/* Standard Deviation */
IplImage* motionDetection::getStandardDeviationFrame(void) {
// Initialize
cvZero(mSum);
for (int i = 0; i < mFrameNumber; ++i) {
// frame[i] <= | frame[i] - Background Model |
cvAbsDiff(mpFrame[i], m_imgBackgroundModel, mTmp8U);
// uchar->float
cvConvert(mTmp8U, mTmp);
// mTmp = mTmp * mTmp
cvPow(mTmp, mTmp, 2.0);
// add mSum += mTmp
cvAdd(mSum, mTmp, mSum);
}
// variance: mTmp <= mSum / (mFrameNumber-1)
for (int i = 0; i < mSize.height; ++i) {
for (int j = 0; j < mSize.width; ++j) {
((float*)(mTmp->imageData + i*mTmp->widthStep))[j] = ((float*)(mSum->imageData + i*mSum->widthStep))[j] / (mFrameNumber - 1);
}
}
// standard deviation
cvPow(mTmp, mTmp, 0.5);
// float->uchar
cvConvert(mTmp, m_imgStandardDeviation);
return m_imgStandardDeviation;
}
void detect_object(IplImage *image, IplImage *pBkImg, IplImage *pFrImg, CvMat *pFrameMat, CvMat *pBkMat, CvMat *pFrMat,int thre_limit)
{
nFrmNum++;
cvCvtColor(image, pFrImg, CV_BGR2GRAY);
cvConvert(pFrImg, pFrameMat);
//高斯滤波
cvSmooth(pFrameMat, pFrameMat, CV_GAUSSIAN, 3, 0, 0);
//当前帧减去背景图像并取绝对值
cvAbsDiff(pFrameMat, pBkMat, pFrMat);
//二值化前景图像
cvThreshold(pFrMat, pFrImg,thre_limit, 255.0, CV_THRESH_BINARY);
/*形态学滤波*/
//IplConvKernel* element = cvCreateStructuringElementEx(2, 2, 0, 0, CV_SHAPE_RECT);
//cvErode(pFrImg, pFrImg,element, 1); // 腐蚀
//delete element;
//element = cvCreateStructuringElementEx(2, 2, 1, 1, CV_SHAPE_RECT);
//cvDilate(pFrImg, pFrImg, element, 1); //膨胀
//delete element;
cvErode(pFrImg, pFrImg,0, 1); // 腐蚀
cvDilate(pFrImg, pFrImg,0, 1); //膨胀
//滑动平均更新背景（求平均）
cvRunningAvg(pFrameMat, pBkMat, 0.004, 0);
//将背景矩阵转化为图像格式，用以显示
cvConvert(pBkMat, pBkImg);
cvShowImage("background", pFrImg);
// cvShowImage("background", pBkImg);
}
void MainWindow::BackgroundDiff()
{
ui->alpha_slider->setEnabled(true);
cvReleaseCapture(&pCapture);
pCapture=cvCaptureFromCAM(0);
// IplImage* pFrame=NULL;
nFrameNum=0;
while(pFrame = cvQueryFrame( pCapture ))
{
nFrameNum++;
//如果是第一帧，需要申请内存，并初始化
if(nFrameNum == 1)
{
pBkImg = cvCreateImage(cvSize(pFrame->width, pFrame->height),IPL_DEPTH_8U,1);
pFrImg = cvCreateImage(cvSize(pFrame->width, pFrame->height), IPL_DEPTH_8U,1);
pBkMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
pFrMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
pFrameMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
//转化成单通道图像再处理
cvCvtColor(pFrame, pBkImg, CV_BGR2GRAY);
cvCvtColor(pFrame, pFrImg, CV_BGR2GRAY);
cvConvert(pFrImg, pFrameMat);
cvConvert(pFrImg, pFrMat);
cvConvert(pFrImg, pBkMat);
}
else
{
cvCvtColor(pFrame, pFrImg, CV_BGR2GRAY);
cvConvert(pFrImg, pFrameMat);
//先做高斯滤波，以平滑图像
cvSmooth(pFrameMat, pFrameMat, CV_GAUSSIAN, 3, 0, 0);
//当前帧跟背景图相减
cvAbsDiff(pFrameMat, pBkMat, pFrMat);
//二值化前景图
cvDilate(pFrMat,pFrMat);
cvErode(pFrMat,pFrMat);
cvThreshold(pFrMat, pFrImg, lowThreshold, 255.0, CV_THRESH_BINARY);
//更新背景
cvRunningAvg(pFrameMat, pBkMat, alpha,0);
//将背景转化为图像格式，用以显示
cvConvert(pBkMat, pBkImg);
pFrame->origin = IPL_ORIGIN_BL;
pFrImg->origin = IPL_ORIGIN_BL;
pBkImg->origin = IPL_ORIGIN_BL;
}
if(27==cvWaitKey(33))
break;
MainWindow::Display(pFrame,pBkImg,pFrImg);
}
}
CVAPI(void) cvParticleStateConfig( CvParticle * p,
const CvSize imsize, CvBox2D std )
{
// config dynamics model
CvMat * dynamicsmat =
cvCreateMat( p->num_states, p->num_states, CV_64F );
cvSetIdentity(dynamicsmat, cvRealScalar(1));
// config random noise standard deviation
// CvRNG rng = cvRNG( time( NULL ) );
double stdarr[] = {
// std.x, std.y,
// std.width, std.height,
std.center.x, std.center.y,
std.size.width, std.size.height,
std.angle
};
CvMat stdmat = cvMat( p->num_states, 1, CV_64FC1, stdarr );
// config minimum and maximum values of states
// lowerbound, upperbound, circular flag (useful for degree)
// lowerbound == upperbound to express no bounding
// const CvSize tsize = cvSize(24.*(imsize.width /160.),
// 30.*(imsize.height/120.));
const CvSize tsize = cvSize(30.*(imsize.width /160.),
36.*(imsize.height/120.));
const float space = // space for rotation of box
(10.*(imsize.width /160.)+ // additional bound
MAX(tsize.width, tsize.height))*0.1415+ // rotated bound
1.0; // additional pixel
double boundarr[] = {
space, imsize.width - (tsize.width +1.) - space, false, // position.x
space, imsize.height -(tsize.height+1.) - space, false, // position.y
12.*(imsize.width/160.), tsize.width , false, // winsize.x
15.*(imsize.height/160.), tsize.height, false, // winsize.y
// 0, 360, true // dtheta + circular flag
-30, 30, false // dtheta
};
CvMat boundmat = cvMat( p->num_states, 3, CV_64FC1, boundarr );
//cvParticleSetDynamics( p, dynamicsmat );
cvConvert(dynamicsmat, p->dynamics);
// cvParticleSetNoise( p, rng, &stdmat );
p->rng = cvRNG( time( NULL ) );
cvConvert(&stdmat,p->std);
// cvParticleSetBound( p, &boundmat );
cvConvert(&boundmat,p->bound);
cvReleaseMat(&dynamicsmat);
}
void ImgAli_common::setTemplate(char *name)
{
IplImage *tmp=cvLoadImage(name,0);
Template=cvCreateMat(tmp->height,tmp->width,CV_64FC1);
cvConvert(tmp,Template);
//cvNamedWindow("1");
//
//cvDrawLine(Template,cvPoint(0,0),cvPoint(50,50),cvScalar(255,0,0));
//cvShowImage("1",Template);
//cvWaitKey();
width=Template->cols;
height=Template->rows;
/*SD_ic=new double**[width];
for (int i=0;i<width;i++)
{
SD_ic[i]=new double *[height];
for (int j=0;j<height;j++)
{
SD_ic[i][j]=new double[shapeDim];
}
}*/
//
setParameters();
}
// Set Pixel Data - Arrays
//--------------------------------------------------------------------------------
void ofxCvFloatImage::operator = ( ofxCvGrayscaleImage& mom ) {
if( mom.width == width && mom.height == height ) {
cvConvert( mom.getCvImage(), cvImage );
} else {
cout << "error in =, images are different sizes" << endl;
}
}
//在图像srcImg上根据contour轮廓画上最小外接椭圆
CvBox2D DrawMinAreaEllipse(IplImage *srcImg,CvSeq *contour,CvScalar color/*=CV_RGB(255,0,0)*/)
{
int count = contour->total; // This is number point in contour
CvPoint center;
CvSize size;
CvBox2D box;
if( count < 6 )
return box;
CvMat* points_f = cvCreateMat( 1, count, CV_32FC2 );
CvMat points_i = cvMat( 1, count, CV_32SC2, points_f->data.ptr );
cvCvtSeqToArray( contour, points_f->data.ptr, CV_WHOLE_SEQ );
cvConvert( &points_i, points_f );
// 椭圆拟合
box = cvFitEllipse2( points_f );
cout<<"拟合的椭圆参数：angle="<<box.angle<<",center=("<<box.center.x<<","
<<box.center.y<<")"<<",size(w,h)=("<<box.size.width<<","<<box.size.height<<")"<<endl;
// 获得椭圆参数
center = cvPointFrom32f(box.center);
size.width = cvRound(box.size.width*0.5)+1;
size.height = cvRound(box.size.height*0.5)+1;
// 画椭圆
cvEllipse(srcImg, center, size,
-box.angle, 0, 360, color, 1, CV_AA, 0);
cvReleaseMat(&points_f);
return box;
}
int main (int argv, char** argc[])
{
int ncell = 0, prev_ncontour = 0, same_count = 0;
////while (!worker->CancellationPending) {
////worker->ReportProgress(50, String::Format(rm->GetString("Progress_Analyze_FoundNCell"), title, ncell));
cvConvert(input_morph, tmp8UC1);
cvClearMemStorage(storage);
int ncontour = cvFindContours(tmp8UC1, storage, &first_con, sizeof(CvContour), CV_RETR_EXTERNAL);
if (ncontour == 0)
break; // finish extract cell
if (ncontour == prev_ncontour) {
cvErode(input_morph, input_morph);
same_count++;
} else
same_count = 0;
prev_ncontour = ncontour;
cur = first_con;
while (cur != nullptr) {
double area = fabs(cvContourArea(cur));
if ((area < 3000.0) || (same_count > 10)) {
int npts = cur->total;
CvPoint *p = new CvPoint[npts];
cvCvtSeqToArray(cur, p);
cvFillPoly(out_single, &p, &npts, 1, cvScalar(255.0)); // move to single
cvFillPoly(input_morph, &p, &npts, 1, cvScalar(0.0)); // remove from input
delete[] p;
ncell++;
}
cur = cur->h_next;
}
////}
}
// Define trackbar callback functon. This function find contours,
// draw it and approximate it by ellipses.
