Wei's profile - activity

It seems to be able to perform faster.

I use calcOpticalFlowPyrLK() to track keypoint in feature ,but the calcOpticalFlowPyrLK() is unstable. How optimize it?
TrackOpticalFlow function:

static vector<KeyPoint> currPoints,predictPoints;
void trackingOpticalFlow(const cv::Mat sceneMat, const vector<KeyPoint> sceneKeyPoint, vector<KeyPoint> *predictKeyPoint){
    if( sceneMat.empty() || sceneKeyPoint.empty()) return;
    vector<uchar> status;
    vector<float> err;
    vector<Point2f> pt1,pt2;
    vector<Point2f> transPt1,transPt2;
    cv::Mat sceneMat_gray;

    cv::Size winSize = cv::Size(9, 9);
    int maxLevel = 21;
    cv::TermCriteria terminationCriteria = cv::TermCriteria( CV_TERMCRIT_ITER + CV_TERMCRIT_EPS, 20, 0.03);

    this->getGrayMat(sceneMat, sceneMat_gray);

    if(m_trackPrev.empty()){
        sceneMat_gray.copyTo(m_trackPrev);
        return;
    }

    if(prevPyr.empty()){
        cv::buildOpticalFlowPyramid(m_trackPrev, prevPyr, winSize, maxLevel, true);
    }

    KeyPoint::convert(sceneKeyPoint, pt1);
    transPt1.clear();
    transPt1 = pt1;

    cv::buildOpticalFlowPyramid(sceneMat_gray, nextPyr, winSize, maxLevel, true);
    cv::calcOpticalFlowPyrLK(prevPyr, nextPyr, transPt1, transPt2, status, err, winSize, maxLevel, terminationCriteria, CV_LKFLOW_PYR_B_READY, 1e-10);

    std::vector<cv::Point2f> trackedPts;
    for (size_t i=0; i<status.size(); i++)
    {
        if (status[i] && transPt1.size()>i && norm(transPt2[i] - transPt1[i]) <= 100)
        {
            trackedPts.push_back(transPt2[i]);
        }
    }

    pt2.clear();
    pt2 = transPt2;


    KeyPoint::convert(pt2, *predictKeyPoint);
    swap(m_trackPrev, sceneMat_gray);
    prevPyr.swap(nextPyr);
}

Input mat in video:

void runTrack(cv::Mat sceneMat){    
   cv::Mat sceneMatGray;
   Mat sceneDescriptor;
   this->getGrayMat(sceneMat, sceneMatGray);
   if(!m_isTrack){
       this->detectFeatures(sceneMatGray, &m_trackPreKeypoints, &sceneDescriptor);

       if(m_trackPreKeypoints.empty() || sceneDescriptor.empty()) {
            m_isDetectObject = false;
            return;
       }else{
            this->matchByKnn(m_patternDescriptor, sceneDescriptor, &m_trackMatches);
       }

       if( m_trackMatches.size() < 4){
            m_isDetectObject = false;
            return;
       }

   }else{

       vector<KeyPoint> outKeypoints;
       outKeypoints.clear();
       this->trackingOpticalFlow(sceneMatGray, m_trackPreKeypoints, &outKeypoints);
       if(outKeypoints.empty()) return;


       for(int i=0;i< m_trackPreKeypoints.size();i++)
             cv::circle(sceneMat, m_trackPreKeypoints[i].pt, 2, CV_RGB(255, 0, 0));

       for(int i=0;i<outKeypoints.size();i++)
              cv::circle(sceneMat, outKeypoints[i].pt, 2, CV_RGB(0, 0, 255));

        m_trackPreKeypoints.swap(outKeypoints);

   }

   cv::Mat homography;
   this->getHomography(m_patternKeypoints, m_trackPreKeypoints, m_trackMatches, &homography);
}

I build framework for OpenCV with opencv_contrib module. OpenCV 3.1 Xcode 7.2.1

I refer http://answers.opencv.org/question/40... . but link is not effective. try:

python build_framework.py --contrib opencv_contrib ios

The error:

