摘要
粗糙集理论能够在保持系统分类不变的基础上,发现系统内的基本知识。利用Kirsch算子进行图像预处理,依据粗糙集的分类原理将火焰图像的高低温区域分离开来,建立特征量提取的初始模型。根据对燃烧特性的分析确定了4个特征变量作为诊断的依据,根据全炉膛火焰图像的特点确定特征量的计算模型,为便于比较研究,对每个特征量均作离散化处理。由于单个特征量只能部分反映燃烧状态,同时,为克服单变量控制的抗干扰性能差、错误率较高的缺点,按照粗糙集的约简原则,在比较均方差σ基础上选用不同的特征量组合作为判断燃烧状态的参数,构建基于粗糙集约简的多变量融合的状态识别规则。实验表明,采用粗糙集方法可以有效地提高处理速度。
Rough sets theory can discover the basic knowledge without changing system classification. After pre- processing with Kirsch arithmetic operator,the flame image is divided into two areas of high and low temperatures by classification means of rough sets theory in order to build the model for characteristics extraction. Four characteristic variables are taken as diagnos to combustion analysis,and their calculation models are determined based on variable is discretized for easy comparison. As single variable can only tic criteria according image identity. Each partially reflect the combustion state,as well as its poor anti-jamming ability and higher state recognition error ratio, different combinations of characteristic variables are taken as combustion diagnostic criteria based on the comparison of mean quadratic error according to the reduction rules of rough sets theory. The experiment shows that the method based on rough sets theory enhances processing speed effectively.
出处
《电力自动化设备》
EI
CSCD
北大核心
2007年第5期84-87,121,共5页
Electric Power Automation Equipment
关键词
粗糙集
图像处理
状态识别
火焰检测
决策表
rough sets
image processing
state recognition
flame detection
decision-making table
