Question

我正在尝试使用Android和Opencv制作货币检测应用程序，我正在使用ORB检测器和knnMatcher如下：

detector = FeatureDetector.create(FeatureDetector.ORB);    FastFeatureDetector.create(FastFeatureDetector.THRESHOLD,false,FastFeatureDetector.TYPE_9_16);
  matcher= BFMatcher.create(BFMatcher.BRUTEFORCE_HAMMING,false);
String fullPath = "/storage/emulated/0/Currency Resources Folder/"+"50-front.jpg";
 Mat blured_src = new Mat();
            Mat blured_test = new Mat();
            Mat resized_test = new Mat();
sizeToScale =1 ;
            Imgproc.resize(mGray,resized_test,new Size(mGray.width()*sizeToScale,mGray.height()*sizeToScale));
Imgproc.medianBlur(img_src,blured_src,3);
            Imgproc.medianBlur(resized_test,blured_test,3);

            Mat canny_src = new Mat();
            Mat canny_test = new Mat();
            Imgproc.Canny(blured_src,canny_src,50,200);
            Imgproc.Canny(blured_test,canny_test,50,200);
            MatOfKeyPoint kp_test = new MatOfKeyPoint();
            MatOfKeyPoint kp_src = new MatOfKeyPoint();
            detector.detect(blured_test,kp_test);
            detector.detect(blured_src,kp_src);
Mat desc_src = new Mat();
        Mat desc_test = new Mat();
        descriptorExtractor.compute(blured_src,kp_src,desc_src);
        descriptorExtractor.compute(blured_test,kp_test,desc_test);
List<MatOfDMatch> matches = new ArrayList<MatOfDMatch>();
        matcher.knnMatch(desc_test,desc_src,matches,2);
LinkedList<DMatch> good_matches = new LinkedList<>();

        for (Iterator<MatOfDMatch> iterator = matches.iterator(); iterator.hasNext();)
        {
            MatOfDMatch next = iterator.next();

            if (next.toArray()[0].distance / next.toArray()[1].distance < .92)
            {
                good_matches.add(next.toArray()[0]);
            }
        }
List<Point> pts1 = new ArrayList<Point>();
        List<Point> pts2 = new ArrayList<Point>();
        for(int i = 0; i<good_matches.size(); i++){
            pts1.add(kp_test.toList().get(good_matches.get(i).queryIdx).pt);
            pts2.add(kp_src.toList().get(good_matches.get(i).trainIdx).pt);
        }
 Mat outputMask = new Mat();
        MatOfPoint2f pts1Mat = new MatOfPoint2f();
        pts1Mat.fromList(pts1);
        MatOfPoint2f pts2Mat = new MatOfPoint2f();
        pts2Mat.fromList(pts2);
if (pts1Mat.toList().size() != 0 && pts2Mat.toList().size() != 0) {
            Mat Homog = Calib3d.findHomography(pts1Mat, pts2Mat, Calib3d.RANSAC, 15, outputMask, 2000, 0.995);

LinkedList<DMatch> better_matches = new LinkedList<DMatch>();
            for (int i = 0; i < good_matches.size(); i++) {
                if (outputMask.get(i, 0)[0] != 0.0) {
                    better_matches.add(good_matches.get(i));
                }
            }
MatOfDMatch matches_mat = new MatOfDMatch();
            matches_mat.fromList(better_matches);
MatOfByte mask = new MatOfByte(outputMask);
            Features2d.drawMatches(mGray, kp_test, img_src, kp_src, matches_mat, output,new Scalar(255,0,0),new Scalar(0,255,255),mask,Features2d.NOT_DRAW_SINGLE_POINTS);
            if (Imgcodecs.imwrite("/storage/emulated/0/Currency Resources Folder/match" + counter + ".jpg", output))
//////        if(SaveImage(output,"/storage/emulated/0/Currency Resources Folder/match"+counter+".jpg"))
            {
                Log.e(tag, "Match image saved");
            }
            Log.e(tag, "matches = " + better_matches.size());


            counter++;

return mRgba ;
}

我正面临着这些问题：

匹配列表总是40和60分的正确和错误对象
匹配器检测我的键盘之类的虚假对象并产生一个匹配的图像
我无法识别匹配器是否已检测到货币纸或它只是检测到一个虚假对象（如我的电脑键盘）
我需要在货币周围放置一个边界并检测它每次

out put图像将更好地描述我想解释的内容

Answer 1

您需要修改检测器的阈值，以使其对背景噪声不太敏感。通过丢弃远距离匹配项，可以进一步提高检测器的性能。

ORB功能匹配检测错误图像

1 个答案: