摘要
回转窑火焰图像的正确分割对于提取回转窑工况参数具有重要的意义。本文研究了多层感知器、径向基函数网络、学习向量量化网络和自组织特征映射网络等四种神经网络在回转窑火焰图像分割中的应用。选取火焰图像的归一化色彩值作为训练样本 ,分别采用有导师学习和无导师学习两种方法进行训练。对实测图像进行分割的结果表明 。
Accurate segmentation of flame image in rotary kiln is very important for the extraction of working parameters. In this paper, four neural networks, i.e. multi layer perception, radial basis function, learning vector quantization and self organizing feature mapping, are used to segment the flame image. Normalized color intensity values are selected as training sample of neural networks. The neural networks are trained by supervised algorithm and unsupervised algorithm respectively. The results of segmenting the actual image show that the proposed approach is feasible.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2001年第1期10-12,16,共4页
Chinese Journal of Scientific Instrument
基金
国家 8 63高技术计划项目 !(863- 51 1 - 9845- 0 0 2 )资助
关键词
图像分割
神经网络
多层感知器
径向基函数
学习向量量化
自组织特征映射
回转窑
Image segmentation Neural networks Multi layered perception Radial basis function Learning vector quantization Self organizing feature mapping Rotary kiln
