Font Size: a A A

Aurora Morphologyand Automatic Classification Of Aurora Images

Posted on:2013-01-27Degree:MasterType:Thesis
Country:ChinaCandidate:S M HanFull Text:PDF
GTID:2248330395956467Subject:Radio Physics
Abstract/Summary:PDF Full Text Request
The morphological characteristics and manifestation of aurora have close relationto the variation of geomagnetic field and various dynamical processes of themagnetospheric boundary layers. So the research of various types of aurora morphologyhas very important theoretical significance and application value. The main work andresults in this paper are summarized as follows:The unit of rayleigh which represent the intensity of aurora is discussed in thispaper, and the conversion formulas between rayleigh and the unit commonly used unitare provided for people to employ conveniently. Keogram image is introduced toanalysis aurora phenomenon and aurora occurrence in visible light and ultraviolet bandrespectively and the Keogram image in ultraviolet band is utilized to observe substormsin the occurrence and development of the whole process.According to the morphology and characteristics of dayside aurora, an auroraclassification method using a multi-level feature representation is proposed in the periodof ground-based aurora image classification feature extraction in visible light band. Themethod can be used to capture both global and local image texture statisticalinformation, so we can get more elaborate texture detail feature. At the same time weused a Minimum-Redundancy Max-Relevance feature selection strategy to selectrepresentative features, which reduce feature redundancy information meanwhilecomputation precision was assured, having great practical value.By analyzing and summing up deeply of a great deal of aurora images provided byPOLAR satellite, we classify ultraviolet aurora images to three kinds of typicalmorphology. According to the special morphology and structure of ultraviolet aurora, inthis paper we incorporate texture features and shapes features to represent aurora images.This method captures both texture and shape information, which can represent imageinformation more comprehensive and accurate. Well characterization capability andSupport Vector Machine excellent classification performance together effectivelyimprove the accuracy rate of classification system of ultraviolet aurora.
Keywords/Search Tags:Aurora Morphology, Keogram Image, Pattern Recognition, Ultraviolet Aurora, Support Vector Machines
PDF Full Text Request
Related items