一种基于亮度值匹配的多光谱影像水体提取方法

A method of water body extraction from multispectral image based on luminance value matching

针对传统水体提取方法对低光谱分辨率影像适应性不强的问题,本文设计了一种基于亮度值匹配的离散粒子群算法。该算法首先通过计算像素各波段亮度值与标准水体各波段亮度值间相似度,得到水体概率,并构建目标函数;然后利用粒子群迭代求解目标函数最大值,通过全局最优值实现水体分类。利用涵盖多种水体分布情况的资源三号四波段多光谱影像数据集进行实验。实验结果表明,本文设计方法能较好实现低光谱分辨率影像水体提取,为低光谱分辨率影像水体精准提取问题提供了部分解决思路。

To solve the problem that the traditional water extraction method is not adaptable to images with low spectral resolution,a discrete particle swarm optimization algorithm based on luminance value matching is designed in this paper.The similarity between the luminance value of each band of image pixel and that of the standard water body is calculated by using this algorithm to obtain the water body probability and build the objective function,whose maximum value is solved by particle swarm optimization iteration,and the water body can be classified using the global optimal value.The multispectral image data of 4 bands of ZY3 satellites covering various water body distributions are used for the experiment,and the results show that this method can realize water body extraction from images with low spectral resolution.