Automatic Detection of Differences Between Two Images

İki Fotoğraf Arasındaki Farkı Algılama

Hüseyin Mert Çalışkan - 210223044

Interactive Image Processing Project with Godot Engine & OpenCV

📄 Research Paper

🎯 Project Overview

This project demonstrates automatic difference detection between two scene images using HSV color space analysis. Built with Godot Engine 3.5 for interactive 3D environment and Python/OpenCV for image processing pipeline.

What It Does

• 3D Scene Navigation: First-person player movement in interactive environment
• Screenshot Capture: Take photos of scene with/without objects
• Automatic Processing: HSV-based difference detection with morphological filtering
• Visual Results: Bounding boxes around detected differences

🛠️ Technology Stack

Frontend (Godot Engine 3.5)

• 3D scene rendering with interactive objects
• First-person camera controls
• Screenshot capture system
• Real-time visualization

Backend (Python + OpenCV)

• HSV color space conversion
• Hue & Saturation channel analysis
• Morphological operations (opening/closing)
• Contour detection and filtering

🎮 How It Works

1. Scene Setup

• Navigate through 3D environment using WASD + mouse
• Scene contains interactive objects that can be toggled on/off

2. Image Capture

# capture.gd - Screenshot system
func capture_pair():
    _save_screenshot("full_scene.png")      # Scene with objects
    interactive.visible = false
    _save_screenshot("background_only.png") # Empty background
    interactive.visible = true

3. Difference Detection

# diff_pipeline.py - Core processing
# HSV conversion and channel analysis
hsv_full = cv2.cvtColor(full, cv2.COLOR_BGR2HSV)
hsv_bg = cv2.cvtColor(bg, cv2.COLOR_BGR2HSV)

# Calculate H & S channel differences
dh = cv2.absdiff(h_full, h_bg)
ds = cv2.absdiff(s_full, s_bg)

# Create binary mask and apply morphology
mask = cv2.bitwise_or(mh, ms)
mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)

4. Result Visualization

• Detected differences marked with red bounding boxes
• Results saved automatically to screenshots/differences_marked.png

📁 Project Structure

godot-image-difference-detection/
├── 📂 scripts/
│   ├── capture.gd              # Screenshot capture system
│   └── Player.gd               # First-person controller
├── 📂 py_scripts/
│   └── diff_pipeline.py        # Image processing pipeline
├── 📂 screenshots/             # Captured images & results
├── 📂 assets/                  # 3D models, textures
├── 📂 scenes/                  # Godot scene files
├── 📂 docs/                    # Research paper
├── project.godot               # Main Godot project
└── requirements.txt            # Python dependencies

🔬 Algorithm Details

HSV Color Space Analysis

• Hue Threshold: 10 (color difference detection)
• Saturation Threshold: 30 (intensity difference detection)
• Channel Combination: OR operation between H and S masks

Morphological Operations

• Opening: 3×3 kernel (noise removal)
• Closing: 7×7 kernel (gap filling)
• Iterations: 1 pass each operation

Contour Filtering

• Minimum Area: 300 pixels
• Aspect Ratio: 0.01 - 5.0 range
• Bounding Box: Red rectangles around valid regions

📸 Sample Results

Background Only	Full Scene	Detected Differences

🎮 Controls

• WASD: Movement
• Mouse: Look around
• Space: Jump
• Screenshot Key: Capture image pair and process

📄 Research Paper

Complete methodology and experimental analysis available in Turkish: 📋 Research Paper (PDF)

🏗️ Built With

• Godot Engine 3.5 - 3D environment and interaction
• Python 3.8 - Image processing backend
• OpenCV 4.x - Computer vision operations
• NumPy - Mathematical operations

👨‍💻 Author

Hüseyin Mert Çalışkan - 210223044
Computer Engineering Student Project

---

This project demonstrates practical application of computer vision techniques in an interactive 3D environment for educational purposes.