一种改进的面向对象最优分割尺度模型

A Modified Optimum Segmentation Scale Model Based on Object Oriented

面向对象的影像提取技术中,分割尺度的选择是一个关键问题。论文在前人研究的基础上,以对象内部的灰度标准差表示对象内部的同质性,以对象与邻域平均差分的绝对值来表示对象间的异质性,并综合考虑了对象与目标地物在形状和光谱上的一致性,提出了一种最优分割尺度计算模型。结果表明,此模型能快速得到最优分割尺度,提高了建筑物提取精度。

In the object oriented image extraction technology,the choice of segmentation scale is a key question.On the basis of previous studies,the internal homogeneity of the object is represented by the gray standard deviation of the object,and the heterogeneity of the object is represented by the absolute value of the mean difference between the object and the neighborhood,and concern the consistency of the object and the actual object in shape and spectrum,an optimal segmentation scale computation model is proposed.The results show this model can quickly get the optimal segmentation scale,improve the accuracy of building extraction.