Fix: undefined reference to `cv::imread

[garfeng]

F:/msys64/mingw32/bin/../lib/gcc/i686-w64-mingw32/10.3.0/../../../../i686-w64-mingw32/bin/ld.exe: ../../lib/libcorrespondence.a(nRobustViewMatching.cc.obj):nRobustViewMatching.cc:(.text$_ZN5libmv14correspondence19nRobustViewMatching11computeDataERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE+0x3b): undefined reference to `cv::imread(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, int)'

The sfm uses imread, but opencv_imgcodecs was not added to it's CMakeLists.txt

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
• The PR is proposed to proper branch
• There is reference to original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake

[alalek]

imread usage should be removed instead.
Generic OpenCV modules should not depend on I/O modules.

[garfeng]

Will anyone fix the issue ?

nRobustViewMatching.cc in sfm still uses imread.

File: opencv_contrib-4.5.3\modules\sfm\src\libmv_light\libmv\correspondence\nRobustViewMatching.cc

Line: 53

Use param of cv::Mat instead ?

/**
 * Compute the data and store it in the class map<string,T>
 *
 * \param[in] filename   The file from which the data will be extracted.
 *
 * \return True if success.
 */
bool nRobustViewMatching::computeData(const string & filename)
{
  cv::Mat im_cv = cv::imread(filename, 0);
  if (im_cv.empty()) {
    LOG(FATAL) << "Failed loading image: " << filename;
    return false;
  }
  else
  {
    libmv::vector<libmv::Feature *> features;
    std::vector<cv::KeyPoint> features_cv;
    m_pDetector->detect( im_cv, features_cv );
    features.resize(features_cv.size());
    for(size_t i=0; i<features_cv.size(); ++i)
      features[i] = new libmv::PointFeature(features_cv[i]);

    cv::Mat descriptors;
    m_pDescriber->compute(im_cv, features_cv, descriptors);

    // Copy data.
    m_ViewData.insert( make_pair(filename,FeatureSet()) );
    FeatureSet & KeypointData = m_ViewData[filename];
    KeypointData.features.resize(descriptors.rows);
    for(int i = 0;i < descriptors.rows; ++i)
    {
      KeypointFeature & feat = KeypointData.features[i];
      descriptors.row(i).copyTo(feat.descriptor);
      *(PointFeature*)(&feat) = *(PointFeature*)features[i];
    }

    DeleteElements(&features);

    return true;
  }
}

[alalek]

Use param of cv::Mat instead

Right.
Computer Vision library should not be limited on work with files only.
Usually data comes from cameras or other sources. cv::Mat is a generic type to handle any input.

---

But it is a separate big enough task ("string" is a part of public API for now so it can't be easily changed for now - breaking change).

[alalek]

On the line 114 there is already included opencv_imgcodecs dependency of the module. No need to duplicate that.

More details are required:

• failed command line via make VERBOSE=1 (without any -j option)

[alalek]

Perhaps we need dependency for libmv library instead of opencv sfm module.

Try to modify this line instead:
https://github.com/opencv/opencv_contrib/blob/4.5.3/modules/sfm/CMakeLists.txt#L98