{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,3]],"date-time":"2026-07-03T07:51:41Z","timestamp":1783065101467,"version":"3.54.6"},"reference-count":28,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2023,2,9]],"date-time":"2023-02-09T00:00:00Z","timestamp":1675900800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Neurorobot."],"abstract":"<jats:sec><jats:title>Introduction<\/jats:title><jats:p>The surface images of steel rails are extremely difficult to detect and recognize due to the presence of interference such as light changes and texture background clutter during the acquisition process.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>To improve the accuracy of railway defects detection, a deep learning algorithm is proposed to detect the rail defects. Aiming at the problems of inconspicuous rail defects edges, small size and background texture interference, the rail region extraction, improved Retinex image enhancement, background modeling difference, and threshold segmentation are performed sequentially to obtain the segmentation map of defects. For the classification of defects, Res2Net and CBAM attention mechanism are introduced to improve the receptive field and small target position weights. The bottom-up path enhancement structure is removed from the PANet structure to reduce the parameter redundancy and enhance the feature extraction of small targets.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>The results show the average accuracy of rail defects detection reaches 92.68%, the recall rate reaches 92.33%, and the average detection time reaches an average of 0.068 s per image, which can meet the real-time of rail defects detection.<\/jats:p><\/jats:sec><jats:sec><jats:title>Discussion<\/jats:title><jats:p>Comparing the improved method with the mainstream target detection algorithms such as Faster RCNN, SSD, YOLOv3 and other algorithms, the improved YOLOv4 has excellent comprehensive performance for rail defects detection, the improved YOLOv4 model obviously better than several others in <jats:italic>P<\/jats:italic><jats:sub><jats:italic>r<\/jats:italic><\/jats:sub>, <jats:italic>R<\/jats:italic><jats:sub><jats:italic>c<\/jats:italic><\/jats:sub>, and F1 value, and can be well-applied to rail defect detection projects.<\/jats:p><\/jats:sec>","DOI":"10.3389\/fnbot.2023.1119896","type":"journal-article","created":{"date-parts":[[2023,2,9]],"date-time":"2023-02-09T09:40:13Z","timestamp":1675935613000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":28,"title":["Research on steel rail surface defects detection based on improved YOLOv4 network"],"prefix":"10.3389","volume":"17","author":[{"given":"Zengzhen","family":"Mi","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ren","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shanshan","family":"Zhao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2023,2,9]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2004.10934.","article-title":"YOLOv4: optimal speed and accuracy of object detection.","author":"Bochkovskiy","year":"2020","journal-title":"arXiv"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.1016\/j.ultras.2020.106304","article-title":"A fast technique for hyper-echoic region separation from brain ultrasound images using patch based thresholding and cubic B-spline based contour smoothing.","volume":"111","author":"Chel","year":"2020","journal-title":"Ultrasonics"},{"key":"B3","doi-asserted-by":"publisher","first-page":"2584","DOI":"10.1109\/IJCNN.2016.7727522","article-title":"Deep convolutional neural networks for detection of rail surface defects","author":"Faghih-Roohi","year":"2016","journal-title":"Proceeding of the 2016 international joint conference on neural networks (IJCNN)"},{"key":"B4","doi-asserted-by":"publisher","first-page":"7935","DOI":"10.1109\/JSEN.2017.2761858","article-title":"A hierarchical extractor-based visual rail surface inspection system.","volume":"17","author":"Gan","year":"2017","journal-title":"IEEE Sens. J."},{"key":"B5","doi-asserted-by":"publisher","first-page":"652","DOI":"10.1109\/TPAMI.2019.2938758","article-title":"Res2Net: a new multi-scale backbone architecture.","volume":"43","author":"Gao","year":"2021","journal-title":"IEEE Trans. Pattern Analy. Mach. Int."},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijleo.2022.168607","article-title":"Non-contact detection of railhead defects and their classification by using convolutional neural network.","volume":"253","author":"Ghafoor","year":"2022","journal-title":"Optik"},{"key":"B7","doi-asserted-by":"publisher","DOI":"10.3390\/s22093467","article-title":"MSFT-YOLO: improved YOLOv5 based on transformer for detecting defects of steel surface.","volume":"22","author":"Guo","year":"2022","journal-title":"Sensors"},{"key":"B8","first-page":"41","article-title":"Damage detection method for rail surface based on multi-level feature fusion.","volume":"42","author":"Han","year":"2021","journal-title":"China Railway Sci."},{"key":"B9","doi-asserted-by":"publisher","first-page":"2011","DOI":"10.1109\/TPAMI.2019.2913372","article-title":"Squeeze-and-excitation networks.","volume":"42","author":"Hu","year":"2020","journal-title":"IEEE Trans. Pattern Analy. Mach. Int."