.y-pt.y)*(center.y-pt.y)));//计算与中心点的大小
if(tmpr1>r)//as the max radius
r=tmpr1;
}
radius=r;
}
void CAIGesture::GetFeature(IplImage* src,CvPoint& center,float radius,
float angle[FeatureNum][10],
float anglecha[FeatureNum][10],
float count[FeatureNum])
{
int width=src->width;
int height=src->height;
int step=src->widthStep/sizeof(uchar);
uchar* data=(uchar*)src->imageData; float R=0.0;
int a1,b1,x1,y1,a2,b2,x2,y2;//the distance of the center to other point
float angle1_tmp[200]={0},angle2_tmp[200]={0},angle1[50]={0},angle2[50]={0};//temp instance to calculate angule
int angle1_tmp_count=0,angle2_tmp_count=0,angle1count=0,angle2count=0,anglecount=0; for(int i=0;i<FeatureNum;i++)//分FeatureNum层进行特征提取（也就是5层）分析
{
R=(i+4)*radius/9;
for(int j=0;j<=3600;j++)
{
if(j<=900)
{
a1=(int)(R*sin(j*3.14/1800));//这个要自己实际画一张图就明白了
b1=(int)(R*cos(j*3.14/1800));
x1=center.x-b1;
y1=center.y-a1;
a2=(int)(R*sin((j+1)*3.14/1800));
b2=(int)(R*cos((j+1)*3.14/1800));
x2=center.x-b2;
y2=center.y-a2;
}
else
{
if(j>900&&j<=1800)
{
a1=(int)(R*sin((j-900)*3.14/1800));
b1=(int)(R*cos((j-900)*3.14/1800));
x1=center.x+a1;
y1=center.y-b1;
a2=(int)(R*sin((j+1-900)*3.14/1800));
b2=(int)(R*cos((j+1-900)*3.14/1800));
x2=center.x+a2;
y2=center.y-b2;
}
else
{
if(j>1800&&j<2700)
{
a1=(int)(R*sin((j-1800)*3.14/1800));
b1=(int)(R*cos((j-1800)*3.14/1800));
x1=center.x+b1;
y1=center.y+a1;
a2=(int)(R*sin((j+1-1800)*3.14/1800));
b2=(int)(R*cos((j+1-1800)*3.14/1800));
x2=center.x+b2;
y2=center.y+a2;
}
else
{
a1=(int)(R*sin((j-2700)*3.14/1800));
b1=(int)(R*cos((j-2700)*3.14/1800));
x1=center.x-a1;
y1=center.y+b1;
a2=(int)(R*sin((j+1-2700)*3.14/1800));
b2=(int)(R*cos((j+1-2700)*3.14/1800));
x2=center.x-a2;
y2=center.y+b2;
}
}
} if(x1>0&&x1<width&&x2>0&&x2<width&&y1>0&&y1<height&&y2>0&&y2<height)
{
if((int)data[y1*step+x1]==255&&(int)data[y2*step+x2]==0)
{
angle1_tmp[angle1_tmp_count]=(float)(j*0.1);//从肤色到非肤色的角度
angle1_tmp_count++;
}
else if((int)data[y1*step+x1]==0&&(int)data[y2*step+x2]==255)
{
angle2_tmp[angle2_tmp_count]=(float)(j*0.1);//从非肤色到肤色的角度
angle2_tmp_count++;
}
}
}
int j=0;
for(j=0;j<angle1_tmp_count;j++)
{
if(angle1_tmp[j]-angle1_tmp[j-1]<0.2)//忽略太小的角度
continue;
angle1[angle1count]=angle1_tmp[j];
angle1count++;
} for(j=0;j<angle2_tmp_count;j++)
{
if(angle2_tmp[j]-angle2_tmp[j-1]<0.2)
continue;
angle2[angle2count]=angle2_tmp[j];
angle2count++;
} for(j=0;j<max(angle1count,angle2count);j++)
{
if(angle1[0]>angle2[0])
{
if(angle1[j]-angle2[j]<7)//忽略小于7度的角度，因为人的手指一般都大于这个值
continue;
angle[i][anglecount]=(float)((angle1[j]-angle2[j])*0.01);//肤色的角度
anglecha[i][anglecount]=(float)((angle2[j+1]-angle1[j])*0.01);//非肤色的角度，例如手指间的角度
anglecount++;
}
else
{
if(angle1[j+1]-angle2[j]<7)
continue;
anglecount++;
angle[i][anglecount]=(float)((angle1[j+1]-angle2[j])*0.01);
anglecha[i][anglecount]=(float)((angle2[j]-angle1[j])*0.01);
}
} if(angle1[0]<angle2[0])
