基于多尺度分割的面向对象分类方法提取冬小麦种植面积

ESTIMATION OF WINTER WHEAT PLANTING AREA USING OBJECT-ORIENTED METHOD BASED ON MULTI-SCALE SEGMENTATION

应用面向对象方法,对遥感图像进行多尺度分割,即首先进行大尺度分割,结合NDVI提取植被信息,将图像分为植被和非植被;然后在植被信息类内再进行小尺度分割,利用NDVI并融入几何特征进一步提取冬小麦种植面积及空间分布。在遥感分类的基础上,将线性数据按宽度缓冲,从分类结果中扣除。将扣除结果与地面样方实测数据对比分析。结果表明,监测结果减轻了传统分类方法的椒盐效应,监测结果与验证样方数据比较精度为94.06%。

Object-oriented method was used for the remote sensing image multi-scale segmentation.The first step was large-scale segmentation,in which the image was classified as vegetation and non-vegetation by NDVI.The second step was small-scale segmentation;the spatial distribution and planting area of winter wheat were extracted by NDVI combined with geometric characters.Based on the remote sensing classification,this paper compared the calculation results with survey sampling results in the same coordinate.The results showed the remote sensing monitoring results had decreased the ＂salt and pepper＂ effect,and the accuracy between remote sensing monitoring results and verification sampling survey results amounted to 94.06%.