论文部分内容阅读
随着卫星遥感影像空间分辨率的不断提高,面向对象的地物信息提取技术发展迅速。图像分割作为面向对象分类的关键步骤之一,其分割尺度的参数设置目前仍以分类者的多次尝试和主观判断为依据,效率较低且分割结果因人而异。本文以WorldView2影像数据为例,结合当前现有的理论和方法,实现了一种计算机可自动进行主成分变换的高分辨率遥感图像全局最优分割尺度选取算法。改进后的算法以主成分变换所得的主成分影像作为图像分割的编辑层,主成分的特征值百分比作为计算异质性参数和分割质量评价值的权重,自动计算当分割尺度从20增至200时分割图像的分割质量评价值(GS),解决了人为确定图像分割编辑层的片面性问题,并利用三次样条插值选取出GS最高值所对应的尺度即为最优分割尺度。结果表明,该最优分割尺度选取方法可有效避免人为确定分割尺度的主观性、片面性和低效性,提升了高分辨率影像分割质量。
With the continuous improvement of spatial resolution of satellite remote sensing images, object-oriented object information extraction technology is developing rapidly. As one of the key steps in object-oriented classification, image segmentation is still based on multiple attempts and subjective judgments of the classifier. The segmentation efficiency is low and the segmentation results vary from person to person. In this paper, taking WorldView2 image data as an example, combined with the current existing theories and methods, a global optimum segmentation scale selection algorithm for high resolution remote sensing images is achieved by computer automatically transforming principal components. The improved algorithm takes the principal component image obtained by the principal component transformation as the editing layer of the image segmentation. The percentage of the eigenvalue of the principal component is used as the weight for calculating the heterogeneity parameter and the segmentation quality evaluation value. When the segmentation scale is increased from 20 to 200 The segmentation quality evaluation value (GS) of the segmentation image is used to solve the problem of one-sidedness of artificially determining the editorial layer of the image segmentation. The optimal segmentation scale is obtained by selecting the corresponding GS maximum value by cubic spline interpolation. The results show that this optimal segmentation method can effectively avoid subjectivity, one-sidedness and inefficiency of the segmentation scale and improve the quality of high-resolution image segmentation.