基于变分贝叶斯ICA的火山灰云检测的研究.pdfVIP

第36 卷 第2 期 红 外 技 术 Vol.36 No.2 2014 年2 月 Infrared Technology Feb. 2014 基于变分贝叶斯 ICA 的火山灰云检测研究 沈 迪，李成范，赵俊娟，尹京苑 （上海大学计算机工程与科学学院，上海 200444 ） 摘要：热红外遥感能够准确、快速地检测出火山灰云信息。针对热红外遥感数据和火山灰云检测特 点，提出了基于变分贝叶斯ICA 的火山灰云检测算法，通过引入贝叶斯网络的条件独立性准则和近 似变分算法，将源信号后验概率分布的计算转化为更容易处理的负变自由能，去逼近真实后验概率 分布的最佳近似值。实验表明该方法取得了85.25%的光谱匹配度，检测精度较高。 关键词：热红外遥感；火山灰云；MODIS 图像；变分贝叶斯；独立分量分析 中图分类号：TN219 文献标识码：A 文章编号：1001-8891(2014)02-0120-05 Volcanic Ash Cloud Detection Based on Variational Bayesian ICA SHEN Di，LI Cheng-fan ，ZHAO Jun-juan ，YIN Jing-yuan (School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China) Abstract ：The thermal infrared remote sensing can be used to accurately and quickly monitor the volcanic ash cloud information. Aiming at the characteristics of thermal infrared remote sensing data and volcanic ash cloud detection, this study proposes a volcanic ash cloud detection method on the basis of variational Bayesian ICA, which proves that with the addition of the independence standards and the variational approximate algorithm of Bayesian networks being put into the Bayesian framework, the computation of posterior probability distribution of source signal can be turned into negative free energy and approach the optimal approximation of real posterior probability distribution. The results indicate that the proposed method has a spectral match of 85.25% and high detection accuracy. Key words ：thermal infrared remote sensing ，volcanic ash cloud ，MODIS image ，variational Bayesian ， independent component analysis （ICA ）