Answers are grounded in this paper and cite their sources.
Computational evaluation demonstrates accurate rail surface defect identification in railway images, indicating feasibility for self-powered inspection systems.
Main tasks in railway inspection includes identifying, locating and classifying rail surface defects.In this paper, we develop a three-stage deep learning based framework for automating multiple analytical tasks in the rail surface defect inspection via using rail images.It is capable of identifying the presence, the sizes with coordinates, and the category of various defects.The first stage employs an autoencoder based generative model to identify the input image containing defects, which then trigger the second stage with a segmentation model for locating defects at the pixel level.Finally, the defect category is classified with segmented and cropped defect regions obtained from the previous stages.The proposed method is evaluated thoroughly with different performance criteria on a real dataset.Moreover, due to limited publicly available datasets in railway inspection, we synthesized a new dataset to further verify the generalizability of the proposed framework.Results of the computational studies validated its accuracy and suitability on a self-powered inspection system in terms of the computational load, performance, and inference time.
No takes yet. Share an insight, caveat, or question.
Tianxin Wang (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: