Pallet Board Detection

This project provides a complete pipeline for detecting horizontal and vertical boards (pallets) from TIFF images. It uses OpenCV, NumPy, and Python to process images, detect edges, extract line structures, group boards, handle fragmented or missing boards, and generate labeled visualizations.

---

Features

• Load and normalize 16-bit TIFF images.
• Detect edges using Sobel operators.
• Extract vertical and horizontal line masks.
• Clean lines using morphological operations.
• Extract horizontal boards, group them into lines, and filter invalid lines.
• Extract vertical boards, merge fragmented boxes, and fill missing gaps.
• Generate labeled images showing detected boards and virtual pallets.
• Supports saving intermediate results and final visualizations per image.

---

Installation

1. Clone this repository:

git clone https://github.com/Vicen-te/pallet-board-detection.git
cd pallet-board-detection

2. Create a Python virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # Linux/macOS
venv\Scripts\activate     # Windows

3. Install required packages:

pip install -r requirements.txt

4. Required libraries:

• opencv-python – Image processing and computer vision operations.
• numpy – Numerical computations, arrays, and image manipulation.
• matplotlib – Visualization of images and plots.

---

Project Structure

pallet-board-detection/
│
├─ src/                       # Source code
│  ├─ image_utils.py          # Loading, normalization, and saving images
│  ├─ edge_detection.py       # Sobel edges, line extraction, and cleaning
│  ├─ horizontal_boards.py    # Extraction and grouping of horizontal boards
│  ├─ vertical_boards.py      # Extraction, merging, and filling vertical boards
│  ├─ pallet_analysis.py      # Helper functions: merge, connect, remove nested boxes
│
├─ config.py                  # Input and output folder definitions
├─ main.py                    # Main pipeline for processing all images
├─ requirements.txt           # Dependencies
└─ README.md                  # Project documentation

---

Usage

1. Place your TIFF images in the INPUT_FOLDER defined in config.py.
2. Run the main script:

python main.py

3. The pipeline will:

• Load and normalize each image from the INPUT_FOLDER.
• Compute Sobel edges in both X and Y directions to detect gradients.
• Extract rough vertical and horizontal masks using intensity thresholds.
• Clean line masks with morphological operations and remove unwanted edges.
• Extract horizontal boards, group them into lines, filter small or invalid lines, and generate labeled images.
• Extract vertical boards, split oversized boxes, remove nested boxes, merge fragmented verticals, and fill missing gaps.
• Combine labeled boards onto the normalized image with horizontal boards in pink and vertical boards in green.
• Save results in a structured folder per image under OUTPUT_FOLDER, including:
  • *_normalized.png – normalized input image:
  • *_horizontal.png – labeled horizontal boards.
  • *_vertical.png – labeled vertical boards.
  • *_combined_boards.png – visualization with both horizontal and vertical boards.

---

Notes

• The project is optimized for grayscale TIFF images, but can be adapted for other formats.
• Adjustable parameters:
  • Morphological kernel sizes (ksize) for connecting fragments.
  • Area thresholds for filtering contours.
  • Tolerances for merging broken vertical boards.
  • Intensity ranges for line extraction.
• Virtual boards (to fill gaps) are marked differently in the labeled images for visual distinction.
• Each processed image has its own subfolder to keep outputs organized.

---

License

MIT License © 2025 Vicente Brisa Saez
GitHub: https://github.com/Vicen-te