一种改进的遥感影像面向对象最优分割尺度计算模型

An Improved Calculation Model of Object-Oriented for the Optimal Segmentation-Scale of Remote Sensing Image

面向对象的影像分析方法能够充分利用高分辨遥感影像信息,有效提取目标信息。分割尺度的选择是面向对象影像分析方法的核心问题。该文采用样本控制的方法,构建了基于面积和周长的分割对象样本一致性评价因子,针对不同的信息提取目标,提出一种改进的遥感影像面向对象最优分割尺度计算模型,并验证了其优越性。

Research for the optimal segmentation scale calculation model has a positive significance to impure the accuracy of object-oriented image interpretation. Aiming at information extracting of different targets,using the methocl of samples-controlling which considering area and perimeter as factors to evaluate consistency between segmented-objects and samples, the quality evaluation function is put forward. This quality evaluation function is an improvement on before,it doesntt only Consider the in- ternal consistency of the segmented-objects and the heterogeneity between the segmented-objects as the factors of quality evaluation of segmentation,but also contains the factor of consistency between the segmented-object and the sample for controlling. And based on improving the existing method of the optimal segmentation scale calculation model, the effectiveness and practicability of the model is verified through two kinds of samples-choosing for the model. Results show that：this model can obtain the optimal segmentation-scale of remote sensing image quickly and efficiently; the choice for sample number and types effects the result of the calculation of the optimal segmentation-scale largely, when information extraction is just for a single target feature type,samples should be all chosen target feature type,and when information extraction is for various types of ground targets, samples should be chosen as the same types with information extraction targets, so the choice of sample＇s number and types should be determined according to the target of information-extraction;the algorithm of the model is simple,and is easy to real- ize,it is a practical algorithm model.