摘要
【目的】以陕西省横山县1990年的TM图像和2013年的OLI图像作为基础数据源,准确获取横山县土地覆盖专题信息,并制作土地利用类型转换图谱。【方法】对图像预处理后,将横山县10种土地利用类型分别归属到不同的专题层,按照提取难度排列各专题层的处理顺序;对于一个专题层针对性的构建特征空间,并辅助专题图件在特征空间中通过全局阈值分割,获取目标土地利用类型的概貌信息;将包含一层概貌信息的二值图像作为掩膜文件与原图像进行求交运算,依据剩余地物之间的光谱特征和形态特征差异选取尺度,对求交结果依次执行面向对象分割、SVM监督分类和基于数学形态学开闭运算的分类后处理操作,以进一步分离出较难识别的背景地物,获取目标土地利用类型的精确空间分布信息;复合各专题层的提取结果,形成一期的土地覆盖分类结果,然后由两期结果制作横山县1990-2013年土地用类型转入、转出图谱。【结果】分层分级提取方法依据目标土地利用类型对象与背景地物对象的光谱特征和形态特征差异,对多光谱遥感图像逐层逐级分解计算,有效减少对光谱特征微弱、混淆地物的漏分误判现象,降低椒盐效应。两期TM/OLI图像提取结果的总体分类精度和Kappa系数分别为82.31%和0.775,81.26%和0.762。【结论】实验所设计的专题信息提取方法能够较为准确、快速地获取横山县土地覆盖专题信息,为解译黄土高原水蚀风蚀交错带TM/OLI图像提供一个新的思路。
Obj ective The aim of this research was to acquire land cover thematic information of Hengshan county and to analyze spectrum of land use types based on TM/OLI multispectral re-mote sensing images in 1990 and 2013.[Method]There were10 kinds of land use types in the Hengshan county and each land use type were divided into one thematic image layer.Processing sequence was arranged according to extraction difficulty.Next,feature space was built based on characteristic bands,such as NDVI,MNDWI and Albedo.So,general thematic information of land cover was derived using threshold segmentation in feature space.Furtherly,accurate land cover information was achieved by object oriented segmentation,SVM supervised classification and mathematical morphology open-close operator in each thematic image.Finally,overlapping each thematic information image layer together produced classification results.Spectrum of land use types in 1990 and 2013 were obtained in Hengshan county.[Results]Higher accurate classifi-cation results were obtained by this new approach.This method increased spectral feature and de-creased misj udgment induced by ground obj ectives.Overall classification accuracy and Kappa coef-ficient in two periods were 82. 31% and 0. 775,81. 26% and 0. 762,respectively.[Conclusion]The thematic information extraction technology noted in this study accurately and rapidly obtained thematic information of land cover in the Heduan country.This provided a new method for TM/OLI images interpretation in the transitional zone of wind-water erosion in the Loess Plateau.
出处
《四川农业大学学报》
CSCD
北大核心
2014年第2期185-193,共9页
Journal of Sichuan Agricultural University
基金
国家自然科学基金项目(编号:41361044
61162025)
关键词
遥感图像
黄土高原水蚀风蚀交错带
分类
特征空间
面向对象分割
remote sensing image
transitional zone of wind-water erosion in the Loess Plateau
classification
feature space
obj ect oriented segmentation
