University of Twente Student Theses

Login

Generalization capabilities and performance analysis of CNNs for pavement crack detection

Mo, T. (2021) Generalization capabilities and performance analysis of CNNs for pavement crack detection.

[img] PDF
2MB
Abstract:Frequent loading may cause pavement cracks, which have hidden dangers to vehicles and pedestrians. Additionally, the cracks will also deteriorate over time thus the cracks should be found out in time. The current crack detection method is to recognize them by naked eyes, but it is time-consuming and non-accurate. Hence, it is necessary to develop automatic crack detection models. Currently, CNN-based road crack detection methods get more popular. The author evaluates the performance, generalization capabilities, and image processing time of state-of-the-art methods on three datasets using ODS, OIS, AP, processing time metrics. And the experiment results show the HED performs better in these three aspects, which reaches above ODS 0.75 in some datasets. The possible reason why HED is better at performance and generalization in crack detection is it outperforms on thin crack identification and robustness. Additionally, it also shows a greater prospect of real-time crack detection.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Programme:Computer Science BSc (56964)
Link to this item:http://purl.utwente.nl/essays/85670
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page