基于机器视觉的晶粒度面积法自动评级

根据国标中晶粒度的定义提出进行晶粒度测量自动评级方法,论述了该方法中应用计算机图像处理技术进行图像尺寸标定、二值化变换、消除噪声、提取晶界和计算晶粒个数,最终计算晶粒度评级参数等过程的算法及实现.提取单像素晶界可解决不同晶粒晶界共用计数难问题;利用链码计算晶粒个数可通过限制链长在计数的同时把第二相、杂质等去除.该评级方法可以避免晶粒度测量评级中人为因素的影响,速度快,精确性高.对于非均匀等轴晶的评级比其他方法如截线法、截圆法等具有更好的适用性.

Automatic Grading of Grain Size by Planimetry Based on Machine Vision

The automatic measure grading method of grain size was put forward based on the definition of grain size in the Chinese Standards.Image dimension calibration,dimension offset from a real picture to a digital image,noise elimination,grain boundary extraction and grains count were performed by using computer image processing techniques.The difficulty of automatic grain counting due to one grain boundary shared by tow adjacent grains can be solved by picking up single-image-element grain boundary.Second phases and impurities can be removed by limiting the length of freeman chain during grain counting.This grading method can avoid man-made mistakes,and has the advantages of high speed and high accuracy.For non-uniform equiaxed crystal,this method is better than others like intercept and so on.