基于图像识别的耐火材料破损率计算方法研究

Method for calculating breakage rate of refractories based on image recognition

针对传统耐火材料抗热震性检测方法所存在的诸如效率低、安全性差、没有过程记录等问题,本文利用图像识别技术并结合机器学习建立了新的检测系统,提出了设定初始误差系数E计算连通域特征分割阈值T的算法和基于骨架线区间划分的破损率计算方法,并借助新检测系统对多种耐火材料样品进行抗热震性测试。结果表明,新检测系统中图像处理算法的分割误差、假阳性率和假阴性率均值分别为3.73%、4.89%和1.82%,耐火材料破损率计算方法计算速度快且计算精度高。新抗热震性检测系统大幅降低了检测工作量并能自动保存所有检测记录。

Aimed at the problems of low efficiency,poor security,and no process records in traditional method for detecting thermal shock resistance of refractories,a new detection system was established based on image recognition technology combined with machine learning,and an algorithm for the connected domain feature segmentation threshold T by setting an initial error coefficient E and a calculation method for the damage rate via the interval division of skeleton line were proposed.Thermal shock resistance tests of multiple refractories were carried out by using this detection system.The results show that the average segmentation error,false positive rate and false negative rate of the proposed image processing algorithm in the new detection system are 3.73%,4.89%and 1.82%,respectively.The proposed method for calculating breakage rate of refractories is faster and more precise.The new thermal shock resistance detection system can greatly reduce workload and automatically save all process records.