A hybrid lightweight encoder-decoder network for automatic bridge crack assessment with real-world interference

•An HLEDNet was presented for concrete bridge crack detection and measurement.•The lightweight design was integrated with HLEDNet backbone architecture.•The influence of annotating handwriting interference was highlighted.•Crack information was retrieved by considering the crack depth information.•T...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Measurement : journal of the International Measurement Confederation Jg. 216; S. 112892
Hauptverfasser: Deng, Jianghua, Lu, Ye, Lee, Vincent C.S.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.07.2023
Schlagworte:
Atrous convolutions
Automated inspection
Concrete bridge cracks
Crack measurement
Encoder-decoder architecture
Structural health monitoring
Automated inspection
Structural health monitoring
Crack measurement
Encoder-decoder architecture
Atrous convolutions
Concrete bridge cracks
ISSN:0263-2241, 1873-412X
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:•An HLEDNet was presented for concrete bridge crack detection and measurement.•The lightweight design was integrated with HLEDNet backbone architecture.•The influence of annotating handwriting interference was highlighted.•Crack information was retrieved by considering the crack depth information.•The unit conversion from pixel to mm was achieved for practical applications. Although many learning-based studies have been conducted to detect cracks, there are still many problems in practice, such as slow inference speed due to a large number of hyperparameters required in network architectures and compromised detection accuracy in different environments. To address these issues, the current study employed a Hybrid Lightweight Encoder-Decoder Network (HLEDNet) as an ad-hoc crack segmentation and measurement system on real-world images captured from various concrete bridges. The proposed HLEDNet model was trained and tested with 3000 annotated images with further extensive data augmentation, which achieved 86.92%, 85.71%, 86.31, and 86.01% in precision, recall, F1 score, and mean intersection over union (mIoU), respectively. A crack measurement module was proposed using combined postprocessing techniques, where the R-squared values of the regression lines in crack length and average crack width are 0.9857 and 0.9925, respectively. Finally, an experimental study was undertaken to convert the crack measuring unit from pixel to millimetre.
ISSN:0263-2241
1873-412X
DOI:10.1016/j.measurement.2023.112892