Search⌘ K
AI Features
Home
Courses
Using Python for Reading and Writing Optical Labels

Reading ArUco Markers and AprilTags

Explore how to read and detect ArUco markers and AprilTags using Python and OpenCV. Understand marker detection, pose estimation, and visualization techniques with and without camera calibration to improve accuracy in computer vision applications.

We'll cover the following...

The Python code we’ll examine is designed to load an image containing ArUco markers, detect these markers, estimate their pose, and then visualize the detected markers and their pose in 3D.

We’ll discuss two versions of the same code. In the first version, we’ll read markers without using a calibrated camera. In the second version, we’ll use a calibrated camera.

Whether we need to use a calibrated camera or not depends on our specific use case.

Reading fiduciary markers with an uncalibrated camera

In the first example, we’ll discuss how to read markers using an ...

Python 3.10.4
import cv2
import numpy as np
def loadSyntheticCalibrationData():
    width = 640
    height = 480
    focal_length = 50 // 0.003
    camera_matrix = np.array([[focal_length, 0, width//2],[0, focal_length, height//2],[0, 0, 1]])
    distortion_coeffs = np.zeros([1,5])
    
    return camera_matrix, distortion_coeffs
def preprocessImage(image):
    processed_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    return processed_image
def detectMarkers(gray_image, marker_dict, detector_params):
    corners, ids, _ = cv2.aruco.detectMarkers(gray_image, marker_dict, parameters=detector_params)
    return corners, ids
def poseEstimation(corners, marker_length, camera_matrix, distortion_coeffs):
    rvecs, tvecs, _ = cv2.aruco.estimatePoseSingleMarkers(corners, marker_length, camera_matrix, distortion_coeffs)
    return rvecs, tvecs
def visualization(image, corners, ids, camera_matrix, distortion_coeffs, rvecs, tvecs, marker_length):
    image_markers = cv2.aruco.drawDetectedMarkers(image.copy(), corners, ids)
    
    if ids is not None:
        for i in range(len(ids)):
            image_markers = cv2.drawFrameAxes(image_markers, camera_matrix, distortion_coeffs, 
                                              rvecs[i], tvecs[i], marker_length)
    return image_markers
image_file = "./resources/marker_images/calibrate_20230614_155453.png"
image = cv2.imread(image_file)
camera_matrix, distortion_coeffs = loadSyntheticCalibrationData()
processed_image = preprocessImage(image)
marker_dict = cv2.aruco.Dictionary_get(cv2.aruco.DICT_5X5_250)
detector_params = cv2.aruco.DetectorParameters_create()
detections, ids = detectMarkers(processed_image, marker_dict, detector_params)
marker_length = .048
rvecs, tvecs = poseEstimation(detections, marker_length, camera_matrix, distortion_coeffs)
result = visualization(image, detections, ids, camera_matrix, distortion_coeffs, rvecs, tvecs, marker_length)
cv2.imwrite(f"./output/marker_detection.png", result)

Lines 1–2: First, we import the ...