钢轨扣件横向偏移特征检测算法研究

王恩鸿柴晓冬钟倩文李立明张乔木

Research on Detection Algorithm of Lateral Offset Characteristics of Rail Fasteners

WANG EnhongCHAI XiaodongZHONG QianwenLI LimingZHANG Qiaomu
摘要:
针对无砟轨道中钢轨扣件发生横向松动、脱离正常工作位置产生偏移的问题,提出一种钢轨扣件横向偏移检测法。首先,该算法为解决传统的扣件图像定位不够精准问题,采用kmeans聚类和类二值算法强化分割前景、背景与轮廓矩特征,实现对采集图像中扣件位置的精准定位;其次,不同于传统扣件特征提取采用复杂语义,提出一种基于机器视觉的轮廓分析方法,通过提取扣件的绝缘帽与螺母的轮廓特征,计算相邻绝缘帽间距和相邻螺母间距,并与安全状态下扣件轮廓特征计算得到的安全距离阈值进行对比,进一步计算偏移量,从而判断扣件是否发生横向松动。结果表明:该算法计算速度快,能够准确地定位弹条位置和偏移量,与传统的识别算法得到扣件的偏移量数据相比准确率显著提高,可达98%。
Abstracts:
Targeting the problem that the fasteners in ballastless track are laterally loosened and off the normal working position, a lateral offset detection method of the track fastener is proposed. Firstly, the algorithm solves the problem that the conventional fastener image positioning is not accurate enough. The k-means clustering and class-like algorithm are used to strengthen the segmentation of foreground, background and contour moment features to achieve accurate positioning of the fastener position in the captured image. Secondly, different from conventional fastener feature extraction using complex semantics, a machine vision based contour analysis method is proposed, which calculates the spacing of adjacent insulating caps and the spacing of adjacent nuts by extracting the contour features of the insulating cap and nut of the fastener. Compared with the safety distance threshold calculated by the fastener profile feature in the safe state, the offset is further calculated to determine whether the fastener is laterally loose. The experimental results show that the algorithm is fast and can accurately locate the position and offset of the spring. Compared with the conventional identification algorithm, the accuracy of the fastener is significantly improved, up to 98%.
引文 / Ref:
王恩鸿,柴晓冬,钟倩文,钢轨扣件横向偏移特征检测算法研究[J].城市轨道交通研究,2021,24(5):142.
WANG Enhong,CHAI Xiaodong,ZHONG Qianwen,Research on Detection Algorithm of Lateral Offset Characteristics of Rail Fasteners[J].Urban mass transit,2021,24(5):142.
论文检索