An integrated method for railway fastener defect detection and geometric parameter measurement using 3D line laser sensor

Railway fasteners are key components that maintain track stability and ensure train operation safety. Automatic detection technologies for fastener defects have been widely adopted, but high-precision measurement of fastener geometric parameters still relies on manual operation, which is associated with low efficiency and significant measurement errors. This paper proposes an integrated method for railway fastener defect detection and geometric parameter measurement based on a 3D line laser sensor. A 3D imaging system is constructed to acquire RGB depth images and corresponding point clouds with a precise one-to-one mapping relationship. The YOLOv8 network is used to detect visual defects and locate intact fastener regions from RGB depth images, which are then mapped to the point clouds to extract fastener point cloud data. The PointNet++ network is adopted to segment fastener components, and the specifications of insulated blocks, the thicknesses of height adjustment pads, and bolt heights are calculated based on the spatial structure of fastener components. Experimental results demonstrate that the YOLOv8 model achieves 97.7% precision and 95.9% recall for visual defect detection, and 99.6% precision and 99.8% recall for intact fastener localization. All critical geometric measurements satisfy the corresponding railway engineering tolerances. 98.7% of HPuR measurements have errors below the 0.5 mm tolerance. HPuIP components with a 10 mm specification step exhibit a maximum measurement error of less than 2.5 mm, well below the 5 mm tolerance. 99% of insulated block measurements achieve errors below 0.5 mm tolerance at horizontal sampling intervals no greater than 0.4 mm. Bolt height measurement achieves no less than 90% precision and 91% recall for severe fastener loosening detection. The system operates at 4.32 km/h, making it suitable for hand-pushed on-site inspection. The proposed method realizes automatic defect detection and high-precision geometric parameter measurement in a unified framework, which can effectively replace manual inspection and significantly improve the intelligence and efficiency of railway track maintenance.