** BUILD FAILED **


The following build commands failed:
    CompileC platforms/ios/ios/build/armv7-iPhoneOS/modules/xfeatures2d/OpenCV.build/Release-iphoneos/opencv_xfeatures2d.build/Objects-normal/armv7/sift.o /Users/admin/Desktop/OpenCV_for_iOS/opencv_contrib/modules/xfeatures2d/src/sift.cpp normal armv7 c++ com.apple.compilers.llvm.clang.1_0.compiler
(1 failure)
ERROR: Command '['xcodebuild', 'IPHONEOS_DEPLOYMENT_TARGET=6.0', 'ARCHS=armv7', '-sdk', 'iphoneos', '-configuration', 'Release', '-parallelizeTargets', '-jobs', '4', '-target', 'ALL_BUILD', 'build']' returned non-zero exit status 65
Traceback (most recent call last):
  File "build_framework.py", line 87, in build
    self._build(outdir)
  File "build_framework.py", line 81, in _build
    self.buildOne(t[0], t[1], mainBD, cmake_flags)
  File "build_framework.py", line 139, in buildOne
    execute(buildcmd + ["-target", "ALL_BUILD", "build"], cwd = builddir)
  File "build_framework.py", line 34, in execute
    retcode = check_call(cmd, cwd = cwd)
  File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/subprocess.py", line 540, in check_call
    raise CalledProcessError(retcode, cmd)
CalledProcessError: Command '['xcodebuild', 'IPHONEOS_DEPLOYMENT_TARGET=6.0', 'ARCHS=armv7', '-sdk', 'iphoneos', '-configuration', 'Release', '-parallelizeTargets', '-jobs', '4', '-target', 'ALL_BUILD', 'build']' returned non-zero exit status 65

The FREAK not find in opencv2/features2D.

OpenCV 3.1 release message for iOS: The iOS framework (opencv2.framework) can now be configured to include both opencv and opencv_contrib functionality.

How to use opencv_contrib ?

Thank. The error is disappeared. But other error: ||=== Build: all in OpenCV (compiler: GNU GCC Compiler) ===| C:\opencv\sources\modules\core\src\system.cpp||In function 'BOOL cv::DllMain(HINSTANCE, DWORD, LPVOID)':| C:\opencv\sources\modules\core\src\system.cpp|1013|warning: no previous declaration for 'BOOL cv::DllMain(HINSTANCE, DWORD, LPVOID)' [-Wmissing-declarations]| C:\opencv\sources\modules\videoio\src\cap_dshow.cpp|122|error: 'struct IEnumPIDMap' has virtual functions and accessible non-virtual destructor [-Werror=non-virtual-dtor]| C:\opencv\sources\modules\videoio\src\cap_dshow.cpp|139|error: 'struct IMPEG2PIDMap' has virtual functions and accessible non-virtual destructor [-Werror=non-virtual-dtor]|

My system is Window 7 64 bit.I want to install in CodeBlocks. However, I try this http://jatb181.blogspot.tw/2014/06/in... ,but I use Cmake to build Opencv3.0 fail. It show not find -lRunTmChk and ||=== Build: all in OpenCV (compiler: GNU GCC Compiler) ===| C:\opencv\sources\modules\core\src\matrix.cpp||In function 'bool cv::ocl_reduce(cv::InputArray, cv::OutputArray, int, int, int, int, int)':| C:\opencv\sources\modules\core\src\matrix.cpp|3536|warning: comparison between signed and unsigned integer expressions [-Wsign-compare]| C:\opencv\sources\modules\core\src\system.cpp||In function 'BOOL cv::DllMain(HINSTANCE, DWORD, LPVOID)':| C:\opencv\sources\modules\core\src\system.cpp|1013|warning: no previous declaration for 'BOOL cv::DllMain(HINSTANCE, DWORD, LPVOID)' [-Wmissing-declarations]| ld.exe||cannot find -lRunTmChk| modules\core\CMakeFiles\opencv_core.dir\build.make|1286|recipe for target 'bin/libopencv_core300.dll' failed| CMakeFiles\Makefile2|1654|recipe for target 'modules/core/CMakeFiles/opencv_core.dir/all' failed| C:\opencvdll\Makefile|148|recipe for target 'all' failed| ||=== Build failed: 4 error(s), 2 warning(s) (5 minute(s), 31 second(s)) ===|

image: