基于互信息F统计量特征选择技术的地基气象云图分类  

Classification of Ground-based Meteorological Cloud Images Based on Feature Selection Technique of Mutual Information F-statistics

作　　者：杨秋良 王钰 杨杏丽[1] 李济洪 YANG Qiu-liang;WANG Yu;YANG Xing-li;LI Ji-hong(School of Mathematical Sciences,Shanxi University,Taiyuan 030006,China;School of Modern Educational Technology,Shanxi University,Taiyuan 030006,China)

机构地区：[1]山西大学数学科学学院,山西太原030006 [2]山西大学现代教育技术学院,山西太原030006

出　　处：《计算机与现代化》2021年第2期18-23,29,共7页Computer and Modernization

基　　金：山西省应用基础研究计划项目(201901D111034,201801D211002);统计与数据科学前沿理论及应用教育部重点实验室开放研究课题(KLATASDS2007);国家自然科学基金资助项目(61806115)。

摘　　要：在地基气象云图的云状(云类)识别研究中,基于局部二值模式(Local Binary Pattern,LBP)描述子的特征选择技术由于它的简单性和有效性成为最通用的方法。然而,LBP特征的高维特性使得云状识别的性能和计算开销不能令人满意。为此,本文提出一种基于互信息构造的F检验统计量的LBP特征选择算法,可以实现高维LBP特征的有效降维,同时保证云状识别的准确性,极大减少了特征选择过程的计算开销。In the study of cloud type recognition of ground-based meteorological cloud images,the feature selection technology based on Local Binary Pattern(LBP)descriptors is the most common method due to its simplicity and effectiveness.However,the high-dimensional property of LBP features makes the performance and computational overhead of cloud type recognition unsatisfactory.To the end,an LBP feature selection algorithm based on F-statistics constructed by mutual information is proposed,which can achieve effective dimensionality reduction of high-dimensional LBP features,while ensuring the accuracy of cloud type recognition,greatly reducing the computational cost of feature selection process.

关 键 词：地基气象云图 高维特征选择 互信息 F统计量 分类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]