基于PCA-ELM的红外多光谱辐射测温

Infrared Multispectral Radiation Temperature Measurement Based on PCA-ELM

在目标发射率未知的情况下,建立一种基于主元分析(PCA)与极限学习机(ELM)相结合的红外多光谱测温方法.分析目标温度与辐射亮度谱的非线性数学模型,确定初始输入向量包含温度估计所需的充分信息;引入PCA方法从输入向量中提取相互独立的主元成分,降低神经网络输入维数;基于ELM网络对样本数据充分学习,最终建立PCA-ELM目标红外测温模型.利用黑体和未知发射率材料涂层目标作为测试目标源,验证该方法的有效性.

In the case of unknown target emissivity,an infrared multispectral radiation temperature measurement method based on principal component analysis(PCA)and extreme learning machine(ELM)is established.The nonlinear mathematic model of target temperature and radiance spectrum is analyzed to find a set of initial input vectors,which include sufficient information to estimate temperature.The PCA method is used to extract the independent principle components in input vectors.This method can also reduce the input dimension for neural network.Based on ELM network,the sample data is sufficiently learned to build the target infrared temperature measurement model by PCA-ELM.The effectiveness of the proposed method is verified by using the blackbody and the coating material with unknown emissivity as test target sources.