Matlab 生成mexwin64文件时遇到的问题

近期下了一个matlab的 人体上半身探测的工具包calvin_upperbody_detector_v1.04，由于其本身是在linux下直接使用的，搬到win7上出现了一点问题。

在使用之前，要将其中的“me_HaarDetectOpenCV.cpp”文件编译为mexw64文件，编译时出现了link问题如下：

[code]me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvCreateImage referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvReleaseImage referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvCreateMemStorage referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvReleaseMemStorage referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvGetSeqElem referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvLoad referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvEqualizeHist referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvReleaseHaarClassifierCascade referenced in function mexFunction 
  me_HaarDetectOpenCV.obj : error LNK2019: unresolved external symbol cvHaarDetectObjects referenced in function mexFunction 
  me_HaarDetectOpenCV.mexw64 : fatal error LNK1120: 9 unresolved externals

首先，由于该cpp使用的是opencv1.0，所以我先将电脑里的opencv2.3.1替换掉，将环境变量和VS2010里的配置也全部换成了opencv1.0，重新mex后还是一样的问题。

这里要提一下的是，我的系统和matlab均为64位版本，不存在不匹配的问题。

然后，我决定放弃opencv1.0，毕竟版本太老，并且与电脑上其他已经写过的程序又不兼容了。换回opencv2.3.1，改回环境变量和VS配置，重启电脑。

最重要的是，将“me_HaarDetectOpenCV.cpp”文件替换掉。替换的代码如下：

#include "mex.h"
#include <string>
#include "opencv2/opencv.hpp"
#include "opencv2/objdetect/objdetect.hpp"
#include "opencv2/imgproc/imgproc.hpp"

void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
    // validate arguments
    if (nrhs < 2) {
        mexErrMsgTxt("Wrong number of input arguments.");
    }
    if (nlhs > 1) {
        mexErrMsgTxt("Too many output arguments.");
    }
    if (!mxIsChar(prhs[0]) || mxGetM(prhs[0])!=1) {
        mexErrMsgTxt("First argument must be a string.");
    }
    if (!mxIsUint8(prhs[1]) || mxGetNumberOfDimensions(prhs[0])!=2) {
        mexErrMsgTxt("Second argument must be a uint8 grayscale image.");
    }

    // get XML cascade file name
    char *xmlfile = mxArrayToString(prhs[0]);
    cv::CascadeClassifier cascade;
    if (!cascade.load(std::string(xmlfile))) {
        mexErrMsgTxt("Failed to load cascade classifier.");
    }
    mxFree(xmlfile);

    // get grayscale image
    mwSize nrows = mxGetM(prhs[1]);
    mwSize ncols = mxGetN(prhs[1]);
    uint8_T *data = reinterpret_cast<uint8_T*>(mxGetData(prhs[1]));

    // copy into an OpenCV mat (there are better ways to do this step!)
    cv::Mat img(nrows, ncols, CV_8UC1, cv::Scalar::all(0));
    for(mwIndex c=0; c<ncols; c++) {
        for(mwIndex r=0; r<nrows; r++) {
            img.at<char>(r,c) = data[r + nrows*c];
        }
    }

    // process image before detection
    cv::equalizeHist(img, img);

    // detect faces
    std::vector<cv::Rect> faces;
    cascade.detectMultiScale(img, faces, 1.1, 4, 0, cv::Size(30,30));

    // return rectangles found to MATLAB
    plhs[0] = mxCreateDoubleMatrix(4, faces.size(), mxREAL);
    double *out = mxGetPr(plhs[0]);
    for(mwIndex i=0; i<faces.size(); i++) {
        out[i+0] = static_cast<double>(faces[i].x);
        out[i+1] = static_cast<double>(faces[i].y);
        out[i+2] = static_cast<double>(faces[i].width);
        out[i+3] = static_cast<double>(faces[i].height);
    }
}

完成了最重要的一步后，紧接着，找到 ”D:\Program Files\MATLAB\R2013b\bin\win64\mexopts\msvc100opts.bat“文件，打开该文件作以下修改：

1.在 set PATH 处添加 D:\Program Files (x86)\opencv\build\x64\vc10\bin;D:\Program Files (x86)\opencv\build\common\tbb\intel64\vc10;

2.在 set INCLUDE 处添加 D:\Program Files (x86)\opencv\build\include;D:\Program Files (x86)\opencv\build\include\opencv;D:\Program Files (x86)\opencv\build\include\opencv2

3.在 set LIB 处添加 D:\Program Files (x86)\opencv\build\x64\vc10\lib;

4.在 set LINKFLAGS 处添加 opencv_calib3d231.lib opencv_contrib231.lib opencv_core231.lib opencv_features2d231.lib opencv_flann231.lib opencv_gpu231.lib opencv_highgui231.lib opencv_imgproc231.lib opencv_legacy231.lib opencv_ml231.lib
opencv_objdetect231.lib opencv_ts231.lib opencv_video231.lib

修改完毕后，保存，重新在matlab中 mex -setup，再次编译需要mex的文件，顺利通过！