angle[i][0]=(float)((angle1[0]+360-angle2[angle2count-1])*0.01);
else
anglecha[i][0]=(float)((angle2[0]+360-angle1[angle1count-1])*0.01); count[i]=(float)anglecount;
angle1_tmp_count=0,angle2_tmp_count=0,angle1count=0,angle2count=0,anglecount=0;
for(j=0;j<200;j++)
{
angle1_tmp[j]=0;
angle2_tmp[j]=0;
}
for(j=0;j<50;j++)
{
angle1[j]=0;
angle2[j]=0;
}
}
}
基本上对于自己使用代码创建的训练库的特征提取函数和基本的肤色检测和连通域的检测的函数的核心代码都已经贴到上面去了。
然后再看一下对于特定的手势识别的文件：
void HandGestureDialog::on_pushButton_StartRecongnise_clicked()
{
if(cam==NULL)
{
QMessageBox::warning (this,tr("Warning"),tr("Please Check Camera !"));
return;
} status_switch = Nothing; status_switch = Recongnise;
}
void HandGestureDialog::StartRecongizeHand (IplImage *img)
{
// Create a string that contains the exact cascade name
// Contains the trained classifer for detecting hand
const char *cascade_name="hand.xml";
// Create memory for calculations
static CvMemStorage* storage = 0;
// Create a new Haar classifier
static CvHaarClassifierCascade* cascade = 0;
// Sets the scale with which the rectangle is drawn with
int scale = 1;
// Create two points to represent the hand locations
CvPoint pt1, pt2;
// Looping variable
int i;
// Load the HaarClassifierCascade
cascade = (CvHaarClassifierCascade*)cvLoad( cascade_name, 0, 0, 0 );
// Check whether the cascade has loaded successfully. Else report and error and quit
if( !cascade )
{
fprintf( stderr, "ERROR: Could not load classifier cascade\n" );
return;
} // Allocate the memory storage
storage = cvCreateMemStorage(0); // Create a new named window with title: result
cvNamedWindow( "result", 1 ); // Clear the memory storage which was used before
cvClearMemStorage( storage ); // Find whether the cascade is loaded, to find the hands. If yes, then:
if( cascade )
{
// There can be more than one hand in an image. So create a growable sequence of hands.
// Detect the objects and store them in the sequence
CvSeq* hands = cvHaarDetectObjects( img, cascade, storage,
1.1, 2, CV_HAAR_DO_CANNY_PRUNING,
cvSize(40, 40) ); // Loop the number of hands found.
for( i = 0; i < (hands ? hands->total : 0); i++ )
{
// Create a new rectangle for drawing the hand
CvRect* r = (CvRect*)cvGetSeqElem( hands, i ); // Find the dimensions of the hand,and scale it if necessary
pt1.x = r->x*scale;
pt2.x = (r->x+r->width)*scale;
pt1.y = r->y*scale;
pt2.y = (r->y+r->height)*scale; // Draw the rectangle in the input image
cvRectangle( img, pt1, pt2, CV_RGB(230,20,232), 3, 8, 0 );
}
} // Show the image in the window named "result"
cvShowImage( "result", img );
cvWaitKey (30);
}
注意该特征文件包含了手掌半握式的手势效果较好：
多谢大家，这么长时间的阅读和浏览，小弟做的很粗糙还有一些地方自已也没有弄明白，希望各位大神批评指教！
opencv 手势识别的更多相关文章
- [手势识别] CNN + OpenCV 手势识别记录
这几天没事,想着再学点一些视觉识别方向的东西,因为之前做了验证码识别,有了机器学习的信心,因此这次打算做个手势识别,参考了很多网上的图像处理方式,中间也遇到了很多问题,最终算是做出来了吧. 1.训练集 ...
- OpenCV——手势识别
使用ANN神经网络训练数据后进行手势识别. #include "header.h" int main() { ; //训练每类图片数量 ; //训练类数3:石头剪刀布 ; ; st ...
