Abstract: As a sort of local sinking air current，microburst may cause disaster result of aviation safety and people’s life and property.Because microburst is characterized by small scope，tremendous intensity，and fast change，it is a challenge to recognize automatically.Current analysis is based on Doppler radar image data in which microburst is featured as interaction between two fuzzy regions closely related to intensity and speed of wind in such regions.An algorithm based on image segmentation is proposed to recognize microburst region automatically.The proposed algorithm recognizes features extracted from history data to determine microburst region.Visual C++ 6.0 is used to develop the proposed algorithm.Experiment shows that accuracy achieves the level as experts do with naked eye.The algorithm is able to discover more covert suspicious region，and enhances precision of atrocious weather forecast.

Key words: microburst, feature extraction, Doppler radar, region recognition

摘要： 下击暴流是一种局部性的下沉气流现象，会对航空以及人们的生命财产安全造成灾难性后果。由于下击暴流具有范围小、强度大、变化快等特点，对它的自动识别是个颇具挑战性的任务。当前对下击暴流的分析是根据多普勒雷达图像数据进行的，图像上下击暴流表现为两个模糊区域间的相互作用，与周边的风速风强关系密切，有明显的特征。提出了一种图像区域分析算法对多普勒雷达图像的下击暴流区域进行自动识别，基于对历史数据集的特征提取，通过特征识别确定下击暴流区域的位置。使用Visual C++ 6.0实现了算法，结果表明算法的精度达到专家肉眼识别水平，能发现较隐蔽的可疑区域，提高了对恶劣天气的预报水平。

关键词: 下击暴流, 特征提取, 多普勒雷达, 区域识别