int getFeatures(Mat &object, Mat &frame, Mat &homography, vector<KeyPoint> &keypoints_object,
vector<KeyPoint> &keypoints_scene, vector<DMatch> &good_matches) {
Ptr<SIFT> detector = SIFT::create();
// Detect features and compute descriptors
Mat descriptors_object, descriptors_scene;
detector->detectAndCompute(object, noArray(), keypoints_object, descriptors_object);
detector->detectAndCompute(frame, noArray(), keypoints_scene, descriptors_scene);
// Match descriptors using FLANN
FlannBasedMatcher matcher;
vector<DMatch> matches;
matcher.match(descriptors_object, descriptors_scene, matches);
// Check if too few matches are found
if (matches.size() <= 4) {
cout << "Error: too few matches were found" << endl;
return -1;
}
// Find minimum and maximum distances between descriptors
double max_dist = 0;
double min_dist = 100;
for (int i = 0; i < descriptors_object.rows; i++) {
double dist = matches[i].distance;
if (dist < min_dist) min_dist = dist;
if (dist > max_dist) max_dist = dist;
}
// If there are sufficient matches, filter to find higher-quality subset
int minMatches = 8;
for (int i = 0; i < descriptors_object.rows; i++) {
if (matches[i].distance < 3*min_dist && matches.size() > minMatches) {
good_matches.push_back(matches[i]);
}
}
// If there are too few good matches, use all matches
if (good_matches.size() <= minMatches) {
for (int i = 0; i < matches.size(); i++) {
if (i < good_matches.size()) {
good_matches[i] = matches[i];
} else {
good_matches.push_back(matches[i]);
}
}
}
vector<Point2f> obj, scene;
for (int i = 0; i < good_matches.size(); i++) {
// Determine keypoints from the matches
obj.push_back(keypoints_object[ good_matches[i].queryIdx ].pt);
scene.push_back(keypoints_scene[ good_matches[i].trainIdx ].pt);
}
// Transform pixel coordinates between images
homography = findHomography(obj, scene, CV_RANSAC);
return good_matches.size();
}
vector<DMatch> GraphicEnd::match( Mat desp1, Mat desp2 )
{
cout<<"GraphicEnd::match two desp"<<endl;
FlannBasedMatcher matcher;
vector<DMatch> matches;
if (desp1.empty() || desp2.empty())
{
return matches;
}
double max_dist = 0, min_dist = 100;
matcher.match( desp1, desp2, matches);
for (int i=0; i<desp1.rows; i++)
{
double dist = matches[ i ].distance;
if (dist < min_dist)
min_dist = dist;
if (dist > max_dist)
max_dist = dist;
}
//return matches;
vector<DMatch> good_matches;
for (size_t i=0; i<matches.size(); i++)
{
if (matches[ i ].distance <= max(4*min_dist, _match_min_dist))
{
good_matches.push_back(matches[ i ]);
}
}
return good_matches;
}
int match(vector<DMatch> &match, frame &f1, frame &f2)
{
vector<DMatch> matches;
FlannBasedMatcher matcher;
matcher.match(f1.desp, f2.desp, matches);
static reader pd("../config/config.ini");
double min_distance = 9999;
double match_threshold = atof(pd.get("match_threshold").c_str());
for (int i = 0; i < matches.size(); ++i)
{
if (matches[i].distance < min_distance)
{
min_distance = matches[i].distance;
}
}
for (int i = 0; i < matches.size(); ++i)
{
if (matches[i].distance < (min_distance * match_threshold))
{
match.push_back(matches[i]);
}
}
return match.size();
}
void main()
{
//Give the names of the images to be registered
const char* imRef_name = "834-r1.png";
const char* imNxt_name = "835-r1.png";
int hessianThresh = 100, ransacThresh = 3;;
Mat mask, H12;
// Read images
Mat img1 = imread(imRef_name, CV_LOAD_IMAGE_GRAYSCALE);
Mat img2 = imread(imNxt_name, CV_LOAD_IMAGE_GRAYSCALE);
Mat img2Out; // Registered image2 wrt image1
// Check to see if images exist
if(img1.empty() || img2.empty())
{
printf("Can’t read one of the images\n");
exit(0);
}
// detecting keypoints
printf("Finding keypoints ... ");
SURF ImgSurf(hessianThresh);
vector<KeyPoint> keypoints1, keypoints2;
ImgSurf(img1, mask, keypoints1);
ImgSurf(img2, mask, keypoints2);
// computing descriptors
SurfDescriptorExtractor extractor;
Mat descriptors1, descriptors2;
extractor(img1,mask,keypoints1,descriptors1,TRUE);
extractor(img2, mask, keypoints2, descriptors2, TRUE);
// Match the points
printf("\nMatching keypoints ... ");
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( descriptors1, descriptors2, matches );
// Extract indices of matched points
vector<int> queryIdxs( matches.size() ), trainIdxs( matches.size() );
for( size_t i = 0; i < matches.size(); i++ )
{
queryIdxs[i] = matches[i].queryIdx;
trainIdxs[i] = matches[i].trainIdx;
}
// Extract matched points from indices
vector<Point2f> points1; KeyPoint::convert(keypoints1, points1, queryIdxs);
vector<Point2f> points2; KeyPoint::convert(keypoints2, points2, trainIdxs);
// Use RANSAC to find the homography
printf("\nComputing homography ... ");
H12 = findHomography( Mat(points2), Mat(points1), CV_RANSAC, ransacThresh );
// Warp the second image according to the homography
warpPerspective(img2, img2Out, H12, cvSize(img2.cols, img2.rows), INTER_LINEAR);
// Write result to file
imwrite("im2reg.png",img2Out);
printf("\nDone!!!.... ");
}
void detectSiftMatchWithOpenCV(const char* img1_path, const char* img2_path, MatrixXf &match) {
Mat img1 = imread(img1_path);
Mat img2 = imread(img2_path);
SiftFeatureDetector detector;
SiftDescriptorExtractor extractor;
vector<KeyPoint> key1;
vector<KeyPoint> key2;
Mat desc1, desc2;
detector.detect(img1, key1);
detector.detect(img2, key2);
extractor.compute(img1, key1, desc1);
extractor.compute(img2, key2, desc2);
FlannBasedMatcher matcher;
vector<DMatch> matches;
matcher.match(desc1, desc2, matches);
match.resize(matches.size(), 6);
cout << "match count: " << matches.size() << endl;
for (int i = 0; i < matches.size(); i++) {
match(i, 0) = key1[matches[i].queryIdx].pt.x;
match(i, 1) = key1[matches[i].queryIdx].pt.y;
match(i, 2) = 1;
match(i, 3) = key2[matches[i].trainIdx].pt.x;
match(i, 4) = key2[matches[i].trainIdx].pt.y;
match(i, 5) = 1;
}
}
bool RelicScn::Match_an_Obj(RelicObj obj)
{
string message;
FlannBasedMatcher matcher;
vector<DMatch> matches;
matcher.match(obj.descriptors, this->descriptors, matches);
vector<DMatch> good_matches = Get_Good_Matches(matches);
//-- Localize the object
std::vector<Point2f> obj_points;
std::vector<Point2f> scn_points;
for (size_t i = 0; i < good_matches.size(); i++)
{
//-- Get the keypoints from the good matches
obj_points.push_back(obj.keypoints[good_matches[i].queryIdx].pt);
scn_points.push_back(this->keypoints[good_matches[i].trainIdx].pt);
}
Mat H = cv::findHomography(obj_points, scn_points, RANSAC);
std::vector<Point2f> obj_corners(4);
obj_corners[0] = cvPoint(0, 0);
obj_corners[1] = cvPoint(obj.img_width-1, 0);
obj_corners[2] = cvPoint(obj.img_width-1, obj.img_height-1);
obj_corners[3] = cvPoint(0, obj.img_height-1);
std::vector<Point2f> possible_obj_corners(4);
perspectiveTransform(obj_corners, possible_obj_corners, H);
BOOST_LOG_TRIVIAL(info) << "原始目标物体大小（像素）： " << contourArea(obj_corners);
BOOST_LOG_TRIVIAL(info) << "检测到的物体大小（像素）： " << contourArea(possible_obj_corners);
this->corners = possible_obj_corners;
double possible_target_area = contourArea(possible_obj_corners);
double whole_scene_area = this->img_gray.rows*this->img_gray.cols;
BOOST_LOG_TRIVIAL(info) << "环境图像大小（像素）： " << whole_scene_area;
double ratio = possible_target_area / whole_scene_area;
BOOST_LOG_TRIVIAL(info) << "检测到的目标占全图比例： " << ratio;
if (ratio>0.03 && ratio<1)
{
for (int i;i < possible_obj_corners.size();i++)
{
if (possible_obj_corners[i].x < 0 || possible_obj_corners[i].y < 0)
{
BOOST_LOG_TRIVIAL(info) << "未能检测到目标物体！";
return false;
}
}
BOOST_LOG_TRIVIAL(info) << "成功检测到目标物体！";
return true;
}
else
{
BOOST_LOG_TRIVIAL(info) << "未能检测到目标物体！";
return false;
}
}
int compare(Mat img_1, Mat img_2) {
//-- Step 1: Detect the keypoints using SURF Detector
int minHessian = 400;
SurfFeatureDetector detector( minHessian );
std::vector<KeyPoint> keypoints_1, keypoints_2;
detector.detect( img_1, keypoints_1 );
detector.detect( img_2, keypoints_2 );
//-- Step 2: Calculate descriptors (feature vectors)
SurfDescriptorExtractor extractor;
Mat descriptors_1, descriptors_2;
extractor.compute( img_1, keypoints_1, descriptors_1 );
extractor.compute( img_2, keypoints_2, descriptors_2 );
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( descriptors_1, descriptors_2, matches );
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors_1.rows; i++ ) {
double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
//-- Draw only "good" matches (i.e. whose distance is less than 2*min_dist,
//-- or a small arbitary value ( 0.02 ) in the event that min_dist is very
//-- small)
//-- PS.- radiusMatch can also be used here.
