基于蓝噪声理论的多尺度遥感森林植被分割

Multi-scale remote sensing forest vegetation segmentation based on the theory of blue noise

针对森林植被分割方法中纹理尺度以及植被纹理描述问题,基于蓝噪声理论,提出一种多尺度的结合灰度、形状以及其他纹理特征构建森林纹理结构基元的分割方法。该方法通过对森林植被典型区域进行快速傅里叶变换,探测区域的蓝噪声特征,并计算森林植被纹理单元的尺度和灰度分布。然后结合区域的灰度、形状和其他纹理特征构建不同尺度下森林纹理结构基元,利用森林纹理结构基元对图像进行提取,获取最终分割结果。实验结果表明,本文提出的算法能够提高植被区域分割的准确性,取得了较好的分割效果。

Aiming at allusion of texture scale and vegetation texture description problem of forest vegetation segmentation method, and based on the blue noise theory, a combination method of multi-scale gray, shapes and textural features is proposed.This method could detect the field of blue noise characteristics via acting FFT to typical area of forest vegetation and calculate scale and gray distribution of forest vegetation segmentation unit simultaneously. The way to obtain final segmentation consequence is to build crown constructing primitive at different scale with combination of gray level, shape and other texture features, then utilizing that constructing primitive to extract original image. The experimental result indicates that arithmetic in the article could advance the veracity of vegetation field with a preferable segmentation effect.