复杂天空云状的特征描述及自动识别研究

项目名称： 复杂天空云状的特征描述及自动识别研究

项目编号： No.41205125

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 大气科学学科

项目作者： 刘磊

作者单位： 中国人民解放军理工大学

项目金额： 24万元

中文摘要： 云状是反映天气变化征兆的重要气象要素，主要依靠人工目测进行观测。随着云的地基观测仪器的不断发展，云状自动识别问题成为研究热点。由于云的复杂多变，一直以来难以找到合适的方法进行特征描述，目前国内外主要利用纹理特征进行单一天空条件下的云状分类。但实际天空常常是多种云状并存的复杂天空，在前期有关研究的基础上，我们尝试开展复杂天空云状识别研究。具体内容包括：（1）研究基于传统模式识别方法的云状识别技术。重点是解决复杂天空条件下有关云的形状、纹理、颜色、轮廓、范围、演变过程等特征描述问题，研究根据这些特征进行复杂天空下的云状识别的方法。（2）研究基于压缩感知的云状识别技术。重点是构造训练样本的超完备基并求解最稀疏的系数解，在此基础上研究利用稀疏矩阵系数实现对复杂天空下的云状识别。

中文关键词： 云状识别；结构特征；压缩感知；；

英文摘要： Cloud type which is one of the important indices for changes in the weather, is mainly observed by human being at present. With the development of ground-based cloud measurement technology, automatic cloud classification is becoming the research hotspot. Some studies focused on the cloud classification on simplex sky conditions, and it's found that clouds are so complicated and changeful that it is hard to find a nice way to extract the effective features. In fact, there is generally more than one type of cloud in the sky at the same time, and we define these as complex sky conditions. Based on the previous work and experiences, we try to do some research on cloud classification on complex sky conditions. Specifically includes:(1)Studies on the traditional pattern recognition methods of cloud classification technology. This content focuses on how to extract effective characteristic factors of cloud shapes, texture, color, contour, range, process of change and some other features on the complex sky conditions. In addition, we will study how to use these features to do cloud classification on complex sky conditions. (2)Studies on cloud classification based on compressed sensing theory. This content focuses on how to construct training samples as an over complete learned dictionary and solve the most sparse coeffic

英文关键词： cloud classification；structure feature；compressed sensing；；