基于DS证据理论的多幅遥感影像融合去噪

发布时间：2018-07-04 06:46

本文选题：遥感影像 + DS证据理论　；参考：《海洋科学进展》2017年03期

【摘要】：针对目前单幅遥感影像去噪方法会过度扼杀影像边缘信息或噪声去除不理想问题,提出了一种基于DS(Dempster-Shafer)证据理论的多幅遥感影像融合去噪方法。DS证据理论结合多源信息,能够有效处理不确定性问题,适用于噪声具有随机性和不确定性情景。本文采用Landsat8卫星获取的上海周边海域遥感数据,依据DS证据理论处理不确定问题的优势,并充分利用多幅遥感影像的有效信息,设计4个模型,即两状态高斯混合模型、两个噪声分析模型、边缘分析模型。这4个模型用以获取每个像素与噪声相关的概率,作为证据理论的4个证据,用于决策去噪方案。结果表明,本文提出的多幅遥感影像融合去噪方法在保证去除噪声的情况下能较好地保持影像边缘和纹理细节信息。
[Abstract]:Aiming at the problem that the single remote sensing image denoising method can kill the edge information or the noise removal is not ideal, this paper proposes a multi-piece remote sensing image fusion denoising method based on DS (Dempster-Shafer) evidence theory. DS evidence theory combines multi-source information. It can deal with the uncertainty problem effectively, and can be applied to the stochastic and uncertain situations of noise. In this paper, the Landsat8 satellite is used to obtain the remote sensing data in the surrounding sea area of Shanghai. According to the advantage of DS evidence theory to deal with the uncertain problem, four models are designed, that is, the two-state Gao Si mixed model, which makes full use of the effective information of multiple remote sensing images. Two noise analysis models, edge analysis models. The four models are used to obtain the noise related probability of each pixel, and as four evidences of evidence theory, they are used for decision denoising scheme. The results show that the proposed multi-image fusion denoising method can keep the image edge and texture details well under the condition of noise removal.
【作者单位】：上海海洋大学信息学院;国家海洋局东海预报中心;
【基金】：上海市科学技术委员会科研计划项目——地方院校能力建设(15590501900) 极地海洋环境监测示范应用系统开发项目——极地海洋环境监测示范应用系统开发(201405031-05) 国家自然科学基金项目——基于多模态深度学习的弱特征多源海洋遥感影像协同分类模型研究(41671431);国家自然科学基金青年基金项目——一种面向多模态遥感信息的质量抽样检验方案研究(41501419)
【分类号】：TP751

【相似文献】