论文部分内容阅读
林分类型的识别是森林资源监测的核心问题之一.为研究多源遥感数据协同的面向对象林分类型分类识别,采用Radarsat-2数据和Quick Bird遥感影像协同进行面向对象分类.在面向对象分类过程中,采用3种分割方案:单独使用Quick Bird遥感影像分割;单独使用Radarsat-2数据分割;Radarsat-2&Quick Bird协同分割.3种分割方案均采用10种分割尺度(25~250,步长25),应用修正的欧式距离3指标评价不同分割方案的分割结果,确定最优分割方案及最优分割尺度.在最优分割结果的基础上,基于地形、高度、光谱及共同特征的不同特征组合,应用带有径向基(RBF)核函数的支持向量机(SVM)分类器进行杉木林、马尾松林、阔叶林3种林分类型识别.结果表明:与单独使用一种数据相比,Radarsat-2数据和Quick Bird遥感影像协同方案在面向对象林分类型分类方面具有优势.Radarsat-2&Quick Bird协同分割方案,以最优尺度参数100进行分割时,分割结果最好.在最优分割结果的基础上,应用两种数据源提取的全部特征进行面向对象林分类型识别的精度最高(总精度为86%,Kappa值为0.86).本研究结果不仅可为多源遥感数据结合进行林分类型识别提供参考和借鉴,而且对于森林资源调查和监测有现实意义.
The classification of forest types is one of the core issues in forest resource monitoring.In order to study the object-oriented classification and recognition of multi-source remote sensing data, Radarsat-2 data and Quick Bird remote sensing images are used to carry out object-oriented classification.On object-oriented In the process of classification, three kinds of segmentation schemes are adopted: Quick Bird Remote Sensing Image Segmentation alone, Radarsat-2 Data Segmentation alone and Radarsat-2 & Quick Bird collaborative segmentation.The three segmentation schemes adopt 10 kinds of segmentation scales (25 ~ 250 steps Based on the optimal segmentation results, based on the different features of topography, height, spectrum and common features , And the support vector machine (SVM) classifier with radial basis function (RBF) kernel function was used to identify three types of Chinese fir plantation, Pinus massoniana plantation and broad-leaved plantation.The results showed that compared with using one kind of data alone , The Radarsat-2 data and the Quick Bird remote sensing image synergetic scheme have the advantages in object-oriented forest classification classification.Radarsat-2 & Quick Bird collaborative segmentation scheme, with the optimal scale The best segmentation result is obtained based on the optimal segmentation results.It has the highest accuracy of object-oriented forest type recognition using all the features extracted from the two data sources (the total accuracy is 86% and the Kappa value is 0.86 The results of this study can not only provide reference and reference for multi-source remote sensing data in combination with forest type identification, but also have practical significance for forest resources investigation and monitoring.