铁路扣件弹条伤损自动检测系统研发与验证
Development and verification of automatic inspection system for high-speed railway fastener

针对铁路扣件缺陷自动识别准确率不高等问题，介绍基于三维激光成像技术的高速铁路扣件自动检测系统，重点阐述基于轨道三维图像的弹条型扣件的检测算法。在扣件检测算法中，以先验知识验证扣件位置的方法保证扣件定位的准确率；实现弹条图像的提取；基于正常弹条图像创建真实模拟折断扣件的虚拟负样本。提取图像特征后，利用经过训练的分类器识别扣件缺失、单侧断裂和双侧断裂的状态。最后，结合室内实尺Ⅲ型板轨道模型的数据验证系统的检测效果。研究结果表明，采用提出的扣件缺陷自动检测算法的识别准确率高达97.3%，该系统具有较高的实用潜力。
Aiming at the low recognition accuracy of detection algorithms for railway fastener defects, a novel high-speed railway fastener automatic inspection system was developed using three-dimension (3D) laser imaging and detection algorithm based on track 3D image for hook-shaped fastener. The detection algorithm has some distinctive characteristics. Firstly, the method of verifying fastener location with prior-knowledge was applied to improve the correct rate of locating fasteners. To the best of our knowledge, this is the first time to extract clip images. Secondly, the algorithm of creating visually defective fastener images was proposed to simulate broken fastener based on intact clip images. After extracting the feature of images, the fastener defective statuses, such as missing, one-side broken and two-side broken were identified by the special classifier after training. The experimental results of this detection algorithm verifying with the data of indoor real-scale model indicate that the detection accuracy of this algorithm is as high as 97.3%, and this inspection system has a high potential for field implementation