面向对象的国产GF-1遥感影像水体信息提取研究

鉴于传统方法在提取高分辨率影像中水体信息的不足,以高分辨率国产卫星GF-1 2m/8m遥感影像为数据源,以特克斯河部分区域为研究区,通过分形网络进化算法实现影像分割并借助ESP尺度分割工具选取特定地物的最优分割尺度,随后使用基于规则的面向对象方法实现了水体信息的精度提取,并与传统的分类方法提取结果进行比较。结果表明:面向对象的总体分类精度达到了94.01%,Kappa系数为0.86;最大似然分类精度为90.00%,Kappa系数为0.80;面向对象的分类方法对国产GF-1影像是可行的,可以避免"椒盐"现象和图斑破碎,且分类精度比传统的方法高,可为国产高分影像研究与应用提供科学参考。

Study of water information extraction based on domestic GF-1 remote sensing image by using object-oriented method

Due to the shortcomings of the traditional method in extracting the high-resolution water body information, the remote sensing image of domestic high-resolution satellite GF-1 2m/8m was used as the data source. Taking Tekes County as the study area, the image was segmented through fractal network evolutionary algorithm and the optimal segmentation scale of specific ground object was selected by ESP scale segmentation tool, and then the accurate extraction of water information is achieved by rule-based and object-oriented approach, and the results are compared with the traditional classification method. The result shows that the accuracy of object-oriented classification is 94.01% and kappa coefficient is 0.86; the accuracy of maximum likelihood classification is 90.00% and the Kappa coefficient is 0.80. So the object-oriented classification of GF-1 image is feasible, which can avoid salt-phenomenon and patch fragmentation, and the accuracy is greatly improved. It provides scientific reference for the research and application of domestic GF-1 images.