如何用开放的简历填充盒子内的盒子

Question

Ok, since no one is answering let me try and change my question, how could one go about trying to identify this boxes?

Picture where i want to detect

to get this:

Picture i want to get

im having trouble since there are white letters inside the box.(i think) Can anyone provide a soluction?

I'm starting out with CV, doing a simple project to learn. I'm trying to get boxes inside boxes and count them. I'm detecting the edges and doing a Thresh binary inv. I'm getting this, which i think is pretty good and i should be able to work with,

picture after processing

after i go on doing a detection of shapes, but i only find one box, te big one.

I'm to the point of thinking: A) the problem is that the boxes inside are not "completed" B) the lines are not a segment

But for what i read, if it dectets the edges it should see it has 4, and detected as a rectangle.

Here is a code

thresh = cv2.threshold(edges, 177, 255, cv2.THRESH_BINARY_INV)[1]

cv2.imshow("Image", thresh)
cv2.waitKey(0)

cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
    cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if imutils.is_cv2() else cnts[1]
sd = ShapeDetector()

after i iterate thourg the edges and try to draw over them in green

And my shape dectector

class ShapeDetector:
def __init__(self):
    pass

def detect(self, c):
    print("ola")
    # initialize the shape name and approximate the contour
    shape = "unidentified"
    peri = cv2.arcLength(c, True)
    approx = cv2.approxPolyDP(c, 0.04 * peri, True)

    if len(approx) == 4:
        # compute the bounding box of the contour and use the
        # bounding box to compute the aspect ratio
        (x, y, w, h) = cv2.boundingRect(approx)
        ar = w / float(h)
        print( "Found one!")
        shape = "rectangulo"

    return shape

Thanks in advance

Answer 1

Instead of cv2.RETR_EXTERNAL (which retrieves only external contours) you should use something else such as cv2.RETR_LIST (simple list with all contours) or cv2.RETR_TREE (all contours in a hierarchy).

See https://docs.opencv.org/3.1.0/d9/d8b/tutorial_py_contours_hierarchy.html

EDIT: RETR_TREE is more complex but gives you the information of what contours are inside other contours, so you can ignore some of them.

If you want perfect rectangles, you should draw the result of the boundingRect() instead of the original contour.

If you find overlapping rectangles, you can also add more conditions to eliminate them (w and h should be between a minimum and a maximum for example).