Where to get the matching information of apap matching points?

[NiaBie]

I'm trying to get the information of matching points and their matches, just like the sample image in the paper.
I have tried using pairwise_matches but the matching seemed incorrectly:

    Mat img1 = images_data[m1].img;
    Mat img2 = images_data[m2].img;
    Mat result = Mat::zeros(max(img1.rows, img2.rows),
                            img1.cols + img2.cols,
                            CV_8UC3);
    // 分割矩阵
    Mat left (result, Rect(0, 0, img1.cols, img1.rows));
    Mat right(result, Rect(img1.cols, 0, img2.cols, img2.rows));
    // 复制矩阵
    img1.copyTo(left);
    img2.copyTo(right);
    for (int i = 0; i < pairwise_matches[pm_index].matches.size(); i ++) {
      int src  = pairwise_matches[pm_index].matches[i].queryIdx;
      int dest = pairwise_matches[pm_index].matches[i].trainIdx;

      Point2 src_p  = images_data[m1].mesh_2d->getVertices()[src];
      Point2 dest_p = images_data[m2].mesh_2d->getVertices()[dest];
      Scalar color(rand() % 256, rand() % 256, rand() % 256);
      circle(result, src_p, 3, color, -1);
      line(result, src_p, dest_p + Point2(img1.cols, 0), color, 1, LINE_AA);
      circle(result, dest_p + Point2(img1.cols, 0), 3, color, -1);
    }
    imwrite(parameter.debug_dir + "matching_points" + to_string(i) + ".png", result);

Could you tell me the right way to do this? Besides, how to display only the good matches(how to judge the quality of a match)?