基于灰度共生矩阵的特征构建及超折射滤除

超折射回波会严重干扰对天气雷达图像中强对流回波的识别。文中从分析超折射回波及强对流回波在雷达反射率图中的分布特点入手，在区域分割的基础上，生成各区域的灰度共生矩阵，将灰度共生矩阵中的元素划分成两个子集，分别用以构建出两个新的特征，即平缓度/跳变性，它们在超折射回波和强对流回波样本之间呈现出显著性差异，配合使用径向速度特征，站在尽量不损失强对流的角度形成能够克服特征缺值的分类决策树。测试结果表明：文中方法较目前业务上普遍使用的模糊逻辑分布式超折射地物识别法，对超折射的滤除率及对强对流云团的保有率更高，特别是文中方法在将对超折射和强对流的识别准确率从96.8%提高到97.9%的前提下，对强对流的滤除率从3.91%降低到0.21%。

Feature Construction and AP Clutter Filtering Based on Gray Lever Co-occurrence Matrix

AP clutter can interfere with strong convection echo’ s identification of weather radar image. It starts with AP clutter and strong convection echo’ s distribution characteristics in the radar reflectivity image. On the basis of region division,generate Gray-Level Co-occurrence Matrix （GLCM）. GLCM elements are divided into two subsets,which build two new gentle/jumping feature. These features show significant differences between AP clutter and strong convection echo. Taking into account without loss strong convection, form classification decision tree under using radial velocity characters and gentle/jumping feature. The results show that AP clutter filtering rate and strong convective cloud clusters retention rate is higher than fuzzy logical algorithm to detect AP clutter in business. Under the premise of recognition accuracy of AP clutter and strong convection echo is raised from 96. 8% to 97. 9%,strong convective cloud clusters filtering rate is lowered from 3. 91% to 0. 21%.