基于改进DBSCAN聚类算法的雷暴单体三维结构识别技术介绍

雷暴是一种严重威胁飞行安全的天气系统，利用地基多普勒天气雷达反射率因子数据和改进的DBSCAN聚类算法对雷暴单体的三维结构识别及特征量计算进行了研究，并在地基平台上对雷暴单体识别算法的有效性进行了验证分析。识别算法核心是将插值后的反射率因子三维网格数据作为输入量，采用多层反射率因子阈值基于改进的DBSCAN聚类方法识别所有等高面上的雷暴分量，并进行结构元素为3×3的腐蚀膨胀运算及雷暴分量特征核心提取，最后基于雷暴分量重叠面积进行垂直关联。结果表明:相对于SCIT算法，雷暴单体识别算法减少了雷暴分量识别的复杂性，可很好地识别任意形状的雷暴单体；使用多层阈值及特征核心提取技术可识别雷暴簇中的雷暴单体；利用腐蚀膨胀技术可解决雷暴单体虚假合并现象。算法可应用于民航机场雷暴的识别和预警。 

INTRODUCTION OF 3D STRUCTURE DETECTION TECHNOLOGY OF THUNDERSTORM CELL BASED ON IMPROVED DBSCAN CLUSTERING ALGORITHM

Thunderstorm is a weather system that seriously threatens flight safety. In this paper, the ground-based Doppler weather radar reflectivity factor data and the improved DBSCAN clustering algorithm are used to study the three-dimensional structure detection and feature quantity calculation of thunderstorm cells, and the effectiveness of the thunderstorm cell detection algorithm is verified and analyzed on the foundation platform. The core of the detection algorithm is to use the interpolated reflectivity factor 3D mesh data as input data, and the multi-layer reflectivity factor threshold is used to detect the thunderstorm components on all levels based on the improved DBSCAN clustering method, and the 3×3 corrosion expansion operation and the thunderstorm component feature core extraction are performed. Finally, vertical correlation is carried out based on the overlapping area of thunderstorm components. The results show that compared with the SCIT algorithm, the thunderstorm single detection algorithm reduces the complexity of thunderstorm component detection and can well detect thunderstorm cells of arbitrary shape. The use of multi-layer threshold and feature core extraction techniques can detect thunderstorm cells in thunderstorm clusters. The use of corrosion expansion technology can solve the false merger of thunderstorms. The algorithm in this paper can be applied to the detection and early warning of thunderstorms in civil aviation airports.