基于机器视觉和SVM的AOD炉终点预测方法  

Prediction method of AOD furnace end point based on machine vision and SVM

作　　者：韩顺杰[1] 尤文[1] 马海涛[1] 孙明超[1] 

机构地区：[1]长春工业大学电气与电子工程学院,吉林长春130012

出　　处：《冶金自动化》2012年第6期20-23,共4页Metallurgical Industry Automation

基　　金：国家科技支撑计划项目(2007BAE17B01)

摘　　要：针对AOD炉冶炼中低碳铬铁合金工艺过程终点判别依靠人工看火操作的难题,提出了利用机器视觉技术模拟传统人工看火过程进行终点判别的方法。利用灰度级提取AOD炉口火焰特征,采用支持向量机(SVM)算法实现图像特征的训练和测试,对比了3种核函数的测试精度。实验结果表明,采用机器视觉技术提取火焰图像特征并与支持向量机结合的方法能够有效识别冶炼终点,并具有较好的识别精度。To consider complexity of manual flame-watch operation in end point discriminant of low carbon ferroehrome alloy smelting process for AOD furnace, a new end point discriminant method based on machine vision technology is presented to simulate traditional manual flame-watch process. AOD mouth flame characteristics is extracted through use of grey level,Support Vector Machine （SVM） al- gorithm is adopted to train and test image characteristics, and test precision of three kinds of kernel function is compared. Test results show that the method based on machine vision and SVM can effec- tively distinguish smelting end point and recognition accuracy is good.

关 键 词：AOD炉 机器视觉 火焰特征 支持向量机 

分 类 号：TF33[冶金工程—冶金机械及自动化]