摘要
通过融合监督分类和非监督分类方法,提出了基于彩色遥感图像的河流提取算法。在监督分类方法中,采用局部傅里叶变换来捕捉图像的不同颜色和纹理特征,用从样本中提取出来的LFT特征集对大间隔最近邻分类器进行训练,把图像分为河流和背景两部分;在非监督分类方法中,采用基于颜色的k-均值分类分割方法,从背景中分离出河流。对两种方法的输出结果进行融合,即可得到最终的河流提取结果。实例表明:该方法能够准确地从彩色遥感图像中进行河流提取。
A river extraction algorithm based on color remote sensing images through fusing the detection outputs of supervised classification method and unsupervised clustering method was proposed. In the supervised classification method,using the local Fourier transform to capture discrimina-tive texture and color representation. Then used the local Fourier transform features extracted from labeled samples to train a large margin nearest neighbor classifier for classifying image pixels into two classes：river and backgrounds. In unsupervised clustering method,k-means clustering was adopted for color-based segmentation to separate river areas from backgrounds. Finally,the outputs of the two methods were fused to obtain detec-tion result. The results show that this method can extract the river from the color remote sensing image accurately.
出处
《人民黄河》
CAS
北大核心
2014年第3期10-12,共3页
Yellow River
基金
国家自然科学基金资助项目(61172181)
关键词
k-均值分类
局部傅里叶变换
图像分割
河流提取
遥感图像
k-means clustering
local Fourier transform
image segmentation
river extraction
remote sensing image