void process_image(int h)
{
CvMemStorage* storage;
CvSeq* contour;
// Create dynamic structure and sequence.
storage = cvCreateMemStorage(0);
contour = cvCreateSeq(CV_SEQ_ELTYPE_POINT, sizeof(CvSeq), sizeof(CvPoint) , storage);
// Threshold the source image. This needful for cvFindContours().
cvThreshold( image03, image02, slider_pos, 255, CV_THRESH_BINARY );
// Find all contours.
cvFindContours( image02, storage, &contour, sizeof(CvContour),
CV_RETR_LIST, CV_CHAIN_APPROX_NONE, cvPoint(0,0));
// Clear images. IPL use.
cvZero(image02);
cvZero(image04);
// This cycle draw all contours and approximate it by ellipses.
for(;contour;contour = contour->h_next)
{
int count = contour->total; // This is number point in contour
CvPoint center;
CvSize size;
CvBox2D box;
// Number point must be more than or equal to 6 (for cvFitEllipse_32f).
if( count < 6 )
continue;
CvMat* points_f = cvCreateMat( 1, count, CV_32FC2 );
CvMat points_i = cvMat( 1, count, CV_32SC2, points_f->data.ptr );
cvCvtSeqToArray( contour, points_f->data.ptr, CV_WHOLE_SEQ );
cvConvert( &points_i, points_f );
// Fits ellipse to current contour.
box = cvFitEllipse2( points_f );
// Draw current contour.
cvDrawContours(image04,contour,CV_RGB(255,255,255),CV_RGB(255,255,255),0,1,8,cvPoint(0,0));
// Convert ellipse data from float to integer representation.
center = cvPointFrom32f(box.center);
size.width = cvRound(box.size.width*0.5);
size.height = cvRound(box.size.height*0.5);
// Draw ellipse.
cvEllipse(image04, center, size,
-box.angle, 0, 360,
CV_RGB(0,0,255), 1, CV_AA, 0);
cvReleaseMat(&points_f);
}
// Show image. HighGUI use.
cvShowImage( "Result", image04 );
}
//--------------------------------------------------------------------------------
void ofxCvColorImage::operator = ( ofxCvFloatImage& mom ) {
if( mom.width == width && mom.height == height ) {
//cvCopy(mom.getCvImage(), cvImage, 0);
//cvConvertScale( mom.getCvImage(), cvImage, 1, 0 );
cvConvert( mom.getCvImage(), cvImage ); // same as above but optimized
} else {
cout << "error in =, images are different sizes" << endl;
}
}
CvMat * icvReadSVHNImages(char * filename, CvMat * response)
{
const int max_samples = response->rows;
const int max_strlen = 64*64;
CvMat * data = cvCreateMat(max_samples,max_strlen,CV_32F);
char datastr[1000];
int ii; CvMat hdr;
CvMat * sample = cvCreateMat(64,64,CV_32F);
CvMat * warp = cvCreateMat(2,3,CV_32F);
for (ii=0;;ii++){
sprintf(datastr,filename,ii+1);
IplImage * img = cvLoadImage(datastr, CV_LOAD_IMAGE_GRAYSCALE);
if (!img || ii==max_samples){break;}
CvMat * mat = cvCreateMat(img->height,img->width,CV_32F);
cvConvert(img,mat);
int nimages = CV_MAT_ELEM(*response,int,ii,0);
float ww = CV_MAT_ELEM(*response,int,ii,3);
float hh = CV_MAT_ELEM(*response,int,ii,4);
float xx = CV_MAT_ELEM(*response,int,ii,1)-ww*.5f;
float yy = CV_MAT_ELEM(*response,int,ii,2)-hh*.5f;
for (int jj=1;jj<nimages;jj++){
float www = CV_MAT_ELEM(*response,int,ii,1+4*jj+2);
float hhh = CV_MAT_ELEM(*response,int,ii,1+4*jj+3);
float xxx = CV_MAT_ELEM(*response,int,ii,1+4*jj+0)-www*.5f;
float yyy = CV_MAT_ELEM(*response,int,ii,1+4*jj+1)-hhh*.5f;
xx = MIN(xx,xxx); yy = MIN(yy,yyy);
ww = MAX(xxx+www-xx,ww);
hh = MAX(yyy+hhh-yy,hh);
}
xx+=ww*.5f;yy+=hh*.5f; float ss = MAX(ww,hh);
cvZero(warp);cvZero(sample);
warp->data.fl[2]=xx-ss*.5f-ss*.15f;
warp->data.fl[5]=yy-ss*.5f-ss*.15f;
warp->data.fl[0]=warp->data.fl[4]=ss*1.3f/64.f; //
icvWarp(mat,sample,warp);
CvScalar avg,sdv;
cvAvgSdv(sample,&avg,&sdv);
cvSubS(sample,avg,sample);
cvScale(sample,sample,1.f/sdv.val[0]);
#if 0 // debug
cvPrintf(stderr,"%f,",warp);
cvAvgSdv(sample,&avg,&sdv); // re-calculate
fprintf(stderr,"avg: %f, sdv: %f\n--\n",avg.val[0],sdv.val[0]);
CV_SHOW(sample); // CV_SHOW(mat);
#endif
memcpy(data->data.fl+max_strlen*ii,sample->data.ptr,max_strlen*sizeof(float));
cvReleaseImage(&img);
cvReleaseMat(&mat);
}
data->rows = ii;
cvReleaseMat(&sample);
cvReleaseMat(&warp);
assert(CV_MAT_TYPE(data->type)==CV_32F);
return data;
}
//--------------------------------------------------------------------------------
void setfilter::operator = ( const ofxCvFloatImage& _mom ) {
//-- for using a floatimage --//
ofxCvFloatImage& mom = const_cast<ofxCvFloatImage&>(_mom);
if( matchingROI(getROI(), mom.getROI()) ) {
cvConvert( mom.getCvImage(), cvImage );
flagImageChanged();
}
else { ofLog(OF_LOG_ERROR, "in =, ROI mismatch"); }
}
//--------------------------------------------------------------------------------
void CPUImageFilter::operator = ( const ofxCvFloatImage& _mom ) {
// cast non-const, no worries, we will reverse any chages
ofxCvFloatImage& mom = const_cast<ofxCvFloatImage&>(_mom);
if( pushSetBothToTheirIntersectionROI(*this,mom) ) {
//cvConvertScale( mom.getCvImage(), cvImage, 1.0f, 0);
cvConvert( mom.getCvImage(), cvImage );
popROI(); //restore prevoius ROI
mom.popROI(); //restore prevoius ROI
flagImageChanged();
} else {
ofLog(OF_LOG_ERROR, "in =, ROI mismatch");
}
}
UINT WINAPI
//DWORD WINAPI
#elif defined(POSIX_SYS)
// using pthread
void *
#endif
ChessRecognition::LineSearchThread(
#if defined(WINDOWS_SYS)
LPVOID
#elif defined(POSIX_SYS)
void *
#endif
Param) {
// mode 2를 통하여 chessboard recognition을 수행하기 위한 thread 함수.
