CAJ | 학술논문

结合阿坝若尔盖县大骨节病典型病区植被分布特点,选用不同时相SPOT4及ETM遥感数据,提出了将较易实现的CART决策树算法与最大似然比分类法有机结合在一起进行植被分类的方法.决策树算法能很好地区分植被大类,分类精度达到96%,但是无法确定区分乔木亚类的阈值;最大似然比法整体分类精度不高,仅为84%,但是针对乔木亚类的分类精度能达到94%,将两种算法综合利用,最终总分类精度达到95.05%,Kappa系数达到0.9016.良好的分类结果不但为研究该区植被覆盖状况与发病率关系提供了很好的一手资料,并且分类算法较易实现,尤其对于新入门者较为实用和快捷.
결합아패약이개현대골절병전형병구식피분포특점,선용불동시상SPOT4급ETM요감수거,제출료장교역실현적CART결책수산법여최대사연비분류법유궤결합재일기진행식피분류적방법.결책수산법능흔호지구분식피대류,분류정도체도96%,단시무법학정구분교목아류적역치;최대사연비법정체분류정도불고,부위84%,단시침대교목아류적분류정도능체도94%,장량충산법종합이용,최종총분류정도체도95.05%,Kappa계수체도0.9016.량호적분류결과불단위연구해구식피복개상황여발병솔관계제공료흔호적일수자료,병차분류산법교역실현,우기대우신입문자교위실용화쾌첩.
This easier method that combined decision tree algorithm and maximum likelihood ratio is proposed aiming at vegetation distribution characteristic at Kashin-Beck disease region of Ruoergai County in Aba,using Spot 4 and ETM data of different time.Decision tree can plot out main category except arbor subturma;the precision of Maximum likelihood ratio is lower,only 84 percent,however,94 percent for arbor subturma.The precision of the class is up to 95.05 percent if it combined with the two methods,Kappa coefficient is 0.9016.The method provides helpful source for studying the relation between vegetation coverage and morbidity,and it is very useful and convenient for beginner.