基于资源三号遥感影像的湿地信息提取

Wetland Information Extraction Based on ZY-3 Remote Sensing Image

基于资源三号高分辨率遥感影像,提出了面向对象的湿地分区提取方法,结合光谱、形状、面积及纹理等多种湿地特征,对大洼县湿地资源进行逐类提取,并对提取结果进行精度评价。该方法利用面向对象的影像分割和合并技术,将同质性较高的像元转化为对象,通过光谱特征和归一化植被指数将湿地分区,然后结合光谱、纹理和形状等特征对湿地进行逐类提取。实验基于资源三号高分辨率遥感影像对大洼县湿地进行遥感提取,总体分类精度达94.03%,kappa系数为0.8617,与传统分类方法相比,提取精度上有一定提高,特别是在芦苇湿地和滩涂湿地的提取精度上有明显提高。

Based on high-resolution remote sensing ZY-3 images,this paper proposed an object-oriented wetland partition extraction method.Combining various wetland characteristics such as spectrum,shape,area and texture,we extracted the Dawa County wetland resources one by one and evaluated the extraction result accuracy.This method uses the object-oriented image segmentation and merging technology to convert the homogeneous pixels into objects,divide the wetland by spectral features and normalized vegetation index,and extract the wetlands one by one by combining spectrum,texture and shape characteristics.The wetland remote sensing extraction experiment in Dawa County achieved the overall classification accuracy of 94.03%and kappa coefficient of 0.8617.Compared with the traditional classification method,the extraction accuracy has been improved,especially in reed wetland and mud wetland.