Abstract: A non-contact roughness measurement method of the face milling surface is studied in this paper，which is based on computer vision theory. The face milling surface image is obtained by the image collection system and is preprocessed. Using Hough transform, the image is rotated to ensure the texture direction is vertical downward. Therefore, we can compute GLCM texture parameters only on this single direction, thus the computing time is saved greatly. The four GLCM texture parameters are extracted as the roughness characteristics of face milling surface. BP neural network is used to model the relations between the GLCM texture parameters and the roughness Ra, and the roughness detection model of face milling surface is established. The experiment results showed that using GLCM based on Hough transform can extract the roughness texture parameters quickly. The BP network model can measure the roughness of face milling surface precisely.

Key words: Surface roughness, hough transform, GLCM, texture feature, BP neural network

摘要： 研究基于计算机视觉的非接触式端铣表面粗糙度检测方法。通过图像采集系统获取端铣工件表面图像，并进行图像预处理。基于Hough变换旋转图像，提取特定方向下GLCM的特征参数，较大程度缩短了计算时间。确定了适于表征端铣表面粗糙度的4个GLCM特征参数。采用BP神经网络算法研究纹理特征参数与粗糙度Ra之间的关系，构建粗糙度BP神经网络检测模型。实验表明，基于Hough变换的GLCM纹理特征参数提取方法能快速实现粗糙度纹理特征的提取，而BP神经网络检测模型具有较好的预测效果，能够满足端铣表面粗糙度测量的精度要求。

关键词: 表面粗糙度, Hough变换, GLCM, 纹理特征, BP神经网络