Abstract:

A non-contact surface roughness inspection method based on machine vision was used to assess different cylindrical grinding surfaces from 01μm to 08μm in different ambient light environment. A new method that counts the pixel numbers whose gray value is larger than sum of the Otsu threshold and a given constant was proposed to represent different roughness in surface inspection.This method was compared with the methods based on standard deviation of gray-level distribution proposed by Luk and the gray-level co-occurrence matrix.Results show that variance sum between inspecting values of the proposed method and its corresponding real value is less than half of the other two ways. 

Key words: surface roughness, machine vision, measurement, threshold

摘要：

对0.1~0.8μm不同粗糙度外圆磨削表面在不同外界环境光下的视觉检测进行了研究，提出一种新的粗糙度评定方法——基于域值的峰值采样法。首先，采用大津阈值法获得随工件图像变动的阈值；在此基础上，增加一常数以获得图像的采样阈值，该常数需在标定过程中进行优化取值；然后，统计采样区域内大于该阈值的点数以获得不同粗糙度表面对应特征量；最后，与基于灰度分布标准差方法及灰度共生矩阵法进行了实验比较，结果表明使用峰值采样法时，检测值同真实值间方差和小于其他两种方法方差和的二分之一。

关键词: 表面粗糙度, 机器视觉, 测量, 阈值