NitinPrasad's profile - activity

Thank you!

I want to study the implementation of BFMatcher to make some modifications.

I have found this file: https://github.com/opencv/opencv/blob..., which contains implementation of ocl_match in BFMatcher, but I am not able to find any implementation for ordinary match. Where can I find the code that executes when BFMatcher.match(...) is called if HAVE_OPENCL is not true?

Thanks!

Where to find implementation of BFMatcher::match I want to study the implementation of BFMatcher to make some modificati

Sorry for the very very late response but yes, I was using OpenCV 3.0 - thanks for the comments! I just fixed my problem.

I'm trying to follow this tutorial:

http://docs.opencv.org/doc/tutorials/...

However I get the error:

undefined reference to `cv::SURF::SURF(double, int, int, bool, bool)'

I looked here

http://stackoverflow.com/questions/16...

but linker is not an option after I try "Project->Properties".

I've also tried using Star Feature Detector instead but this does not return any keypoints.

@Guanta, if you too believe that this is a bug, would there be some better place to file an official ticket?

I'm running into issues that lead me to believe that Hamming matcher is not working as expected (or at least, not as I expect). Consider the following code (please run it if you have the appropriate packages installed):

int main( int argc, char** argv ) {

const int rows = 2;
const int cols = 3;
const int bitl = cols * 8;

BruteForceMatcher<Hamming> matcher;
std::vector<DMatch> matches;

Mat mat1 = cv::Mat::zeros(rows, cols, CV_8U);
std::bitset<bitl>* ptr1 = (std::bitset<bitl>*) (mat1.data + (rows - 1) * mat1.step[0]);
Mat mat2 = cv::Mat::zeros(rows, cols, CV_8U);
std::bitset<bitl>* ptr2 = (std::bitset<bitl>*) (mat2.data + (rows - 1) * mat2.step[0]);

ptr1->set(0, 0);
ptr2->set(0, 1);

std::cout << mat1 << std::endl << std::endl;
  std::cout << mat2 << std::endl << std::endl;

  matcher.match(mat1, mat2, matches);

  for (int i = 0; i < (int) matches.size(); i++ ) {
    std::cout << matches[i].distance << std::endl;
  }

  std::cout << std::endl << std::endl;

  ptr1->set(0, 1);
  ptr2->set(0, 0);

  std::cout << mat1 << std::endl << std::endl;
  std::cout << mat2 << std::endl << std::endl;

  matcher.match(mat1, mat2, matches);

  for (int i = 0; i < (int) matches.size(); i++ ) {
    std::cout << matches[i].distance << std::endl;
  }
}

This gives the following output:

 [0, 0, 0;
  0, 0, 0]

[0, 0, 0;
 1, 0, 0]

0
0


[0, 0, 0;
 1, 0, 0]

[0, 0, 0;
 0, 0, 0]

0
1

which is, of course, nonsensical because the Hamming distance should be symmetric. Is there some subtlety in the way the Hamming matcher works that I am missing? Thank you all

EDIT: The example transferred to current OpenCV 2.4.9-version gives same output

const int rows = 2;
const int cols = 3;

cv::BFMatcher matcher(cv::NORM_HAMMING);
std::vector<cv::DMatch> matches;

cv::Mat mat1 = cv::Mat::zeros(rows, cols, CV_8U);
cv::Mat mat2 = cv::Mat::zeros(rows, cols, CV_8U);

mat2.at<uchar>(1,0) = 1;
std::cout << mat1 << "\n\n" << mat2 << "\n\n";

matcher.match(mat1, mat2, matches);

for (size_t i = 0; i < matches.size(); i++ ) { 
    std::cout << matches[i].distance << std::endl;
}   

std::cout << std::endl << std::endl;

mat2.at<uchar>(1,0) = 0;
mat1.at<uchar>(1,0) = 1;
std::cout << mat1 << "\n\n" << mat2 << "\n\n";

matcher.match(mat1, mat2, matches);

for (size_t i = 0; i < matches.size(); i++ ) { 
    std::cout << matches[i].distance << std::endl;
}

I'm running on a machine (Intel i5) that runs SSE2, but when I run OpenCV code the CV_SSE2 sections never run. Is there an easy way to enable this without re installing?

I'm running Ubuntu 13.04 and Eclipse CDT, but I can change to command line if necessary.

Thanks

I recently found a paper with source code avaiable so I've been trying to get this to run; I believe I installed OpenCV itself correctly but when I try to use CMake to generate the proper makefiles for the code I'm trying to run, I get the following error.

        CMake Error at CMakeLists.txt:15 (find_package):
      Found package configuration file:

        /home/nitin/Documents/Research/SIFT/code/opencv/release/OpenCVConfig.cmake

      but it set OpenCV_FOUND to FALSE so package "OpenCV" is considered to be
      NOT FOUND.

-- Configuring incomplete, errors occurred!

Can someone shed some light on this situation? The only line in that cmake file that changes OpenCV_FOUND is:

if(ANDROID AND OpenCV_ANDROID_NATIVE_API_LEVEL GREATER ANDROID_NATIVE_API_LEVEL)
  ...
  set(OpenCV_FOUND "OpenCV_FOUND-NOTFOUND")
  ...
endif()