java - OpenCV FeatureDetector 从文件读取后似乎没有设置各种属性

Question

以下是使用 Java 和 OpenCV 3.0 检测彩色皮肤图像中的 Blob /痣的代码

private void initLibService(){

    String opencvpath = System.getProperty("user.dir") + File.separator + "lib" + File.separator;
    System.load(opencvpath + Core.NATIVE_LIBRARY_NAME + ".dll");

    blobDetector = FeatureDetector.create(FeatureDetector.SIMPLEBLOB);

    try {
        File temp = File.createTempFile("tempFile",".tmp");
        String settings="%YAML:1.0\nthresholdStep: " + Integer.parseInt(ResourceUtil.getProperty("MOLE_THRESHOLD_STEPS")) + "\nminThreshold: "+ Integer.parseInt(ResourceUtil.getProperty("MOLE_UPPER_INTENSITY_THRESOLD")) + "\n";
        System.out.println("\nFeature detector setting data: " + settings + "\n\n");

        FileWriter writer=new FileWriter(temp,false);
        writer.write(settings);
        writer.close();

        blobDetector.read(temp.getPath());

        temp.deleteOnExit();
      }
     catch (  IOException e) {
        e.printStackTrace();
      }
    }

public static Mat detectBlob(String imagePath, FeatureDetector blobDetector) {

    // Read image
    Mat im = Imgcodecs.imread(imagePath);

    MatOfKeyPoint keypoints = new MatOfKeyPoint();

    blobDetector.detect(im, keypoints);
    System.out.println("keypoints: " + keypoints.toList());

    Mat im_with_keypoints = new Mat();
    Features2d.drawKeypoints(im, keypoints, im_with_keypoints, new Scalar(0, 255, 0), Features2d.DRAW_RICH_KEYPOINTS);

    return im_with_keypoints;
}

如果我注释行 blobDetector.read(temp.getPath()); ，则附加的皮肤图像将在一些痣周围用圆圈标记(可能正在使用 featureDetector 的某些默认属性值) ，但如果我不这样做，图像上就不会标记任何内容。

Answer 1

您需要指定文件中的所有属性，否则检测器将无法正确加载。

您的文件应如下所示(默认值):

%YAML:1.0
thresholdStep: 10.
minThreshold: 50.
maxThreshold: 220.
minRepeatability: 2
minDistBetweenBlobs: 10.
filterByColor: 1
blobColor: 0
filterByArea: 1
minArea: 25.
maxArea: 5000.
filterByCircularity: 0
minCircularity: 8.0000001192092896e-001
maxCircularity: 3.4028234663852886e+038
filterByInertia: 1
minInertiaRatio: 1.0000000149011612e-001
maxInertiaRatio: 3.4028234663852886e+038
filterByConvexity: 1
minConvexity: 9.4999998807907104e-001
maxConvexity: 3.4028234663852886e+038