std::vector< DMatch > good_matches;
for( int i = 0; i < descriptors_1.rows; i++ ) {
if( matches[i].distance <= max(2 * min_dist, 0.02) ) {
good_matches.push_back( matches[i]);
}
}
// Mat img_matches;
// drawMatches( img_1, keypoints_1, img_2, keypoints_2,
// good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
// vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
// //-- Show detected matches
// imshow( "Good Matches", img_matches );
waitKey(0);
return good_matches.size();
}
int PlayedCard::computeSurfGoodMatches(vector<KeyPoint> keypoints_1, vector<KeyPoint> keypoints_2, Mat descriptors_1, Mat descriptors_2) {
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
vector< DMatch > matches;
matcher.match(descriptors_1, descriptors_2, matches);
filterMatchesByAbsoluteValue(matches, 0.125);
filterMatchesRANSAC(matches, keypoints_1, keypoints_2);
return (int)matches.size();
}
int main(int argc, char* argv[])
{
VideoCapture camera;
camera.set(CV_CAP_PROP_FRAME_WIDTH, WIDTH);
camera.set(CV_CAP_PROP_FRAME_HEIGHT, HEIGHT);
camera.open(0);
//checkOpenCL();
Ptr<FeatureDetector> detector = FeatureDetector::create("STAR");
//Ptr<DescriptorExtractor> extractor = DescriptorExtractor::create("FREAK");
BriefDescriptorExtractor extractor;
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
Mat descriptor[2];
int k = 0;
camera >> image;
detector->detect(image, keypoint[1]);
extractor.compute(image, keypoint[1], descriptor[1]);
for (bool loop = true; loop; )
{
switch (waitKey(10))
{
case 'q':
loop = false;
break;
}
camera >> image;
if (image.empty())
break;
// detect features
detector->detect(image, keypoint[k % 2]);
extractor.compute(image, keypoint[k % 2], descriptor[k % 2]);
try {
matcher.match(descriptor[0], descriptor[1], matches);
}
catch (Exception ex)
{
printf("%s", ex.msg);
}
printf("%d\n", keypoint[k % 2].size());
for (int i = 0; i < keypoint[k % 2].size(); i++)
{
Point2f pt = keypoint[k % 2][i].pt;
circle(image, Point(pt.x, pt.y), 3, Scalar(0, 0, 255));
}
k++;
imshow("image", image);
}
return 0;
}
/**
* @brief featuredetector::testFeatures
* FLANN based matching.
* Algorithm is taken from
* http://docs.opencv.org/doc/tutorials/features2d/feature_flann_matcher/feature_flann_matcher.html
* @return Matched points image.
*/
Mat featuredetector::testFeatures(){
cv::Mat im1, im2;
//Grayscale the images.
if(_image1.channels() == 3)
cv::cvtColor(_image1,im1, CV_BGR2GRAY);
else _image1.copyTo(im1);
if(_image2.channels() == 3)
cv::cvtColor(_image2,im2, CV_BGR2GRAY);
else _image2.copyTo(im2);
int minH = 100; // (should be around ~100)
Ptr<xfeatures2d::SURF> detector = xfeatures2d::SURF::create(minH);
detector->setHessianThreshold(minH);
std::vector<KeyPoint> keypoints1, keypoints2;
Mat descriptors1, descriptors2;
detector->detectAndCompute( im1, Mat(), keypoints1, descriptors1 );
detector->detectAndCompute( im2, Mat(), keypoints2, descriptors2 );
FlannBasedMatcher matcher;
std::vector<DMatch> matches;
matcher.match( descriptors1, descriptors2, matches );
double maxDist = 0; double minDist = 80;
for( int i = 0; i < descriptors1.rows; i++ )
{ double dist = matches[i].distance;
if( dist < minDist ) minDist = dist;
if( dist > maxDist ) maxDist = dist;
}
std::vector< DMatch > goodMatches;
for( int i = 0; i < descriptors1.rows; i++ )
{ if( matches[i].distance <= max(2*minDist, 0.02) )
{ goodMatches.push_back( matches[i]); }
}
Mat matchMat;
drawMatches( im1, keypoints1, im2, keypoints2,
goodMatches, matchMat, Scalar::all(-1), Scalar::all(-1),
vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
imshow( "Matches", matchMat );
return matchMat;
}
/*
* @function main
* @brief Main function
*/
int flann( int argc, char** argv )
{
if( argc != 3 )
{ readme(); return -1; }
Mat img_1 = imread( argv[1], IMREAD_GRAYSCALE );
Mat img_2 = imread( argv[2], IMREAD_GRAYSCALE );
if( !img_1.data || !img_2.data )
{ std::cout<< " --(!) Error reading images " << std::endl; return -1; }
//-- Step 1: Detect the keypoints using SURF Detector, compute the descriptors
int minHessian = 400;
Ptr<SURF> detector = SURF::create();
detector->setHessianThreshold(minHessian);
std::vector<KeyPoint> keypoints_1, keypoints_2;
Mat descriptors_1, descriptors_2;
detector->detectAndCompute( img_1, Mat(), keypoints_1, descriptors_1 );
detector->detectAndCompute( img_2, Mat(), keypoints_2, descriptors_2 );
//-- Step 2: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( descriptors_1, descriptors_2, matches );
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors_1.rows; i++ )
{ double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist );
printf("-- Min dist : %f \n", min_dist );
//-- Draw only "good" matches (i.e. whose distance is less than 2*min_dist,
//-- or a small arbitary value ( 0.02 ) in the event that min_dist is very
//-- small)
//-- PS.- radiusMatch can also be used here.
std::vector< DMatch > good_matches;
for( int i = 0; i < descriptors_1.rows; i++ )
{ if( matches[i].distance <= max(2*min_dist, 0.02) )
{ good_matches.push_back( matches[i]); }
}
//-- Draw only "good" matches
Mat img_matches;
drawMatches( img_1, keypoints_1, img_2, keypoints_2,
good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
//-- Show detected matches
imshow( "Good Matches", img_matches );
for( int i = 0; i < (int)good_matches.size(); i++ )
{ printf( "-- Good Match [%d] Keypoint 1: %d -- Keypoint 2: %d \n", i, good_matches[i].queryIdx, good_matches[i].trainIdx ); }
waitKey(0);
return 0;
}
vector<DMatch> GraphicEnd::match( vector<PLANE>& p1, vector<PLANE>& p2 )
{
cout<<"GraphicEnd::match two planes"<<endl;
FlannBasedMatcher matcher;
vector<DMatch> matches;
cv::Mat des1(p1.size(), 4, CV_32F), des2(p2.size(), 4, CV_32F);
for (size_t i=0; i<p1.size(); i++)
{
pcl::ModelCoefficients c = p1[i].coff;
float m[1][4] = { c.values[0], c.values[1], c.values[2], c.values[3] };
Mat mat = Mat(1,4, CV_32F, m);
mat.row(0).copyTo( des1.row(i) );
}
for (size_t i=0; i<p2.size(); i++)
{
pcl::ModelCoefficients c = p2[i].coff;
float m[1][4] = { c.values[0], c.values[1], c.values[2], c.values[3] };
Mat mat = Mat(1,4, CV_32F, m);
mat.row(0).copyTo( des2.row(i) );
}
matcher.match( des1, des2, matches);
return matches;
double max_dist = 0, min_dist = 100;
for (int i=0; i<des1.rows; i++)
{
double dist = matches[ i ].distance;
if (dist < min_dist)
min_dist = dist;
if (dist > max_dist)
max_dist = dist;
}
vector<DMatch> good_matches;
for (size_t i=0; i<matches.size(); i++)
{
if (matches[ i ].distance <= 3*min_dist)
{
good_matches.push_back(matches[ i ]);
}
}
return good_matches;
}
//Realiza el Matching entre puntos
void computeMatching(Mat& img1, Mat& img2,vector<KeyPoint>& keypoints1,vector<KeyPoint>& keypoints2, vector<DMatch>& matches ){
// computing descriptors
#if _SURF_
SurfDescriptorExtractor extractor;
#else if _SIFT_
SiftDescriptorExtractor extractor;
#endif
Mat descriptors1, descriptors2;
extractor.compute(img1, keypoints1, descriptors1);
extractor.compute(img2, keypoints2, descriptors2);
FlannBasedMatcher matcher;
matcher.match(descriptors1,descriptors2,matches);
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors1.rows; i++ ){
double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist );
printf("-- Min dist : %f \n", min_dist );
//-- Draw only "good" matches (i.e. whose distance is less than 2*min_dist )
//-- PS.- radiusMatch can also be used here.
std::vector< DMatch > good_matches;
for( int i = 0; i < descriptors1.rows; i++ ){
if( matches[i].distance < 2*min_dist ){
good_matches.push_back( matches[i]);
}
}
//-- Draw only "good" matches
Mat img_matches;
drawMatches( img1, keypoints1, img2, keypoints2,
good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
//-- Show detected matches
imshow( "Good Matches", img_matches );
}
void findTopFiveFLANNMatches(Mat hqDesc, vector<Mat>* keyframeDesc, vector<vector< DMatch >>* matchVec, vector<int>* matchIndices){
FlannBasedMatcher matcher;
int index = 0;
//Calculate matches between high quality image and
for (vector<Mat>::iterator it = keyframeDesc->begin(); it != keyframeDesc->end(); ++it){
vector< DMatch > matches;
//calculate initial matches
Mat kfDesc = *it;
matcher.match(hqDesc, kfDesc, matches);
//determine good matches
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for (int i = 0; i < hqDesc.rows; i++)
{
double dist = matches[i].distance;
if (dist < min_dist) min_dist = dist;
if (dist > max_dist) max_dist = dist;
}
std::vector< DMatch > good_matches;
for (int i = 0; i < hqDesc.rows; i++)
{
if (matches[i].distance <= max(2 * min_dist, 0.02))
{
good_matches.push_back(matches[i]);
}
}
matchVec->push_back(good_matches);
index++;
}
//pickTopFive
pickTopFive(matchVec, matchIndices);
index = 0;
}
int main(int argc, char** argv)
{
std::vector<KeyPoint> keypoints_1;
vector<vector<KeyPoint>>keypoints = vector<vector<KeyPoint>>();
//////////////////////////////////////////////////////////////////////////
Ptr<FeatureDetector> detector = xfeatures2d::SIFT::create();
Ptr<DescriptorExtractor> extractor = xfeatures2d::SIFT::create();
//Ptr<DescriptorMatcher> matcher = new FlannBasedMatcher();
FlannBasedMatcher* matcher = new FlannBasedMatcher();
//////////////////////////////////////////////////////////////////////////
Mat image, descriptor, homography;
Mat posters[7], descriptors[7];
vector<DMatch> matches = vector<DMatch>();
//detects and extracts the local features
openImage("poster_test.jpg", image);
detector->detect(image, keypoints_1);
extractor->compute(image, keypoints_1, descriptor);
for (int i = 0;i < 7;i++)
{
vector<KeyPoint> keypoint;
openImage("poster" + std::to_string(i + 1) + ".jpg", posters[i]);
detector->detect(posters[i], keypoint);
extractor->compute(posters[i], keypoint, descriptors[i]);
keypoints.push_back(keypoint);
}
matcher->match(descriptor, descriptors[5], matches);
filterMatchesByAbsoluteValue(matches, 90);
homography = filterMatchesRANSAC(matches, keypoints_1, keypoints[5]);
showResult(image, keypoints_1, posters[5], keypoints[5], matches, homography);
return 0;
}
std::vector< DMatch > matcher (Mat descripteur1,Mat descripteur2){
//renvoie les indices des "bons" points uniquement
//on recupere les zones en communs
FlannBasedMatcher matcher;
std::vector<DMatch> matches;
if(descripteur1.type()!=CV_32F) {
descripteur1.convertTo(descripteur1, CV_32F);//pb de types -< le match prend des descriptors float et non binaires
}
if(descripteur2.type()!=CV_32F) {
descripteur2.convertTo(descripteur2, CV_32F);
}
matcher.match( descripteur1, descripteur2, matches );
double dmax = 0; double dmin= 1000;
//calcul des distances max et min entre les matches
for( int i = 0; i < matches.size(); i++ )
{
double d = matches[i].distance;
if( d<dmin )
dmin=d;
if( d>dmax)
dmax=d;
}
//on veut ne garder que les "bonnes zones communes" -> définir "bonnes zones" !!!
