特征匹配（提取和匹配描述符FeatureDetector&DescriptorExtractor）

void featureMatch() {

    cv::TickMeter meter;

    meter.start();

    Mat image1 = imread(MYPICTUREPATH "lena.jpg");

    Mat image2 = imread(MYPICTUREPATH "lena_face.jpg");

    //Mat image1 = imread(MYPICTUREPATH "test.jpg");

    //Mat image2 = imread(MYPICTUREPATH "test_templ.jpg");

    //! 1.检测关键点

    Ptr<AKAZE> akaze=AKAZE::create();

    vector<KeyPoint> keypoints1, keypoints2;

    akaze->detect(image1, keypoints1);

    akaze->detect(image2, keypoints2);

    qDebug() << "1.检测关键点";

    //! 2.计算/提取描述符

    Mat descriptor1, descriptor2;

    akaze->compute(image1, keypoints1, descriptor1);

    akaze->compute(image2, keypoints2, descriptor2);

    qDebug() << "2.计算/提取描述符";

    //! 3.描述符匹配

    //! 注意：若create()的实参选择不正确，将报出如下错误：

    //! OpenCV: terminate handler is called! The last OpenCV error is:

    //! OpenCV(4.6.0) Error: Unsupported format or combination of formats (> type=0

    //! > ) in buildIndex_, file D:\Qt-OpenCV\OpenCV-Source\modules\flann\src\miniflann.cpp, line 336

    Ptr<DescriptorMatcher> descMatcher = DescriptorMatcher::create(DescriptorMatcher::FLANNBASED);

    //Ptr<DescriptorMatcher> descMatcher = DescriptorMatcher::create(DescriptorMatcher::BRUTEFORCE);

    qDebug() << "3.描述符匹配";

    //! 4.调用match函数

    vector<DMatch> matches;

    descMatcher->match(descriptor1, descriptor2, matches);

    qDebug() << "4.调用match函数";

    //! 5.drawMatches可以用来自动创建用于显示的适当输出结果

    Mat dispImg;

    drawMatches(image1, keypoints1, image2, keypoints2, matches, dispImg);

    /*

    drawMatches(image1,

                keypoints1,

                image2,

                keypoints2,

                matches,

                dispImg,

                Scalar(0,255,0), // green for matched

                Scalar::all(-1), // unmatched color (default)

                vector<char>(),  // empty mask

                DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS);

    */

    qDebug() << "5.drawMatches可以用来自动创建用于显示的适当输出结果";

    imshow("Image-match", dispImg);　　

    //! 6.滤除不想要的匹配，以获得更好匹配

    vector<DMatch> goodMatches;

    double matchThresh = 0.01;//匹配阈值是通过试错法找到的

    for (int i = 0; i < descriptor1.rows; ++i) {

        if(matches[i].distance < matchThresh)

            goodMatches.push_back(matches[i]);

    }

    drawMatches(image1, keypoints1, image2, keypoints2, goodMatches, dispImg);

    imshow("Image-good-match", dispImg);

    qDebug() << goodMatches.size();

    qDebug() << "6.滤除不想要的匹配，以获得更好匹配";

    //! 7.根据好匹配对关键点进行筛选，

    //! 然后将这些筛选的关键点送入findHomography函数，

    //! 以获得所需的变换矩阵或者单应性变化

    vector<Point2f> goodP1, goodP2;

    for (size_t i = 0; i < goodMatches.size(); ++i) {

        goodP1.push_back(keypoints1[goodMatches[i].queryIdx].pt);

        goodP2.push_back(keypoints2[goodMatches[i].trainIdx].pt);

    }

    Mat homoChange = findHomography(goodP1, goodP2);

    //! 使用刚刚找到的单应性变化矩阵对匹配点应用透视变换。

    //! 要做到这一点需要构造对应于第一个图像的四个角点，然后应用该变换。

    //! 最后简单绘制连接四个结果点的四条线。

    vector<Point2f> corners1(4), corners2(4);

    corners1[0] = Point2f(0, 0);

    corners1[1] = Point2f(image1.cols - 1, 0);

    corners1[2] = Point2f(image1.cols - 1, image1.rows - 1);

    corners1[3] = Point2f(0, image1.rows - 1);

    perspectiveTransform(corners1, corners2, homoChange);

    image2.copyTo(dispImg);

    line(dispImg, corners2[0], corners2[1], Scalar::all(255), 2);

    line(dispImg, corners2[1], corners2[2], Scalar::all(255), 2);

    line(dispImg, corners2[2], corners2[3], Scalar::all(255), 2);

    line(dispImg, corners2[3], corners2[0], Scalar::all(255), 2);　　

    meter.stop();

    qDebug() << "耗时：" << meter.getTimeMilli() << "毫秒";//

}