TY - CHAP
T1 - Artificial intelligence for classification of railway earth retaining structures
AU - Islam, Md Zahidul
AU - Leo, Chin Jian
AU - Zou, Ju Jia
AU - Liyanapathirana, Samanthika
AU - Hu, Pan
AU - Xiao, Bo
AU - Yuen, Stanley
PY - 2026
Y1 - 2026
N2 - Earth retaining structures (ERS) play a crucial role in the infrastructure of railway networks. It supports and protects railway networks from soil and rock erosion. However, different types of defects can form on ERS that can cause slope failure, soil erosion, and degradation of its structural integrity. Timely repair of these defects is essential. However, before detecting defects, it is important to classify ERS, as the occurrence and nature of defects often depend on the specific type of ERS. This study proposes the use of artificial intelligence (AI), specifically improved ConvNeXt V2, for automatically classifying three common types of railway ERS captured with GPS metadata from Transport for New South Wales (TfNSW) railway networks as a proof of concept. Classifying ERS is a necessary step prior to defect detection because defects of ERS are dependent on its type. The study also extracts the GPS coordinates of the classified ERS from the captured data to enable the inspector to locate and conduct further inspection of that ERS.
AB - Earth retaining structures (ERS) play a crucial role in the infrastructure of railway networks. It supports and protects railway networks from soil and rock erosion. However, different types of defects can form on ERS that can cause slope failure, soil erosion, and degradation of its structural integrity. Timely repair of these defects is essential. However, before detecting defects, it is important to classify ERS, as the occurrence and nature of defects often depend on the specific type of ERS. This study proposes the use of artificial intelligence (AI), specifically improved ConvNeXt V2, for automatically classifying three common types of railway ERS captured with GPS metadata from Transport for New South Wales (TfNSW) railway networks as a proof of concept. Classifying ERS is a necessary step prior to defect detection because defects of ERS are dependent on its type. The study also extracts the GPS coordinates of the classified ERS from the captured data to enable the inspector to locate and conduct further inspection of that ERS.
KW - Classification
KW - ConvNeXt V2
KW - Earth Retaining Structures (ERS)
KW - GPS metadata
KW - Railway
UR - https://www.scopus.com/pages/publications/105039299118
UR - https://go.openathens.net/redirector/westernsydney.edu.au?url=https://doi.org/10.1007/978-3-032-20885-9_10
U2 - 10.1007/978-3-032-20885-9_10
DO - 10.1007/978-3-032-20885-9_10
M3 - Chapter
AN - SCOPUS:105039299118
SN - 9783032208842
T3 - Lecture Notes in Civil Engineering
SP - 79
EP - 86
BT - 3rd International Conference on Geomechanics and Geoenvironmental Engineering: Proceedings of iCGMGE 2025
A2 - Hu, Pan
A2 - Nguyen, Giang
A2 - Leo, Chin
A2 - Wong, Henry
A2 - Liyanapathirana, Samanthika
PB - Springer
CY - Switzerland
T2 - 3rd International Conference on Geomechanics and Geoenvironmental Engineering, iCGMGE 2025
Y2 - 26 November 2025 through 28 November 2025
ER -