Deep Architecture for High-Speed Railway Insulator Surface Defect Detection: Denoising Autoencoder With Multitask Learning

The insulator is an important catenary component that maintains the insulation between the catenary and earth. Due to the long-term impact of railway vehicles and the environment, defects in the insulator are inevitable. Recently, automatic catenary inspection using computer vision and pattern recog...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on instrumentation and measurement Jg. 68; H. 8; S. 2679 - 2690
Hauptverfasser: Kang, Gaoqiang, Gao, Shibin, Yu, Long, Zhang, Dongkai
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.08.2019
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Artificial neural networks
Automation
Catenary
Computer vision
deep denoising autoencoder (DDAE)
defect detection
Defects
Feature extraction
High speed rail
high-speed railway
Image classification
Image segmentation
Impact strength
Inspection
Insulation
insulator
Insulators
Monte Carlo methods
multitask learning (MTL)
Neural networks
Noise reduction
Pattern recognition
Rail transportation
Surface defects
ISSN:0018-9456, 1557-9662
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The insulator is an important catenary component that maintains the insulation between the catenary and earth. Due to the long-term impact of railway vehicles and the environment, defects in the insulator are inevitable. Recently, automatic catenary inspection using computer vision and pattern recognition has been introduced to improve the safety of railway operation. However, achieving full automation of insulator defect detection is still very challenging due to the visual complexity of defects and the small number of defective insulators. To overcome these problems, this paper proposes a novel insulator surface defect detection system using a deep convolutional neural network (CNN). The proposed system consists of two stages. First, a Faster R-CNN network is adopted to localize the key catenary components, and the image areas that contain the insulators are obtained. Then, the classification score and anomaly score are determined from a deep multitask neural network that is composed of a deep material classifier and a deep denoising autoencoder. The defect state is determined by analyzing the classification score and anomaly score. Experiments of the catenary insulator defect detection along the Hefei-Fuzhou high-speed railway line indicate that the system can achieve high detection accuracy.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0018-9456
1557-9662
DOI:10.1109/TIM.2018.2868490