std::vector<DMatch> bonMatches;
for (int i=0;i<matches.size();i++)
{
if (matches[i].distance<=2*dmin) //2*distance minimale est fixé arbitrairement
bonMatches.push_back(matches[i]);
}
return bonMatches;
}
//static void matchDescriptors( const Mat& queryDescriptors, const vector<Mat>& trainDescriptors,
//vector<DMatch>& matches, FlannBasedMatcher& descriptorMatcher )
static void matchDescriptors( const Mat& queryDescriptors, const vector<Mat>& trainDescriptors,
vector<DMatch>& matches, FlannBasedMatcher& descriptorMatcher, const vector<Mat>& trainImages, const vector<string>& trainImagesNames )
{
cout << "< Set train descriptors collection in the matcher and match query descriptors to them..." << endl;
descriptorMatcher.add( trainDescriptors );
descriptorMatcher.train();
descriptorMatcher.match( queryDescriptors, matches );
CV_Assert( queryDescriptors.rows == (int)matches.size() || matches.empty() );
cout << "Number of matches: " << matches.size() << endl;
cout << ">" << endl;
for( int i = 0; i < trainDescriptors.size(); i++){
std::vector< std::vector< DMatch> > matches2;
std::vector< DMatch > good_matches;
descriptorMatcher.knnMatch( queryDescriptors, trainDescriptors[i], matches2, 2);
CV_Assert( queryDescriptors.rows == (int)matches2.size() || matches2.empty() );
for (int j = 0; j < matches2.size(); ++j){
const float ratio = 0.8; // As in Lowe's paper; can be tuned
if (matches2[j][0].distance < ratio * matches2[j][1].distance){
good_matches.push_back(matches2[j][0]);
}
}
cout << "currentMatchSize : " << good_matches.size() << endl;
}
}
std::vector<DMatch> Match::vecMatches(Mat * img1, Mat * img2,
Mat &descriptors_object, vector<KeyPoint> &keypoints_object,
vector<KeyPoint> &keypoints_scene)
{
Mat img_object = *img1;
Mat img_scene = *img2;
/*
if( !img_object.data || !img_scene.data )
{ std::cout<< " --(!) Error reading images " << std::endl; return -1; }
*/
//-- Step 1: Detect the keypoints using SURF Detector
int minHessian = 400;
SurfFeatureDetector detector(minHessian);
//std::vector<KeyPoint> keypoints_object, keypoints_scene;
detector.detect(img_object, keypoints_object); //TODO: use the third argument here? from previous match?
detector.detect(img_scene, keypoints_scene);
//-- Step 2: Calculate descriptors (feature vectors)
SurfDescriptorExtractor extractor;
Mat /*descriptors_object,*/ descriptors_scene;
extractor.compute(img_object, keypoints_object, descriptors_object);
extractor.compute(img_scene, keypoints_scene, descriptors_scene);
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector<DMatch> matches;
matcher.match(descriptors_object, descriptors_scene, matches);
return matches;
}
void trainKNN(){
cout << "(1) Loading Training Set ...\n";
//read dictionary
FileStorage fsDict("dictionary.yml", FileStorage::READ);
fsDict["vocabulary"] >> dictionary;
fsDict.release();
//read training data
Mat trainingData;
FileStorage fs("training_set.yml", FileStorage::READ);
fs["training_set"] >> trainingData;
fs.release();
Mat test_features;
cout << "(2) Loading Training Classes\n";
vector<int> trainingClasses = readClass("training_classes.txt");
float trng_class_array[trainingClasses.size()]; //copy contents of vector to float array
std::copy(trainingClasses.begin(), trainingClasses.end(), trng_class_array);
Mat trainingClasses_mat = Mat(1, trainingClasses.size(), CV_32FC1, &trng_class_array);
vector<int> testingClasses = readClass("testing_classes.txt");
float testingClasses_array[testingClasses.size()]; //copy contents of vector to float array
std::copy(testingClasses.begin(), testingClasses.end(), testingClasses_array);
Mat testingClasses_mat = Mat(1, testingClasses.size(), CV_32FC1, &testingClasses_array);
testingClasses_mat = testingClasses_mat.t();
Mat predicted(testingClasses_mat.rows, 1, CV_32F);
int K = 1;
cout << "(3) Initializing knn classifier \n";
cv::KNearest knn(trainingData, trainingClasses_mat.t(), cv::Mat(), false, K);
///==============================================================================================
vector<string> filenames = readFile("test_files.txt", "../CS296/data/testing_images/");
for(int i=0; i<filenames.size();i++){
string file = filenames.at(1);
Mat img1 = imread(file.c_str(), CV_LOAD_IMAGE_GRAYSCALE );
Mat img1_equalized;
cout << "\nProcessing image: " << file << " - " << img1.rows << "x" << img1.cols << " iteration - " << i << endl;
img1_equalized = img1.clone();
equalizeHist( img1, img1_equalized); // histogram equalization
cout << "\n\n(1) Image histogram equalized ... ";
Mat img1_sift = img1_equalized.clone();
Mat sift_descriptors, sift_dest;
vector<KeyPoint> keypoints;
//SiftDescriptorExtractor detector_sift;
//detector_sift.detect(img1_sift, keypoints);
//detector_sift.compute(img1_sift, keypoints,sift_descriptors);
SIFT sift(50,3,0.004);
sift(img1_equalized, img1_equalized, keypoints, sift_descriptors, false);
drawKeypoints(img1_sift, keypoints, sift_dest, Scalar::all(-1), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
cout << "\n\n(2) SIFT Descriptor for image computed:\n";
cout << "Keypoints detected: " << keypoints.size() << endl;
cout << "Feature Vector: " << sift_descriptors.size();
//lbp<unsigned char>(img1_equalized, dest_lbp, radius_lbp, neighbor_lbp); /// Local Binary Pattern
Mat dest_lbp;
orig_lbp<unsigned char>(img1_equalized, dest_lbp);
cout << "\n\n(3) LBP Computed Image: " << dest_lbp.size() <<endl;
Mat dest_lbp_out = dest_lbp.clone();
drawKeypoints(dest_lbp_out, keypoints, dest_lbp_out, Scalar::all(-1), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
//obtain patch values per keypoint (8x8)
Mat patch;
Mat lbp_descriptors;
int patchSize = 8;
Size size = Size(patchSize , patchSize); //square patchSize x patchSize region (8x8 patch)
for(int i=0; i<keypoints.size();i++){
Point2f center(keypoints.at(i).pt.x,keypoints.at(i).pt.y);
getRectSubPix(dest_lbp,size,center,patch);
patch = patch.reshape(0,1); //flatten matrix to 1-D Vector
lbp_descriptors.push_back(patch);
}
cout << "LBP Descriptors computed ... " << endl;
cout << "LBP Vector: [" << lbp_descriptors.rows << " x" << lbp_descriptors.cols << "]" << endl;
lbp_descriptors.convertTo(lbp_descriptors,CV_32FC1);
Mat sift_lbp_features;
hconcat(sift_descriptors, lbp_descriptors, sift_lbp_features);
//==============================================================================
//create a nearest neighbor matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( sift_lbp_features, dictionary, matches );
cout << "Number of matches: " << matches.size() << endl;
float bins[dictionary.rows] = {0.0};
for (int i =0; i<matches.size(); i++){
bins[matches.at(i).trainIdx]= bins[matches.at(i).trainIdx] + 1; //update number of bins
}
Mat norm_bins(1,dictionary.rows,CV_32F,&bins);
normalize( norm_bins, norm_bins, 0, 1, NORM_MINMAX, -1, Mat() );
predicted.at<float>(i,0) = knn.find_nearest(norm_bins, K);
cout << "\n\n Predicted class: " << predicted.at<float>(i,0) << " - " << testingClasses_mat.at<float>(i,0) << endl;
}
//=======================================================================================
// plot_binary(testData, prediction, "Predictions Backpropagation");
//store the vocabulary
FileStorage fs2("predicted.yml", FileStorage::WRITE);
fs2 << "predicted" << predicted;
fs2.release();
/* Mat outImage (img1.rows, img1.cols * 2, CV_8UC1);
Mat c1(outImage, Rect(0, 0, img1.cols, img1.rows));
Mat c2(outImage, Rect(img1.cols, 0, img1.cols, img1.rows));
img1.copyTo(c1);
img1_equalized.copyTo(c2);
imshow("SIFT", sift_dest);
imshow("LBP", dest_lbp_out);
imshow("Input Image", outImage);*/
}
/**
* @function main
*/
int main(int, char **argv)
{
image1 = imread(argv[1], 1);
image2 = imread(argv[2], 1);
rows = image1.rows;
cols = image1.cols;
namedWindow("image1", WINDOW_AUTOSIZE);
imshow("image1", image1);
namedWindow("image2", WINDOW_AUTOSIZE);
imshow("image2", image2);
Mat image1_gray;
Mat image2_gray;
/// Converts an image from one color space to another.
cvtColor(image1, image1_gray, COLOR_BGR2GRAY);
cvtColor(image2, image2_gray, COLOR_BGR2GRAY);
/// Detector parameters
int blockSize = 2;
int apertureSize = 3;
double k = 0.04;
/// Detecting corners
/*
void ocl::cornerHarris(const oclMat& src, oclMat& dst, int blockSize, int ksize, double k, int bordertype=cv::BORDER_DEFAULT)
src – Source image. Only CV_8UC1 and CV_32FC1 images are supported now.
dst – Destination image containing cornerness values. It has the same size as src and CV_32FC1 type.
blockSize – Neighborhood size
ksize – Aperture parameter for the Sobel operator
k – Harris detector free parameter
bordertype – Pixel extrapolation method. Only BORDER_REFLECT101, BORDER_REFLECT, BORDER_CONSTANT and BORDER_REPLICATE are supported now.