ChessRecognition *_TChessRecognition = (ChessRecognition *)Param;
_TChessRecognition->_LineSearchBased = new LineSearchBased();
IplImage *_TGray = cvCreateImage(cvSize(_TChessRecognition->_Width, _TChessRecognition->_Height), IPL_DEPTH_8U, 1);
while (_TChessRecognition->_EnableThread != false) {
// 연산에 필요한 이미지를 main으로부터 복사해 옴.
_TChessRecognition->_ChessBoardDetectionInternalImageProtectMutex.lock();
if (_TChessRecognition->_ChessBoardDetectionInternalImage->nChannels != 1)
cvConvert(_TChessRecognition->_ChessBoardDetectionInternalImage, _TGray);
else
cvCopy(_TChessRecognition->_ChessBoardDetectionInternalImage, _TGray);
_TChessRecognition->_ChessBoardDetectionInternalImageProtectMutex.unlock();
// 복사한 이미지를 실제 연산에 사용.
_TChessRecognition->_ChessBoardDetectionInternalImageProtectMutex.lock();
_TChessRecognition->_LineSearchBased->ChessLineSearchProcess(_TGray, &_TChessRecognition->_CP);
_TChessRecognition->_ChessBoardDetectionInternalImageProtectMutex.unlock();
Sleep(2);
}
// 이미지 할당 해제.
cvReleaseImage(&_TGray);
//cvReleaseImage(&img_process);
delete _TChessRecognition->_LineSearchBased;
#if defined(WINDOWS_SYS)
_endthread();
#elif defined(POSIX_SYS)
#endif
_TChessRecognition->_EndThread = true;
return 0;
}
/*
Calculates a least-squares planar homography from point correspondeces.
@param pts array of points
@param mpts array of corresponding points; each pts[i], i=0..n-1, corresponds to mpts[i]
@param n number of points in both pts and mpts; must be at least 4
@return Returns the 3 x 3 least-squares planar homography matrix that
transforms points in pts to their corresponding points in mpts or NULL if
fewer than 4 correspondences were provided
*/
CvMat* lsq_homog( CvPoint2D64f* pts, CvPoint2D64f* mpts, int n )
{
CvMat* H, * A, * B, X;
double x[9];//数组x中的元素就是变换矩阵H中的值
int i;
//输入点对个数不够4
if( n < 4 )
{
fprintf( stderr, "Warning: too few points in lsq_homog(), %s line %d\n",
__FILE__, __LINE__ );
return NULL;
}
//将变换矩阵H展开到一个8维列向量X中，使得AX=B，这样只需一次解线性方程组即可求出X，然后再根据X恢复H
/* set up matrices so we can unstack homography into X; AX = B */
A = cvCreateMat( 2*n, 8, CV_64FC1 );//创建2n*8的矩阵，一般是8*8
B = cvCreateMat( 2*n, 1, CV_64FC1 );//创建2n*1的矩阵，一般是8*1
X = cvMat( 8, 1, CV_64FC1, x );//创建8*1的矩阵，指定数据为x
H = cvCreateMat(3, 3, CV_64FC1);//创建3*3的矩阵
cvZero( A );//将A清零
//由于是展开计算，需要根据原来的矩阵计算法则重新分配矩阵A和B的值的排列
for( i = 0; i < n; i++ )
{
cvmSet( A, i, 0, pts[i].x );//设置矩阵A的i行0列的值为pts[i].x
cvmSet( A, i+n, 3, pts[i].x );
cvmSet( A, i, 1, pts[i].y );
cvmSet( A, i+n, 4, pts[i].y );
cvmSet( A, i, 2, 1.0 );
cvmSet( A, i+n, 5, 1.0 );
cvmSet( A, i, 6, -pts[i].x * mpts[i].x );
cvmSet( A, i, 7, -pts[i].y * mpts[i].x );
cvmSet( A, i+n, 6, -pts[i].x * mpts[i].y );
cvmSet( A, i+n, 7, -pts[i].y * mpts[i].y );
cvmSet( B, i, 0, mpts[i].x );
cvmSet( B, i+n, 0, mpts[i].y );
}
//调用OpenCV函数，解线性方程组
cvSolve( A, B, &X, CV_SVD );//求X，使得AX=B
x[8] = 1.0;//变换矩阵的[3][3]位置的值为固定值1
X = cvMat( 3, 3, CV_64FC1, x );
cvConvert( &X, H );//将数组转换为矩阵
cvReleaseMat( &A );
cvReleaseMat( &B );
return H;
}
int main() {
IplImage *frame, *accumulator=NULL;
cv_dc1394_init();
while (cv_sdl_process_events()) {
frame = cv_dc1394_capture_yuv(1)[0];
if (!accumulator)
accumulator = cvCreateImage(cvGetSize(frame), 32, 3);
cvRunningAvg(frame, accumulator, 0.1);
cvConvert(accumulator, frame);
cv_sdl_show_yuv_image(frame);
}
}
/*
// Getting feature pyramid
//
// API
// int getFeaturePyramid(IplImage * image, const filterObject **all_F,
const int n_f,
const int lambda, const int k,
const int startX, const int startY,
const int W, const int H, featurePyramid **maps);
// INPUT
// image - image
// OUTPUT
// maps - feature maps for all levels
// RESULT
// Error status
*/
int getFeaturePyramid(IplImage * image, CvLSVMFeaturePyramid **maps)
{
IplImage *imgResize;
float step;
int numStep;
int maxNumCells;
int W, H;
if(image->depth == IPL_DEPTH_32F)
{
imgResize = image;
}
else
{
imgResize = cvCreateImage(cvSize(image->width , image->height) ,
IPL_DEPTH_32F , 3);
cvConvert(image, imgResize);
}
W = imgResize->width;
H = imgResize->height;
step = powf(2.0f, 1.0f / ((float)LAMBDA));
maxNumCells = W / SIDE_LENGTH;
if( maxNumCells > H / SIDE_LENGTH )
{
maxNumCells = H / SIDE_LENGTH;
}
numStep = (int)(logf((float) maxNumCells / (5.0f)) / logf( step )) + 1;
allocFeaturePyramidObject(maps, numStep + LAMBDA);
getPathOfFeaturePyramid(imgResize, step , LAMBDA, 0,
SIDE_LENGTH / 2, maps);
getPathOfFeaturePyramid(imgResize, step, numStep, LAMBDA,
SIDE_LENGTH , maps);
if(image->depth != IPL_DEPTH_32F)
{
cvReleaseImage(&imgResize);
}
return LATENT_SVM_OK;
}
/*
Calculates a planar homography from point correspondeces using the direct
linear transform. Intended for use as a ransac_xform_fn.
@param pts array of points
@param mpts array of corresponding points; each pts[i], i=0..n-1,
corresponds to mpts[i]
@param n number of points in both pts and mpts; must be at least 4
@return Returns the 3x3 planar homography matrix that transforms points
in pts to their corresponding points in mpts or NULL if fewer than 4
correspondences were provided
*/
CvMat* dlt_homog( CvPoint2D64f* pts, CvPoint2D64f* mpts, int n )
{
CvMat* H, * A, * VT, * D, h, v9;
double _h[9];
int i;
if( n < 4 )
return NULL;
/* set up matrices so we can unstack homography into h; Ah = 0 */
A = cvCreateMat( 2*n, 9, CV_64FC1 );
cvZero( A );
for( i = 0; i < n; i++ )
{
cvmSet( A, 2*i, 3, -pts[i].x );
cvmSet( A, 2*i, 4, -pts[i].y );
cvmSet( A, 2*i, 5, -1.0 );
cvmSet( A, 2*i, 6, mpts[i].y * pts[i].x );
cvmSet( A, 2*i, 7, mpts[i].y * pts[i].y );
cvmSet( A, 2*i, 8, mpts[i].y );
cvmSet( A, 2*i+1, 0, pts[i].x );
cvmSet( A, 2*i+1, 1, pts[i].y );
cvmSet( A, 2*i+1, 2, 1.0 );
cvmSet( A, 2*i+1, 6, -mpts[i].x * pts[i].x );
cvmSet( A, 2*i+1, 7, -mpts[i].x * pts[i].y );
cvmSet( A, 2*i+1, 8, -mpts[i].x );
}
D = cvCreateMat( 9, 9, CV_64FC1 );
VT = cvCreateMat( 9, 9, CV_64FC1 );
cvSVD( A, D, NULL, VT, CV_SVD_MODIFY_A + CV_SVD_V_T );
v9 = cvMat( 1, 9, CV_64FC1, NULL );
cvGetRow( VT, &v9, 8 );
h = cvMat( 1, 9, CV_64FC1, _h );
cvCopy( &v9, &h, NULL );
h = cvMat( 3, 3, CV_64FC1, _h );
H = cvCreateMat( 3, 3, CV_64FC1 );
cvConvert( &h, H );
cvReleaseMat( &A );
cvReleaseMat( &D );
cvReleaseMat( &VT );
return H;
}
static
void ShowResult()
{
switch (g_effectId)
{
case EFFECT01: cvConvert(g_pEffect01->GetResult(), g_pImage); break;
case EFFECT02: cvConvert(g_pEffect02->GetResult(), g_pImage); break;
case EFFECT03: cvConvert(g_pEffect03->GetResult(), g_pImage); break;
case EFFECT05: cvConvert(g_pEffect05->GetResult(), g_pImage); break;
case EFFECT09: cvConvert(g_pEffect09->GetResult(), g_pImage); break;
case EFFECT10: cvConvert(g_pEffect10->GetResult(), g_pImage); break;
case EFFECT15: cvConvert(g_pEffect15->GetResult(), g_pImage); break;
case EFFECT19: cvConvert(g_pEffect19->GetResult(), g_pImage); break;
default:
cvZero(g_pImage);
fprintf(stderr, "Err: Not available!\n");
break;
}
cvShowImage("Lens Flare", g_pImage);
}
/*
Calculates a least-squares planar homography from point correspondeces.
