期刊论文

Self-extraction of water bodies from remote sensing image based on graph cuts

中文

计算机工程与设计

1000-7024

2019

40

5

1413-1417,1423

10.16208/j.issn1000-7024.2019.05.039

本校为其他机构

为实现遥感图像水体的自提区,结合图片灰度值的分布,做出水体灰度值与非水体灰度值均服从高斯分布的假设,融合高斯混合模型,提出一种基于图割的图像分割算法,解决了传统图割算法需要人工标注的问题,能准确地在遥感图片中提取水体区域。对通过高斯滤波处理后的真实遥感图像进行实验,采用EM算法计算高斯混合模型的相关参数,利用高斯混合模型计算图割中t-link的权重,采用增广路径法解决最大流最小割问题对图像进行分割。对比阈值分离相关算法,融合高斯混合模型的图割算法具有较高的准确率,省去传统图割算法中人工标记这一环节。

According to the distribution of pixel gray value,an assumption that gray values of water and non-water follow Gaus- sian distribution respectively was made,and an algorithm of image segmentation based on GMM was presented to achieve the purpose of self-extraction of water bodies in microwave remote sensing images.The real remote sensing images which were processed using Gauss filter were tested.The EM algorithm was used to calculate the relevant parameters of the Gaussian mixture model and the model was used to calculate the weight of t-link in Graph Cuts.The augmented path method was used to...