*/
Mat image1dst;
Mat image2dst;
image1dst = Mat::zeros(image1.size(), CV_32FC1);
image2dst = Mat::zeros(image2.size(), CV_32FC1);
cornerHarris(image1_gray, image1dst, blockSize, apertureSize, k, BORDER_DEFAULT);
cornerHarris(image2_gray, image2dst, blockSize, apertureSize, k, BORDER_DEFAULT);
int threshHarris = 100;
/// Normalizing
/*
void normalize(InputArray src, OutputArray dst, double alpha=1, double beta=0, int norm_type=NORM_L2, int dtype=-1, InputArray mask=noArray() )
src – input array.
dst – output array of the same size as src .
alpha – norm value to normalize to or the lower range boundary in case of the range normalization.
beta – upper range boundary in case of the range normalization; it is not used for the norm normalization.
normType – normalization type (see the details below).
dtype – when negative, the output array has the same type as src; otherwise, it has the same number of channels as src and the depth =CV_MAT_DEPTH(dtype).
mask – optional operation mask.
*/
Mat image1dst_norm;
Mat image2dst_norm;
normalize(image1dst, image1dst_norm, 0, 255, NORM_MINMAX, CV_32FC1, Mat());
normalize(image2dst, image2dst_norm, 0, 255, NORM_MINMAX, CV_32FC1, Mat());
/*
On each element of the input array, the function convertScaleAbs performs three operations sequentially: scaling, taking an absolute value, conversion to an unsigned 8-bit type:
*/
Mat image1dst_norm_scaled;
Mat image2dst_norm_scaled;
convertScaleAbs(image1dst_norm, image1dst_norm_scaled);
convertScaleAbs(image2dst_norm, image2dst_norm_scaled);
KeyPoint kp;
for (int j = 0; j < image1dst_norm.rows; j++) {
for (int i = 0; i < image1dst_norm.cols; i++) {
if ((int) image1dst_norm.at < float >(j, i) > threshHarris) {
kp.pt.x = (float) j;
kp.pt.y = (float) i;
//necessaire je ne sais pas pk
kp.size = 100.0;
keypoints1.push_back(kp);
}
}
}
for (int j = 0; j < image2dst_norm.rows; j++) {
for (int i = 0; i < image2dst_norm.cols; i++) {
if ((int) image2dst_norm.at < float >(j, i) > threshHarris) {
kp.pt.x = (float) j;
kp.pt.y = (float) i;
//necessaire je ne sais pas pk
kp.size = 100.0;
keypoints2.push_back(kp);
}
}
}
BriefDescriptorExtractor briefDesc(64);
Mat descriptors1, descriptors2;
briefDesc.compute(image1, keypoints1, descriptors1);
briefDesc.compute(image2, keypoints2, descriptors2);
//Ptr<DescriptorMatcher> matcher = new FlannBasedMatcher(DescriptorMatcher::create("Flann"));
FlannBasedMatcher matcher;
Mat descriptorAuxKp1;
Mat descriptorAuxKp2;
vector < int >associateIdx;
for (int i = 0; i < descriptors1.rows; i++) {
//on copie la ligne i du descripteur, qui correspond aux différentes valeurs données par le descripteur pour le Keypoints[i]
descriptors1.row(i).copyTo(descriptorAuxKp1);
//ici on va mettre que les valeurs du descripteur des keypoints de l'image 2 que l'on veut comparer aux keypoints de l'image1 en cours de traitement
descriptorAuxKp2.create(0, 0, CV_8UC1);
//associateIdx va servir à faire la transition entre les indices renvoyés par matches et ceux des Keypoints
associateIdx.erase(associateIdx.begin(), associateIdx.end());
for (int j = 0; j < descriptors2.rows; j++) {
float p1x = keypoints1[i].pt.x;
float p1y = keypoints1[i].pt.y;
float p2x = keypoints2[j].pt.x;
float p2y = keypoints2[j].pt.y;
float distance = sqrt(pow((p1x - p2x), 2) + pow((p1y - p2y), 2));
//parmis les valeurs dans descriptors2 on ne va garder que ceux dont les keypoints associés sont à une distance définie du keypoints en cours, en l'occurence le ieme ici.
if (distance < 10) {
descriptorAuxKp2.push_back(descriptors2.row(j));
associateIdx.push_back(j);
}
}
//ici on ne matche qu'un keypoints de l'image1 avec tous les keypoints gardés de l'image 2
matcher.add(descriptorAuxKp1);
matcher.train();
matcher.match(descriptorAuxKp2, matches);
//on remet à la bonne valeur les attributs de matches
for (int idxMatch = 0; idxMatch < matches.size(); idxMatch++) {
//on a comparer le keypoints i
matches[idxMatch].queryIdx = i;
//avec le keypoints2 j
matches[idxMatch].trainIdx = associateIdx[matches[idxMatch].trainIdx];
}
//on concatene les matches trouvés pour les points précedents avec les nouveaux
matchesWithDist.insert(matchesWithDist.end(), matches.begin(), matches.end());
}
//ici on trie les matchesWithDist par distance des valeurs des descripteurs et non par distance euclidienne
nth_element(matchesWithDist.begin(), matchesWithDist.begin() + 24, matchesWithDist.end());
// initial position
// position of the sorted element
// end position
Mat imageMatches;
Mat matchesMask;
drawMatches(image1, keypoints1, // 1st image and its keypoints
image2, keypoints2, // 2nd image and its keypoints
matchesWithDist, // the matches
imageMatches, // the image produced
Scalar::all(-1), // color of the lines
Scalar(255, 255, 255) //color of the keypoints
);
namedWindow(matches_window, CV_WINDOW_AUTOSIZE);
imshow(matches_window, imageMatches);
imwrite("resultat.png", imageMatches);
/// Create a window and a trackbar
namedWindow(transparency_window, WINDOW_AUTOSIZE);
createTrackbar("Threshold: ", transparency_window, &thresh, max_thresh, interface);
interface(0, 0);
waitKey(0);
return (0);
}
int main(int argc, char** argv)
{
if(argc != 3)
{
return -1;
}
Mat img_object = imread(argv[1],1);
Mat img_scene = imread(argv[2],1);
//String ImageName = (argv[3]);
if(!img_object.data || !img_scene.data)
{
cout << "--(!)Error leyendo imagenes " << endl;
return -1;
}
int minHessian = 400;
SurfFeatureDetector detector(minHessian);
vector<KeyPoint>keypoints_object, keypoints_scene;
detector.detect( img_object, keypoints_object );
detector.detect( img_scene, keypoints_scene );
//-- Step 2: Calculate descriptors (feature vectors)
Mat descriptors_object, descriptors_scene;
SurfDescriptorExtractor extractor;
extractor.compute( img_object, keypoints_object, descriptors_object );
extractor.compute( img_scene, keypoints_scene, descriptors_scene );
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
vector< DMatch > matches;
matcher.match( descriptors_object, descriptors_scene, matches );
//////////////////////////////////////////////////////////////////////////////
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors_object.rows; i++ )
{
double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist );
printf("-- Min dist : %f \n", min_dist );
cout << "Initializing Good Matches" << endl;
//-- Draw only "good" matches (i.e. whose distance is less than 3*min_dist )
vector< DMatch > good_matches;
for( int i = 0; i < descriptors_object.rows; i++ )
{
if( matches[i].distance < 3*min_dist )
{
good_matches.push_back( matches[i]);
}
}
Mat img_matches;
namedWindow("Good Matches & Object detection", 0);
drawMatches( img_object, keypoints_object, img_scene, keypoints_scene,
good_matches, img_matches, Scalar::all(-1), Scalar(0,0,255),
vector<char>() );
//////////////////////////////////////////////////////////////////////////////
cout << "Finding Object " << endl;
//-- Localize the object
vector<Point2f> obj;
vector<Point2f> scene;
for( int i = 0; i < good_matches.size(); i++ )
{
//-- Get the keypoints from the good matches
obj.push_back( keypoints_object[ good_matches[i].queryIdx ].pt );
scene.push_back( keypoints_scene[ good_matches[i].trainIdx ].pt );
}
Mat H = findHomography( obj, scene, CV_LMEDS );
//-- Get the corners from the image_1 ( the object to be "detected" )
vector<Point2f> obj_corners(4);
obj_corners[0] = cvPoint(0,0);
obj_corners[1] = cvPoint( img_object.cols, 0);
