基于雷达反射率图像特征的冰雹暴雨识别

为了实现冰雹暴雨天气的识别与分类,提出了一种基于雷达反射率图像特征的自动识别方法.对雷达回波反射率图像中冰雹回波区域和暴雨回波区域的图像特征进行提取,通过分析冰雹暴雨间单一特征的差异性和不同特征之间的分类互补性,确定了识别冰雹暴雨的有效图像特征(包括强度特征和纹理特征).将提取出的样本有效特征与探空数据(0℃和~20℃温度层高度)结合,利用粗糙集理论进行数据挖掘,进而建立了冰雹暴雨天气的客观识别模型.通过对362个测试样本的测试与统计,冰雹击中率达到93.29%,暴雨的击中率达到89.27%,并且两者均具有较低的误警率.实验结果与传统PUP系统比较,表明利用雷达反射率图像特征实现对冰雹暴雨天气的识别与分类具有较好的效果.

Recognition of hail and rainstorm based on the radar reflectivity image features

Based on radar reflectivity image features, an automatic recognition method is proposed to identify the hail and rainstorm. We extract the image features of hail echo areas and rainstorm echo areas from radar reflectivity images. By analyzing both the differences in single feature between hail and rainstorm and the classified complementarity among different features, we determine the effective image features, including the intensity and texture features, to identify the hail and rainstorm. The hail and rainstorm objective recognition model can be established through the data mining of the extracted sample features and sounding data by using the Rough Set Theory. Through the test and identification of the 362 test samples, the hit ratio of hail reaches 93.29% and the hit ratio of rainstorm reaches 89.07%. The false alarm ratios of hail and rainstorms can be also at a low level. Compared with those from the PUP system, the experimental results from the present system show that it has a good effect to identify and classify hail and rainstorm by using the radar reflectivity image features.
Keywords：features extractionRough Set Theorydata mining