Related Experiment Video
Updated: Jul 5, 2025
14:55
Evaluating the Effect of Roadside Parking on a Dual-Direction Urban Street
Published on: January 20, 2023
3.3K
Automatic Damage Detection of Pavement through DarkNet Analysis of Digital, Infrared, and Multi-Spectral Dynamic
Hyungjoon Seo1, Yunfan Shi2, Lang Fu1
1Department of Civil and Environmental Engineering, University of Liverpool, Liverpool L69 7WW, UK.
Sensors (Basel, Switzerland)
|January 23, 2024
Summary
This study automatically classified road pavement damage using DarkNet on 13,500 images. Digital images achieved 97.4% accuracy, showing the potential for automated road inspection and maintenance.
Area of Science:
- Road safety engineering
- Computer vision for infrastructure monitoring
- Pavement management systems
Background:
- Road pavement safety requires automated damage assessment and repair.
- Pavement surfaces contain diverse features beyond cracks, including manholes and markings.
- Distinguishing between damage and non-damage elements is crucial for accurate analysis.
Purpose of the Study:
- To develop an automated system for classifying road pavement conditions.
- To evaluate the effectiveness of different imaging techniques (digital, IR, MSX) for damage detection.
- To introduce a method for analyzing crack directionality.
Main Methods:
- Collected 13,500 digital, Infrared (IR), and Multispectral (MSX) images of road pavements.
- Utilized the DarkNet deep learning model for automatic classification of nine distinct categories.
- Applied a two-dimensional wavelet transform to determine crack directionality.
Main Results:
- DarkNet achieved high classification accuracy: 97.4% for digital images, 80.1% for IR, and 91.1% for MSX.
- MSX images, an enhancement of IR, performed better than IR but slightly lower than digital images.
- The study demonstrated the feasibility of automated classification and crack direction analysis.
Conclusions:
- Automated classification using DarkNet is effective for identifying road pavement features and damages.
- Digital images offer the highest accuracy, while MSX images provide a valuable complementary dataset.
- Crack directionality detection using wavelet transforms can enhance future pavement inspection research.
