面向对象和DEM的ZY-1 02C影像湿地提取研究

本文运用面向对象分类与DEM数据相结合的方法,对资源卫星一号02C卫星遥感影像进行湿地提取。探索了基于对象与DEM信息的提取技术,在02C影像湿地的提取应用,对研究我国国产卫星在湿地监测和保护方面有重要的意义。研究结果表明：（1）面向对象的遥感影像信息提取方法,可同时兼顾影像光谱信息及空间信息,适用于02C影像的湿地提取,精度得到明显提高;（2）基于对象与DEM信息的提取方法,使沼泽地与草地相混淆的现象明显减轻,湿地分类精度进一步提高,该方法适用于高分辨率遥感影像的湿地提取研究;（3）基于对象与DEM信息提取的水田、水体、沼泽地及河滩的精度,分别为88.46%、97.44%、86.96%和83.33%,满足资源卫星一号02C遥感影像对湿地进行监测和保护的需要。

Wetland Information Extraction of ZY-1 02C Image Based on the Object and DEM

With the application of high resolution remote sensing image, the object-oriented technology has been developing rapidly in high resolution remote sensing image information extraction. Many scholars are interested in conducting research and application in this field, but the study of ZY-1 02C satellite imagery is rare. This paper uses the method of combining object-oriented classification and DEM data, to extract the wetland information of ZY-1 02C satellite image, and explores the application of this method in wetland information extraction from ZY-1 02C satellite imagery. It is of great significance to the research on the application of domestic satellite in wetland monitoring and protection. The results showed that: (1) ZY-1 02C satellite imagery has relatively high resolution, rich spatial information and large local heterogeneity. In addition, its spectral characteristics among various types of wetland are similar. The proposed object-oriented remote sensing image information extraction considers both the image’s spectral information and spatial information. It is applied to ZY-1 02C satellite imagery for the wetland information extraction, and the precision of its classification result is significantly higher than the pixel-based method. (2) This article has explored the method of combining object-oriented classification and DEM. The DEM data was taken as an additional band of image and was associated with the three original bands from the image to be involved in the segmentation, and the segmented objects are classified. The confusion between marsh and grass lands during the extraction is reduced, and the accuracy of wetland type classification is further improved. This method is suitable for wetland information extraction from high resolution remote sensing image. (3) The information extraction results based on object-oriented classification and DEM showed that, the precisions of paddy fields, water body, marsh and tidal flats are 97.44%, 86.96%, 88.46% and 88.46% respectively, which satisfy the wetland monitoring and protection requirements of 02C remote sensing image.