Identifying contour area of vehicle registration inside a car

I'm trying to identify the contour area of a german vehicle registration. I followed some tutorials and had the idea to use cv2 findContours in combination with HoughLinesP . The challenge is that the registration is held by human that sits inside a car.

1. The hands breaks the contour of the registration
2. The car has other contours like the registration e.g. a car radio
3. The registration reaches over the boundary of the image.

When I crop the image based on the line that i created I get something from the radio instead of the registration. Because the lines of the registration are not connected.

I would would really appreciate if someone could point me into the right direction.

original image:

Canny:

image with lines:

croped area:

def crop_image(dilated):
    binary = cv2.threshold(dilated, 150, 255, cv2.THRESH_BINARY)[1]
    cv2.imshow("binary", binary)

    edges = cv2.Canny(binary, 150, 84, apertureSize=3)
    cv2.imshow("canny", edges)
    dilted_lines = dilated
    lines = cv2.HoughLinesP(edges, 1, 3.14 / 180, 100, minLineLength=15, maxLineGap=500)
    for line in lines:
        l = line[0]
        cv2.line(dilted_lines, (l[0], l[1]), (l[2], l[3]),
                 (0, 255, 0), 4)

    cv2.imshow("lines", dilted_lines)

    contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    sortedCnt = sorted(contours, key=lambda x: cv2.contourArea(x))
    x, y, w, h = cv2.boundingRect(sortedCnt[-1])
    cv2.imshow("crop", dilated[y:y + h, x:x + w])