基于图像灰度梯度特征的钢轨表面缺陷检测

Rail surface defects detection based on gray scale gradient characteristics of image

利用机器视觉技术检测钢轨表面缺陷时,检测结果的准确性易受光照变化、钢轨表面反射不均和锈迹等因素的影响。为此,提出了基于图像灰度梯度特征的钢轨表面缺陷检测方法。首先,在设计钢轨表面缺陷检测装置的基础上,分析了钢轨图像中不同区域的灰度和梯度特征;然后,基于双边滤波思想设计了背景平滑滤波器,利用局部灰度和梯度变化信息自适应调整不同特征区域的平滑程度,对原图像平滑得到背景图像;最后,将原图像与背景图像差分,通过对差分图像阈值分割并利用连通区域标记法,实现钢轨表面缺陷检测。实验结果表明,该方法可以在凸显图像缺陷部分的同时,有效减弱光照变化、钢轨表面反射不均和锈迹的影响,对不同轨道环境下的疤痕和裂纹缺陷均取得了较好的检测效果,缺陷漏检率和误检率分别为5.79%和6.84%,具有一定的实用价值。

When the machine vision technology is adopted to detect rail surface defects,the accuracy of results is easily influenced by the change of illumination,uneven reflection of rail surface,rust,etc. Therefore,this paper proposes a detection method for rail surface defection based on gray scale gradient characteristics of image. Firstly,the gray scale and gradient characteristics of different regions in the rail image are analyzed with the proposed defects detection device. Then,the background smoothing filter is designed based on the idea of bilateral filtering. By using local gray scale and gradient change information to adjust the smoothing degrees of different feature region adaptively,the original image is smoothed to generate a background image. Finally,image difference is made. The detection of rail surface defects is realized by setting threshold for the differential image and using connected-component labeling method. The experimental results show that the defects region in the image can be highlighted. Meanwhile,the proposed method can reduce the impact of illumination change,uneven reflection of rail surface,rust,etc. Both scar and crack defects in different track environments have been detected effectively. The undetected rate and false detection rate are 5. 79% and 6. 84%,respectively,which mean that the proposed method has certain practical value.