基于卷积神经网络的非多普勒激光雷达测风系统

Non-doppler lidar wind measurement system based on convolutional neural network

提出了一种基于卷积神经网络技术的非多普勒激光雷达测风系统。该系统利用半导体感光元件(charge coupled device,CCD)拍摄气溶胶颗粒物的激光雷达后向散射图,针对不同风速气溶胶颗粒物的运动轨迹特征,实现定量的风速测量。卷积神经网络对经过图像预处理的气溶胶颗粒物运动轨迹图,进行特征提取并生成风速测量模型,第100次训练样本和验证样本的准确率分别为0.92和0.93。利用生成的风速测量模型对测试样本进行实验,准确率达到0.84。这种低成本、操作简便的非多普勒激光雷达测风系统,能够解决当前多普勒频移测风激光雷达成本高的痛点,具有很强的现实意义。

A Non-Doppler lidar wind measurement system based on convolutional neural network technology is proposed.The system uses(charge coupled device,CCD)to capture the lidar backscatter map of aerosol particles,and realizes quantitative wind speed measurement according to the movement trajectory of aerosol particles at different wind speeds.The convolutional neural network extracts the features of the aerosol particle movement trajectory after lidar image preprocessing and generates a wind speed measurement model.The accuracy of the 100 th training sample and verification sample are 0.92 and 0.93,respectively.The wind speed measurement model is used on the test sample,and the accuracy rate reaches 0.84.This low-cost,easy-to-operate non-Doppler lidar wind measurement system can effectively reduce the cost of wind measurement and has strong practical significance.