},{"key":"B10","doi-asserted-by":"publisher","DOI":"10.3390\/s21103483","article-title":"HADF-crowd: a hierarchical attention-based dense feature extraction network for single-image crowd counting.","volume":"21","author":"Ilyas","year":"2021","journal-title":"Sensors"},{"key":"B11","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2209.02976","article-title":"YOLOv6: a single-stage object detection framework for industrial applications.","author":"Li","year":"2022","journal-title":"arXiv"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46448-0_2","article-title":"SSD: single shot multibox detextor","author":"Liu","year":"2016","journal-title":"Computer vision \u2013 ECCV 2016. ECCV 2016. Lecture notes in computer science"},{"key":"B13","doi-asserted-by":"publisher","first-page":"324","DOI":"10.3788\/LOP202158.2212001","article-title":"Rail surface defect detection based on image enhancement and improved cascade R-CNN.","volume":"58","author":"Luo","year":"2021","journal-title":"Laser Optoelect. Prog."},{"key":"B14","doi-asserted-by":"publisher","first-page":"5806","DOI":"10.1109\/TITS.2021.3058635","article-title":"Detection for rail surface defects via partitioned edge feature.","volume":"23","author":"Ni","year":"2021","journal-title":"IEEE Trans. Int. Trans. Syst."},{"key":"B15","doi-asserted-by":"crossref","first-page":"3099","DOI":"10.1109\/ICSMC.2004.1400815","article-title":"Background subtraction techniques: a review","author":"Piccardi","year":"2004","journal-title":"Proceeding of the 2004 IEEE international conference on systems, man and cybernetics (IEEE Cat. No.04CH37583)"},{"key":"B16","doi-asserted-by":"publisher","first-page":"2550","DOI":"10.1016\/j.trpro.2022.06.294","article-title":"Critical analysis of RCF rail defects classification.","volume":"63","author":"Popovi\u0107","year":"2022","journal-title":"Trans. Res. Proc."},{"key":"B17","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1804.02767","article-title":"YOLOv3: an incremental improvement.","author":"Redmon","year":"2021","journal-title":"arxiv"},{"key":"B18","doi-asserted-by":"publisher","first-page":"6615","DOI":"10.3233\/JIFS-210475","article-title":"Automatic road crack detection and classification using multi-tasking faster RCNN.","volume":"41","author":"Sekar","year":"2021","journal-title":"J. Int. Fuzzy Syst."},{"key":"B19","doi-asserted-by":"publisher","first-page":"45","DOI":"10.23919\/ICACT.2018.8323642","article-title":"Detection of rail surface defects based on CNN image recognition and classification","author":"Shang","year":"2018","journal-title":"Proceeding of the 2018 20th International Conference on Advanced Communication Technology (ICACT)"},{"key":"B20","first-page":"82","article-title":"Vision detection of tunnel cracks based on local image texture calculation.","volume":"40","author":"Wang","year":"2018","journal-title":"J. China Railway Soc."},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01234-2_1","article-title":"CBAM: convolutional block attention module","author":"Woo","year":"2018","journal-title":"Computer vision \u2013 ECCV 2018. ECCV 2018. Lecture notes in computer science"},{"key":"B22","doi-asserted-by":"publisher","DOI":"10.1016\/j.measurement.2021.110351","article-title":"Research on defect detection of high-speed rail based on multi-frequency excitation composite electromagnetic method.","volume":"187","author":"Xu","year":"2022","journal-title":"Measurement"},{"key":"B23","doi-asserted-by":"publisher","first-page":"649","DOI":"10.1109\/TIP.2018.2810539","article-title":"A novel retinex based enhancement algorithm considering noise","author":"Yu","year":"2017","journal-title":"Proceeding of the 2017 IEEE\/ACIS 16th international conference on computer and information science (ICIS)"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1772","DOI":"10.3788\/OPE.20162407.1772","article-title":"Rail image segmentation based on Otsu threshold method.","volume":"24","author":"Yuan","year":"2016","journal-title":"Optics Precis. Eng."},{"key":"B25","doi-asserted-by":"publisher","first-page":"4389","DOI":"10.1016\/j.ceramint.2020.10.065","article-title":"Review of non-destructive testing methods for defect detection of ceramics.","volume":"47","author":"Zhao","year":"2021","journal-title":"Ceramics Int."},{"key":"B26","doi-asserted-by":"publisher","DOI":"10.1117\/12.2037345","article-title":"An improved vein image segmentation algorithm based on SLIC and niblack threshold method","author":"Zhou","year":"2013","journal-title":"proceeding of the 2013 International conference on optical instruments and technology: optoelectronic imaging and processing technology"},{"key":"B27","doi-asserted-by":"publisher","DOI":"10.3390\/rs13091633","article-title":"Forest 3D reconstruction and individual tree parameter extraction combining close-range photo enhancement and feature matching.","volume":"13","author":"Zhu","year":"","journal-title":"Remote Sens."},{"key":"B28","doi-asserted-by":"publisher","first-page":"2778","DOI":"10.1109\/ICCVW54120.2021.00312","article-title":"TPH-YOLOv5: improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios","author":"Zhu","year":"","journal-title":"Proceeding of the 2021 IEEE\/CVF international conference on computer vision workshops (ICCVW)"}],"container-title":["Frontiers in Neurorobotics"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fnbot.2023.1119896\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,2,9]],"date-time":"2023-02-09T09:40:23Z","timestamp":1675935623000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fnbot.2023.1119896\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,2,9]]},"references-count":28,"alternative-id":["10.3389\/fnbot.2023.1119896"],"URL":"https:\/\/doi.org\/10.3389\/fnbot.2023.1119896","relation":{},"ISSN":["1662-5218"],"issn-type":[{"value":"1662-5218","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,2,9]]},"article-number":"1119896"}}