中文摘要：

基于多普勒激光雷达，结合大气流场理论实现对下击暴流、侧风切变、低空急流及顺逆风切变四种低空风切变的三维模拟仿真，建立雷达图像样本库，解决了风场数据难以获得及样本库难以建立的难题。基于图像处理方法实现风切变类型识别，针对同类样本差异较大的难题，采用多特征提取方法提取图像的形状特征和纹理特征进行组合识别。同时，采用基于SOM-SVM的二次识别算法买现了对缺失信息的图像的有效识别，提高了整体识别率。与传统的风切变研究相比，所用方法忽略气象因素，完全转化为图像处理问题，实现了低空风切变的类型识别，识别率达到97％。

英文摘要：

Based on Doppler lidar and airflow field theory, the three-dimensional simulation of the microburst, side wind shear, low level jet stream and tailwind-or-headwind shear was accomplished. The lidar scan database was established to solve the problem of wind field data acquisition. Because of the obvious differences in the same sample, a multi-feature extraction method was adopted. The shape features and texture features of images was extracted and combined to recognize the type of wind shear based on image processing. A tow-step recognition algorithm based on SOM-SVM is adopted to recognize the images with missing information effectively. Compared with the traditional study of wind shear, the proposed method neglects the meteorological factors which is completely transformed into the study of image processing. The recognition rate with this method is up to 97%.