论文部分内容阅读
在遥感影像分割分类中,种子区域生长算法是一种常见的分割算法.传统的种子区域生长算法只能提取单一连续的、纹理简单的目标地物,而对具有复杂纹理和多光谱特征的遥感影像,分割时存在分割效果差、不能同时有效地提取多个地物的问题.针对以上问题,本文提出了一种改进的面向对象的自动多种子区域生长算法.该方法适用于同时提取多个目标地物,且分割效果好.该方法首先使用一种改进的中值滤波对影像进行平滑处理,使目标内部一致性更高,同时保留纹理信息.然后通过一定的准则进行自动种子选取并进行生长,最后对生长后的区域进行碎斑合并处理,最终得到多种对象的分割结果.本文采用三组不同大小的1 m空间分辨率的航空影像进行实验,通过与分水岭以及传统单种子区域生长算法的多组实验对比,发现该方法可以面向全局对象,自动选取覆盖各种地物类型的种子,同时对多种地物目标进行分割处理,可为后续面向对象影像分析和应用提供可靠的数据基础.
In the remote sensing image segmentation classification, the seed region growing algorithm is a common segmentation algorithm.Traditional seed region growing algorithm can only extract a single continuous, simple texture of the target object, and remote sensing with complex texture and multi-spectral features Image and segmentation, the segmentation problem is poor and can not effectively extract multiple features at the same time.According to the above problems, this paper proposes an improved object-oriented automatic multi-sub-region growth algorithm.This method is suitable for simultaneous extraction of multiple And the segmentation effect is good.This method first uses an improved median filter to smooth the image to make the internal consistency of the target higher while preserving the texture information.Then the automatic seed selection is carried out by certain criteria And finally the segmentation results of the multiple objects are obtained.Three groups of aerial images with different spatial resolution of 1 m are used in this paper to study the effects of watershed and traditional single seed area growth The algorithm of multi-group experiments and found that the method can be oriented to global objects, automatically select the coverage of various Types of seeds, while a variety of ground object segmentation process, provides a reliable data base for the object oriented image analysis and subsequent application.