Parking Space Detection System

Overview

This project implements a computer vision-based system for detecting and monitoring parking space availability. The system supports two detection methods:

1. Traditional CV Method: Using OpenCV image processing techniques (thresholding, contours)
2. ML-Based Method: Using YOLOv8 deep learning model for vehicle detection

Features

• Interactive Parking Space Marking: Click and drag to define parking spaces
• Real-time Space Monitoring: Track and display free/occupied spaces
• Dual Detection Methods:
  • OpenCV-based processing with adaptive parameters
  • YOLOv8-based vehicle detection with IoU calculation
• Persistence: Parking space positions are saved between sessions
• Adjustable Parameters: Fine-tune detection with trackbars

Installation

1. Clone this repository:

   git clone https://github.com/pranjayyelkotwar/Cars-Parking-Detector.git
   cd parking-space-detection

2. Install the required dependencies:

   pip install -r requirements.txt

3. Download the YOLOv8 model (if using ML-based detection):

   # The model will be downloaded automatically when first using the ML method
   # or download manually from https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8n.pt

4. Place a video file named car_parking.mp4 in the project directory or update the VIDEO_PATH in config.py.

Usage

Running the Application

Run with traditional OpenCV-based detection (default):

python main.py

Run with YOLO-based ML detection:

python main.py --method ml

Workflow

1. First Run - Marking Mode:

   • The application starts in marking mode if no parking spaces are defined
   • Click and drag to define rectangular parking spaces
   • Press 'q' or ESC to switch to detection mode

2. Detection Mode:

   • The system will analyze the video frame-by-frame
   • Green rectangles indicate free spaces, red rectangles indicate occupied spaces
   • The count of free/total spaces is displayed in the top left
   • Adjust parameters with the trackbars (for OpenCV method)

Parameter Tuning

The OpenCV method uses three parameters that can be adjusted in real-time:

• Val1: Block size for adaptive thresholding
• Val2: Constant subtracted from the mean
• Val3: Kernel size for median blur

Project Structure

parking_space/
├── main.py              # Main application entry point
├── parking_logic.py     # Core detection and space management functions
├── video_processing.py  # Video handling and frame preprocessing
├── config.py            # Configuration parameters
├── CarParkPos           # Binary file storing parking space positions
├── yolov8n.pt          # YOLOv8 model file (downloaded when needed)
├── requirements.txt     # Project dependencies
└── README.md            # This documentation

Configuration

You can modify the following parameters in config.py:

# File paths
VIDEO_PATH = 'car_parking.mp4'  # Path to input video
PARKING_FILE = 'CarParkPos'     # File to store parking spaces

# Detection parameters
PARKING_THRESHOLD = 500         # Pixel count threshold for occupancy

# UI Parameters
WINDOW_NAME = "Vals"            # Name of parameter window
WINDOW_SIZE = (640, 240)        # Size of parameter window

# Pre-processing parameters
GAUSSIAN_BLUR_KERNEL = (3, 3)   # Kernel size for Gaussian blur
GAUSSIAN_BLUR_SIGMA = 1         # Sigma for Gaussian blur
DILATION_KERNEL = (3, 3)        # Kernel size for dilation
DILATION_ITERATIONS = 1         # Number of dilation iterations

Technical Details

OpenCV-based Detection

The traditional method uses these steps:

1. Convert frame to grayscale
2. Apply Gaussian blur
3. Apply adaptive thresholding
4. Count non-zero pixels in each parking space
5. Determine occupancy based on pixel count threshold

YOLOv8-based Detection

The ML-based method:

1. Detects vehicles in the entire frame using YOLOv8
2. For each parking space, calculates overlap with detected vehicles
3. Determines occupancy based on Intersection over Union (IoU)

Requirements

See requirements.txt for the complete list of dependencies.

License

MIT License

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.