基于纹理特征的地基云分类识别研究

云是一种重要的天气现象,云观测对当时天气系统的观测具有很强的指导意义,同时云的变化趋势也是预测未来天气的一项重要指标?目前对遥感卫星云图的研究取得了较丰富的成果,但地基云图分类识别研究取得的成果则比较有限,地基云图的研究一直是模式识别中的难点?采用标准气象站地基云图数据,选取了浓积云?积雨云?雨层云等5类常见的云类,并将Gabor小波变换应用于地基云图的纹理特征数据提取,运用实验法选取了Gabor小波变换的最优参数?将地基云图纹理特征数据运用到最短距离和包含BP神经网络的陀螺分类器构造成的地基云图分类系统,从而实现了对地基云图的分类?实验结果表明,该方案不仅能够有效地运用于地基云图分类,而且分类效果也是比较理想的?

Ground-based cloud classification and recognition based on textural feature

Cloud is an important weather phenomenon.Cloud observations have great directive significance on the observation of weather system.The variation trend of clouds is also an important indicator to predict weather in the future.At present the study of satellite cloud imagery has achieved great success.But the research of ground-based cloud is not perfect;it is difficult work in pattern recognition.In this paper,the ground-based cloud data is from a standard meteorological station.We select five categories of clouds.Gabor wavelet transform is used in texture features extraction of ground-based nephogram.Experimental method is used for selecting the optimal parameters of Gabor wavelet transform.The ground-based nephogram classifier is constructed by neural network which is through the shortest distance and gyroclassifier. Experimental results show that this scheme achieves good results in ground-based nephogram classification.