Foreign object detection for railway ballastless trackbeds: A semisupervised learning method. (28th February 2022)

• Record Type: Journal Article
• Authors: Chen, Zhengxing; Wang, Qihang; Yu, Tianle; Zhang, Min; Liu, Qibin; Yao, Jidong; Wu, Yanhua; Wang, Ping; He, Qing
• Is Part Of: Measurement. Volume 190(2022)
• Publication Date: 2022-02-28
• Subjects: Foreign object detection -- Ballastless trackbed -- Semisupervised learning -- Deep learning -- Convolutional neural networks; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.110757
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Detect foreign objects on ballastless trackbed with semisupervised learning. Develop an enhanced deep SVDD model for the detection. Devise a Mask R-CNN algorithm to segment and extract rail and fastener areas. Analyze the performance of deep SVDD with many image preprocessing methods. Abstract: This paper proposes a semisupervised algorithm for detecting foreign objects in ballastless beds based on the improved deep SVDD (Support Vector Data Description) algorithm. First, we use the improved Mask R-CNN algorithm to extract the rail and fastener areas in images, assuming that no foreign object exists in the rail and fastener areas. Second, we deepen the backbone network of the deep SVDD to enhance its ability to extract deep semantics from complex images. We perform pure color coverage processing with different colors and mean blur processing with different blur kernels on the rails and fastener regions extracted by the improved Mask R-CNN. The results show that the AUC (Area Under the Curve) of our improved deep SVDD algorithm is 89.23% and improves the AUC compared to that of the original model by 11.09%.