@param pts array of points
@param mpts array of corresponding points; each pts[i], i=1..n, corresponds
to mpts[i]
@param n number of points in both pts and mpts; must be at least 4
@return Returns the 3 x 3 least-squares planar homography matrix that
transforms points in pts to their corresponding points in mpts or NULL if
fewer than 4 correspondences were provided
*/
CvMat* lsq_homog( CvPoint2D64f* pts, CvPoint2D64f* mpts, int n )
{
CvMat* H, * A, * B, X;
double x[9];
int i;
if( n < 4 )
{
fprintf( stderr, "Warning: too few points in lsq_homog(), %s line %d\n",
__FILE__, __LINE__ );
return NULL;
}
/* set up matrices so we can unstack homography into X; AX = B */
A = cvCreateMat( 2*n, 8, CV_64FC1 );
B = cvCreateMat( 2*n, 1, CV_64FC1 );
X = cvMat( 8, 1, CV_64FC1, x );
H = cvCreateMat(3, 3, CV_64FC1);
cvZero( A );
for( i = 0; i < n; i++ )
{
cvmSet( A, i, 0, pts[i].x );
cvmSet( A, i+n, 3, pts[i].x );
cvmSet( A, i, 1, pts[i].y );
cvmSet( A, i+n, 4, pts[i].y );
cvmSet( A, i, 2, 1.0 );
cvmSet( A, i+n, 5, 1.0 );
cvmSet( A, i, 6, -pts[i].x * mpts[i].x );
cvmSet( A, i, 7, -pts[i].y * mpts[i].x );
cvmSet( A, i+n, 6, -pts[i].x * mpts[i].y );
cvmSet( A, i+n, 7, -pts[i].y * mpts[i].y );
cvmSet( B, i, 0, mpts[i].x );
cvmSet( B, i+n, 0, mpts[i].y );
}
cvSolve( A, B, &X, CV_SVD );
x[8] = 1.0;
X = cvMat( 3, 3, CV_64FC1, x );
cvConvert( &X, H );
cvReleaseMat( &A );
cvReleaseMat( &B );
return H;
}
CvMat* myCvGetRotationMatrix(CvPoint2D32f center, CvMat* matrix)
{
double m[2][3];
CvMat M =cvMat(2,3,CV_64FC1, m);
double alpha,beta;
double angle=CV_PI;
alpha= cos(angle);
beta= sin(angle);
m[0][0]=alpha;
m[0][1]=beta;
m[0][2]=(1-alpha)*center.x - beta*center.y;
m[1][0]=-beta;
m[1][1]=alpha;
m[1][2]=beta*center.x + (1-alpha)* center.y;
cvConvert(&M,matrix);
return matrix;
}
bool optimizeDepthMap()
{
cvErode(uImage,uImage,0,2); //Smoothen the User Map as well
cvDilate(uImage,uImage,0,2);
CvScalar depthMean=cvAvg(dImage,uImage); //Get teh Average Depth Value of the User Pixels
cvNot(uImage,uImage); //Invert the user pixels to paint the rest of the image with average user depth
//viewImage(dImage);
cvSet(dImage,depthMean,uImage);
IplImage* tempImage=cvCreateImage(dSize,IPL_DEPTH_8U,1);
cvConvertScale(dImage,tempImage,1.0/256);
cvSmooth(tempImage,tempImage,CV_GAUSSIAN,7);//Perform Gaussian Smoothing, depth map is optimized.
cvConvert(tempImage,dImage);
cvScale(dImage,dImage,256);
cvSet(dImage,cvScalar(0),uImage);
//viewImage(dImage);
//cvSmooth(dImage,dImage,CV_GAUSSIAN,gaussian_m,gaussian_n,gaussian_e);//Perform Gaussian Smoothing, depth map is optimized.
cvNot(uImage,uImage);
cvReleaseImage(&tempImage);
return true;
}
int diff2_main( int argc, char** argv )
{
//声明IplImage指针
IplImage* pFrame = NULL;
IplImage* pFrImg = NULL;
IplImage* pBkImg = NULL;
CvMat* pFrameMat = NULL;
CvMat* pFrMat = NULL;
CvMat* pBkMat = NULL;
CvCapture* pCapture = NULL;
int nFrmNum = 0;
//创建窗口
cvNamedWindow("video", 1);
cvNamedWindow("background",1);
cvNamedWindow("foreground",1);
//使窗口有序排列
cvMoveWindow("video", 30, 0);
cvMoveWindow("background", 360, 0);
cvMoveWindow("foreground", 690, 0);
if( !(pCapture = cvCaptureFromAVI("bike.avi")))
{
//pCapture = cvCaptureFromCAM(-1))
fprintf(stderr, "Can not open camera.\n");
return -2;
}
//逐帧读取视频
while(pFrame = cvQueryFrame( pCapture ))
{
nFrmNum++;
//如果是第一帧，需要申请内存，并初始化
if(nFrmNum == 1)
{
pBkImg = cvCreateImage(cvSize(pFrame->width, pFrame->height), IPL_DEPTH_8U,1);
pFrImg = cvCreateImage(cvSize(pFrame->width, pFrame->height), IPL_DEPTH_8U,1);
pBkMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
pFrMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
pFrameMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
//转化成单通道图像再处理
cvCvtColor(pFrame, pBkImg, CV_BGR2GRAY);
cvCvtColor(pFrame, pFrImg, CV_BGR2GRAY);
cvConvert(pFrImg, pFrameMat);
cvConvert(pFrImg, pFrMat);
cvConvert(pFrImg, pBkMat);
}
else
{
cvCvtColor(pFrame, pFrImg, CV_BGR2GRAY);
cvConvert(pFrImg, pFrameMat);
//高斯滤波先，以平滑图像
//cvSmooth(pFrameMat, pFrameMat, CV_GAUSSIAN, 3, 0, 0);
//当前帧跟背景图相减
cvAbsDiff(pFrameMat, pBkMat, pFrMat);
//二值化前景图
cvThreshold(pFrMat, pFrImg, 60, 255.0, CV_THRESH_BINARY);
//进行形态学滤波，去掉噪音
//cvErode(pFrImg, pFrImg, 0, 1);
//cvDilate(pFrImg, pFrImg, 0, 1);
//更新背景
cvRunningAvg(pFrameMat, pBkMat, 0.003, 0);
//将背景转化为图像格式，用以显示
cvConvert(pBkMat, pBkImg);
//显示图像
cvShowImage("video", pFrame);
cvShowImage("background", pBkImg);
cvShowImage("foreground", pFrImg);
//如果有按键事件，则跳出循环
//此等待也为cvShowImage函数提供时间完成显示
//等待时间可以根据CPU速度调整
if( cvWaitKey(20) >= 0 )
{
break;
}
}
}
cvWaitKey();
//销毁窗口
cvDestroyWindow("video");
cvDestroyWindow("background");
cvDestroyWindow("foreground");
//释放图像和矩阵
cvReleaseImage(&pFrImg);
cvReleaseImage(&pBkImg);
cvReleaseMat(&pFrameMat);
cvReleaseMat(&pFrMat);
cvReleaseMat(&pBkMat);
cvReleaseCapture(&pCapture);
return 0;
}
int getFeaturePyramid(IplImage * image, CvLSVMFeaturePyramid **maps,
const int bx, const int by)
{
IplImage *imgResize;
float step;
unsigned int numStep;
unsigned int maxNumCells;
unsigned int W, H;
if (image->depth == IPL_DEPTH_32F)
{
imgResize = image;
}
else
{
imgResize = cvCreateImage(cvSize(image->width, image->height),
IPL_DEPTH_32F, 3);
cvConvert(image, imgResize);
}
W = imgResize->width;
H = imgResize->height;
step = powf(2.0f, 1.0f / ((float) Lambda));
maxNumCells = W / Side_Length;
if (maxNumCells > H / Side_Length)
{
maxNumCells = H / Side_Length;
}
numStep = (int) (logf((float) maxNumCells / (5.0f)) / logf(step)) + 1;
allocFeaturePyramidObject(maps, numStep + Lambda);
#ifdef PROFILE
TickMeter tm;
tm.start();
cout << "(featurepyramid.cpp)getPathOfFeaturePyramid START " << endl;
#endif
uploadImageToGPU1D(imgResize);
getPathOfFeaturePyramidGPUStream(imgResize, step , Lambda, 0,
Side_Length / 2, bx, by, maps);
getPathOfFeaturePyramidGPUStream(imgResize, step, numStep, Lambda,
Side_Length , bx, by, maps);
cleanImageFromGPU1D();
#ifdef PROFILE
tm.stop();
cout << "(featurepyramid.cpp)getPathOfFeaturePyramid END time = "
<< tm.getTimeSec() << " sec" << endl;
#endif
if (image->depth != IPL_DEPTH_32F)
{
cvReleaseImage(&imgResize);
}
return LATENT_SVM_OK;
}
int int_cvtcolor(char *fname)
{
