峰丛洼地农作物面向对象信息提取规则集

摘要

喀斯特地区典型峰丛洼地具有地类零碎、农作物种植分散、信息难以提取的特点，针对这些特点，该文利用面向对象分类技术，综合运用纹理、光谱和几何特征，构建规则集，对该地区农作物进行提取。通过用RapidEye 影像对南宁城郊喀斯特地区典型峰丛洼地的农作物提取试验，印证该方法能得到较好的提取效果，总精度达到87．89％，Kappa 系数为0．8612。利用其创建规则集的方法，在 SPOT-5影像中再次试验并进行验证，水稻和甘蔗信息提取总体精度达0．8703，Kappa 系数为0．854。%In the typical peak cluster depression of karst area,terrain is broken with scattered crops to extract.According to these characteristics,this paper integrated texture,spectrum and geometrical characteristics and built a set of rules to extract the region of crops.By the experiment in Nanning,it was found the total accuracy is 87.89%,and the Kappa is 0.8612.It was further tested on the SPOT-5 image that by using the created rule sets,the overall accuracy of extracted rice and sugar cane is 0.8703,and the Kappa is 0.854,respectively.