面向对象的喀斯特地区石漠化遥感信息提取研究——以贵州省大方地区为例

OBJECT-ORIENTED REMOTE SENSING INFORMATION EXTRACTION METHOD FOR ROCKY DESERTIFICATION IN KARST AREA——A CASE STUDY OF DAFANG COUNTY,GUIZHOU

针对现有基于像素的监督和非监督分类方法在地质环境复杂、地形起伏较大、阴影明显的喀斯特石漠化地区难以满足石漠化信息提取精度要求的问题,采用基于纹理特征数据和地形数据辅助面向对象方法进行喀斯特地区石漠化信息的提取。该方法首先依据石漠化分布在TM/ETM+影像面积大小不均匀的特征,利用纹理和地形因子计算最优分割参数进行多尺度分割;然后根据植被覆盖率、岩石裸露率以及坡度因子构建石漠化分级指标;最后参照石漠化分级标准、光谱信息以及纹理特征等建立的分类规则提取喀斯特地区石漠化信息。选取贵州省石漠化严重的大方县时序TM/ETM+影像进行石漠化信息提取试验,结果表明:与基于像素的监督分类和非监督分类方法相比,基于面向对象的分类可以有效地减少因复杂地形导致石漠化信息提取结果 "椒盐化"现象,提取精度明显优于基于像素的监督分类和非监督分类方法。

The existing pixel-based supervised and unsupervised classification methods can′t meet the requirements of rocky desertification information extraction accuracy in karst rocky desertification area under the circumstances of complicated geological environment,large topographic relief and obvious shadows.In order to improve the accuracy of remote sensing image information extraction,texture feature data and topographic data are used to assist the object oriented method in the rocky desertification information extraction in karst rocky desertification area.Firstly,based on the characteristics of rocky desertification with uneven image sizes in TM/ETM+,the optimal segmentation parameters are calculated using texture and terrain factors to conduct multi-scale segmentation.Secondly,the grading indexes of rocky desertification are established based on vegetation coverage rates,rock exposure rates and slope factors.Finally,according to the grading rules of rocky desertification,spectral information and texture features,the information of rocky desertification in Karst area is extracted.The temporal TM/ETM+images of rocky desertification areas in DaFang,Guizhou,are selected for rocky desertification information extraction.The results show that comparing with pixel-based supervised classification and unsupervised classification methods,the object-oriented classification technology can effectively reduce the“salt and pepper phenomenon”caused by complicated topography,and the extraction accuracy is much better.