Truck Measurement System – Detailed Project Report

Project Overview

This project is a computer vision system designed to automatically detect trucks in images, classify their type, estimate their real-world dimensions (width and height), and measure logos on trucks. It combines deep learning (YOLOv8), image processing (OpenCV), and a user-friendly web interface (Streamlit).

Main Features

• Truck Detection: Finds trucks in uploaded images using YOLOv8.
• Truck Classification: Identifies the type of truck (Semi Trailer, Box Truck, Cube Van, Sprinter Van, Cargo Van).
• Real-world Measurement: Estimates truck width and height in meters using known reference heights for each truck type.
• Automatic Logo Measurement: Upload a logo template image; the app finds and measures the logo on the truck using OpenCV template matching.
• Interactive Web App: Simple Streamlit interface for uploading images and viewing results instantly.
• Command Line Support: Process images and save annotated results via CLI.

File Structure

truck-measurement-system/
│
├── truck_measurement/           # Core package
│   ├── __init__.py              # Package exports
│   ├── main.py                  # CLI entry point
│   ├── detector.py              # YOLO detection logic
│   ├── classifier.py            # Truck classification logic
│   ├── measurement.py           # Measurement calculations
│   ├── visualizer.py            # Drawing overlays and annotations
│   └── utils.py                 # Logging and validation
│
├── tests/                       # Unit tests for all modules
├── examples/                    # Example scripts/images
├── app.py                       # Streamlit web app
├── run_web_app.py               # App launcher script
├── requirements.txt             # Python dependencies
├── setup.py                     # Installer and packaging
├── README.md                    # Full documentation
├── README_SIMPLE.md             # Simple, layman README
└── yolov8m.pt                   # YOLO model weights (auto-downloaded if missing)

How the System Works

1. Truck Detection:

   • The YOLOv8 model scans the uploaded image and finds trucks.
   • The largest or most confident detection is selected for measurement.

2. Truck Classification:

   • The system uses the detected truck's bounding box size and aspect ratio to classify its type.
   • Each truck type has a known reference height (e.g., Semi Trailer ≈ 4.0m).

3. Measurement Calculation:

   • The pixel height of the detected truck is used to calculate a pixel-to-meter scale.
   • The truck's width and height in pixels are converted to meters using this scale.

4. Logo Measurement (Automatic):

   • You upload a cropped logo image as a template.
   • The app uses OpenCV's template matching to find the logo in the truck image.
   • The logo's pixel size is converted to meters using the truck's scale.

5. Web App Interface:

   • Upload a truck image and (optionally) a logo template.
   • The app displays the detected truck, its type, real-world size, and logo measurements.
   • Results are shown instantly with annotated images.

How to Run

1. Set up Python (3.8+) and activate your virtual environment.
2. Install dependencies:

   pip install -r requirements.txt
   

3. Run the web app:

   python run_web_app.py
   

   or

   streamlit run app.py
   

4. Upload your images and logo template in the app.

Screenshots

Add your screenshots here to show the app in action!

Web Interface

Detection Results

Technical Details

• YOLOv8 is used for truck detection (weights auto-downloaded if missing).
• OpenCV is used for image processing and logo template matching.
• Streamlit provides the web interface.
• Pillow and NumPy are used for image handling and calculations.

Development & Customization

• You can add more vehicle types or change reference heights in classifier.py and measurement.py.
• The code is modular and easy to extend for new features (e.g., video detection, batch processing).
• Unit tests are provided in the tests/ folder for reliability.

Author

• Rroopesh Hari
• Data Science @ Oklahoma State University

Support

For questions, help, or suggestions, open an issue on GitHub or contact the author directly.

---