基于面向对象变化向量分析法的遥感影像森林变化检测

发布时间：2018-05-03 09:14

本文选题：面向对象 + 变化向量分析(CVA)　；参考：《国土资源遥感》2017年03期

【摘要】：为探讨用于森林资源数据库更新的森林变化空间信息采集方法,以林地变化频繁快速、变化图斑多且小的广西上思县局部区域为研究区,以资源三号(ZY-3)和高分一号(GF-1)高空间分辨率卫星遥感图像和小班专题图为数据源,采用面向对象的变化向量分析(change vector analysis,CVA)方法,基于马氏距离、欧氏距离和相对误差距离度量变化强度,通过目标函数确定最佳检测阈值,以小班为单元进行森林变化检测。结果表明,用欧氏距离、马氏距离检测的森林变化结果都不甚理想,漏检率和误检率高,总体精度较低,Kappa系数较小;用相对误差距离检测的结果较好,漏检率(21.0%)和误检率(32.5%)均最小,总体精度最高(89.6%),Kappa系数最大(0.664);误检测的图斑多为成林地和无林地(建设用地、林区道路等),各个变化类型都出现了少量漏检图斑。
[Abstract]:In order to study the method of collecting spatial information of forest change used for updating forest resource database, the study area is local area of Shangsi County, Guangxi, where woodland changes frequently and rapidly, and the change map is many and small. Taking the high spatial resolution satellite remote sensing images and subclass thematic maps as data sources, using the method of object oriented change vector analysis (CVA), and based on the Markov distance, the paper introduces the method of "ZY-3" and "GF-1" (GF-1) as data sources, and adopts the method of object oriented change vector analysis (OVA), which is based on Markov distance. Euclidean distance and relative error distance are used to measure the change intensity, and the optimal detection threshold is determined by objective function, and forest change detection is carried out with small class as unit. The results show that the results of forest change detected by Euclidean distance and Markov distance are not satisfactory, the rate of missing detection and false detection is high, the total precision is lower and the Kappa coefficient is lower, and the result of using relative error distance is better. The miss detection rate (21.0%) and the false detection rate (32.5%) were the smallest, the overall accuracy was the highest (89.6%) and the Kappa coefficient was the largest (0.664m), and most of the map spots were forest land and no forest land (construction land, forest road, etc.), and a small number of miss check map spots appeared in each change type.
【作者单位】：广西大学林学院;广西林业勘测设计院;
【基金】：广西林业科学研究项目“森林变化遥感信息自动检测与提取”(编号:201423)资助
【分类号】：S771.8;TP751

【相似文献】