potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 08:13:04 +04:00
Code Packages Releases Wiki Activity

fixed #2580, #2210. some work on #2025.

Browse Source 
modified SIFT to 1) double image before finding keypoints, 2) use floating-point internally instead of 16-bit integers, 3) set the keypoint response to the abs(interpolated_DoG_value). step 1) increases the number of detected keypoints significantly and together with 2) and 3) it improves some detection benchmarks. On the other hand, the stability of the small keypoints is lower, so the rotation and scale invariance tests now struggle a bit. In 2.5 need to make this feature optional and add some more intelligence to the algorithm.

added test that finds a planar object using SIFT.
This commit is contained in:
Vadim Pisarevsky
2013-01-29 19:38:56 +04:00
parent ab8d92e1b8
commit c69312ea0d
5 changed files with 207 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/nonfree/test/test_features2d.cpp
+57 -1
View File
@@ -40,6 +40,7 @@
//M*/
#include "test_precomp.hpp"
#include "opencv2/calib3d/calib3d.hpp"
using namespace std;
using namespace cv;
@@ -1085,4 +1086,59 @@ TEST(Features2d_BruteForceDescriptorMatcher_knnMatch, regression)
Ptr<DescriptorExtractor> s = DescriptorExtractor::create("SURF");
ASSERT_STREQ(s->paramHelp("extended").c_str(), "");
}
*/
*/
class CV_DetectPlanarTest : public cvtest::BaseTest
{
public:
CV_DetectPlanarTest(const string& _fname, int _min_ninliers) : fname(_fname), min_ninliers(_min_ninliers) {}
protected:
void run(int)
{
Ptr<Feature2D> f = Algorithm::create<Feature2D>("Feature2D." + fname);
if(f.empty())
return;
string path = string(ts->get_data_path()) + "detectors_descriptors_evaluation/planar/";
string imgname1 = path + "box.png";
string imgname2 = path + "box_in_scene.png";
Mat img1 = imread(imgname1, 0);
Mat img2 = imread(imgname2, 0);
if( img1.empty() || img2.empty() )
{
ts->printf( cvtest::TS::LOG, "missing %s and/or %s\n", imgname1.c_str(), imgname2.c_str());
ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
return;
}
vector<KeyPoint> kpt1, kpt2;
Mat d1, d2;
f->operator()(img1, Mat(), kpt1, d1);
f->operator()(img1, Mat(), kpt2, d2);
vector<DMatch> matches;
BFMatcher(NORM_L2, true).match(d1, d2, matches);
vector<Point2f> pt1, pt2;
for( size_t i = 0; i < matches.size(); i++ ) {
pt1.push_back(kpt1[matches[i].queryIdx].pt);
pt2.push_back(kpt2[matches[i].trainIdx].pt);
}
Mat inliers, H = findHomography(pt1, pt2, RANSAC, 10, inliers);
int ninliers = countNonZero(inliers);
if( ninliers < min_ninliers )
{
ts->printf( cvtest::TS::LOG, "too little inliers (%d) vs expected %d\n", ninliers, min_ninliers);
ts->set_failed_test_info( cvtest::TS::FAIL_INVALID_TEST_DATA );
return;
}
}
string fname;
int min_ninliers;
};
TEST(Features2d_SIFTHomographyTest, regression) { CV_DetectPlanarTest test("SIFT", 80); test.safe_run(); }
//TEST(Features2d_SURFHomographyTest, regression) { CV_DetectPlanarTest test("SURF", 80); test.safe_run(); }