The inherent relationships about surface texture features and roughness of ground Si3N4 were investigated through the gray level co-occurrence matrix(GLCM) , based on surface images captured by digital microscope. The sampling offset and total gray levels of GLCM were 4 and 64 respectively. And then 4 matrices concerning 0,45, 90° and 135° were built up to acquire the average of each texture feature. Six features such as contrast and entropy were set as main texture parameters according to the correlation analysis, and the relations between texture features and roughness parameters were discussed later. Furthermore, the multiple nonlinear regression technique was applied to establish the forecasting models of texture features and roughness. The result shows that the deviation of predicted roughness and measured result is less than 0. 25, which reveals a good predictive effect of these models.%通过数字显微镜获取Si3N4磨削表面图像,运用灰度共生矩阵(GLCM)分析表面纹理特征与粗糙度的内在关系.根据实验确定GLCM采样点间距、总灰度级,按4个方向建立GLCM并计算纹理特征参数均值.利用变量相关分析,确定对比度等6个参数为Si3N4磨削表面主要纹理特征,并探讨各纹理参数与Ra,Ry,S,Tp的变化规律,从而定性评估Si3N4磨削表面粗糙度.采用多元非线性回归方法研究纹理特征参数与粗糙度评定指标的联系,构建4个回归预测模型.结果表明:模型计算值与实测值偏差小于0.25,具有较好的预测效果.
展开▼
