基于ESS模型的云状自动识别方法

Automatic cloud recognition method based on ESS model

为了提高云图自动识别的准确率,提出了一种新的云状自动识别方法。基于湍流标度理论,将湍流的标度特征应用于云图的自动识别和分类中,对云图的灰度数据进行扩展自相似(ESS)模型标度分析,提取云图的标度指数特征,利用不同云系的标度特征识别云图。选用支持向量机作为分类器,对波状云、层状云、积状云、卷云和晴空5种云图进行识别。研究结果表明,通过提取ESS模型标度特征进行典型云状识别的准确率接近或超过90%。由于具有较强的显著性,基于湍流标度理论提取云图特征对云状识别方法是一个很好的补充。

In order to improve the accuracy of cloud classification, a new method was proposed based on the statistical scaling turbulence theory. Extended self-similarity（ESS） model was applied to cloud classification and recognition. Feature extraction model was used to analyze these time-series data converted by original image.The support vector machine was chosen as the classifier to distinguish waveform, stratiform, cumuliform, cirriform clouds and clear sky. The results show that the accuracy of single cloud species recognition is near or above 90%. Thus, features extracted by ESS model are significant parameters in cloud classification. This model is a complement to other cloud recognition methods.