基于转移矩阵深度挖掘的橡胶林高分遥感变化分析

A Deep Analysis Method of Rubber Plantations Change Analysis From High Resolution Remote Sensing Image Basing on Transition Matrix

为了更加精细的分析橡胶林变化情况,文章建立一套适用于高分辨率遥感影像的橡胶林变化信息深度挖掘分析方法。首先选择融合多特征的C5.0决策树方法对橡胶林进行分类,并以两期分类结果图像为基础进行地物变化转移矩阵的计算,分别计算出地物间的转移变化、地物净变化量和交换变化量以及类别间的转换规则。方法运用于海南省东方市广坝农场内的一块橡胶林种植区进行试验,在2010年5月6日至2015年7月4日期间,区域内的橡胶林新增量为6.51%,减少量为10.17%,净变化量为3.66%,交换变化量为13.02%,总变化量为16.68%,整体上橡胶林在减少,水体和裸地有增加的趋势。试验的结果有力的证明了方法的有效性和实用性。

In order to precisely analyze the change of rubber plantations,a fit change analysis method basing highresolution remote sensing images is proposed.Firstly,precise classification of rubber plantations by the use of multi-featuresintegrated C5.0decision tree is done;Secondly,land objects change transition matrix basing on two precisely classifiedimages is computed too;the transition change,net change,and transfer change of land objects are computed,and thetransition rules analysis of land objects are done at last.The method is experimented in a rubber plantation area locatedGuangba Farm,Dongfang City,Hainan Island,China.During period between May6,2010and July4,2015,experimentalresults are shown as follows:new increase of rubber plantations is6.51%,while the new decrease of rubber plantations is10.17%,the net change,transfer change and total change of rubber plantations are3.66%,13.02%,and16.68%respectively;the rubber plantations has a tendency of decreasing while the water and bare area have a tendency of increasing.Theexperimental results prove the feasibility and efficiency of our method.