fuction that returns most occuring parallel lines in set of lines.

cv_tools.py

def parallel_lines(line_array, min_lines_num=5):
    """ Returns numpy_array with only min_lines_num parallel lines left.
    returns parallel lines with most occurring slopes
    :param line_array: np.array generated via cv2.HoughTransformP()
    :param min_lines_num: amount of lines that will be analyzed, pass odd numbers.
    :return: line array with only parallel lines or array with only first line
    """

    def calc_slope(p1_x, p1_y, p2_x, p2_y):
        def round_down(num, divisor):
            num[num < 0] *= -1
            return num - num % divisor
        delta_y = p2_y - p1_y
        delta_x = p2_x - p1_x
        # masking to prevent zero division error
        delta_x[delta_x == 0] = 1
        return round_down(delta_y / delta_x, 10)

    array_len = len(line_array)
    # if len of incoming array is 1 just return it
    split = min(min_lines_num, array_len if array_len % 2 else array_len - 1)
    if split == 1:
        return line_array
    
    # numpy fancy indexing to get array of all values in column
    x1, y1, x2, y2 = [line_array[:array_len, 0, x] for x in (0, 1, 2, 3)]
    slope = calc_slope(x1, y1, x2, y2)
    slopes = Counter(slope)

    # if slopes are same for every element in array just return array with only first element
    slopes_values = slopes.values()
    if min(slopes_values) == max(slopes_values):
        return line_array[:1]

    # get indexes of most occurring slope and use "numpy magic" to return array based on those indexes
    key = slopes.most_common(1)[0][0]
    item_index = np.where(slope == key)
    return line_array[item_index]