obj_corners[2] = cvPoint( img_object.cols, img_object.rows);
obj_corners[3] = cvPoint( 0, img_object.rows);
vector<Point2f> scene_corners(4);
perspectiveTransform( obj_corners, scene_corners, H);
//-- Draw lines between the corners (the mapped object in the scene - image_2 )
line( img_matches, scene_corners[0] + Point2f( img_object.cols, 0),
scene_corners[1] + Point2f( img_object.cols, 0), Scalar(0, 255, 0), 4 );
line( img_matches, scene_corners[1] + Point2f( img_object.cols, 0),
scene_corners[2] + Point2f( img_object.cols, 0), Scalar( 0, 255, 0), 4 );
line( img_matches, scene_corners[2] + Point2f( img_object.cols, 0),
scene_corners[3] + Point2f( img_object.cols, 0), Scalar( 0, 255, 0), 4 );
line( img_matches, scene_corners[3] + Point2f( img_object.cols, 0),
scene_corners[0] + Point2f( img_object.cols, 0), Scalar( 0, 255, 0), 4 );
//-- Show detected matches
imwrite("../data/Example.jpg",img_matches );
imshow( "Good Matches & Object detection", img_matches );
cout << "Write image " << endl;
waitKey(0);
return 0;
}
int main(int argc, char** argv) {
gval_debug_init();
// get options
int showhelp = 0;
unsigned int n_cluster = 0;
int opt;
while ((opt = getopt(argc, argv, "k:h")) != -1) {
switch (opt) {
case 'h':
showhelp = 1;
break;
case 'k':
n_cluster = atoi(optarg);
break;
default:
showhelp = 1;
break;
}
}
if (showhelp || n_cluster <= 0 || argc - optind < 2) {
fprintf(stderr, "Usage: %s -k n_cluster input_dir output\n", argv[0]);
return EXIT_SUCCESS;
}
path p(argv[optind]);
if (!exists(p) || !is_directory(p)) {
fprintf(stderr, "%s is not a directory\n", argv[optind]);
return EXIT_FAILURE;
}
BOWKMeansTrainer bow(n_cluster);
directory_iterator dir_end;
for (directory_iterator i(p); i != dir_end; i++) {
if (i->path().extension() == DESCRIPTORS_EXT) {
FILE* in = fopen(i->path().c_str(), "r");
assert(in);
int counter = 0;
int nempty = 0;
Mat* desc = (Mat*) gval_read_cvmat(in);
while (desc != NULL) {
counter++;
if (!desc->empty()) {
nempty++;
bow.add(desc->clone());
}
gval_free_cvmat(desc);
desc = (Mat*) gval_read_cvmat(in);
}
fclose(in);
fprintf(stderr, "Read from file %s (%d/%d)\n",
i->path().c_str(), nempty, counter);
}
}
fprintf(stderr, "Clustering (%d descriptors, %d clusters)...",
bow.descripotorsCount(), n_cluster);
Mat voc = bow.cluster();
fprintf(stderr, " Done\n");
fprintf(stderr, "Counting document frequency...");
int* dfcounter = (int*) calloc(n_cluster, sizeof(int));
vector<Mat> all = bow.getDescriptors();
FlannBasedMatcher matcher;
matcher.add(vector<Mat>(1, voc));
for (vector<Mat>::const_iterator it = all.begin();
it != all.end(); it++) {
vector<int> ct(n_cluster, 0);
vector<DMatch> matches;
matcher.match(*it, matches);
for (vector<DMatch>::const_iterator jt = matches.begin();
jt != matches.end(); jt++) {
assert(jt->trainIdx >= 0 && jt->trainIdx < n_cluster);
ct[jt->trainIdx] = 1;
}
for (int j = 0; j < n_cluster; j++) {
dfcounter[j] += ct[j];
}
}
int dtotal = all.size();
fprintf(stderr, " Done\n");
FILE* out = fopen(argv[optind + 1], "w");
assert(out);
gval_write_cvmat(&voc, out);
fwrite(dfcounter, sizeof(int), n_cluster, out);
fwrite(&dtotal, sizeof(int), 1, out);
// debug
fprintf(stderr, "total:%d\n", dtotal);
for (int j = 0; j < n_cluster; j++) {
fprintf(stderr, "%d:%d\n", j, dfcounter[j]);
}
fclose(out);
fprintf(stderr, "Written to %s\n", argv[optind + 1]);
free(dfcounter);
return EXIT_SUCCESS;
}
/**
* @function main
* @brief Main function
*/
int main( int argc, char** argv )
{
// if( argc != 3 )
// { readme(); return -1; }
Mat img_1 = imread( "/home/mac/Documents/PROJECT/SIFT/a.JPG", CV_LOAD_IMAGE_GRAYSCALE );
Mat img_2 = imread( "/home/mac/Documents/PROJECT/SIFT/r.JPG", CV_LOAD_IMAGE_GRAYSCALE );
if( !img_1.data || !img_2.data )
{ std::cout<< " --(!) Error reading images " << std::endl; return -1; }
//-- Step 1: Detect the keypoints using SURF Detector
int minHessian = 400;
//SurfFeatureDetector detector( minHessian );
cv::SiftFeatureDetector detector;
std::vector<KeyPoint> keypoints_1, keypoints_2;
detector.detect( img_1, keypoints_1 );
detector.detect( img_2, keypoints_2 );
//-- Step 2: Calculate descriptors (feature vectors)
SurfDescriptorExtractor extractor;
Mat descriptors_1, descriptors_2;
extractor.compute( img_1, keypoints_1, descriptors_1 );
extractor.compute( img_2, keypoints_2, descriptors_2 );
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( descriptors_1, descriptors_2, matches );
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors_1.rows; i++ )
{ double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist );
printf("-- Min dist : %f \n", min_dist );
//-- Draw only "good" matches (i.e. whose distance is less than 2*min_dist,
//-- or a small arbitary value ( 0.02 ) in the event that min_dist is very
//-- small)
//-- PS.- radiusMatch can also be used here.
std::vector< DMatch > good_matches;
for( int i = 0; i < descriptors_1.rows; i++ )
{ if( matches[i].distance <= max(2*min_dist, 0.02) )
{ good_matches.push_back( matches[i]); }
}
//-- Draw only "good" matches
Mat img_matches;
drawMatches( img_1, keypoints_1, img_2, keypoints_2,
good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
//-- Show detected matches
imshow( "Good Matches", img_matches );
for( int i = 0; i < (int)good_matches.size(); i++ )
{ printf( "-- Good Match [%d] Keypoint 1: %d -- Keypoint 2: %d \n", i, good_matches[i].queryIdx, good_matches[i].trainIdx ); }
waitKey(0);
return 0;
}
void imageCallback(const sensor_msgs::ImageConstPtr& msg)
{
if (count1 ==0)
{
cv_bridge::CvImagePtr cv_ptr;
try
{
cv_ptr = cv_bridge::toCvCopy(msg, sensor_msgs::image_encodings::BGR8);
}
catch (cv_bridge::Exception& e)
{
ROS_ERROR("cv_bridge exception: %s", e.what());
return;
}
img_1 = cv_ptr->image;
imageToDraw = img_1;
count1++;
img_tot=img_1;
}
else
{
cv_bridge::CvImagePtr cv_ptr;
try
{
cv_ptr = cv_bridge::toCvCopy(msg, sensor_msgs::image_encodings::BGR8);
}
catch (cv_bridge::Exception& e)
{
ROS_ERROR("cv_bridge exception: %s", e.what());
return;
}
img_2 = cv_ptr->image;
if( !img_1.data || !img_2.data )
{ std::cout<< " --(!) Error reading images " << std::endl;}
// canny detection
//dst.create(img_1.size(), img_1.type() );
namedWindow( window_name, CV_WINDOW_AUTOSIZE );
/// Create a Trackbar for user to enter threshold
createTrackbar( "Min Threshold:", window_name, &lowThreshold, max_lowThreshold, CannyThreshold1 );
/// Show the image
CannyThreshold1(0, 0);
CannyThreshold2(0, 0);
//img_1 = detected_edges_1;
//img_2 = detected_edges_2;
//-- Step 1: Detect the keypoints using SURF Detector
int minHessian = 400;
SurfFeatureDetector detector( minHessian );
std::vector<KeyPoint> keypoints_1, keypoints_2;
detector.detect( img_1, keypoints_1 );
detector.detect( img_2, keypoints_2 );
//-- Step 2: Calculate descriptors (feature vectors)
SurfDescriptorExtractor extractor;
Mat descriptors_1, descriptors_2;
extractor.compute( img_1, keypoints_1, descriptors_1 );
extractor.compute( img_2, keypoints_2, descriptors_2 );
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( descriptors_1, descriptors_2, matches );
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors_1.rows; i++ )
{ double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist );
printf("-- Min dist : %f \n", min_dist );
//-- Draw only "good" matches (i.e. whose distance is less than 2*min_dist,
//-- or a small arbitary value ( 0.02 ) in the event that min_dist is very
//-- small)
//-- PS.- radiusMatch can also be used here.
