Good ArUco Marker Detection in DICT4X4 but not in DICT5x5

Question

I am trying to detect ArUco Markers to recognise some real moving objects, to write my Degree Thesis. Fisrt, I tried with some 4x4 ArUco Markers, i got good results, then i tried with an image that i got from my camera. This image contains 5x5 ArUco Markers, applying same code, except changing Dictionaries, 4x4 per 5x5, why is only detecting 4x4 markers ?

Results: 4x4 Markers detected 1

5x5 Markers not detected 2

import cv2
import numpy as np
from cv2 import aruco
import camera_calibration

path='/home/alejandrocruzado/Documents/Imaginni/aruco4.png'
path_5 = '/home/alejandrocruzado/Documents/Imaginni/aruco5.jpg'
path5x5 = '/home/alejandrocruzado/Documents/Imaginni/5x5.jpg'
distCoeffs=[0.0189223469433419, -0.0206788674793396,0.003225513523750, 0.001510000668961]
cameraMatrix=[[600.8293, 0, 330.2756],[0, 601.3519, 225.9791],[0, 0, 1.0000]]

def MarkerDetection4x4(path):
    img = cv2.imread(path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    aruco_dict = cv2.aruco_Dictionary.get(aruco.DICT_4X4_1000)
    parameters = cv2.aruco_DetectorParameters.create()
    corners, ids, rejectedImgPoints = cv2.aruco.detectMarkers(gray, aruco_dict, parameters=parameters)
    cv2.aruco.drawDetectedMarkers(img,corners,ids,)
    cv2.imshow('Image', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()
    rvecs, tvecs, _objPoints = cv2.aruco.estimatePoseSingleMarkers(corners, 0.045, cameraMatrix, distCoeffs)
    cv2.aruco.drawAxis(img, cameraMatrix, distCoeff, rvecs, tvecs, 0.1)


def MarkerDetection5x5(path):
    img = cv2.imread(path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)    
    aruco_dict = cv2.aruco_Dictionary.get(aruco.DICT_5X5_250)
    parameters = cv2.aruco_DetectorParameters.create()
    corners, ids, rejectedImgPoints = cv2.aruco.detectMarkers(gray, aruco_dict, parameters=parameters)
    cv2.aruco.drawDetectedMarkers(img,corners,ids,)
    cv2.imshow('Image', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()
    rvecs, tvecs, _objPoints = cv2.aruco.estimatePoseSingleMarkers(corners, 0.045, cameraMatrix, distCoeffs)
    cv2.aruco.drawAxis(img, cameraMatrix, distCoeff, rvecs, tvecs, 0.1)


if __name__ == "__main__":

    #MarkerDetection4x4(path)
    MarkerDetection5x5(path5x5)

Thanks

Please edit your question to include a question. "Doesn't work" is not a question. — mypetlion, Apr 18 '18 at 18:23
"SO, where theses get written." Come on guys, it's that easy. Easy! — Stefan Dragnev, Apr 18 '18 at 20:04

zardosht · Answer 1 · 2018-10-26T15:37:47.977

You can draw lines around the rejected marker candidates using this code:

for rejected in rejected_img_points:
     rejected = rejected.reshape((4, 2))
     cv2.line(img, tuple(rejected[0]), tuple(rejected[1]), (0, 0, 255), thickness=2)
     cv2.line(img, tuple(rejected[1]), tuple(rejected[2]), (0, 0, 255), thickness=2)
     cv2.line(img, tuple(rejected[2]), tuple(rejected[3]), (0, 0, 255), thickness=2)
     cv2.line(img, tuple(rejected[3]), tuple(rejected[0]), (0, 0, 255), thickness=2)

If you see that the marker images are correctly detected but are rejected, then possibly you are using a different dictionary for your detection from the dictionary that was used to draw those markers. The marker dictionary has changed.

Create a new set of 5x5 markers using the following code and try again:

marker_id = 5
# The image size for marker should be a 
# multiple of (number of bits + 2 * border) or 
# very large compared to number of bits  
marker_size = 140   
marker_dictionary = cv2.aruco.getPredefinedDictionary(cv2.aruco.DICT_5X5_50)
marker_imge = np.zeros((marker_size, marker_size), np.uint8)
marker_img = cv2.aruco.drawMarker(marker_dictionary, marker_id, marker_size, marker_img, 1)

Good ArUco Marker Detection in DICT4X4 but not in DICT5x5

1 Answers1