# Automatic Crack Detection in Built Infrastructure Using Unmanned Aerial Vehicles

**Authors:** Manh Duong Phung, Van Truong Hoang, Tran Hiep Dinh, Quang Ha  
**arXiv:** [1707.09715](https://arxiv.org/abs/1707.09715)  
**Date:** 2017-07-31  
**Summary page:** https://www.summarizepaper.com/en/paper/1707.09715v1/

## Key points

- Comprehensive approach to crack detection in built infrastructure using UAVs
- Two-stage process: data collection and crack detection
- Data collection stage involves creating a 3D model of the structure using laser scanners
- Geometric properties extracted from the model to generate waypoints for UAV navigation and image capture
- Images stitched together from overlapping fields of view
- Crack detection involves histogram analysis and peak detection techniques on the resulting image
- Image clustering based on histogram enables identification of potential cracks using locally adaptive thresholds
- Automated process improves inspection time and minimizes safety hazards
- Prototypical system developed and experiments conducted to evaluate the approach
- Experimental results demonstrate effectiveness in detecting cracks in built infrastructure
- Detailed description of an automated crack detection system utilizing UAVs for data collection and employing histogram analysis for crack identification
- Proposed approach shows promise for improving health monitoring of built infrastructure by reducing inspection time and enhancing safety measures.

## Summary

This paper presents a comprehensive approach to crack detection in built infrastructure using unmanned aerial vehicles (UAVs). The authors propose a two-stage process, consisting of data collection and crack detection. In the data collection stage, a 3D model of the structure is created using laser scanners. From this model, geometric properties are extracted to generate waypoints for the UAV to navigate and capture images of the structure. The obtained images are then stitched together from overlapping fields of view. For crack detection, histogram analysis and peak detection techniques are employed on the resulting image. By clustering the image based on its histogram, potential cracks can be identified using locally adaptive thresholds. This automated process significantly improves inspection time while minimizing safety hazards. To evaluate their approach, the authors have developed a prototypical system and conducted experiments. The experimental results demonstrate the effectiveness of their method in detecting cracks in built infrastructure. Overall, this paper provides a detailed description of an automated crack detection system that utilizes UAVs for data collection and employs histogram analysis for crack identification. The proposed approach shows promise for improving health monitoring of built infrastructure by reducing inspection time and enhancing safety measures.