Pywayne Stereo Tag Matcher

This module matches AprilTags detected in stereo camera pairs.

Quick Start

from pywayne.cv.stereo_tag_matcher import StereoTagMatcher
from pathlib import Path

# Initialize matcher with custom colors
matcher = StereoTagMatcher(
    target_height=600,
    line_color=(0, 0, 255),  # Red
    all_tag_color=(0, 255, 0),  # Green
    common_tag_color=(0, 255, 255)  # Yellow
)

# Process stereo pair
left_img = Path('left.png')
right_img = Path('right.png')
matched_info, stitched = matcher.process_pair(left_img, right_img, show=True)

# Save result
if stitched is not None:
    import cv2
    cv2.imwrite('stereo_result.png', stitched)

Initialization

matcher = StereoTagMatcher(
    target_height=600,      # Fixed height for resizing
    line_color=(0, 0, 255),   # Custom line color (BGR)
    line_thickness=2,
    box_thickness=2,
    all_tag_color=(0, 255, 0),
    common_tag_color=(0, 255, 255)
)

Input

Parameter	Type	Description
image1_input	str, Path, or np.ndarray	Left camera image
image2_input	str, Path, or np.ndarray	Right camera image
show	bool	Display stitched result with cv2.imshow

Output

Returned Dictionary

{
    "tag_id": {
        "cam1_center": (x, y),      # Left image center
        "cam1_corners": [(x1, y1), ...], # Left image corners
        "cam2_center": (x, y),      # Right image center
        "cam2_corners": [(x1, y1), ...]  # Right image corners
    },
    ...
}

Only tags found in both images are included in the output.

Visualization

The stitched image displays:

• All tags - Green boxes (BGR: 0, 255, 0)
• Common tags - Yellow boxes (BGR: 0, 255, 255)
• Connection lines - Red lines connecting common tag centers (BGR: 0, 0, 255)

Use Cases

• Stereo camera calibration - Match common tags to calibrate stereo cameras
• Robot vision - Identify shared landmarks for navigation
• Augmented reality - Track common fiducial markers

Requirements

• cv2 (OpenCV) - Image processing and display
• numpy - Array operations
• pywayne.cv.apriltag_detector - AprilTag detection

Notes

• Images are resized to target_height for consistent annotation
• Tag coordinates are scaled proportionally based on image dimensions
• Supports both grayscale and BGR color input images