论文部分内容阅读
建立一种基于RBF神经网络的目标红外辐射亮度建模方法,进而实现对目标光谱发射率的估计。通过FTIR光谱仪测量目标表面3~14μm波段的红外辐射特性,亮度光谱会受到二氧化碳、水蒸气等的吸收及大气辐射的干扰。文中首先结合红外传输理论选择有效学习样本;然后基于RBF网络对样本进行充分学习,建立目标红外辐射亮度模型;利用所建模型估计大气吸收和杂散干扰波段的亮度,最终计算出较完整的目标光谱发射率。黑体测试结果与理论发射率比较,最大相对误差为1.5%。测温验证的结果也表明文中所建的RBF神经网络可以有效地对目标光谱发射率进行估计。
A method of target infrared radiation brightness modeling based on RBF neural network is established to estimate the target spectral emissivity. The FTIR spectrometer measures the infrared radiation in the band of 3 ~ 14μm on the target surface. The intensity spectrum will be disturbed by the absorption of carbon dioxide, water vapor and atmospheric radiation. Firstly, the effective learning samples are selected based on the theory of infrared transmission. Then the samples are fully studied based on the RBF network, and the target infrared radiation brightness model is established. By using the model to estimate the atmospheric absorption and spurious interference bands, the more complete target Spectral emissivity. Compared with the theoretical emissivity, the maximum relative error is 1.5%. The results of temperature measurement also show that RBF neural network built in this paper can effectively estimate the target spectral emissivity.