Hough Space Transformation for Car License Plate Detection

Hough Space Transformation implemented from scratch.

• EdgeDetector.py includes

  • canny(): edge detection method
  • contour(): extracts possible coordinates for candidates areas
  • reduce_noise(): removes unnecessary edges

• Hough.py includes the methods related to creating the accumulator matrix, finding local maximas, finding intersections and drawing the lines.

  • hough_space(): creates hough space
  • find_local_maximas(): extracts local extremums on hough space
  • extract_hough_lines(): de-houghes the polar coordinates and creates lines
  • intersection(): finds intersection point of lines
  • segmented_intersections(): searches possibility for intersections between vertical and horizontal lines
  • mark_intersections(): marks intersection points
  • extract_vertical_horizontal(): groups vertical and horizontal lines in hough space according to theta
  • choose_plate_corners(): eliminates candidate intersection points and find bottom left and top right
  • calculate_iou(): calculate iou between prediction and ground truth
  • draw(): draws identified lines on the image

• visualize.py includes operations for reading images and visualizing the results.

  • extract_metadata(): extracts ground truth metadata from .xml files
  • execute(): calls required commands and plots the results

Execution

python run.py  'path_to_root_folder' 'path_to_image_folder' 'path_to_annotation_folder'