基于国产高分卫星面向对象城市地物最优尺度选择及评价研究

Optimal Scale Selection and Evaluation for Urban Object Classification Based on GF-1 and GF-2 Satellite Imagery

针对面向对象分类中的分割尺度问题,本文以国产高分一号卫星(GF-1)、高分二号卫星(GF-2)影像为数据源,选取沈阳市两处城区影像作为实验数据,以典型城市地物为分割对象,选择分割质量函数法、均值方差法、RMAS指数法进行尺度分割,并利用欧几里得2指数(ED2)评价其最优尺度,从而确定不同影像不同地物对应的最优尺度分割方法与尺度。

In terms of the issues of segmentation scale in object-oriented classification, this study selected remote sensing images ofGF-1 and GF-2 as data sources. The urban area in Shenyang was selected as the experimental area. The aims of this study are to select and evaluate the optimal segmentation scale for urban features. Segmentation quality function method,mean variance method, andthe Ratio of Mean Diff. to Neighbors （ABS） to Standard Deviation （RMAS） index method were used to select the optimal scale for different urban objects. Furthermore, the Euclidean Distance 2 （ED2） method was applied to evaluate the optimal segmentation scale forcomparing with the reference object.