模糊免疫网络算法在数字图像火焰监测中的应用  

The Application of Fuzzy Immune Network Algorithm to Flame Monitoring

作　　者：郭建民[1] 刘石[1] 姜凡[1] 

机构地区：[1]中国科学院工程热物理研究所,北京市海淀区100080

出　　处：《中国电机工程学报》2007年第2期1-5,共5页Proceedings of the CSEE

基　　金：国家863高技术基金项目(2003AA529290)~~

摘　　要：提出了一种基于模糊免疫网络算法对火焰数字图像进行分类的研究方法。该算法由2部分构成:免疫网络算法与模糊聚类算法。模糊免疫网络算法是非监督学习算法,它不需要了解类别的先验知识,能够随着燃烧状况的变化实现动态聚类,同时该算法克服了传统模糊聚类算法须事先确定聚类数的缺陷。利用现场所获得的火焰图像,运用数字图像处理技术提取其特征量,对其进行分类研究,通过观察火焰图像类别的变化来判断燃烧状态是否发生变化。试验结果证明了该方法能有效地判断燃烧状态的改变。The method based on fussy immune network algorithm was presented, by which the flame digital images can be classified better. The algorithm involves two parts, immune network algorithm and fussy clustering algorithm. Fussy immune network algorithm is unsupervised algorithm and it need not to learn the transcendental knowledge of image classification. Furthermore, dynamic clustering can be carried out along with the movement of combustion status by the algorithm. The method proposed overcomes the weakness of iraditional fuzzy clustering that must determine the class number in advance. The eigenvectors of flame image, which obtained from plant trial, has been extracted by digital image technology and classified, and the change of combustion status of flame was judged by recognizing the image classification. The result shows that the method has very high recognition accuracy to judge changes of combustion status of flame.

关 键 词：热能动力工程 模糊免疫网络 数字图像 聚类 燃烧诊断 

分 类 号：TK311[动力工程及工程热物理—热能工程]