abdulfatir / pyanpd

Automatic Number Plate Detection for Python using OpenCV

pyANPD

pyANPD is a high recall automatic number plate detector based on this Stack Overflow answer. For details of the algorithm, check the answer on Stack Overflow. The detector is robust to orientation. The precision decreases and recall increases if the edge_density threshold is decreased.

Dependencies

1. OpenCV
2. Numpy

Usage

python pyANPR.py <image_file_path>

This outputs an image with the same name with -detected appended.

How to tweak it for your dataset?

Change aspect_ratio_range (Range of Aspect Ratio for Accepted Rectangles), area_range (Range of Area for Accepted Rectangles), and se_shape (Shape of Structuring Element for Morphological Closing) to the taste of your own dataset.

Results

pyanpd's Issues

TypeError: Expected cv::UMat for argument 'src'

Failure with OpenCV 4

Using opencv-python-4.1.0.25, the following error is generated:

$ python pyANPD.py image.jpg 
Traceback (most recent call last):
  File "pyANPD.py", line 152, in <module>
    o1 = process_image(path, 0, type='rect')
  File "pyANPD.py", line 79, in process_image
    _,contours,_=cv2.findContours(gray, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
ValueError: need more than 2 values to unpack

This does not happen with OpenCV 3.4.5.20, i.e. use
pip install opencv-python==3.4.5.20

TypeError: Expected cv::UMat for argument 'src'

Traceback (most recent call last):
File "D:\Computer Vision\pyANPD-master\pyANPD.py", line 152, in
o1 = process_image(path, 0, type='rect')
File "D:\Computer Vision\pyANPD-master\pyANPD.py", line 73, in process_image
gray = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY)
TypeError: Expected cv::UMat for argument 'src'

TypeError: Expected cv::UMat for argument 'src'

Add -blur option

Would be great if the library had a direct -blur option, i.e. rather than just highlight the selected regions (which works well), it blurred them instead.

src data type = 17 is not supported

TypeError Traceback (most recent call last)
in ()
1 t1 = time.time()
----> 2 o1 = process_image(path, 0, type='rect')
3 #cv2.imwrite('%s-detected.png' % path[:path.rfind('.')], o1)
4
5 print ('Time taken: %d ms'%((time.time()-t1)*1000))

in process_image(name, debug, **options)
7 raw_image = cv2.imread(name,1)
8 input_image = np.copy(raw_image)
----> 9 gray = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY)
10 gray = enhance(gray)
11 gray = cv2.GaussianBlur(gray, (5,5), 0)

I am working to extract name plate using this code using 1-detection.png image but getting this error .
please verify and send the results

problem

HI!
How can I deal with this problem?
I use the report picture to run the code
but it have the problem!

Display All contours

Hello!

How to display all possible contours?
i.e minAreaRect()s are shown in orange and the one which satisfies our criteria is in green.