A small computer-vision toolkit that detects ArUco markers, performs a planar (homography) transform, and tracks colored objects (e.g., a player and a ball) in video. The transformed coordinates are overlaid onto a soccer-field background and basic in-game events (goal/throw-in/goal-kick) are inferred from positions.
Built with OpenCV and NumPy
python -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txtmain_code requires ArUco markers with IDs 0–3 and writes a 2100×2970 rectified image:
python main_code path/to/input.jpg path/to/output.png-
ArUco detection using OpenCV’s cv2.aruco (DICT_4X4_50).
-
Homography / perspective transform from the camera view into a fixed 640×416 pitch plane.
-
Color-based segmentation with interactive HSV trackbars for two objects (player & ball).
-
Coordinate smoothing via a Gaussian kernel to stabilize trajectories.
-
Event logic (GOAL/THROW-IN/GOAL KICK) from transformed ball coordinates.
-
testing.py -> Track a yellow “player” and orange “ball” in a video via HSV + contours, use ArUco markers to compute a homography, project their positions onto a soccer-field image to draw motion trails, and display/save the live results
-
main_code -> Detects ArUco markers with IDs 0–3 in an image, computes a homography to a fixed 2100×2970 plane, and saves the rectified result
-
minimum_bounding_rect -> Threshold yellow regions, find contours, pick the largest one whose rotated box has an aspect ratio between 2 and 3.5, draw that box on the image, and display it
-
erode 'n dilate -> Interactively threshold a yellow color range with HSV trackbars, generate its mask, visualize the original/mask, and show the effects of erosion and dilation on both
-
detect_arucos -> Detect four specific ArUco markers (IDs 0–3), map their corners, warp the image onto a 2100×2970 plane, save the rectified result and print detection/transform timings
Additionally: Photos of items and tools used