- 手势识别控制pygame精灵
步骤: 编写简易pygame精灵游戏(只实现键盘上下左右控制) 解决opencv手势识别核心问题 上述2部分对接上 pygame部分我们只加载个背景,然后里面放1只乌龟精灵,用键盘的上下左右键来控制, ...
- OpenCV——写手势识别碰到的各种错误
最近写opencv的手势识别时,真的碰到了好多好多程序运行时的错误. 比如: 位置冲突: findcontours函数的使用错误: 各种符号的加载错误: 这种符号加载的错误,一般用网上各种方法就可以解 ...
- 模式识别开发之项目---基于opencv的手势识别
我使用OpenCV2.4.4的windows版本+Qt4.8.3+VS2010的编译器做了一个手势识别的小程序. 本程序主要使到了Opencv的特征训练库和最基本的图像处理的知识,包括肤色检测等等. ...
- Python C++ OpenCV TensorFlow手势识别(1-10) 毕设 定制开发
Python C++ OpenCV TensorFlow手势识别(1-10) 毕设 支持定制开发 (MFC,QT, PyQt5界面,视频摄像头识别) QQ: 3252314061 效果如下:
- 开源项目（4-2）手势识别-Keras/Theano/OpenCV实现的CNN手势识别
https://github.com/asingh33/CNNGestureRecognizer 我提供了两种捕获模式: 二进制模式:在这里我首先将图像转换为灰度,然后应用高斯模糊效果和自适应阈值滤波 ...
- [OpenCV] Samples 13: opencv_version
cv::CommandLineParser的使用. I suppose CommandLineParser::has("something") should be true whe ...
- opencv 61篇
(一)--安装配置.第一个程序 标签: imagebuildincludeinputpathcmd 2011-10-21 16:16 41132人阅读 评论(50) 收藏 举报 分类: OpenCV ...
随机推荐
- AngularJS 拦截器和应用例子(转)
$httpAngularJS 的 $http 服务允许我们通过发送 HTTP 请求方式与后台进行通信.在某些情况下,我们希望可以俘获所有的请求,并且在将其发送到服务端之前进行操作.还有一些情况是,我们 ...
- HDU5731 Solid Dominoes Tilings 状压dp+状压容斥
题意:给定n,m的矩阵,就是求稳定的骨牌完美覆盖,也就是相邻的两行或者两列都至少有一个骨牌 分析:第一步: 如果是单单求骨牌完美覆盖,请先去学基础的插头dp(其实也是基础的状压dp)骨牌覆盖 hiho ...
- duilib入门之贴图描述、类html文本描述、动态换肤、Dll插件、资源打包
转载自duilib入门文档 贴图描述: Duilib的表现力丰富很大程度上得益于贴图描述的简单强大.Duilib的贴图描述分为简单模式和复杂模式两种. 简单模式使用文件名做为贴图描述内容,在这种方式下 ...
- springmvc里面的中文乱码问题
如果是以get方法提交的表单,则可以在comcat服务器的server.xml文件里面设置 <Connector connectionTimeout="20000" port ...
- flash recovery area配置
检查数据库是否开启闪回: SQL> select flashback_on from v$database; FLASHBACK_ON ----------------------------- ...
- .net高级技术（class0515)
本次课程中讲的有的东西都是根据初学者的认知规律进行了调整,并不是严谨的,比如很多地方在多AppDomain条件下很多说法就不对了,但是说严谨了大家就晕了,因此继续不严谨的讲吧. 很多面试题都在这阶段的 ...
- C#冒泡排序详解
今天写一简单的冒泡排序,带有详细的中文注释,新手一定要看看! 因为这是找工作面试时经常 笔试 要考的题目. using System; using System.Collections.Generic ...
- SVM原理简介
本文只是简单介绍一下SVM的理论框架,想要详细了解当中细节问题处理可以参看后续章节或者网上各种详细资料.推荐Andrew Ng的斯坦福大学机器学习课程. 年代中期发展起来的基于统计学习理论的一种机器学 ...
- javadoc注释规范
javadoc做注释 一. Java 文档 // 注释一行 /* ...... */ 注释若干行 /** ...... */ 注释若干行,并写入 javadoc 文档 通常这种注释的多行写法如下: / ...
- hdfs[命令] dfsadmin
Usage: java DFSAdminNote: Administrative commands can only be run as the HDFS superuser. [-report] [ ...