Deep Learning for Concrete Crack Detection and Measurement
Problem
Detecting and measuring cracks in concrete structures automatically, in practical millimetre units rather than pixels, where traditional image processing struggles.
Why it matters
Undetected worsening cracks can cause catastrophic structural failures with costly human and economic consequences.
What the AI did
A three-stage pipeline used a custom CNN for crack classification, a U-Net for segmentation, and a novel laser calibration method to measure crack width in millimetres.
Results
99.22% classification accuracy; 96.54% segmentation accuracy; crack-width MAE value truncated in abstract
Impact
Evidence supports high-accuracy automated crack detection and millimetre-scale measurement on the studied data.
Record
| Venue | Metrology |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Mthabisi Adriano Nyathi, Jiping Bai, Ian David Wilson |
| Citations | 55 (OpenAlex, at import time) |
| References | 43 |
| Model or method | Custom CNN and U-Net |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/metrology4010005 |
| Open access | gold |
| Keywords | deep learning, crack detection, CNN, U-Net, segmentation, laser calibration |
Importance rating
significant — thousands to millions affected
Abstract states undetected worsening cracks can cause catastrophic failures with costly implications for human life and the economy; broad infrastructure safety relevance.
More in Engineering & Materials
| # | Paper | Year | Citations | Importance |
|---|---|---|---|---|
| 1 | Scaling deep learning for materials discovery | 2023 | 1,471 |
8
|
| 2 | Application of Machine Learning in Material Synthesis and Property Prediction | 2023 | 165 |
7
|
| 3 | Machine learning-guided discovery of ionic polymer electrolytes for lithium metal batteries | 2023 | 115 |
7
|
| 4 | Machine Learning for Structural Health Monitoring of Aerospace Structures: A Review | 2025 | 44 |
7
|
| 5 | Power System Transient Stability Assessment Based on Machine Learning Algorithms and Grid Topology | 2023 | 42 |
7
|