card_detector.py

import cv2
import numpy as np
import math
from screeninfo import get_monitors

"""
This is the first attempt of identifying MTG cards using only classical computer vision technique.
Most of the processes are similar to the process used in opencv_dnn.py, but it instead tries to use 
Hough transformation to identify straight edges of the card.
However, there were difficulties trying to associate multiple edges into a rectangle, as some of them 
either didn't show up or was too short to intersect.
There were also no method to dynamically adjust various threshold, even finding all the edges were
very conditional.
"""

def detect_a_card(img, thresh_val=80, blur_radius=None, dilate_radius=None, min_hyst=80, max_hyst=200,
                  min_line_length=None, max_line_gap=None, debug=False):
    dim_img = (len(img[0]), len(img)) # (width, height)
    # Intermediate variables

    # Default values
    if blur_radius is None:
        blur_radius = math.floor(min(dim_img) / 100 + 0.5) // 2 * 2 + 1  # Rounded to the nearest odd
    if dilate_radius is None:
        dilate_radius = math.floor(min(dim_img) / 67 + 0.5)
    if min_line_length is None:
        min_line_length = min(dim_img) / 10
    if max_line_gap is None:
        max_line_gap = min(dim_img) / 10

    thresh_radius = math.floor(min(dim_img) / 20 + 0.5) // 2 * 2 + 1  # Rounded to the nearest odd

    img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # Median blur better removes background textures than Gaussian blur
    img_blur = cv2.medianBlur(img_gray, blur_radius)
    # Truncate the bright area while detecting the border
    img_thresh = cv2.adaptiveThreshold(img_blur, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
                                       cv2.THRESH_BINARY_INV, thresh_radius, 20)
    #_, img_thresh = cv2.threshold(img_blur, thresh_val, 255, cv2.THRESH_TRUNC)

    # Dilate the image to emphasize thick borders around the card
    kernel_dilate = np.ones((dilate_radius, dilate_radius), np.uint8)
    #img_dilate = cv2.dilate(img_thresh, kernel_dilate, iterations=1)
    img_dilate = cv2.erode(img_thresh, kernel_dilate, iterations=1)

    img_contour = img_dilate.copy()
    _, contours, _ = cv2.findContours(img_contour, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
    img_contour = cv2.cvtColor(img_contour, cv2.COLOR_GRAY2BGR)
    img_contour = cv2.drawContours(img_contour, contours, -1, (128, 128, 128), 1)
    card_found = contours is not None
    print(len(contours))
    print([len(contour) for contour in contours])

    # find the biggest area
    c = max(contours, key=cv2.contourArea)

    x, y, w, h = cv2.boundingRect(c)
    # draw the book contour (in green)
    img_contour = cv2.drawContours(img_contour, [c], -1, (0, 255, 0), 1)

    # Canny edge - low minimum hysteresis to detect glowed area,
    # and high maximum hysteresis to compensate for high false positives.
    img_canny = cv2.Canny(img_dilate, min_hyst, max_hyst)
    #img_canny = img_dilate
    # Apply Hough transformation to detect the edges
    detected_lines = cv2.HoughLinesP(img_dilate, 1, np.pi / 180, threshold=60,
                                     minLineLength=min_line_length,
                                     maxLineGap=max_line_gap)
    card_found = detected_lines is not None
    print(len(detected_lines))

    if card_found:
        if debug:
            img_hough = cv2.cvtColor(img_dilate.copy(), cv2.COLOR_GRAY2BGR)
            for line in detected_lines:
                x1, y1, x2, y2 = line[0]
                cv2.line(img_hough, (x1, y1), (x2, y2), (0, 0, 255), 1)
    elif not debug:
        print('Hough couldn\'t find any lines')

    # Debug: display intermediate results from various steps
    if debug:
        img_blank = np.zeros((len(img), len(img[0]), 3), np.uint8)
        img_thresh = cv2.cvtColor(img_thresh, cv2.COLOR_GRAY2BGR)
        img_dilate = cv2.cvtColor(img_dilate, cv2.COLOR_GRAY2BGR)
        #img_canny = cv2.cvtColor(img_canny, cv2.COLOR_GRAY2BGR)
        if not card_found:
            img_hough = img_blank

        # Append all images together
        img_row_1 = np.concatenate((img, img_thresh), axis=1)
        img_row_2 = np.concatenate((img_contour, img_hough), axis=1)
        img_result = np.concatenate((img_row_1, img_row_2), axis=0)

        # Resize the final image to fit into the main monitor's resolution
        screen_size = get_monitors()[0]
        resize_ratio = max(len(img_result[0]) / screen_size.width, len(img_result) / screen_size.height, 1)
        img_result = cv2.resize(img_result, (int(len(img_result[0]) // resize_ratio),
                                             int(len(img_result) // resize_ratio)))
        cv2.imshow('Result', img_result)
        cv2.waitKey(0)

    # TODO: output meaningful data
    return card_found

def main():
    img_test = cv2.imread('data/li38_handOfCards.jpg')
    card_found = detect_a_card(img_test,
                               #dilate_radius=5,
                               #thresh_val=100,
                               #min_hyst=40,
                               #max_hyst=160,
                               #min_line_length=50,
                               #max_line_gap=100,
                               debug=True)
    if card_found:
        return
    return
    for dilate_radius in range(1, 6):
        for min_hyst in range(50, 91, 10):
            for max_hyst in range(180, 119, -20):
                print('dilate_radius=%d, min_hyst=%d, max_hyst=%d: ' % (dilate_radius, min_hyst, max_hyst),
                      end='', flush=True)
                card_found = detect_a_card(img_test, dilate_radius=dilate_radius,
                                           min_hyst=min_hyst, max_hyst=max_hyst, debug=True)
                if card_found:
                    print('Card found')
                else:
                    print('Not found')

if __name__ == '__main__':
    main()