char * cvt_str_code = NULL; //input string code
int cvt_code; //opencv cvt_color code
IplImage * src_image = NULL;
IplImage * src_f32_image = NULL;
IplImage * dst_f32_image = NULL;
IplImage * dst_f64_image = NULL;
int m2, n2, l2;
//check the number of input and output parameters
CheckRhs(2,2);
CheckLhs(1,1);
GetRhsVar(2, "c", &m2, &n2, &l2);
cvt_str_code = cstk(l2);
//load the input image
src_image = Mat2IplImg(1);
// check if input image is correctly loaded
if(src_image == NULL)
{
sciprint("%s error: can't load the input image.\r\n", fname);
return -1;
}
//check the number of channels
if(src_image->nChannels != 3)
{
sciprint("%s error: The input image must be 3-channel image.\r\n", fname);
cvReleaseImage(&src_image);
return -1;
}
//this function is called
//from scilab macros rgb2hsv/hsv2rgb/...
//the input image can only be double float.
if(src_image->depth != IPL_DEPTH_64F)
{
sciprint("%s error: The input image must be double type.\r\n", fname);
cvReleaseImage(&src_image);
return -1;
}
if( strcmp(cvt_str_code, "rgb2hsv") == 0)
cvt_code = CV_BGR2HSV; //the channel order of IplImage in SIVP is BGR
else if( strcmp(cvt_str_code, "hsv2rgb") == 0)
cvt_code = CV_HSV2BGR;
else if( strcmp(cvt_str_code, "rgb2ycrcb") == 0)
cvt_code = CV_BGR2YCrCb;
else if( strcmp(cvt_str_code, "ycrcb2rgb") == 0)
cvt_code = CV_YCrCb2BGR;
else
{
sciprint("%s error: unsupport convertion code %s.\r\n", fname, cvt_str_code);
cvReleaseImage(&src_image);
return -1;
}
//create images needed
src_f32_image = cvCreateImage(cvGetSize(src_image),
IPL_DEPTH_32F,
3);
dst_f32_image = cvCreateImage(cvGetSize(src_image),
IPL_DEPTH_32F,
3);
dst_f64_image = cvCreateImage(cvGetSize(src_image),
IPL_DEPTH_64F,
3);
if(src_f32_image == NULL ||
dst_f32_image == NULL ||
dst_f64_image == NULL)
{
sciprint("Error: create image error.\r\n");
cvReleaseImage(&src_image);
cvReleaseImage(&src_f32_image);
cvReleaseImage(&dst_f32_image);
cvReleaseImage(&dst_f64_image);
return -1;
}
cvConvert(src_image, src_f32_image);
cvCvtColor(src_f32_image, dst_f32_image, cvt_code);
cvConvert(dst_f32_image, dst_f64_image);
IplImg2Mat(dst_f64_image, 3);
//send the result
LhsVar(1)=3;
cvReleaseImage(&src_image);
cvReleaseImage(&src_f32_image);
cvReleaseImage(&dst_f32_image);
cvReleaseImage(&dst_f64_image);
return 0;
}
int main()
{
/*读配置文件*/
FILE* fp;
char readin[100],file[100];
int value1=0, value2=0, value3=0,value4=0,value5=0;
fopen_s(&fp, "configure/configure.txt", "r");
fscanf_s(fp, "video: %s", readin, sizeof(readin));
sprintf(file, readin);
memset(readin, 0, sizeof(readin));
fgets(readin, sizeof(readin),fp);
fscanf_s(fp, "value1: %d", &value1, sizeof(int));
memset(readin, 0, sizeof(readin));
fgets(readin, sizeof(readin), fp);
fscanf_s(fp, "value2: %d", &value2, sizeof(int));
memset(readin, 0, sizeof(readin));
fgets(readin, sizeof(readin), fp);;
fscanf_s(fp, "value3: %d", &value3, sizeof(int));
memset(readin, 0, sizeof(readin));
fgets(readin, sizeof(readin), fp);
fscanf_s(fp, "value4: %d", &value4, sizeof(int));
memset(readin, 0, sizeof(readin));
fgets(readin, sizeof(readin), fp);
fscanf_s(fp, "value5: %d", &value5, sizeof(int));
printf("%s %d %d %d\n", file,value1,value2,value3);
/*声明IplImage指针*/
IplImage *image0 = NULL; //原始帧
IplImage *image = NULL; //当前帧
IplImage *image_pass = NULL; //上一帧
IplImage *res = NULL; //帧差
IplImage *res0 = NULL; //帧差
IplImage *pFrame = NULL;
IplImage *pFrImg = NULL;
IplImage *pBkImg = NULL;
/*声明CvMat指针*/
CvMat* pFrameMat = NULL;
CvMat* pFrMat = NULL;
CvMat* pBkMat = NULL;
CvMat* IndexMat = NULL;
/*声明caputer指针*/
CvCapture *capture = NULL;
capture = cvCaptureFromFile(file);
image0 = cvQueryFrame(capture);
nFrmNum++;
// 创建窗口
//cvNamedWindow("video", 1);
cvNamedWindow("background", 1);
cvNamedWindow("tracking", 1);
// 排列窗口
//cvMoveWindow("video", 30, 80);
cvMoveWindow("background",10, 100);
cvMoveWindow("tracking", 660, 100);
/*移动物体的队列*/
movingObject *curr_head = new movingObject();
movingObject *prev_head = new movingObject();
movingObject *p_obj = new movingObject();
movingObject *share_head = new movingObject();
curr_head->next = NULL;
curr_head->share_next = NULL;
prev_head->next = NULL;
prev_head->share_next = NULL;
share_head->next = NULL;
CvScalar color[7] = { { 0, 0, 0 }, { 0, 255, 0 }, {255, 0, 0 }, { 255, 201, 14 }, { 255, 0, 255 }, { 0, 166, 0 }, {121,255,121} };
image = cvCreateImage(cvSize(640,360), IPL_DEPTH_8U, 3);
cvResize(image0, image, CV_INTER_LINEAR);
int image_width = image->width;
int image_height = image->height;
image_pass = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3);
res = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1);
res0 = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3);
pBkImg = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1);
pFrImg = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1);
pFrame = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1);
pBkMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
pFrMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
pFrameMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1);
IndexMat = cvCreateMat(image_height, image_width, CV_32FC1);
cvCopy(image, image_pass, NULL);
/*背景*/
cvCvtColor(image, pBkImg, CV_BGR2GRAY);
cvConvert(pBkImg, pBkMat);
while (1)
{
image0 = cvQueryFrame(capture);
if (!image0) break;
cvResize(image0, image, CV_INTER_LINEAR);
/*高斯平滑*/
cvSmooth(image, image, CV_GAUSSIAN, 3, 0, 0);
/*运动目标检测*/
detect_object(image, pBkImg, pFrImg, pFrameMat, pBkMat, pFrMat,value1);
/*帧差法优化*/
frame_dif(image, image_pass, res,res0, pFrImg,pFrame,value2);
cvConvert(pFrame, pFrMat);
/*计算连通区域和质心*/
computeObject(pFrMat, image_width, image_height,IndexMat, curr_head,value3);
/*画出质心*/
p_obj = curr_head;
while (p_obj->next != NULL)
{
p_obj = p_obj->next;
cvRectangle(image, cvPoint(p_obj->x - 1, p_obj->y - 1), cvPoint(p_obj->x + 1, p_obj->y + 1), cvScalar(0, 0, 255), 2, 8, 0);
}
if (nFrmNum == 2)
{
movingObject* q = NULL;
computeObject(pFrMat, image_width, image_height, IndexMat, prev_head,value3);
}
/*画出跟踪框*/
if (nFrmNum > 2)
{
movingObject *q = NULL, *p = NULL;
movingObject *last=curr_head;
for (q = curr_head->next; q; )
{
int close = 0;
