Encoder-Decoder Architectures for Crack Detection on Surfaces: A Deep Learning Approach

The building and maintenance of structures like roads, expressways, buildings, skyscrapers, and so forth requires a sizable workforce as well as a sizable financial investment in the fields of civil engineering and construction. Following completion, it is necessary to regularly check for deformatio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Ingénierie des systèmes d'Information Jg. 30; H. 2; S. 495 - 503
Hauptverfasser: Sakshi, Vijay, Richa, Lodhi, Sachin, Noonia, Ajit, Berar, Gagandeep, Kumar, Ajay
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Edmonton International Information and Engineering Technology Association (IIETA) 01.02.2025
Schlagworte:
Accuracy
Algorithms
Cost control
Cracks
Datasets
Deep learning
Deformation
Encoders-Decoders
Fractures
Localization
Machine learning
Neural networks
Nuclear power plants
Real time
Research methodology
Road maintenance
Skyscrapers
ISSN:1633-1311, 2116-7125
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The building and maintenance of structures like roads, expressways, buildings, skyscrapers, and so forth requires a sizable workforce as well as a sizable financial investment in the fields of civil engineering and construction. Following completion, it is necessary to regularly check for deformations such as fractures, the structure's outermost layer peeling off, rusting, etc. To maintain the safety of both human life and the building itself, such deformations must therefore be continuously monitored and repaired. We present a neural network-based method to find cracks in such structures in addition to physical inspection. Using 2500 images as training data, the model had an accuracy of 95.0%; on the validation set, it had a mean IoU score of 83%. The proposed method also demonstrates superior performance, achieving a 15% increase in prediction accuracy when compared to state-of-the-art methods, thereby illustrating its worth in real-time applications.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1633-1311
2116-7125
DOI:10.18280/isi.300221