std::vector< DMatch > good_matches;
std::vector<KeyPoint> good_keypoints_1, good_keypoints_2;
for( int i = 0; i < descriptors_1.rows; i++ )
{ if( matches[i].distance <= max(3*min_dist, 0.0) )
{ good_matches.push_back( matches[i]);
good_keypoints_1.push_back(keypoints_1[i]);
good_keypoints_2.push_back(keypoints_2[i]);
}
}
//-- Draw only "good" matches
Mat img_matches;
drawMatches( img_1, keypoints_1, img_2, keypoints_2,
good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
//-- Show detected matches
imshow( "Good Matches", img_matches );
//float deltax, deltay;
float deltax[(int)good_matches.size()], deltay[(int)good_matches.size()];
float sum_deltax=0, sum_deltay=0;
for( int i = 0; i < (int)good_matches.size(); i++ )
{
// printf( "-- Good Match [%d] Keypoint 1: %d -- Keypoint 2: %d \n", i, good_matches[i].queryIdx, good_matches[i].trainIdx );
// cout << "-- Good Match Keypoint 1:" << keypoints_1[good_matches[i].queryIdx].pt.x << endl;
// cout << "-- Good Match Keypoint 2:" << keypoints_2[good_matches[i].trainIdx].pt.x << endl;
deltax[i] = keypoints_2[good_matches[i].trainIdx].pt.x - keypoints_1[good_matches[i].queryIdx].pt.x;
deltay[i] = keypoints_2[good_matches[i].trainIdx].pt.y - keypoints_1[good_matches[i].queryIdx].pt.y;
sum_deltax=+deltax[i];
sum_deltay=+deltay[i];
}
float av_deltax = sum_deltax/(int)good_matches.size();
float av_deltay = sum_deltay/(int)good_matches.size();
float av_deltax2 = 0, av_deltay2 =0;
cout << "before: av_deltax " << av_deltax << " av_deltay " << av_deltay << endl;
int count2=0;
for( int i = 0; i < (int)good_matches.size(); i++ )
{
if ((abs(deltax[i]-av_deltax) < 50 ) & (abs(deltay[i]-av_deltay)<50))
{
av_deltax2 =+deltax[i];
av_deltay2 =+deltay[i];
count2++;
}
}
if (count2>0)
{av_deltax2 = -av_deltax2/count2;
av_deltay2 = -av_deltay2/count2;
cout << "after: av_deltax " << av_deltax << " av_deltay " << av_deltay << endl;
}
else
{cout << "ATTENZIONE:keeping the old value " << endl;
av_deltax2=0;
av_deltay2=0;}
Pt_new.x=Pt_old.x+av_deltax2;
Pt_new.y=Pt_old.y+av_deltay2;
cout << "Pt_new.x " << Pt_new.x << " Pt_old.x " << Pt_old.x << endl;
line(imageToDraw, Pt_new, Pt_old, Scalar(255, 255, 255), 2);
imshow( "Trajectory", imageToDraw );
//Stitcher stitcher = Stitcher::createDefault();
//Mat rImg;
//vector< Mat > vImg;
//vImg.push_back(img_1);
//vImg.push_back(img_2);
//Stitcher::Status status = stitcher.stitch(vImg, rImg);
//if (Stitcher::OK == status)
//imshow("Stitching Result",rImg);
//else
//printf("Stitching fail.");
std::vector< Point2f > obj;
std::vector< Point2f > scene;
for( int i = 0; i < good_matches.size(); i++ )
{
//-- Get the keypoints from the good matches
obj.push_back( keypoints_1[ good_matches[i].queryIdx ].pt );
scene.push_back( keypoints_2[ good_matches[i].trainIdx ].pt );
}
// Find the Homography Matrix
Mat H = findHomography( obj, scene, CV_RANSAC );
// Use the Homography Matrix to warp the images
cv::Mat result;
warpPerspective(img_1,result,H,cv::Size(img_1.cols+img_tot.cols,img_1.rows));
cv::Mat half(result,cv::Rect(0,0,img_tot.cols,img_tot.rows));
img_tot.copyTo(half);
imshow( "Result", result );
Pt_old = Pt_new;
img_1 = img_2;
waitKey(11);
}
}
//SURF Extraction, with Flann Matcher
void surfAlgo (int, void*)
{
int key=0, i=0;
CvMemStorage *storage = cvCreateMemStorage(0);
CvSeq *imageKeyPoints=0, *imageDescriptors = 0;
vector<KeyPoint> keyPoints_orig, keyPoints_trans;
Mat descriptor_orig, descriptor_trans;
double max_dist=0;double min_dist=100;
std::vector < DMatch > good_matches;
FlannBasedMatcher matcher;
std::vector<DMatch>matches;
SurfFeatureDetector detector(1500);
SurfDescriptorExtractor extractor;
cvtColor (src_2,src_trans,CV_BGR2GRAY);
cvtColor (src_1,src_orig,CV_BGR2GRAY);
detector.detect(src_orig, keyPoints_orig);
detector.detect(src_trans, keyPoints_trans);
extractor.compute(src_orig, keyPoints_orig, descriptor_orig);
extractor.compute(src_trans, keyPoints_trans, descriptor_trans);
matcher.match(descriptor_orig, descriptor_trans, matches);
for (int count=0;count<descriptor_orig.rows;count++)
{
double dist=matches[i].distance;
if( dist < min_dist ) min_dist=dist;
if (dist > max_dist) max_dist = dist;
}
for(int count = 0; count < descriptor_orig.rows; count++)
{
if( matches[count].distance < 3*min_dist )
{
good_matches.push_back( matches[count]);
}
}
//drawKeypoints(src_trans, keyPoints_trans, src_trans, Scalar(0,0,255), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
//drawKeypoints(src_orig, keyPoints_orig, src_orig, Scalar(255,0,0), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
//drawMatches ( src_orig, keyPoints_orig, src_trans, keyPoints_trans, matches, src_trans);
drawMatches( src_orig, keyPoints_orig, src_trans, keyPoints_trans,good_matches,src_new,Scalar(0,0,255),Scalar(0,0,255),vector<char>(),DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS);
std::vector<Point2f> obj;
std::vector<Point2f> scene;
for (int count=0; count<good_matches.size();count++)
{
obj.push_back(keyPoints_trans[ matches[count].queryIdx].pt);
scene.push_back(keyPoints_orig[ matches[count].trainIdx].pt);
}
/*
Code work in progress
Mat H = findHomography( obj, scene, CV_RANSAC );
std::vector<Point2f> object_corners(4);
std::vector<Point2f> scene_corners(4);
object_corners[0] = cvPoint (0,0);
object_corners[1] = cvPoint (src_trans.cols,0);
object_corners[2]= cvPoint (src_trans.cols,src_trans.rows);
object_corners[3] = cvPoint (0,src_trans.rows);
perspectiveTransform( object_corners, scene_corners, H);
line (src_new, scene_corners[0]+Point2f(src_trans.cols,0), scene_corners[1]+Point2f(src_trans.cols,0),Scalar(0,255,0),4);
line (src_new, scene_corners[1]+Point2f(src_trans.cols,0), scene_corners[2]+Point2f(src_trans.cols,0),Scalar(0,255,0),4);
line (src_new, scene_corners[2]+Point2f(src_trans.cols,0), scene_corners[3]+Point2f(src_trans.cols,0),Scalar(0,255,0),4);
line (src_new, scene_corners[3]+Point2f(src_trans.cols,0), scene_corners[0]+Point2f(src_trans.cols,0),Scalar(0,255,0),4);
*/
imwrite("/Users/brainwave/Desktop/testrun1.jpg",src_new);
resize(src_new,src_new,Size(),0.3,0.3,CV_INTER_LINEAR);
imshow(window_name, src_new);
}
int main(int argc, char **argv) {
if (argc != 3) {
printUsage();
return -1;
}
cv::Rect finalRect;
vector<cv::Rect> boundingRects;
vector<Point2f> tlPoints;
vector<Point2f> brPoints;
vidFileName = string(argv[1]);
featFileName = string(argv[2]);
// Get BOINC resolved file paths.
#ifdef _BOINC_APP_
string resolved_vid_path;
string resolved_feat_path;
int retval = boinc_resolve_filename_s(vidFileName.c_str(), resolved_vid_path);
if (retval) {
cerr << "Error, could not open file: '" << vidFileName << "'" << endl;
cerr << "Resolved to: '" << resolved_vid_path << "'" << endl;
return false;
}
vidFileName = resolved_vid_path;
retval = boinc_resolve_filename_s(featFileName.c_str(), resolved_feat_path);
if (retval) {
cerr << "Error, could not open file: '" << featFileName << "'" << endl;
cerr << "Resolved to : '" << resolved_feat_path << "'" << endl;
return false;
}
featFileName = resolved_feat_path;
#endif
CvCapture *capture = cvCaptureFromFile(vidFileName.c_str());
Mat descriptors_file;
FileStorage infile(featFileName, FileStorage::READ);
if (infile.isOpened()) {
read(infile["Descriptors"], descriptors_file);
infile.release();
} else {
cout << "Feature file " << featFileName << " does not exist." << endl;
exit(-1);
}
cout << "Descriptors: " << descriptors_file.size() << endl;
int framePos = cvGetCaptureProperty(capture, CV_CAP_PROP_POS_FRAMES);
int total = cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_COUNT);
double fps = cvGetCaptureProperty(capture, CV_CAP_PROP_FPS);
int framesInThreeMin = (int)fps*180;
#ifdef _BOINC_APP_
boinc_init();
#endif
cerr << "Video File Name: " << vidFileName << endl;
cerr << "Feature File Name: " << featFileName << endl;
cerr << "Frames Per Second: " << fps << endl;
cerr << "Frame Count: " << total << endl;
cerr << "Number of Frames in Three Minutes: " << framesInThreeMin << endl;
// cerr << "<slice_probabilities>" << endl;
checkpoint_filename = "checkpoint.txt";
if(read_checkpoint()) {
if(checkpointVidFileName.compare(vidFileName)!=0 || checkpointFeatFileName.compare(featFileName)!=0) {
cerr << "Checkpointed video or feature filename was not the same as given video or feature filename... Restarting" << endl;
} else {
cerr << "Continuing from checkpoint..." << endl;
}
} else {
cerr << "Unsuccessful checkpoint read" << endl << "Starting from beginning of video" << endl;
}
skipNFrames(capture, percentages.size() * framesInThreeMin);
framePos = cvGetCaptureProperty(capture, CV_CAP_PROP_POS_FRAMES);
cerr << "Starting at Frame: " << framePos << endl;
long start_time = time(NULL);
while ((double)framePos/total < 1.0) {
if (framePos % 10 == 0) {
cout << "FPS: " << framePos/((double)time(NULL) - (double)start_time) << endl;
}
//cout << framePos/total << endl;
Mat frame(cvarrToMat(cvQueryFrame(capture)));
framePos = cvGetCaptureProperty(capture, CV_CAP_PROP_POS_FRAMES);
SurfFeatureDetector detector(minHessian);
vector<KeyPoint> keypoints_frame;
detector.detect(frame, keypoints_frame);
SurfDescriptorExtractor extractor;
Mat descriptors_frame;
extractor.compute(frame, keypoints_frame, descriptors_frame);
cout << "keypoints detected: " << keypoints_frame.size() << endl;
for (int i = 0; i < keypoints_frame.size(); i++) {
cout << "\t" << keypoints_frame[i].pt.x << ", " << keypoints_frame[i].pt.y << " -- " << keypoints_frame[i].angle << " : " << keypoints_frame[i].size << " -- " << keypoints_frame[i].response << endl;
cout << "\t\t(" << descriptors_frame.rows << ", " << descriptors_frame.cols << ") ";
for (int j = 0; j < descriptors_frame.cols; j++) {
cout << " " << descriptors_frame.at<float>(i,j);
}
cout << endl;
}
cout << endl;
// Find Matches
FlannBasedMatcher matcher;
vector<DMatch> matches;
matcher.match(descriptors_frame, descriptors_file, matches);
double max_dist = 0;
double min_dist = 100;
double avg_dist = 0;
for (int i=0; i<matches.size(); i++) {
double dist = matches[i].distance;
if(dist < min_dist) min_dist = dist;
if(dist > max_dist) max_dist = dist;
}
//cout << "Max dist: " << max_dist << endl;
cout << "Min dist: " << min_dist << endl;
vector<DMatch> good_matches;
for (int i=0; i<matches.size(); i++) {
if (matches[i].distance <= 0.18 && matches[i].distance <= 2.0*min_dist) {
good_matches.push_back(matches[i]);
avg_dist += matches[i].distance;
}
}
if (good_matches.size() > 0) {
avg_dist = avg_dist/good_matches.size();
cout << "Avg dist: " << avg_dist << endl;
}
// Localize object.
vector<Point2f> matching_points;
vector<KeyPoint> keypoints_matches;
for (int i=0; i<good_matches.size(); i++) {
keypoints_matches.push_back(keypoints_frame[good_matches[i].queryIdx]);
matching_points.push_back(keypoints_frame[good_matches[i].queryIdx].pt);
}
// Code to draw the points.