share_head->share_next = NULL;
for (p = prev_head->next; p; p = p->next)
{
int dist = cal_dist(p->x, p->y, q->x, q->y);
if (dist <= value5)
{
close++;
p->share_next = share_head->share_next;
share_head->share_next= p;
}
}
if (close == 1)
{
if (share_head->share_next->track == 1) //已被用,删掉当前结点
{
last->next = q->next;
movingObject* t=q;
q = q->next;
delete t;
continue;
}
q->label = (share_head->share_next)->label;
cvRectangle(image, q->points[0], q->points[1], color[q->label],2);
q->move = q->x - (share_head->share_next)->x; //两帧位移
share_head->share_next->track = 1;
q->track = 0;
q->keeptime = 0;
last = q;
q = q->next;
}
else if (close==0)
{
//生成新标签
q->label = ++label_num;
cvRectangle(image, q->points[0], q->points[1], color[q->label],2);
q->track = 0;
q->keeptime = 0;
last = q;
q = q->next;
}
else if (close>1)
{
movingObject* t = share_head->share_next;
while ( t != NULL)
{
if (t->track==1)
t = t->share_next;
else break;
}
if (t==NULL) //全部跟踪完毕
{
last->next = q->next;
t = q;
q = q->next;
delete t;
continue;
}
//重用当前队列的这一object
q->label = t->label;
q->move = t->move;
q->area = t->area;
q->x = t->x + t->move;
q->y = t->y;
q->points[0].x = t->points[0].x+t->move;
q->points[0].y = t->points[0].y;
q->points[1].x = t->points[1].x + t->move;
q->points[1].y = t->points[1].y;
q->track = 0;
t->track = 1;
q->keeptime = 0;
cvRectangle(image, q->points[0], q->points[1], color[q->label],2);
t = t->share_next;
while (t)
{
if (t->track == 1)
{
t = t->share_next;
continue;
}
movingObject* newobject = new movingObject();
newobject->area = t->area;
newobject->label = t->label;
newobject->move = t->move;
newobject->next = q->next;
q->next = newobject;
q = newobject;
newobject->points[0].x = t->points[0].x + t->move;
newobject->points[0].y = t->points[0].y;
newobject->points[1].x = t->points[1].x + t->move;
newobject->points[1].y = t->points[1].y;
newobject->x = t->x + t->move;
newobject->y = t->y;
newobject->track = 0;
newobject->keeptime = 0;
cvRectangle(image, newobject->points[0], newobject->points[1], color[newobject->label],2);
t->track = 1; //已跟踪这一个prev目标
t = t->share_next;
}
last = q;
q = q->next;
}
}//end for
detect_hiding(prev_head, image_width, curr_head,value4);
}//end if
cvShowImage("tracking", image);
release_link(prev_head,share_head);
prev_head->next = curr_head->next;
curr_head->next = NULL;
int ctrl;
if ( (ctrl=cvWaitKey(20)) == 27)//ESC退出
{
break;
}
else if (ctrl == 32) //空格暂停
{
while ((ctrl=cvWaitKey(0))!=13)//回车继续
{
if (ctrl == 27)
exit(0);
continue;
}
}
}
cvReleaseImage(&res);
cvReleaseImage(&res0);
cvReleaseImage(&image_pass);
cvReleaseImage(&pBkImg);
cvReleaseImage(&pFrImg);
cvReleaseImage(&pFrame);
cvReleaseCapture(&capture);
return 1;
}
void CvEM::kmeans( const CvVectors& train_data, int nclusters, CvMat* labels,
CvTermCriteria termcrit, const CvMat* centers0 )
{
CvMat* centers = 0;
CvMat* old_centers = 0;
CvMat* counters = 0;
CV_FUNCNAME( "CvEM::kmeans" );
__BEGIN__;
CvRNG rng = cvRNG(-1);
int i, j, k, nsamples, dims;
int iter = 0;
double max_dist = DBL_MAX;
termcrit = cvCheckTermCriteria( termcrit, 1e-6, 100 );
termcrit.epsilon *= termcrit.epsilon;
nsamples = train_data.count;
dims = train_data.dims;
nclusters = MIN( nclusters, nsamples );
CV_CALL( centers = cvCreateMat( nclusters, dims, CV_64FC1 ));
CV_CALL( old_centers = cvCreateMat( nclusters, dims, CV_64FC1 ));
CV_CALL( counters = cvCreateMat( 1, nclusters, CV_32SC1 ));
cvZero( old_centers );
if( centers0 )
{
CV_CALL( cvConvert( centers0, centers ));
}
else
{
for( i = 0; i < nsamples; i++ )
labels->data.i[i] = i*nclusters/nsamples;
cvRandShuffle( labels, &rng );
}
for( ;; )
{
CvMat* temp;
if( iter > 0 || centers0 )
{
for( i = 0; i < nsamples; i++ )
{
const float* s = train_data.data.fl[i];
int k_best = 0;
double min_dist = DBL_MAX;
for( k = 0; k < nclusters; k++ )
{
const double* c = (double*)(centers->data.ptr + k*centers->step);
double dist = 0;
for( j = 0; j <= dims - 4; j += 4 )
{
double t0 = c[j] - s[j];
double t1 = c[j+1] - s[j+1];
dist += t0*t0 + t1*t1;
t0 = c[j+2] - s[j+2];
t1 = c[j+3] - s[j+3];
dist += t0*t0 + t1*t1;
}
for( ; j < dims; j++ )
{
double t = c[j] - s[j];
dist += t*t;
}
if( min_dist > dist )
{
min_dist = dist;
k_best = k;
}
}
labels->data.i[i] = k_best;
}
}
if( ++iter > termcrit.max_iter )
break;
CV_SWAP( centers, old_centers, temp );
cvZero( centers );
cvZero( counters );
// update centers
for( i = 0; i < nsamples; i++ )
{
const float* s = train_data.data.fl[i];
k = labels->data.i[i];
double* c = (double*)(centers->data.ptr + k*centers->step);
for( j = 0; j <= dims - 4; j += 4 )
{
double t0 = c[j] + s[j];
double t1 = c[j+1] + s[j+1];
c[j] = t0;
c[j+1] = t1;
t0 = c[j+2] + s[j+2];
t1 = c[j+3] + s[j+3];
c[j+2] = t0;
c[j+3] = t1;
}
for( ; j < dims; j++ )
c[j] += s[j];
counters->data.i[k]++;
}
if( iter > 1 )
max_dist = 0;
for( k = 0; k < nclusters; k++ )
{
double* c = (double*)(centers->data.ptr + k*centers->step);
if( counters->data.i[k] != 0 )
{
double scale = 1./counters->data.i[k];
for( j = 0; j < dims; j++ )
c[j] *= scale;
}
else
{
const float* s;
for( j = 0; j < 10; j++ )
{
i = cvRandInt( &rng ) % nsamples;
if( counters->data.i[labels->data.i[i]] > 1 )
break;
}
s = train_data.data.fl[i];
for( j = 0; j < dims; j++ )
c[j] = s[j];
}
if( iter > 1 )
{
double dist = 0;
const double* c_o = (double*)(old_centers->data.ptr + k*old_centers->step);
for( j = 0; j < dims; j++ )
{
double t = c[j] - c_o[j];
dist += t*t;
}
if( max_dist < dist )
max_dist = dist;
}
}
if( max_dist < termcrit.epsilon )
break;
}
cvZero( counters );
for( i = 0; i < nsamples; i++ )
counters->data.i[labels->data.i[i]]++;
// ensure that we do not have empty clusters
for( k = 0; k < nclusters; k++ )
if( counters->data.i[k] == 0 )
for(;;)
{
i = cvRandInt(&rng) % nsamples;
j = labels->data.i[i];
if( counters->data.i[j] > 1 )
{
labels->data.i[i] = k;
counters->data.i[j]--;
counters->data.i[k]++;
break;
}
}
__END__;
cvReleaseMat( ¢ers );
cvReleaseMat( &old_centers );
cvReleaseMat( &counters );
}
void CvEM::init_auto( const CvVectors& train_data )
{
CvMat* hdr = 0;
const void** vec = 0;
CvMat* class_ranges = 0;
CvMat* labels = 0;
CV_FUNCNAME( "CvEM::init_auto" );
