Open CV - 与图像匹配的视频帧

Question

我正在尝试将视频帧的关键点与我上传的图像进行比较。这是我的代码。 img参数是上传的图片，pixels_color是来自框架的像素，高度和宽度也来自框架。

int Image::match(ofImage img){

    cv::SurfFeatureDetector detector(400);
    vector<cv::KeyPoint> keypoints1, keypoints2;
    cv::Mat img1(height, width, CV_8UC3, pixels_color);
    cv::Mat img2(img.getHeight(), img.getWidth(), CV_8UC3, img.getPixels());

    detector.detect(img1, keypoints1);
    detector.detect(img2, keypoints2);

    cv::SurfDescriptorExtractor extractor;
    cv::Mat descriptors1, descriptors2;
    extractor.compute(img1, keypoints1, descriptors1);
    extractor.compute(img2, keypoints2, descriptors2);

    cv::BruteForceMatcher<cv::L2<float>> matcher;
    vector<cv::DMatch> matches;
    matcher.match(descriptors1, descriptors2, matches);

    return matches.size();
}

问题是返回始终是keypoints1的大小。我不知道为什么，除非我的keypoints2的大小为0，否则返回值将始终是keypoints1的大小。

如果ofImage img与框架有关，则无关紧要，它将始终返回keypoints1的大小。这是一个例子：

Answer 1

我仍然不明白发生了什么，所以我采用了不同的方法，我使用FlannBasedMatcher进行规范化，允许我实现我想要的目标：

int Image::match(ofImage img){

    bool result = false;
    float neares_neighbor_distance_ratio = 0.7f;

    cv::SurfFeatureDetector detector(400);
    vector<cv::KeyPoint> keypoints1, keypoints2;
    cv::SurfDescriptorExtractor extractor;
    cv::Mat descriptors1, descriptors2;

    cv::Mat img1(height, width, CV_8UC3, pixels_color);
    cv::Mat img2(img.getHeight(), img.getWidth(), CV_8UC3, img.getPixels());

    detector.detect(img1, keypoints1);
    detector.detect(img2, keypoints2);

    extractor.compute(img1, keypoints1, descriptors1);
    extractor.compute(img2, keypoints2, descriptors2);

    cv::FlannBasedMatcher matcher;
    vector<vector<cv::DMatch>> matches; 
    matcher.knnMatch(descriptors1, descriptors2, matches, 2);

    vector<cv::DMatch> good_matches;
    good_matches.reserve(matches.size());  

    for(size_t i = 0; i < matches.size(); ++i)
    { 
        if(matches[i].size() < 2)
            continue;

        const cv::DMatch &m1 = matches[i][0];
        const cv::DMatch &m2 = matches[i][1];

        if(m1.distance <= neares_neighbor_distance_ratio * m2.distance)        
            good_matches.push_back(m1);     
    }

    return good_matches.size();
}

这个方法有点大，但我觉得它很有用。