摘要
根据耕地在作物全生长季内光谱特征变化频繁,其他地物光谱特征相对稳定或有规律变化的区别,基于多时相TM图像,结合相关地物的变化检测成果,提出了一种耕地信息提取方法,即首先利用改进的归一化差值水体指数(mordified normalized difference water index,MNDWI)剔除水体,然后利用基于分类后验概率变化向量分析方法剔除其他非耕地植被及人工地面,再利用面积阈值结合众数滤波方法去除噪声,最终得到耕地信息。将该方法在北京市通州区进行试用,总体识别精度达到93.3%,得到的耕地信息对修正原有耕地矢量图上的错分和漏分有重要作用。
This paper proposed a method for extracting cropland based on multi - temporal TM images. Modified normalized difference water index was used to water body extraction ; through the change vector analysis of posterior probability space, the information of grassland, artificial surface and woodland was acquired. Cropland result of Tongzhou District of Beijing extracted by the method was assessed by vector data from historical high resolution images. It is shown that the cropland result has high accuracy and can update historical data. Based on precision evaluation, the authors have found that the overall accuracy and Kappa coefficient of classification image are both higher than 90%. The error direction analysis shows that the possibility of cropland omission is small.
出处
《国土资源遥感》
CSCD
北大核心
2013年第4期166-173,共8页
Remote Sensing for Land & Resources
基金
高分辨率对地观测系统重大专项(编号:234008)
关键词
多时相
耕地提取
TM图像
变化向量分析
multi - temporal
cropland information extraction
TM image
change vector analysis