__BEGIN__;
int nclusters = params.nclusters, nsamples = train_data.count, dims = train_data.dims;
int i, j;
if( nclusters == nsamples )
{
CvMat src = cvMat( 1, dims, CV_32F );
CvMat dst = cvMat( 1, dims, CV_64F );
for( i = 0; i < nsamples; i++ )
{
src.data.ptr = train_data.data.ptr[i];
dst.data.ptr = means->data.ptr + means->step*i;
cvConvert( &src, &dst );
cvZero( covs[i] );
cvSetIdentity( cov_rotate_mats[i] );
}
cvSetIdentity( probs );
cvSet( weights, cvScalar(1./nclusters) );
}
else
{
int max_count = 0;
CV_CALL( class_ranges = cvCreateMat( 1, nclusters+1, CV_32SC1 ));
if( nclusters > 1 )
{
CV_CALL( labels = cvCreateMat( 1, nsamples, CV_32SC1 ));
kmeans( train_data, nclusters, labels, cvTermCriteria( CV_TERMCRIT_ITER,
params.means ? 1 : 10, 0.5 ), params.means );
CV_CALL( cvSortSamplesByClasses( (const float**)train_data.data.fl,
labels, class_ranges->data.i ));
}
else
{
class_ranges->data.i[0] = 0;
class_ranges->data.i[1] = nsamples;
}
for( i = 0; i < nclusters; i++ )
{
int left = class_ranges->data.i[i], right = class_ranges->data.i[i+1];
max_count = MAX( max_count, right - left );
}
CV_CALL( hdr = (CvMat*)cvAlloc( max_count*sizeof(hdr[0]) ));
CV_CALL( vec = (const void**)cvAlloc( max_count*sizeof(vec[0]) ));
hdr[0] = cvMat( 1, dims, CV_32F );
for( i = 0; i < max_count; i++ )
{
vec[i] = hdr + i;
hdr[i] = hdr[0];
}
for( i = 0; i < nclusters; i++ )
{
int left = class_ranges->data.i[i], right = class_ranges->data.i[i+1];
int cluster_size = right - left;
CvMat avg;
if( cluster_size <= 0 )
continue;
for( j = left; j < right; j++ )
hdr[j - left].data.fl = train_data.data.fl[j];
CV_CALL( cvGetRow( means, &avg, i ));
CV_CALL( cvCalcCovarMatrix( vec, cluster_size, covs[i],
&avg, CV_COVAR_NORMAL | CV_COVAR_SCALE ));
weights->data.db[i] = (double)cluster_size/(double)nsamples;
}
}
__END__;
cvReleaseMat( &class_ranges );
cvReleaseMat( &labels );
cvFree( &hdr );
cvFree( &vec );
}
void CvEM::init_em( const CvVectors& train_data )
{
CvMat *w = 0, *u = 0, *tcov = 0;
CV_FUNCNAME( "CvEM::init_em" );
__BEGIN__;
double maxval = 0;
int i, force_symm_plus = 0;
int nclusters = params.nclusters, nsamples = train_data.count, dims = train_data.dims;
if( params.start_step == START_AUTO_STEP || nclusters == 1 || nclusters == nsamples )
init_auto( train_data );
else if( params.start_step == START_M_STEP )
{
for( i = 0; i < nsamples; i++ )
{
CvMat prob;
cvGetRow( params.probs, &prob, i );
cvMaxS( &prob, 0., &prob );
cvMinMaxLoc( &prob, 0, &maxval );
if( maxval < FLT_EPSILON )
cvSet( &prob, cvScalar(1./nclusters) );
else
cvNormalize( &prob, &prob, 1., 0, CV_L1 );
}
EXIT; // do not preprocess covariation matrices,
// as in this case they are initialized at the first iteration of EM
}
else
{
CV_ASSERT( params.start_step == START_E_STEP && params.means );
if( params.weights && params.covs )
{
cvConvert( params.means, means );
cvReshape( weights, weights, 1, params.weights->rows );
cvConvert( params.weights, weights );
cvReshape( weights, weights, 1, 1 );
cvMaxS( weights, 0., weights );
cvMinMaxLoc( weights, 0, &maxval );
if( maxval < FLT_EPSILON )
cvSet( weights, cvScalar(1./nclusters) );
cvNormalize( weights, weights, 1., 0, CV_L1 );
for( i = 0; i < nclusters; i++ )
CV_CALL( cvConvert( params.covs[i], covs[i] ));
force_symm_plus = 1;
}
else
init_auto( train_data );
}
CV_CALL( tcov = cvCreateMat( dims, dims, CV_64FC1 ));
CV_CALL( w = cvCreateMat( dims, dims, CV_64FC1 ));
if( params.cov_mat_type == COV_MAT_GENERIC )
CV_CALL( u = cvCreateMat( dims, dims, CV_64FC1 ));
for( i = 0; i < nclusters; i++ )
{
if( force_symm_plus )
{
cvTranspose( covs[i], tcov );
cvAddWeighted( covs[i], 0.5, tcov, 0.5, 0, tcov );
}
else
cvCopy( covs[i], tcov );
cvSVD( tcov, w, u, 0, CV_SVD_MODIFY_A + CV_SVD_U_T + CV_SVD_V_T );
if( params.cov_mat_type == COV_MAT_SPHERICAL )
cvSetIdentity( covs[i], cvScalar(cvTrace(w).val[0]/dims) );
else if( params.cov_mat_type == COV_MAT_DIAGONAL )
cvCopy( w, covs[i] );
else
{
// generic case: covs[i] = (u')'*max(w,0)*u'
cvGEMM( u, w, 1, 0, 0, tcov, CV_GEMM_A_T );
cvGEMM( tcov, u, 1, 0, 0, covs[i], 0 );
}
}
__END__;
cvReleaseMat( &w );
cvReleaseMat( &u );
cvReleaseMat( &tcov );
}
void
icvCrossCorr( const CvArr* _img, const CvArr* _templ, CvArr* _corr,
CvPoint anchor, double delta, int borderType )
{
// disable OpenMP in the case of Visual Studio,
// otherwise the performance drops significantly
#undef USE_OPENMP
#if !defined _MSC_VER || defined CV_ICC
#define USE_OPENMP 1
#endif
const double block_scale = 4.5;
const int min_block_size = 256;
cv::Ptr<CvMat> dft_img[CV_MAX_THREADS];
cv::Ptr<CvMat> dft_templ;
std::vector<uchar> buf[CV_MAX_THREADS];
int k, num_threads = 0;
CvMat istub, *img = (CvMat*)_img;
CvMat tstub, *templ = (CvMat*)_templ;
CvMat cstub, *corr = (CvMat*)_corr;
CvSize dftsize, blocksize;
int depth, templ_depth, corr_depth, max_depth = CV_32F,
cn, templ_cn, corr_cn, buf_size = 0,
tile_count_x, tile_count_y, tile_count;
img = cvGetMat( img, &istub );
templ = cvGetMat( templ, &tstub );
corr = cvGetMat( corr, &cstub );
if( CV_MAT_DEPTH( img->type ) != CV_8U &&
CV_MAT_DEPTH( img->type ) != CV_16U &&
CV_MAT_DEPTH( img->type ) != CV_32F &&
CV_MAT_DEPTH( img->type ) != CV_64F )
CV_Error( CV_StsUnsupportedFormat,
"The function supports only 8u, 16u and 32f data types" );
if( !CV_ARE_DEPTHS_EQ( img, templ ) && CV_MAT_DEPTH( templ->type ) != CV_32F )
CV_Error( CV_StsUnsupportedFormat,
"Template (kernel) must be of the same depth as the input image, or be 32f" );
if( !CV_ARE_DEPTHS_EQ( img, corr ) && CV_MAT_DEPTH( corr->type ) != CV_32F &&
CV_MAT_DEPTH( corr->type ) != CV_64F )
CV_Error( CV_StsUnsupportedFormat,
"The output image must have the same depth as the input image, or be 32f/64f" );
if( (!CV_ARE_CNS_EQ( img, corr ) || CV_MAT_CN(templ->type) > 1) &&
(CV_MAT_CN( corr->type ) > 1 || !CV_ARE_CNS_EQ( img, templ)) )
CV_Error( CV_StsUnsupportedFormat,
"The output must have the same number of channels as the input (when the template has 1 channel), "
"or the output must have 1 channel when the input and the template have the same number of channels" );
depth = CV_MAT_DEPTH(img->type);
cn = CV_MAT_CN(img->type);
