摘要
提出了一种卫星图像河流检测的融合算法,包含两方面要素,一是基于无监督聚类k-means算法利用颜色信息将河流从图像背景中分隔出来;二是运用提出的基于主元分析(PCA)的局部傅里叶变换(PLFT)算法实现图像纹理图像特征信息检测,然后利用有监督的大间隔近邻(LMNN)分类算法将图像像素分为河流和背景。在决策层次上,融合这两种方法来提取图像中的河流信息,融合算法在保留这两类方法的优势的同时又摈弃了各单一方法的不足之处。选取大量实验时拍摄的卫星河流图像进行研究,结果表明上述方法更加清晰完整地检测出卫星图像中的河流信息。
The fusion algorithm for satellite image river detection proposed in this paper contains two factors,one is separating the river from the image background with color information based on unsupervised clustering k-means algorithm;the other is detecting the image texture feature information by the proposed PCA-based local Fourier transform(PLFT)algorithm,and then dividing the pixels of the images into river and background with use of the supervised large margin nearest neighbor(LMNN)classifier.At the decision level,through fusing the outputs of the above two methods to detect the river information in the satellite images,the fusion algorithm retains the advantages of these two methods while abandoning the inadequacies of each single method.A large number of satellite images taken from the experiment were selected for research,of which the results show that the proposed method can detect river information in satellite images more clearly and completely.
作者
汤振鹏
陈劲
TANG Zhen-peng;CHEN Jin(Jiangmen Power Supply Bureau,Guangdong Grid Co.,Jiangmen Guangdong 529000,China)
出处
《计算机仿真》
北大核心
2019年第3期45-49,86,共6页
Computer Simulation
基金
国家自然科学基金项目(61602182)