项目名称: 基于边缘探测细胞神经网络的雹云判识模型理论研究
项目编号: No.11461063
项目类型: 地区科学基金项目
立项/批准年度: 2015
项目学科: 数理科学和化学
项目作者: 李国东
作者单位: 新疆财经大学
项目金额: 40万元
中文摘要: 边缘探测细胞神经网络(EDCNN)是细胞神经网络(Cellular Neural Network, 简称CNN)的重要分支,已在图像和视频信号处理,复杂系统建模等领域得到广泛应用。EDCNN模板的鲁棒性、拓广和应用的研究具有重要的理论价值和应用前景. 本研究旨在通过数学分析的方法结合数值模拟研究EDCNN动力学机理,研究EDCNN的鲁棒性问题,建立相应的EDCNN模板功能的设计定理,为CNN的优化设计提供定量的解析判据,丰富CNN理论。 并研究采用EDCNN技术对云团雷达跟踪图像进行实时探测,提取变化的纹理信息,研究云团雷达图像的降雹信息。构造出云团生命发展变化的特征向量。通过小波变换、傅立叶变换和矩阵分析等方法对该特征向量进行数据分析与数据挖掘,并利用聚类分析等数学方法对特征向量进行分类,建立雹云判识指标的数学模型。最后编写雹云识别软件,辅助气象预报人员对冰雹天气进行短时预警.
中文关键词: 数据挖掘;统计计算;细胞神经网络;纹理识别
英文摘要: Edge detection of cellular neural network(EDCNN) is an important branch of cellular neural network, has been widely used in the fields of image and video signal processing, complex system modeling. And the research of robustness, extension and application of EDCNN template has an important theoretical value and application prospect. The research aims to study dynamics mechanism and robustness problems of EDCNN by using the methods of mathematical analysis in combination with numerical simulation and then establish the corresponding template function which provides quantitative analysis criterion for optimization design of CNN and enriches the theory of CNN.And the research processing maily inlcude three steps :detect the cloud radar images in real time,extract textural information of the changing clouds and summarizes the hail clouds radar information.In a word ,The research mainly constructs feature vector of change rule of the clouds life.Furtherly the research is to establish the mathematical model as the indicator to identify the hail clouds by using the mathmatical methods such as wavelet transform ,fourier transform ,matrix analysis methods cluster analysis.Finally,the research end up to writing the recognition software of the hail clouds to help the weather forecasters on short-term warning.
英文关键词: Data mining;statistical calculation;CNN;Veins analysis