Research on Accurate Extraction Method of RS Image of Cultivated Land in Guanzhong Plain
[Purposes]This paper aims to assist in the dynamic monitoring and management of farmland protection in the Guanzhong Plain region,and efficiently obtain large-scale farmland image distribution data.[Methods]Taking Lintong District,Xi'an City,Shaanxi Province as the research area,and based on the SNAP platform and ENVI processing software,the maximum likelihood classification method in su-pervised classification is used to identify and extract the remote sensing data of remote sensing imagery of Sentinel 2's high-resolution farmland RS(Remote Sensing),and obtain geographic information such as spatial distribution and area of farmland.[Findings]The total area of cultivated land extracted is 411.57 square kilometers,mainly distributed in the northern part of Lintong District,including Xiangq-iao,Jiaokou,Dongyang and other streets,as well as in the gentle areas of Hezhai and Lingkou streets in the central and eastern parts.It is similar to the official statistical area,with an error of only 0.92%.The overall classification accuracy of the extraction is 96.15%,and the Kappa coefficient is 0.94,which meets the accuracy requirements.[Conclusions]The maximum likelihood classification method is more suitable for extracting cultivated land results and actual data,which proves that the maximum likelihood classification method has a high matching degree in the recognition and detection of actual cultivated land types,and can accurately achieve land use type recognition.
maximum likelihood classificationSentinel-2 high-resolution remote sensing imagesculti-vated landENVI
NETL
NSTL
万方数据
