In this paper we studied the textural features extraction , remote sensing images classification and bp neural network techniques and their applications in the meteorological problems such as recognition of the cloud cluster feature , cloud - drift wind retrieval and heavy rain process analysis etc . to the question of the low precise recognition of satellite images by using spectral features , the proposed approach assumes to perform a multiple analysis based on an advisable decision - making model by first developing a mixed pixel model which was based on the textural features of images , and then improving the recognition intelligence 本文对模式识别领域中的图像纹理特征提取、遥感图像分类、 bp神经网络与纹理特征组合分类等方法,以及它们在云团属性识别、云迹风反演和暴雨过程分析等气象问题中的应用作了研究。针对过去利用图像光谱亮度特征进行识别分析气象卫星图像准确度不高的问题,本文提出了发展混合像元的分解模型,以图像的纹理特征为基础,提高图像识别的智能水平,以实现在分析决策模型的支持下,快速准确的复合分析的解决方案。