基于遥感数据光谱和空间特征的边缘提取方法

An Edge Extraction Method for Integrating Spectral and Spatial Features of Remote Sensing Image

传统的边缘检测算子(Sobel、Robert、Prewitt、Kirsch、Gauss Laplac等算子)主要是通过图像空域特征微分,建立不同结构的模板完成高分辨率遥感图像的边缘提取。而高分辨率遥感数字图像包括了空间域和光谱域两种信息,借助这两种特征信息提高提取城市边缘的信息精度已经成为当前算法开发的基本思路。本研究采用了"象元替换"的思想设计光谱分解和边缘检测的算子模板步骤,综合了图像的光谱特征和空间特征信息。研究结果表明,这种方法有效地提高了城市遥感数据边缘信息的提取精度,同时还具有方法简便、计算速度快的特点。

Traditional edge detection kernels such as Sobel kernel work through different templates of differential coefficients in spatial territory. Spatial information as well as spectral information is included in the remote sensing image of high resolution. So, it will be a essencial thoughtway of current method for city edge information extraction integrating spatial and spectral features of image. The processing was designed for edge information extraction with the idea of pixel swap and spectral analysis, integrating spectral and spatial features of image.The results prove its effectiveness and show that the precision of city remote sensing image edge information extraction is improved. Simultaneously, its a easy way to understand and have a high rate.