templ_depth = CV_MAT_DEPTH(templ->type);
templ_cn = CV_MAT_CN(templ->type);
corr_depth = CV_MAT_DEPTH(corr->type);
corr_cn = CV_MAT_CN(corr->type);
CV_Assert( corr_cn == 1 || delta == 0 );
max_depth = MAX( max_depth, templ_depth );
max_depth = MAX( max_depth, depth );
max_depth = MAX( max_depth, corr_depth );
if( depth > CV_8U )
max_depth = CV_64F;
/*if( img->cols < templ->cols || img->rows < templ->rows )
CV_Error( CV_StsUnmatchedSizes,
"Such a combination of image and template/filter size is not supported" );*/
if( corr->rows > img->rows + templ->rows - 1 ||
corr->cols > img->cols + templ->cols - 1 )
CV_Error( CV_StsUnmatchedSizes,
"output image should not be greater than (W + w - 1)x(H + h - 1)" );
blocksize.width = cvRound(templ->cols*block_scale);
blocksize.width = MAX( blocksize.width, min_block_size - templ->cols + 1 );
blocksize.width = MIN( blocksize.width, corr->cols );
blocksize.height = cvRound(templ->rows*block_scale);
blocksize.height = MAX( blocksize.height, min_block_size - templ->rows + 1 );
blocksize.height = MIN( blocksize.height, corr->rows );
dftsize.width = cvGetOptimalDFTSize(blocksize.width + templ->cols - 1);
if( dftsize.width == 1 )
dftsize.width = 2;
dftsize.height = cvGetOptimalDFTSize(blocksize.height + templ->rows - 1);
if( dftsize.width <= 0 || dftsize.height <= 0 )
CV_Error( CV_StsOutOfRange, "the input arrays are too big" );
// recompute block size
blocksize.width = dftsize.width - templ->cols + 1;
blocksize.width = MIN( blocksize.width, corr->cols );
blocksize.height = dftsize.height - templ->rows + 1;
blocksize.height = MIN( blocksize.height, corr->rows );
dft_templ = cvCreateMat( dftsize.height*templ_cn, dftsize.width, max_depth );
#ifdef USE_OPENMP
num_threads = cvGetNumThreads();
#else
num_threads = 1;
#endif
for( k = 0; k < num_threads; k++ )
dft_img[k] = cvCreateMat( dftsize.height, dftsize.width, max_depth );
if( templ_cn > 1 && templ_depth != max_depth )
buf_size = templ->cols*templ->rows*CV_ELEM_SIZE(templ_depth);
if( cn > 1 && depth != max_depth )
buf_size = MAX( buf_size, (blocksize.width + templ->cols - 1)*
(blocksize.height + templ->rows - 1)*CV_ELEM_SIZE(depth));
if( (corr_cn > 1 || cn > 1) && corr_depth != max_depth )
buf_size = MAX( buf_size, blocksize.width*blocksize.height*CV_ELEM_SIZE(corr_depth));
if( buf_size > 0 )
{
for( k = 0; k < num_threads; k++ )
buf[k].resize(buf_size);
}
// compute DFT of each template plane
for( k = 0; k < templ_cn; k++ )
{
CvMat dstub, *src, *dst, temp;
CvMat* planes[] = { 0, 0, 0, 0 };
int yofs = k*dftsize.height;
src = templ;
dst = cvGetSubRect( dft_templ, &dstub, cvRect(0,yofs,templ->cols,templ->rows));
if( templ_cn > 1 )
{
planes[k] = templ_depth == max_depth ? dst :
cvInitMatHeader( &temp, templ->rows, templ->cols, templ_depth, &buf[0][0] );
cvSplit( templ, planes[0], planes[1], planes[2], planes[3] );
src = planes[k];
planes[k] = 0;
}
if( dst != src )
cvConvert( src, dst );
if( dft_templ->cols > templ->cols )
{
cvGetSubRect( dft_templ, dst, cvRect(templ->cols, yofs,
dft_templ->cols - templ->cols, templ->rows) );
cvZero( dst );
}
cvGetSubRect( dft_templ, dst, cvRect(0,yofs,dftsize.width,dftsize.height) );
cvDFT( dst, dst, CV_DXT_FORWARD + CV_DXT_SCALE, templ->rows );
}
tile_count_x = (corr->cols + blocksize.width - 1)/blocksize.width;
tile_count_y = (corr->rows + blocksize.height - 1)/blocksize.height;
tile_count = tile_count_x*tile_count_y;
#if defined _OPENMP && defined USE_OPENMP
#pragma omp parallel for num_threads(num_threads) schedule(dynamic)
#endif
// calculate correlation by blocks
for( k = 0; k < tile_count; k++ )
{
#ifdef USE_OPENMP
int thread_idx = cvGetThreadNum();
#else
int thread_idx = 0;
#endif
int x = (k%tile_count_x)*blocksize.width;
int y = (k/tile_count_x)*blocksize.height;
int i, yofs;
CvMat sstub, dstub, *src, *dst, temp;
CvMat* planes[] = { 0, 0, 0, 0 };
CvMat* _dft_img = dft_img[thread_idx];
uchar* _buf = buf_size > 0 ? &buf[thread_idx][0] : 0;
CvSize csz = { blocksize.width, blocksize.height }, isz;
int x0 = x - anchor.x, y0 = y - anchor.y;
int x1 = MAX( 0, x0 ), y1 = MAX( 0, y0 ), x2, y2;
csz.width = MIN( csz.width, corr->cols - x );
csz.height = MIN( csz.height, corr->rows - y );
isz.width = csz.width + templ->cols - 1;
isz.height = csz.height + templ->rows - 1;
x2 = MIN( img->cols, x0 + isz.width );
y2 = MIN( img->rows, y0 + isz.height );
for( i = 0; i < cn; i++ )
{
CvMat dstub1, *dst1;
yofs = i*dftsize.height;
src = cvGetSubRect( img, &sstub, cvRect(x1,y1,x2-x1,y2-y1) );
dst = cvGetSubRect( _dft_img, &dstub,
cvRect(0,0,isz.width,isz.height) );
dst1 = dst;
if( x2 - x1 < isz.width || y2 - y1 < isz.height )
dst1 = cvGetSubRect( _dft_img, &dstub1,
cvRect( x1 - x0, y1 - y0, x2 - x1, y2 - y1 ));
if( cn > 1 )
{
planes[i] = dst1;
if( depth != max_depth )
planes[i] = cvInitMatHeader( &temp, y2 - y1, x2 - x1, depth, _buf );
cvSplit( src, planes[0], planes[1], planes[2], planes[3] );
src = planes[i];
planes[i] = 0;
}
if( dst1 != src )
cvConvert( src, dst1 );
if( dst != dst1 )
cvCopyMakeBorder( dst1, dst, cvPoint(x1 - x0, y1 - y0), borderType );
if( dftsize.width > isz.width )
{
cvGetSubRect( _dft_img, dst, cvRect(isz.width, 0,
dftsize.width - isz.width,dftsize.height) );
cvZero( dst );
}
cvDFT( _dft_img, _dft_img, CV_DXT_FORWARD, isz.height );
cvGetSubRect( dft_templ, dst,
cvRect(0,(templ_cn>1?yofs:0),dftsize.width,dftsize.height) );
cvMulSpectrums( _dft_img, dst, _dft_img, CV_DXT_MUL_CONJ );
cvDFT( _dft_img, _dft_img, CV_DXT_INVERSE, csz.height );
src = cvGetSubRect( _dft_img, &sstub, cvRect(0,0,csz.width,csz.height) );
dst = cvGetSubRect( corr, &dstub, cvRect(x,y,csz.width,csz.height) );
if( corr_cn > 1 )
{
planes[i] = src;
if( corr_depth != max_depth )
{
planes[i] = cvInitMatHeader( &temp, csz.height, csz.width,
corr_depth, _buf );
cvConvertScale( src, planes[i], 1, delta );
}
cvMerge( planes[0], planes[1], planes[2], planes[3], dst );
planes[i] = 0;
}
else
{
if( i == 0 )
cvConvertScale( src, dst, 1, delta );
else
{
if( max_depth > corr_depth )
{
cvInitMatHeader( &temp, csz.height, csz.width,
corr_depth, _buf );
cvConvert( src, &temp );
src = &temp;
}
cvAcc( src, dst );
}
}
}
}
}
