Research and implementation of image preprocessing technology for rail fasteners based on OpenCV
Abstract
As the critical connecting components of railway lines, the operational status of rail fasteners directly impacts railway safety and track stability. Machine vision-based fastener condition detection technology has become a research hotspot due to its non-contact operation and real-time efficiency advantages. However, field-collected images are prone to interference from lighting fluctuations and background noise, resulting in information redundancy and low targetbackground contrast that hinder subsequent feature extraction. To address this challenge, this study designs a streamlined image preprocessing workflow for rail fasteners using OpenCV to eliminate redundant information and enhance target features. By selecting 1,000 field images under varying operational conditions as samples and combining subjective evaluations with objective metrics including PSNR and SSIM, the effectiveness of this approach was validated. Experimental results demonstrate that post-processing improves PSNR by 35.2% and average SSIM by 0.25 compared to original images. Compared to single preprocessing algorithms, the PSNR and SSIM values increased by 19.3% and 0.11 respectively, providing high-quality input data for subsequent processes with significant practical engineering value.