Opencv Image segmentation for licence plate

Question

I working on a project that will read licence plates, my plan is

1. Converting the image to grayscale for better performance
2. Use Histogram equalizer to bring out the licence plate characters
3. Blur the image to remove some noise
4. Use adaptive threshold to binaries the image
5. Use open and close morphology
6. Detect the rectangular bounding box for the licence plate

Well, the issue is: my code is not so good, the result is so bad I cannot detect the rectangle, below is my code:

Imgproc.cvtColor(image, image, Imgproc.COLOR_BGR2GRAY, 0);
    Imgproc.blur(image, image, new Size(3, 3));
    Imgproc.equalizeHist(image, image);
    Mat openElem = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(1, 1));
    Mat closeElem = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(1, 1));
    Imgproc.adaptiveThreshold(image, image, 225, Imgproc.ADAPTIVE_THRESH_GAUSSIAN_C, Imgproc.THRESH_BINARY_INV, 11, 2);

The input image:

The output image:

If anyone with experience would help I will appreciate

Answer 1

In the first is better to detect the plate position in the image, just follow steps below:

1. Convert to gray

   cvCvtColor(image, grayScale, CV_BGR2GRAY);

2. Do sobel

   Mat sobel = new Mat(grayScale.size(), CvType.CV_16S); cvSobel(grayScale, sobel, 2, 0, 7); Mat temp = new Mat(sobel.size(), CvType.CV_8UC1); convertScaleAbs(sobel, temp, 0.00390625, 0);

3. Do threshold

   cvThreshold(sobel, threshold, 0, 255, CV_THRESH_BINARY| CV_THRESH_OTSU);

4. Do morphology

   Mat kernal = cvCreateStructuringElementEx(3,1, 1, 0, CV_SHAPE_RECT); cvDilate(threshold, erode_dilate, kernal, 2);//X vErode(erode_dilate, erode_dilate, kernal, 4);//X cvDilate(erode_dilate, erode_dilate, kernal, 2);//X

   kernal = cvCreateStructuringElementEx(1, 3, 0, 1, CV_SHAPE_RECT); cvErode(erode_dilate, erode_dilate, kernal, 1);// Y cvDilate(erode_dilate, erode_dilate, kernal, 2);

5. Now you are able to detect the rectangle in the image:

6. Then you can process the plate for OCR

Hope it helps you!

Answer 2

Ok so this is what I came up with

Imgproc.cvtColor(image, image, Imgproc.COLOR_BGR2GRAY, 0);
    originalFrameGrayScale = image.clone();

    Mat morph = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(9, 3));
    Imgproc.morphologyEx(image, image, Imgproc.MORPH_TOPHAT, morph);
    Imgproc.Sobel(image, image, -1, 2, 0);
    Imgproc.GaussianBlur(image, image, new Size(5,5), 3,3);
    Imgproc.morphologyEx(image, image, Imgproc.MORPH_CLOSE, morph);
    Imgproc.threshold(image, image, 200, 255, Imgproc.THRESH_OTSU);
    Vector<Rect> rectangles = detectionContour(image);

And find the greatest contour with height, width and area matching certain values enter image description here