Camera Posture Estimation Using A ChArUco Board

Posted on Edited on In ,

Preparation

ChAruco is an integrated marker, which combines a chessboard with an aruco marker. The code is also very similar to the code in our previous blog aruco board.

Coding

The code can be found at OpenCV Examples. And the code in the first two subsections are exactly the same as what’s written in our previous blogs. We’ll neglect the first two subsections ever since.

First of all

Exactly the same as in previous blogs.

Secondly

Exactly the same as in previous blogs.

Thirdly

Dictionary aruco.DICT_6X6_1000 is integrated with a chessboard to construct a grid charuco board. The experimenting board is of size 5X7, which looks like:

charuco.DICT_6X6_1000.board57 
1
aruco_dict = aruco.Dictionary_get( aruco.DICT_6X6_1000 )

After having this aruco board marker printed, the edge lengths of this chessboard and aruco marker (displayed in the white cell of the chessboard) are to be measured and stored in two variables squareLength and markerLength, which are used to create the 5X7 grid board.

1
2
3
squareLength = 40   # Here, our measurement unit is centimetre.
markerLength = 30   # Here, our measurement unit is centimetre.
board = aruco.CharucoBoard_create(5, 7, squareLength, markerLength, aruco_dict)

Meanwhile, create aruco detector with default parameters.

1
arucoParams = aruco.DetectorParameters_create()

Finally

Now, let’s test on a video stream, a .mp4 file.

1
2
videoFile = "charuco_board_57.mp4"
cap = cv2.VideoCapture(videoFile)

Then, we calculate the camera posture frame by frame:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
while(True):
    ret, frame = cap.read() # Capture frame-by-frame
    if ret == True:
        frame_remapped = cv2.remap(frame, map1, map2, cv2.INTER_LINEAR, cv2.BORDER_CONSTANT)    # for fisheye remapping
        frame_remapped_gray = cv2.cvtColor(frame_remapped, cv2.COLOR_BGR2GRAY)

        corners, ids, rejectedImgPoints = aruco.detectMarkers(frame_remapped_gray, aruco_dict, parameters=arucoParams)  # First, detect markers
        aruco.refineDetectedMarkers(frame_remapped_gray, board, corners, ids, rejectedImgPoints)

        if ids != None: # if there is at least one marker detected
            charucoretval, charucoCorners, charucoIds = aruco.interpolateCornersCharuco(corners, ids, frame_remapped_gray, board)
            im_with_charuco_board = aruco.drawDetectedCornersCharuco(frame_remapped, charucoCorners, charucoIds, (0,255,0))
            retval, rvec, tvec = aruco.estimatePoseCharucoBoard(charucoCorners, charucoIds, board, camera_matrix, dist_coeffs)  # posture estimation from a charuco board
            if retval == True:
                im_with_charuco_board = aruco.drawAxis(im_with_charuco_board, camera_matrix, dist_coeffs, rvec, tvec, 100)  # axis length 100 can be changed according to your requirement
        else:
            im_with_charuco_left = frame_remapped

        cv2.imshow("charucoboard", im_with_charuco_board)

        if cv2.waitKey(2) & 0xFF == ord('q'):
            break
    else:
        break

The drawn axis is just the world coordinators and orientations estimated from the images taken by the testing camera. At the end of the code, we release the video capture handle and destroy all opening windows.

1
2
cap.release()   # When everything done, release the capture
cv2.destroyAllWindows()