基于图像区域划分的轨道缺陷自动检测技术研究

Research on rail defects automatic detection technology based on image region partition

轨道表面缺陷检测是保障铁路运输安全的重要手段之一,本文设计了基于机器视觉的轨道缺陷自动检测系统,并对钢轨表面缺陷提取技术进行了研究。改进了最大类间方差自适应阈值分割算法,提出了基于轨道峰区检测的自适应二值图像投影法快速提取钢轨表面区域;最后,采用内部点掏空法和链码跟踪算法获取并存储缺陷轮廓信息,实现了钢轨表面缺陷区域的自动检测。实验结果表明:本文所采用的方法可快速定位钢轨区域,并自动准确提取缺陷图像,平均每幅图像耗时11 ms,从而为后续缺陷的测量和识别奠定基础。

Rail surface defects detection is one of the most important approaches to ensure the safety of railway transportation.A rail surface defect automatic inspecting system based on machine vision has been designed and the rail surface defect region extraction techniques are investigated.Otsu adaptive threshold segmentation algorithm is improved and the rail surface region is extracted by a method of adaptive binary image projection based on peak district detecting.Finally,the method of hollowing out interior point and chain code tracking algorithm are used to obtain and store the defect contour information.Using this method,the rail surface defect region can be automatically detected.The experiment results indicate that the proposed method can quickly locate the region of rail surface and accurately and automatically extract the defect image.The processing of each image consumes an average time of 11 ms,which shows that the proposed method has a practical application prospect.