Mat frame_points;
#ifdef GUI
drawKeypoints(frame, keypoints_matches, frame_points, Scalar::all(-1), DrawMatchesFlags::DEFAULT);
#endif
//Get bounding rectangle.
if (matching_points.size() == 0) {
Point2f tlFrame(0, 0);
Point2f brFrame(frame.cols, frame.rows);
tlPoints.push_back(tlFrame);
brPoints.push_back(brFrame);
} else {
cv::Rect boundRect = boundingRect(matching_points);
//Calculate mean.
Mat mean;
reduce(matching_points, mean, CV_REDUCE_AVG, 1);
double xMean = mean.at<float>(0,0);
double yMean = mean.at<float>(0,1);
//Calculate standard deviation.
vector<int> xVals;
vector<int> yVals;
for (int i=0; i<matching_points.size(); i++) {
xVals.push_back(matching_points[i].x);
yVals.push_back(matching_points[i].y);
}
double xStdDev = standardDeviation(xVals, xMean);
double yStdDev = standardDeviation(yVals, yMean);
Point2f tlStdPoint(xMean-xStdDev/2, yMean+yStdDev/2);
Point2f brStdPoint(xMean+xStdDev/2, yMean-yStdDev/2);
cv::Rect stdDevRect(tlStdPoint, brStdPoint);
#ifdef GUI
color = Scalar(0, 0, 255); // Blue, Green, Red
rectangle(frame_points, boundRect.tl(), boundRect.br(), color, 2, 8, 0);
color = Scalar(0, 255, 255); // Blue, Green, Red
rectangle(frame_points, stdDevRect.tl(), stdDevRect.br(), color, 2, 8, 0);
#endif
boundingRects.push_back(boundRect);
tlPoints.push_back(boundRect.tl());
brPoints.push_back(boundRect.br());
}
if (tlPoints.size() != 0) {
// Calculate mean rectangle.
Mat tlMean;
Mat brMean;
reduce(tlPoints, tlMean, CV_REDUCE_AVG, 1);
reduce(brPoints, brMean, CV_REDUCE_AVG, 1);
Point2f tlPoint(tlMean.at<float>(0,0), tlMean.at<float>(0,1));
Point2f brPoint(brMean.at<float>(0,0), brMean.at<float>(0,1));
cv::Rect averageRect(tlPoint, brPoint);
// Calculate median rectangle.
vector<int> tlxVals;
vector<int> tlyVals;
vector<int> brxVals;
vector<int> bryVals;
for (int i=0; i<tlPoints.size(); i++) {
tlxVals.push_back(tlPoints[i].x);
tlyVals.push_back(tlPoints[i].y);
brxVals.push_back(brPoints[i].x);
bryVals.push_back(brPoints[i].y);
}
int tlxMedian;
int tlyMedian;
int brxMedian;
int bryMedian;
tlxMedian = quickMedian(tlxVals);
tlyMedian = quickMedian(tlyVals);
brxMedian = quickMedian(brxVals);
bryMedian = quickMedian(bryVals);
//cout << "lt Median: " << tlxMedian << "," << tlyMedian << endl;
//cout << "br Median: " << brxMedian << "," << bryMedian << endl;
Point2i tlMedianPoint(tlxMedian, tlyMedian);
Point2i brMedianPoint(brxMedian, bryMedian);
cv::Rect medianRect(tlMedianPoint, brMedianPoint);
#ifdef GUI
color = Scalar(255, 0, 0); // Blue, Green, Red
rectangle(frame_points, averageRect.tl(), averageRect.br(), color, 2, 8, 0);
color = Scalar(0, 255, 0);
rectangle(frame_points, medianRect.tl(), medianRect.br(), color, 2, 8, 0);
#endif
finalRect = averageRect;
}
// Check for frames in three minutes mark.
framesInThreeMin = 20;
if (framePos != 0 && framePos % framesInThreeMin == 0.0) {
double probability;
if (tlPoints.empty() && brPoints.empty()) {
probability = 0.0;
} else {
double frameDiameter = sqrt(pow((double)frame.cols, 2) * pow((double)frame.rows, 2));
double roiDiameter = sqrt(pow((double)finalRect.width, 2) * pow((double)finalRect.height, 2));
probability = 1-(roiDiameter/(frameDiameter*0.6));
}
if (probability < 0) probability = 0.0;
percentages.push_back(probability);
#ifndef _BOINC_APP_
cout << "Min Dist: " << min_dist << endl;
cout << probability << endl;
#endif
boundingRects.clear();
tlPoints.clear();
brPoints.clear();
}
// Update percent completion and look for checkpointing request.
#ifdef _BOINC_APP_
boinc_fraction_done((double)framePos/total);
if(boinc_time_to_checkpoint()) {
cerr << "checkpointing" << endl;
write_checkpoint();
boinc_checkpoint_completed();
}
#endif
#ifdef GUI
imshow("SURF", frame_points);
if(cvWaitKey(15)==27) break;
#endif
}
cerr << "<slice_probabilities>" << endl;
for (int i=0; i<percentages.size(); i++) cerr << percentages[i] << endl;
cerr << "</slice_probabilities>" << endl;
#ifdef GUI
cvDestroyWindow("SURF");
#endif
cvReleaseCapture(&capture);
#ifdef _BOINC_APP_
boinc_finish(0);
#endif
return 0;
}
/** @function main */
int main( )
{
// Apply SURF method to match images by camera in motion
Mat img_object = imread("/home/hailong/Pictures/house_contour.png");//first input image location
Mat img_scene = imread( "/home/hailong/Pictures/house_simulation_contour.png" );//second input image location
if( !img_object.data || !img_scene.data )
{ std::cout<< " --(!) Error reading images " << std::endl; return -1; }
//Detect the keypoints using SURF Detector
int minhessian = 400;
SurfFeatureDetector detector( minhessian );
std::vector<KeyPoint> keypoints_object, keypoints_scene;
detector.detect( img_object, keypoints_object );
detector.detect( img_scene, keypoints_scene );
//Calculate descriptors (feature vectors)
SurfDescriptorExtractor extractor;
Mat descriptors_object, descriptors_scene;
extractor.compute( img_object, keypoints_object, descriptors_object );
extractor.compute( img_scene, keypoints_scene, descriptors_scene );
//Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( descriptors_object, descriptors_scene, matches );
double max_dist = 0; double min_dist = 100;
//Quick calculation of max and min distances between keypoints
for( int i = 0; i < descriptors_object.rows; i++ )
{ double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist );
printf("-- Min dist : %f \n", min_dist );
// Draw only "good" matches (i.e. whose distance is less than 3*min_dist )
std::vector< DMatch > good_matches;
for( int i = 0; i < descriptors_object.rows; i++ )
{ if( matches[i].distance < 3*min_dist )
{ good_matches.push_back( matches[i]); }
}
Mat img_matches;
drawMatches( img_object, keypoints_object, img_scene, keypoints_scene,
good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS );
//Localize the object
std::vector<Point2f> obj;
std::vector<Point2f> scene;
for( int i = 0; i < good_matches.size(); i++ )
{
//Get the keypoints from the good matches
obj.push_back( keypoints_object[ good_matches[i].queryIdx ].pt );
scene.push_back( keypoints_scene[ good_matches[i].trainIdx ].pt );
}
Mat H = findHomography( obj, scene, CV_RANSAC );
//Get the corners from the image_1 ( the object to be "detected" )
std::vector<Point2f> obj_corners(4);
obj_corners[0] = cvPoint(0,0); obj_corners[1] = cvPoint( img_object.cols, 0 );
obj_corners[2] = cvPoint( img_object.cols, img_object.rows ); obj_corners[3] = cvPoint( 0, img_object.rows );
std::vector<Point2f> scene_corners(4);
perspectiveTransform( obj_corners, scene_corners, H);
//Draw lines between the corners (the mapped object in the scene - image_2 )
line( img_matches, scene_corners[0] + Point2f( img_object.cols, 0), scene_corners[1] + Point2f( img_object.cols, 0), Scalar(0, 255, 0), 4 );
line( img_matches, scene_corners[1] + Point2f( img_object.cols, 0), scene_corners[2] + Point2f( img_object.cols, 0), Scalar( 0, 255, 0), 4 );
line( img_matches, scene_corners[2] + Point2f( img_object.cols, 0), scene_corners[3] + Point2f( img_object.cols, 0), Scalar( 0, 255, 0), 4 );
line( img_matches, scene_corners[3] + Point2f( img_object.cols, 0), scene_corners[0] + Point2f( img_object.cols, 0), Scalar( 0, 255, 0), 4 );
//Show detected matches
imshow( "SURF Match Results", img_matches );
waitKey(0);
return 0;
}
//
// Following an example from
// http:// ramsrigoutham.com/2012/11/22/panorama-image-stitching-in-opencv/
//
void calcHomographyFeature(const Mat& image1, const Mat& image2)
{
static const char* difffeat = "Difference feature registered";
Mat gray_image1;
Mat gray_image2;
// Convert to Grayscale
if(image1.channels() != 1)
cvtColor(image1, gray_image1, CV_RGB2GRAY);
else
image1.copyTo(gray_image1);
if(image2.channels() != 1)
cvtColor(image2, gray_image2, CV_RGB2GRAY);
else
image2.copyTo(gray_image2);
//-- Step 1: Detect the keypoints using SURF Detector
int minHessian = 400;
SurfFeatureDetector detector(minHessian);
std::vector<KeyPoint> keypoints_object, keypoints_scene;
detector.detect(gray_image1, keypoints_object);
detector.detect(gray_image2, keypoints_scene);
//-- Step 2: Calculate descriptors (feature vectors)
SurfDescriptorExtractor extractor;
Mat descriptors_object, descriptors_scene;
extractor.compute(gray_image1, keypoints_object, descriptors_object);
extractor.compute(gray_image2, keypoints_scene, descriptors_scene);
