基于多尺度分析的地基云图自动识别的研究

Research on Recognition of Ground-Based Cloud Images Based on Multi-Scale Analysis

研究地基云图自动识别问题。地基云图由于受光照旋转等影响,传统方法提取的特征难以准确描述图像特征,导致地基云图识别的精度较低,还无法达到实际应用的要求。为了提高地基云图识别精度,提出采用多尺度分析的地基云图自动识别算法。首先将地基云图划分为若干个大小相同互不重叠的子图,然后采用三种不同尺度的权重局部二值模式提取每一幅子图的纹理特征,最后将子图的纹理特征顺序排列获取最终鉴别特征。通过对积雨云,高积云和层积云三类地基云图进行分类识别,实验结果表明改进方法可以增强云图的局部特征描述能力,有效提高地基云图分类的精度。

This paper studied the problem of ground-based cloud images recognition. Due to the effects of light and rotation, It is difficult for the traditional method to describe the features of images accurately, resulting in lower recognition of ground-based images. In the paper, cloud images were first ivided into a number of non-overlapping sub-graphs with the same size, then three different scales of weights local binary pattern were used to extract the texture feature of each parcel sub-graph. Finally, the texture features of each sub-graph were combined to obtain the final identification feature set. The recognition experimental results of cumulonimbus, altocumulus and stratocumulus show that the proposed method can get better partial characterization capabihty of cloud images and improve the accu- racy of the classification of ground-based cloud images.