【摘要】 研究提取耕地的方法是为了提高耕地提取的高效性,直接目视解译时山体阴影和水体难以区分,建设用地和未种植的耕地难以区分,本文主要以2015年的Landsat 8遥感影像,借助辅助数据DEM及MODIS-NDVI等多源数据基于eCognition平台应用多阈值分割、多尺度分割将影像进行分割成对象,再分别用阈值法、归一化水体指数(NDWI)、增强的指数型建筑指数(EIBI)等将坡度大于25度的地区、水体、建设用地剔除。最后利用影像的波段特点及MODIS-NDVI值进行多次提取研究区耕地,以达到较高精度的分类方法。利用总体面积精度及空间分布进行精度评价。其总体面积精度达到98.84%,且提取出的耕地分布情况符合作物的种植规律。更多还原

【Abstract】 The method of extracting cultivated land is to improve the efficiency of cultivated land extraction and direct visual interpretation of mountain shadows and water is difficult to distinguish.In this paper,the Landsat 8 remote sensing image in 2015,multisource data such as auxiliary data DEM and MODIS NDVI are applied to multi-threshold and multi-scale segmentation based on eCognition platform to segment images into objects,then using the threshold method,NDWI,EIBI to remove areas,waters and construction sites with slopes that are greater than 25 degrees.Finally,image band features and MODIS NDVI values are used for multiple extraction of cultivated Land in order to achieve a higher accuracy of the classification method.By using the area of uniformity,spatial distribution of accuracy evaluation,the accuracy of the total area reached 98.84%,and the distribution of cultivated land was in accordance with the law of crop cultivation.更多还原