基于同态系统滤波的高分辨率遥感图像河流信息提取  

River information extraction from high resolution remote sensing image based on homomorphic system filtering

作　　者：杨宜菩 杨帆[1] 潘国峰[1,2] 张慧敏[1] 

机构地区：[1]河北工业大学电子信息工程学院,天津300401 [2]河北工业大学微电子研究所,天津300401

出　　处：《计算机应用》2016年第1期248-253,281,共7页journal of Computer Applications

基　　金：国家科技重大专项(2009ZX02308-004);河北省高等学校科学研究项目(Z2014088)~~

摘　　要：为了解决在高分辨率遥感图像中易受同物异谱现象影响而导致河流信息提取不完整问题,依据河流区域在遥感图像中具有特定频率纹理信息这一特点,提出一种基于同态系统滤波的高分辨率遥感图像河流信息提取方法。该算法通过对遥感图像的频谱分析,确定河流信息在频域中的特征标识,采用一维剖面线加窗短时分析河流定位技术实现了全自动、快速全幅遥感图像河流定位和宽度估计,在同态系统下设计了低通滤波器实现对低频河流信息的提取。该算法采用"高分一号"遥感影像作为实验数据来源,相对于传统利用光谱信息的提取方法具有较高的提取精度,Kappa系数达0.85以上,并且实现全自动提取。实验结果表明所提算法能够快速、有效地提取复杂地貌区域内的河流信息。Concerning the incomplete extraction of river information easily affected by synonyms spectrum phenomenon from the high resolution remote sensing image, considering the specific frequency texture information of the river region, a river information extraction method from the high resolution remote sensing image based on homomorphic system filtering was proposed. The spectral analysis of remote sensing image was used by the proposed algorithm to define the feature identification of river information in the frequency domain. The one-dimensional profile line adding-window short-time analysis river location technology was adopted to realize river location and width estimation of the remote sensing image automatically and quickly.Finally, the low-pass filter under the homomorphic system was designed to extract low frequency river information. The GF-1image was used as experimental data. The extraction accuracy of the proposed algorithm was higher than the traditional method based on spectral information. The Kappa coefficient reached 0. 85, and the extraction process was automatic. The experimental results show that the proposed algorithm can extract river information from complex geomorphology region quickly and effectively.

关 键 词：高分辨率遥感图像 河流信息提取 加窗短时分析 同态系统滤波 高分一号卫星 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置] P236[自动化与计算机技术—控制科学与工程]