Real-time biscuit tile image segmentation method based on edge detection

Area of Science:

• Computer Vision
• Image Processing
• Industrial Automation

Background:

• Automating ceramic tile inspection traditionally relies on human operators.
• Existing computer vision methods struggle to meet the real-time demands of tile production lines.
• Accurate segmentation of tile pixels from the background is crucial for automated visual inspection.

Purpose of the Study:

• To propose a novel real-time Biscuit Tile Segmentation (BTS) method.
• To enable automated visual inspection of ceramic tiles on a production line.
• To develop a segmentation technique that meets strict real-time constraints.

Main Methods:

• The Biscuit Tile Segmentation (BTS) method employs signal change detection and contour tracing algorithms.
• The BTS method is optimized for parallel execution on Graphics Processing Unit (GPU) devices.
• The approach focuses on distinguishing tile pixels from background elements in captured images.

Main Results:

• The BTS method achieves real-time performance, suitable for high-speed production lines.
• The proposed BTS method demonstrates superior performance compared to 2D threshold-based, 1D edge detection, and contour-based methods.
• The segmentation accuracy is sufficient for practical application in industrial settings.

Conclusions:

• The developed BTS method effectively automates the visual inspection of biscuit tiles.
• The GPU implementation ensures the method meets the real-time requirements of ceramic tile production.
• The BTS method offers a significant improvement over existing techniques for tile segmentation in industrial environments.