Longer Vision Technology

Github Blog

0%

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()

Posted on Edited on In ,

Preparation

Today, let’s test on an aruco board, instead of a single marker or a diamond marker. Again, you need to make sure your camera has already been calibrated. In the coding section, it’s assumed that you can successfully load the camera calibration parameters.

Coding

The code can be found at OpenCV Examples.

First of all

We need to ensure cv2.so is under our system path. cv2.so is specifically for OpenCV Python.

1
2
import sys
sys.path.append('/usr/local/python/3.5')

Then, we import some packages to be used.

1
2
3
4
import os
import cv2
from cv2 import aruco
import numpy as np

Secondly

Again, we need to load all camera calibration parameters, including: cameraMatrix, distCoeffs, etc. :

1
2
3
4
calibrationFile = "calibrationFileName.xml"
calibrationParams = cv2.FileStorage(calibrationFile, cv2.FILE_STORAGE_READ)
camera_matrix = calibrationParams.getNode("cameraMatrix").mat()
dist_coeffs = calibrationParams.getNode("distCoeffs").mat()

If you are using a calibrated fisheye camera like us, two extra parameters are to be loaded from the calibration file.

1
2
r = calibrationParams.getNode("R").mat()
new_camera_matrix = calibrationParams.getNode("newCameraMatrix").mat()

Afterwards, two mapping matrices are pre-calculated by calling function cv2.fisheye.initUndistortRectifyMap() as (supposing the images to be processed are of 1080P):

1
2
image_size = (1920, 1080)
map1, map2 = cv2.fisheye.initUndistortRectifyMap(camera_matrix, dist_coeffs, r, new_camera_matrix, image_size, cv2.CV_16SC2)

Thirdly

In our test, the dictionary aruco.DICT_6X6_1000 is adopted as the unit pattern to construct a grid board. The board is of size 5X7, which looks like:

aruco.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 particular aruco marker and the distance between two neighbour markers are to be measured and stored in two variables markerLength and markerSeparation, which are used to create the 5X7 grid board.

1
2
3
markerLength = 40   # Here, our measurement unit is centimetre.
markerSeparation = 8   # Here, our measurement unit is centimetre.
board = aruco.GridBoard_create(5, 7, markerLength, markerSeparation, 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. We first load the video file and initialize a video capture handle.

1
2
videoFile = "aruco\_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
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
            im_with_aruco_board = aruco.drawDetectedMarkers(frame_remapped, corners, ids, (0,255,0))
            retval, rvec, tvec = aruco.estimatePoseBoard(corners, ids, board, camera_matrix, dist_coeffs)  # posture estimation from a diamond
            if retval != 0:
                im_with_aruco_board = aruco.drawAxis(im_with_aruco_board, camera_matrix, dist_coeffs, rvec, tvec, 100)  # axis length 100 can be changed according to your requirement
        else:
            im_with_aruco_board = frame_remapped

        cv2.imshow("arucoboard", im_with_aruco_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()

Posted on Edited on In ,

Preparation

Very similar to our previous post Camera Posture Estimation Using A Single aruco Marker, you need to make sure your camera has already been calibrated. In the coding section, it’s assumed that you can successfully load the camera calibration parameters.

Coding

The code can be found at OpenCV Examples.

First of all

We need to ensure cv2.so is under our system path. cv2.so is specifically for OpenCV Python.

1
2
import sys
sys.path.append('/usr/local/python/3.5')

Then, we import some packages to be used.

1
2
3
4
import os
import cv2
from cv2 import aruco
import numpy as np

Secondly

Again, we need to load all camera calibration parameters, including: cameraMatrix, distCoeffs, etc. :

1
2
3
4
calibrationFile = "calibrationFileName.xml"
calibrationParams = cv2.FileStorage(calibrationFile, cv2.FILE_STORAGE_READ)
camera_matrix = calibrationParams.getNode("cameraMatrix").mat()
dist_coeffs = calibrationParams.getNode("distCoeffs").mat()

If you are using a calibrated fisheye camera like us, two extra parameters are to be loaded from the calibration file.

1
2
r = calibrationParams.getNode("R").mat()
new_camera_matrix = calibrationParams.getNode("newCameraMatrix").mat()

Afterwards, two mapping matrices are pre-calculated by calling function cv2.fisheye.initUndistortRectifyMap() as (supposing the images to be processed are of 1080P):

1
2
image_size = (1920, 1080)
map1, map2 = cv2.fisheye.initUndistortRectifyMap(camera_matrix, dist_coeffs, r, new_camera_matrix, image_size, cv2.CV_16SC2)

Thirdly

The dictionary aruco.DICT_6X6_250 is to be loaded. Although current OpenCV provides four groups of aruco patterns, 4X4, 5X5, 6X6, 7X7, etc., it seems OpenCV Python does NOT provide a function named drawCharucoDiamond(). Therefore, we have to refer to the C++ tutorial Detection of Diamond Markers. And, we directly use this particular diamond marker in the tutorial:

aruco.DICT_6X6_250.diamond

1
aruco_dict = aruco.Dictionary_get( aruco.DICT_6X6_250 )

After having this aruco diamond marker printed, the edge lengths of this particular diamond marker are to be measured and stored in two variables squareLength and markerLength.

1
2
squareLength = 40   # Here, our measurement unit is centimetre.
markerLength = 25   # Here, our measurement unit is centimetre.

Meanwhile, create aruco detector with default parameters.

1
arucoParams = aruco.DetectorParameters_create()

Finally

This time, let’s test on a video stream, a .mp4 file. We first load the video file and initialize a video capture handle.

