基于阈值分割与决策树的SAR影像水体信息提取

目前我国的GF-3 SAR数据可实现全天时全天候的对地观测,已服务于海洋、减灾、水利、气象等多个领域。改进了基于阈值分割法与决策树的GF-3 SAR影像水体信息提取方法,首先对GF-3 SAR影像进行基本处理;再采用KI二值化阈值分割法进行图像分割;然后通过构建知识决策树模型来提取水体信息,为了提高精度,采用GDEM数据进行地表建模,提取山体阴影;最后利用空间分析功能将地形建模阴影图与提取的水体范围进行匹配,去除山体阴影,进而获得水体信息的精确范围。通过混淆矩阵计算得到水体信息提取的总体精度为89.22%,Kappa系数为0.71,精度优于基于光学GF-2号影像的水体指数法提取结果。整个流程人工干预少,具有自动化更强、效率更高的优势。

Water Body Information Extraction of SAR Images Based on Threshold Segmentation and Decision Tree

At present,China’s GF-3 SAR data can realize all-weather Earth observation,and have served in the ocean,disaster reduction,water conservancy,meteorology and other fields.In this paper,we improved the GF-3 SAR image water information extraction method based on threshold segmentation and decision tree.After basic processing of GF-3 SAR images,we used KI binarization threshold segmentation method to segment images,and built a knowledge decision tree model to extract water information.To improve the accuracy of experimental results,we used GDEM data for surface modeling,and extracted mountain shadows.Finally,we used the function of spatial analysis to match the terrain modeling shadow map with the extracted water body range,so as to remove the image of mountain shadow,and obtained the accurate range of water body information.The total accuracy of water body information is 89.22%and the Kappa coefficient is 0.71.The accuracy of the experimental result is better than that of the water body index method of optical GF-2 image.The whole process has less manual intervention,stronger automation and higher efficiency.