基于eCogition面向对象分类方法的水稻种植面积提取

摘要

As one of food production bases of Hengxian ,Guangxi ,the rice planting area in Baihe county accounts for half of the arable land area of the town .Using remote sensing technology to ex-tract rice planting area has important significance for evaluation of production capacity and advantage . This research is based on eCognition platform ,using multi - threshold ,multi -scale segmentation algorithm for segmentation of high resolution RapidEye image ,using the multi-threshold segmenta-tion ,NDVI ,RVI to eliminate the effect of forest ,water ,construction land respectively .By using the nearest neighbor method for object classification ,random sampling ,area of consistency and spatial distribution are carried out for accuracy evaluation .The area of consensus reaches 94% and the extrac-tion lives up to distribution of rice .%百合镇作为广西壮族自治区横县的粮食生产基地之一，水稻种植面积占到全镇耕地面积的一半，用遥感技术提取其水稻种植面积对于评价生产能力及优势具有重要意义。本文基于eCognition平台，运用多阈值、多尺度分割算法对RapidEye高分辨率影像进行对象分割，分别用多阈值分割、ndvi、rvi剔除林地、水体、建筑地的影响。利用最邻近法对对象分类，用野外实测点随机抽样、面积一致性、空间分布进行精度评价，其中面积一致性达到94％，提取的水稻分布符合自然规律。