为什么 OpenCV 斑点检测不起作用？

Question

我正在将 EmguCV 用于 C#，无论我尝试什么配置，我都无法弄清楚为什么我的斑点检测不起作用。这是我的代码：

public static void Process(string videoPath, string folderSavePath)
{
    // Create video capture with memory safety
    using VideoCapture videoCapture = new(videoPath);

    // Get properties of video
    int fps = (int)videoCapture.Get(CapProp.Fps);
    int frameCount = (int)videoCapture.Get(CapProp.FrameCount);
    Size size = new(videoCapture.Width, videoCapture.Height);
    Size processSize = size / 4;

    // Determine new path for video output
    string videoName = Path.GetFileNameWithoutExtension(videoPath) + "_out" + ".mp4";
    string videoSavePath = Path.Combine(folderSavePath, videoName);

    // Create video writer with memory safety
    using VideoWriter videoWriter = new(videoSavePath, VideoWriter.Fourcc('H', '2', '6', '4'), fps, size, true);
    
    // Set up background subtraction
    BackgroundSubtractorMOG2 backgroundSubtractor = new();
    backgroundSubtractor.History = 30;
    backgroundSubtractor.VarThreshold = 40;

    // Set up blob detection
    SimpleBlobDetectorParams blobDetectorParams = new()
    {
        FilterByArea = true,
        FilterByCircularity = false,
        FilterByConvexity = false,
        FilterByInertia = false,
        MinArea = 1,
        MaxArea = 100_000
    };
    SimpleBlobDetector blobDetector = new(blobDetectorParams);
    int blobs = 0;

    // Read frames from video capture
    Mat[] frames = new Mat[frameCount];
    Mat sourceFrame = new();
    while (videoCapture.Read(sourceFrame))
    {
        // Resize frame
        Mat downsizedFrame = new();
        CvInvoke.Resize(sourceFrame, downsizedFrame, processSize, interpolation: Inter.Linear);

        // Subtract the background
        Mat foregroundFrame = new();
        backgroundSubtractor.Apply(downsizedFrame, foregroundFrame);

        // Remove noise
        Mat noiselessFrame = new();
        CvInvoke.MedianBlur(foregroundFrame, noiselessFrame, 5);

        // Convert to binary video
        Mat binaryFrame = new Mat();
        CvInvoke.Threshold(noiselessFrame, binaryFrame, 127, 255, ThresholdType.Binary);

        // Blob detection
        VectorOfKeyPoint keypoints = new VectorOfKeyPoint();
        blobDetector.Detect(binaryFrame, keypoints);

        // Draw blobs
        Mat blobsFrame = new Mat();
        Features2DToolbox.DrawKeypoints(binaryFrame, keypoints, blobsFrame, new Bgr(0, 255, 0));
        blobs += keypoints.Size;
        
        // Resize again
        Mat upsizedFrame = new();
        CvInvoke.Resize(blobsFrame, upsizedFrame, size, interpolation: Inter.Cubic);

        // Write frame to video
        videoWriter.Write(upsizedFrame);
    }

    Debug.Print($"Detected {blobs} blobs.");
}

令人沮丧的是视频输出清晰的形状。我的背景扣除和噪音消除似乎工作得非常完美。然而，斑点检测仍然失败。

顺便说一句，我知道我正在向二值图像绘制绿色关键点，我相信这实际上可能不会绘制。那可以忽略不计。一旦它工作，我会将它们绘制到我的非二进制框架之一。我一直在使用打印语句来确定是否找到任何斑点，而不是实际的输出视频本身。

Answer 1

原来我只需要使用

blobDetector.DetectRaw(binaryFrame, keypoints);