基于Madaline网络的气溶胶消光系数反演算法

A Retrieve Method for Aerosol Extinction Coefficient Based on Madaline Networks

运用激光雷达监测气溶胶是大气环境监测的一项重要内容,通过激光雷达方程可以反演得到气溶胶消光系数,并进而获得气溶胶的其他特性。然而传统方法在反演气溶胶消光系数时需要很多假设,使得反演精度受到很大限制。提出了一种利用多层自适应线性(Madaline)人工神经网络来反演气溶胶消光系数的方法,通过对网络进行训练,可由激光雷达回波信号直接反演气溶胶消光系数,从而可有效避免传统方法的诸多假设。对比实验表明该方法使反演精度大大提高,获得了很好的反演结果。

Aerosol detecting using lidar is an important part of atmosphere monitoring. Many aerosol properties are gained on the base of the aerosol extincition coefficient received through lidar equation. The traditional method restricts the precision of the aerosol extinction coefficient because of many hypotheses. A new procedure was developed, based on the Madaline artificial neural network theory to retrieve the aerosol extinction coefficient. Through the trained network, the aerosol extinction coefficient could be retrieved from the lidar echo signal directly, consequently the new method dispensed with those hypotheses in the traditional method. According to the comparison of the different test results, the new method improved the precision of retrieval greatly and good result was obtained.