论文部分内容阅读
针对实时条件下中红外波段平均大气透过率的计算,提出了一种基于贝叶斯正则化BP神经网络的方法。利用BP神经网络良好的非线性拟合特点,建立大气参数与中红外平均透过率之间的关系模型,从而可以准确迅速地得到计算结果。此网络模型是以实测温度、压强、湿度和气溶胶的后向散射系数作为输入向量,分别以水蒸气和CO2吸收透过率、气溶胶散射透过率和大气透过率作为输出。仿真结果表明:在相同的大气参数下,本方法的计算结果与期望值之间的相对误差较小,且远小于经验公式法,验证了本方法的可行性与有效性。因此,本方法对大气透过率的准确地快捷计算提供了有益的借鉴。
Aiming at the calculation of average atmospheric transmittance of mid-infrared band under real-time conditions, a Bayesian regularized BP neural network method is proposed. By using the good non-linear fitting characteristic of BP neural network, a model of the relationship between atmospheric parameters and the average transmittance of mid-infrared can be established so that the calculation result can be obtained accurately and rapidly. The network model takes the measured temperature, pressure, humidity and backscatter coefficient of aerosol as input vectors, and takes the water vapor and CO2 absorption transmissivity, aerosol scattering transmissivity and atmospheric transmissivity as output respectively. The simulation results show that under the same atmospheric parameters, the relative error between the calculated result and the expected value is smaller than the empirical formula method, which verifies the feasibility and effectiveness of the method. Therefore, this method provides a useful reference for accurate and quick calculation of atmospheric transmittance.