Fine-grained pixel-level crack mapping in earthen heritage structures

Area of Science:

• Civil Engineering
• Computer Science
• Heritage Conservation

Background:

• Heritage structures require reliable methods for detecting surface cracks to ensure preservation.
• Existing methods for crack detection can be inefficient or lack accuracy, posing challenges for heritage management.

Purpose of the Study:

• To propose an entirely automated framework for real-time surface crack detection and analysis in heritage structures.
• To develop an integrated system for efficient and accurate structural assessment using high-resolution imagery.

Main Methods:

• Image preprocessing and skeleton-based analysis were employed.
• A lightweight deep learning model was utilized for crack detection in controlled and real environments.
• An application was developed for image uploading and live camera input.

Main Results:

• The framework demonstrated highly accurate and consistent performance in real-time, on-field testing.
• The system successfully extracted structural information and detected cracks efficiently.
• The model proved effective in both controlled and real-world scenarios.

Conclusions:

• The proposed automated framework offers a scalable solution for structural assessment of heritage sites.
• The system enables rationalized decision-making for maintenance and heritage management.
• The developed tool supports practical site inspections with real-time crack analysis.