//-- Step 3: Matching descriptor vectors using FLANN matcher
FlannBasedMatcher matcher;
std::vector<DMatch> matches;
matcher.match(descriptors_object, descriptors_scene, matches);
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for(int i = 0; i < descriptors_object.rows; i++)
{
double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
if( dist > max_dist ) max_dist = dist;
}
//-- Use only "good" matches (i.e. whose distance is less than 3*min_dist)
std::vector<DMatch> good_matches;
for(int i = 0; i < descriptors_object.rows; i++) {
if(matches[i].distance < 3*min_dist) {
good_matches.push_back( matches[i]);
}
}
std::vector< Point2f > obj;
std::vector< Point2f > scene;
for(size_t i = 0; i < good_matches.size(); i++)
{
//-- Get the keypoints from the good matches
obj.push_back( keypoints_object[ good_matches[i].queryIdx ].pt );
scene.push_back( keypoints_scene[ good_matches[i].trainIdx ].pt );
}
// Find the Homography Matrix
Mat H = findHomography( obj, scene, CV_RANSAC );
// Use the Homography Matrix to warp the images
Mat result;
Mat Hinv = H.inv();
warpPerspective(image2, result, Hinv, image1.size());
cout << "--- Feature method\n" << H << endl;
Mat imf1, resf;
image1.convertTo(imf1, CV_64FC3);
result.convertTo(resf, CV_64FC3);
showDifference(imf1, resf, difffeat);
}
bool RelicDetect::Match(RelicDetect obj,RelicDetect scn)
{
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match(obj.descriptors, scn.descriptors, matches);
double max_dist = 0; double min_dist = 100;
//-- Quick calculation of max and min distances between keypoints
for (int i = 0; i < obj.descriptors.rows; i++)
{
double dist = matches[i].distance;
if (dist < min_dist) min_dist = dist;
if (dist > max_dist) max_dist = dist;
}
printf("-- Max dist : %f \n", max_dist);
printf("-- Min dist : %f \n", min_dist);
//-- Draw only "good" matches (i.e. whose distance is less than 3*min_dist )
std::vector< DMatch > good_matches;
for (int i = 0; i < obj.descriptors.rows; i++)
{
if (matches[i].distance <= 3 * min_dist)
{
good_matches.push_back(matches[i]);
}
}
max_dist = 0;min_dist = 100;double total_min_dist = 0;
for (int i = 0; i < good_matches.size(); i++)
{
double dist = good_matches[i].distance;
total_min_dist += dist;
if (dist < min_dist) min_dist = dist;
if (dist > max_dist) max_dist = dist;
}
printf("-- good matches Max dist : %f \n", max_dist);
printf("-- good matches Min dist : %f \n", min_dist);
printf("-- good matches total Min dist : %f \n", total_min_dist);
cout << "-- good matches size " << good_matches.size() << endl;
cout << "-- dist per match" << total_min_dist / (double)good_matches.size() << endl;
Mat img_matches;
drawMatches(obj.img_color, obj.keypoints, scn.img_color, scn.keypoints,
good_matches, img_matches, Scalar::all(-1), Scalar::all(-1),
std::vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS);
//imshow("matches", img_matches);
//-- Localize the object
std::vector<Point2f> obj_points;
std::vector<Point2f> scn_points;
for (size_t i = 0; i < good_matches.size(); i++)
{
//-- Get the keypoints from the good matches
obj_points.push_back(obj.keypoints[good_matches[i].queryIdx].pt);
scn_points.push_back(scn.keypoints[good_matches[i].trainIdx].pt);
}
Mat H = cv::findHomography(obj_points, scn_points, RANSAC);
cout << "H:" << endl;
for (int i = 0;i < H.rows;i++)
{
for (int j = 0;j < H.cols;j++)
{
cout << H.at<double>(i, j) << " ";
}
cout << endl;
}
//-- Get the corners from the image_1 ( the object to be "detected" )
std::vector<Point2f> obj_corners(4);
obj_corners[0] = cvPoint(0, 0);
obj_corners[1] = cvPoint(obj.img_color.cols, 0);
obj_corners[2] = cvPoint(obj.img_color.cols, obj.img_color.rows);
obj_corners[3] = cvPoint(0, obj.img_color.rows);
std::vector<Point2f> scene_corners(4);
perspectiveTransform(obj_corners, scene_corners, H);
cout << "object area" << contourArea(obj_corners) << endl;
cout << "scene detected area" << contourArea(scene_corners) << endl;
auto scene_area = contourArea(scene_corners);
//-- Draw lines between the corners (the mapped object in the scene - image_2 )
line(img_matches, scene_corners[0] + Point2f(obj.img_color.cols, 0), scene_corners[1] + Point2f(obj.img_color.cols, 0), Scalar(0, 255, 0), 4);
line(img_matches, scene_corners[1] + Point2f(obj.img_color.cols, 0), scene_corners[2] + Point2f(obj.img_color.cols, 0), Scalar(0, 255, 0), 4);
line(img_matches, scene_corners[2] + Point2f(obj.img_color.cols, 0), scene_corners[3] + Point2f(obj.img_color.cols, 0), Scalar(0, 255, 0), 4);
line(img_matches, scene_corners[3] + Point2f(obj.img_color.cols, 0), scene_corners[0] + Point2f(obj.img_color.cols, 0), Scalar(0, 255, 0), 4);
//-- Show detected matches
imshow("Good Matches & Object detection", img_matches);
waitKey(0);
if (scene_area>1000)
{
return true;
}
else
{
return false;
}
}
void buildTrainingDescriptors(){
Mat training_features;
FileStorage fs("dictionary.yml", FileStorage::READ);
fs["vocabulary"] >> dictionary;
fs.release();
vector<string> filenames = readFile("files.txt", "../CS296/data/training_images/");
for(int i=0; i<filenames.size();i++){
string file = filenames.at(i);
Mat img1 = imread(file.c_str(), CV_LOAD_IMAGE_GRAYSCALE );
Mat img1_equalized;
cout << "\nProcessing image: " << file << " - " << img1.rows << "x" << img1.cols << " iteration - " << i << endl;
img1_equalized = img1.clone();
equalizeHist( img1, img1_equalized); // histogram equalization
cout << "\n\n(1) Image histogram equalized ... ";
Mat img1_sift = img1_equalized.clone();
Mat sift_descriptors, sift_dest;
vector<KeyPoint> keypoints;
//SiftDescriptorExtractor detector_sift;
//detector_sift.detect(img1_sift, keypoints);
//detector_sift.compute(img1_sift, keypoints,sift_descriptors);
SIFT sift(2000,3,0.004);
sift(img1_equalized, img1_equalized, keypoints, sift_descriptors, false);
drawKeypoints(img1_sift, keypoints, sift_dest, Scalar::all(-1), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
cout << "\n\n(2) SIFT Descriptor for image computed:\n";
cout << "Keypoints detected: " << keypoints.size() << endl;
cout << "Feature Vector: " << sift_descriptors.size();
//lbp<unsigned char>(img1_equalized, dest_lbp, radius_lbp, neighbor_lbp); /// Local Binary Pattern
Mat dest_lbp;
orig_lbp<unsigned char>(img1_equalized, dest_lbp);
cout << "\n\n(3) LBP Computed Image: " << dest_lbp.size() <<endl;
Mat dest_lbp_out = dest_lbp.clone();
drawKeypoints(dest_lbp_out, keypoints, dest_lbp_out, Scalar::all(-1), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
//obtain patch values per keypoint (8x8)
Mat patch;
Mat lbp_descriptors;
int patchSize = 8;
Size size = Size(patchSize , patchSize); //square patchSize x patchSize region (8x8 patch)
for(int i=0; i<keypoints.size();i++){
Point2f center(keypoints.at(i).pt.x,keypoints.at(i).pt.y);
getRectSubPix(dest_lbp,size,center,patch);
patch = patch.reshape(0,1); //flatten matrix to 1-D Vector
lbp_descriptors.push_back(patch);
}
cout << "LBP Descriptors computed ... " << endl;
cout << "LBP Vector: [" << lbp_descriptors.rows << " x" << lbp_descriptors.cols << "]" << endl;
lbp_descriptors.convertTo(lbp_descriptors,CV_32FC1);
Mat sift_lbp_features;
hconcat(sift_descriptors, lbp_descriptors, sift_lbp_features);
cout << "\nTotal Image feature vector: [ image " << " - " << sift_lbp_features.rows << " key points x " << sift_lbp_features.cols << " features]" << endl;
// total_features.push_back(sift_lbp_features);
// cout << "Total features: " << total_features.rows << endl << endl;
cout << "===================================================================\n\n";
//prepare BOW descriptor extractor from the dictionary
//create a nearest neighbor matcher
FlannBasedMatcher matcher;
std::vector< DMatch > matches;
matcher.match( sift_lbp_features, dictionary, matches );
cout << "Number of matches: " << matches.size() << endl;
float bins[dictionary.rows] = {0.0};
for (int i =0; i<matches.size(); i++){
bins[matches.at(i).trainIdx]= bins[matches.at(i).trainIdx] + 1; //update number of bins
}
Mat norm_bins(1,dictionary.rows,CV_32F,&bins);
normalize( norm_bins, norm_bins, 0, 1, NORM_MINMAX, -1, Mat() );
training_features.push_back(norm_bins);
}
//store the vocabulary
FileStorage fs2("training_set.yml", FileStorage::WRITE);
fs2 << "training_set" << training_features;
fs2.release();
}