1
2
videoFile = "aruco_diamond.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
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

        if ids != None: # if there is at least one marker detected
            diamondCorners, diamondIds = aruco.detectCharucoDiamond(frame_remapped_gray, corners, ids, squareLength/markerLength)   # Second, detect diamond markers
            if len(diamondCorners) >= 1:    # if there is at least one diamond detected
                im_with_diamond = aruco.drawDetectedDiamonds(frame_remapped, diamondCorners, diamondIds, (0,255,0))
                rvec, tvec = aruco.estimatePoseSingleMarkers(diamondCorners, squareLength, camera_matrix, dist_coeffs)  # posture estimation from a diamond
                im_with_diamond = aruco.drawAxis(im_with_diamond, camera_matrix, dist_coeffs, rvec, tvec, 100)  # axis length 100 can be changed according to your requirement
        else:
            im_with_diamond = frame_remapped

        cv2.imshow("diamondLeft", im_with_diamond)   # display

        if cv2.waitKey(2) & 0xFF == ord('q'):   # press 'q' to quit
            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()

Posted on Edited on In ,

Preparation

Before start coding, you need to ensure your camera has already been calibrated. (Camera calibration is covered in our blog as well.) In the coding section, it’s assumed that you can successfully load the camera calibration parameters.

Coding

The code can be found at OpenCV Examples.

First of all

We need to ensure cv2.so is under our system path. cv2.so is specifically for OpenCV Python.

1
2
import sys
sys.path.append('/usr/local/python/3.5')

Then, we import some packages to be used.

1
2
3
4
import os
import cv2
from cv2 import aruco
import numpy as np

Secondly

We now load all camera calibration parameters, including: cameraMatrix, distCoeffs, etc. For example, your code might look like the following:

1
2
3
4
calibrationFile = "calibrationFileName.xml"
calibrationParams = cv2.FileStorage(calibrationFile, cv2.FILE_STORAGE_READ)
camera_matrix = calibrationParams.getNode("cameraMatrix").mat()
dist_coeffs = calibrationParams.getNode("distCoeffs").mat()

Since we are testing a calibrated fisheye camera, two extra parameters are to be loaded from the calibration file.

1
2
r = calibrationParams.getNode("R").mat()
new_camera_matrix = calibrationParams.getNode("newCameraMatrix").mat()

Afterwards, two mapping matrices are pre-calculated by calling function cv2.fisheye.initUndistortRectifyMap() as (supposing the images to be processed are of 1080P):

1
2
image_size = (1920, 1080)
map1, map2 = cv2.fisheye.initUndistortRectifyMap(camera_matrix, dist_coeffs, r, new_camera_matrix, image_size, cv2.CV_16SC2)

Thirdly

A dictionary is to be loaded. Current OpenCV provides four groups of aruco patterns, 4X4, 5X5, 6X6, 7X7, etc. Here, aruco.DICT_6X6_1000 is randomly selected as our example, which looks like:

aruco.DICT_6X6_1000

1
aruco_dict = aruco.Dictionary_get( aruco.DICT_6X6_1000 )

After having this aruco square marker printed, the edge length of this particular marker is to be measured and stored in a variable markerLength.

1
markerLength = 20 # Here, our measurement unit is centimetre.

Meanwhile, create aruco detector with default parameters.

1
arucoParams = aruco.DetectorParameters_create()

Finally

Estimate camera postures. Here, we are testing a sequence of images, rather than video streams. We first list all file names in sequence.

1
2
3
imgDir = "imgSequence"  # Specify the image directory
imgFileNames = [os.path.join(imgDir, fn) for fn in next(os.walk(imgDir))[2]]
nbOfImgs = len(imgFileNames)

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
for i in range(0, nbOfImgs):
    img = cv2.imread(imgFileNames[i], cv2.IMREAD_COLOR)
    imgRemapped = cv2.remap(img, map1, map2, cv2.INTER_LINEAR, cv2.BORDER_CONSTANT) # for fisheye remapping
    imgRemapped_gray = cv2.cvtColor(imgRemapped, cv2.COLOR_BGR2GRAY)    # aruco.etectMarkers() requires gray image
    corners, ids, rejectedImgPoints = aruco.detectMarkers(imgRemapped_gray, aruco_dict, parameters=arucoParams) # Detect aruco
    if ids != None: # if aruco marker detected
        rvec, tvec = aruco.estimatePoseSingleMarkers(corners, markerLength, camera_matrix, dist_coeffs) # For a single marker
        imgWithAruco = aruco.drawDetectedMarkers(imgRemapped, corners, ids, (0,255,0))
        imgWithAruco = aruco.drawAxis(imgWithAruco, camera_matrix, dist_coeffs, rvec, tvec, 100)    # axis length 100 can be changed according to your requirement
    else:   # if aruco marker is NOT detected
        imgWithAruco = imgRemapped  # assign imRemapped_color to imgWithAruco directly

    cv2.imshow("aruco", imgWithAruco)   # display
    
    if cv2.waitKey(2) & 0xFF == ord('q'):   # if 'q' is pressed, quit.
        break

The drawn axis is just the world coordinators and orientations estimated from the images taken by the testing camera.

Posted on Edited on In

Hi, everyone. This is Nobody from Longer Vision Technology. I come back to life, at least, half life. And finally, I decided to write something, either useful, or useless. Hope my blogs will be able to help some of the pure researchers, as well as the students, in the field of Computer Vision & Machine Vision. By the way, our products will be put on sale soon. Keep an eye on our blogs please. Thank you…