高光谱遥感影像分解模型研究

项目名称： 高光谱遥感影像分解模型研究

项目编号： No.61271408

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 无线电电子学、电信技术

项目作者： 王毅

作者单位： 中国地质大学（武汉）

项目金额： 75万元

中文摘要： 高光谱图像分解能够有效揭示地物几何结构和纹理特征，为遥感影像信息提取与识别奠定良好基础。如何有效分离几何结构、纹理和噪声来进行图像处理与分析和准确提取与识别地物信息是高光谱遥感影像分解研究中的核心难题。本课题拟以"图像分解-非线性尺度空间滤波-信息融合"为研究主线，采用偏微分方程、小波变换和形态学滤波等技术，发展高光谱遥感影像分解模型。并且，以基于扩散张量的各向异性扩散模型为基础，建立结构/纹理分量病态问题求解通用偏微分方程模型，并通过不同的条件约束，来进行图像恢复与去噪、修复、内插和分割等。此外，拟对结构/纹理图像进行分类，并结合光谱特征和逻辑模型，提出高光谱影像结构、纹理和光谱信息融合方法，目的是准确提取和识别地物，提高影像分类精度。本课题的研究成果有望为高光谱遥感影像的应用分析提供可靠的信息来源，并为后续处理奠定基础，具有较高的实际应用价值。

中文关键词： 高光谱；遥感；图像去噪；偏微分方程；图像分解

英文摘要： Hyperspectral image decomposition can reveal geometrical structure and texture features and provide a good foundation for remote sensing image information extraction and recognition. How to effectively separate geometrical structure, texture and noise to perform image processing and analysis, and to accurately extract and recognize of material information are core problems in the study of hyperspectral remote sensing image decomposition. The main threads of the project are focus on image decomposition, nonlinear scale space filtering and information fusion. We plan to use partial differential equations, wavelet transform and morphological filtering techniques to develop a hyperspectral remote sensing image decomposition model. And we plan to present a general partial differential equation model to solve ill-posed problems for decomposed structure and texture components, based on a diffusion tensor driven anisotropic diffusion model. Moreover, different constrained conditions are to added in the model to perform image restoration and denoising, inpainting, interpolation and segmentation. In addition, we plan to perform image classification for decomposed structure and texture images and present an integration method of structure, texture and spectral information by using a logic model. The aim of the method is to

英文关键词： hyperspectral；remote sensing；image denoising；partial differential equation；image decomposition