Computer Vision untuk Robot Soccer

Computer vision adalah "mata" robot yang memungkinkan robot memahami lingkungan sekitar melalui kamera. Dalam RoboCup, vision system bertanggung jawab untuk mendeteksi bola, garis lapangan, gawang, dan robot lain.

Mengapa Penting?

Vision Pipeline Overview

Pipeline ini penting karena setiap tahap membatasi noise dan mengubah data ke bentuk yang bisa dipakai robot untuk mengambil keputusan. Ringkasnya:

• Preprocess menstabilkan input (denoise, resize, koreksi warna) agar deteksi lebih konsisten.
• Detection mengekstrak objek penting (bola, garis, gawang) dari citra.
• 3D Projection mengubah posisi pixel ke koordinat lapangan agar bisa dipakai untuk navigasi dan gerak.

Learning Tracks

Recommended Learning Path

Tools & Libraries

Primary Tools

ROS 2 Packages

# Image transport
sudo apt install ros-humble-image-transport

# CV Bridge (OpenCV ↔ ROS)
sudo apt install ros-humble-cv-bridge

# Image pipeline
sudo apt install ros-humble-image-pipeline

Quick Start: Ball Detection

Contoh sederhana deteksi bola oranye menggunakan OpenCV:

import cv2
import numpy as np

def detect_ball(image):
    """Detect orange ball using HSV color filtering."""
    # Convert BGR to HSV
    hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)

    # Define orange color range
    lower_orange = np.array([5, 100, 100])
    upper_orange = np.array([15, 255, 255])

    # Create mask
    mask = cv2.inRange(hsv, lower_orange, upper_orange)

    # Find contours
    contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL,
                                    cv2.CHAIN_APPROX_SIMPLE)

    if contours:
        # Get largest contour (assume it's the ball)
        largest = max(contours, key=cv2.contourArea)
        (x, y), radius = cv2.minEnclosingCircle(largest)

        if radius > 10:  # Minimum size threshold
            return int(x), int(y), int(radius)

    return None

# Usage
cap = cv2.VideoCapture(0)
while True:
    ret, frame = cap.read()
    if not ret:
        break

    result = detect_ball(frame)
    if result:
        x, y, r = result
        cv2.circle(frame, (x, y), r, (0, 255, 0), 2)
        cv2.circle(frame, (x, y), 5, (0, 0, 255), -1)

    cv2.imshow('Ball Detection', frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

Challenges in RoboCup

1. Lighting Variations

Lapangan RoboCup memiliki pencahayaan yang bisa berubah:

• Natural light vs artificial light
• Shadows dari robot/arena
• Specular reflections dari lantai

Solusi: Adaptive thresholding, histogram equalization, ML-based detection

2. Motion Blur

Robot bergerak cepat menyebabkan motion blur:

• Ball tracking saat walking
• Head movement untuk scanning

Solusi: Higher FPS camera, motion compensation, predictive tracking

3. Occlusion

Objek bisa terhalang:

• Ball behind opponent robot
• Partial line visibility

Solusi: Multi-frame tracking, probabilistic models

4. Real-time Constraint

Processing harus cepat (~30 fps minimum):

• Limited compute on robot
• Battery constraints

Solusi: Efficient algorithms, GPU acceleration, region of interest

Performance Metrics

Resources

Documentation

• OpenCV Python Tutorials
• PyTorch Vision Tutorial
• ROS 2 Image Pipeline

Papers & Research

• B-Human Team Report - State-of-the-art SPL vision
• RoboCup Symposium Papers
• YOLO Paper - Original YOLO architecture

Video Tutorials

• OpenCV Course - freeCodeCamp
• PyTorch for Computer Vision

GitHub Repositories

• Ultralytics YOLO
• B-Human Code Release
• soccer_vision ROS package

Next Steps

1. Pemula: Mulai dengan Image Processing Basics
2. Geometry: Pelajari Camera Calibration
3. Detection: Lanjut ke Ball Detection dan Field Detection
4. Stability: Tambahkan Object Tracking
5. Deployment: Tutup dengan ROS 2 Integration dan Debugging Playbook