基于认知理论的高分辨率PolSAR图像目标解译算法研究
发布时间：2018-07-24 21:41
【摘要】：在当前大气污染日益严重,雾霾天气多发情况下,传统借助光学传感器成像的遥感系统的使用受到诸多限制,而天气状况等无碍极化合成孔径雷达(Polarimetric Synthetic Aperture Radar,简称Pol SAR)全天时、全天候、全极化模式地获取地物信息,被广泛运用于军事及民用研究领域,其发展受到广泛关注。Pol SAR系统通过获得不同极化状态下目标的散射信息,实现地物信息的轻松获取,如目标的物理特性、空间分布等。但是,高分辨率Pol SAR图像中,目标散射特性更加复杂,同一地物的不同部分可能呈现出不同的散射特性,低分辨率下的解译技术不满足当下Pol SAR系统图像处理要求。基于人类的认知机理及已有的认知模型,建立高分辨率Pol SAR图像目标解译算法。人类对图像的认知是目前最高水平的图像解译机制,主要归功于三个特点:一是视觉认知过程是由整体到局部的;二是认知过程分层进行,确保运行过程的高效及高精度;三是知识和经验是认知顺利进行的先决条件。深入研究人类的图像认知机制,并将其引入Pol SAR图像目标解译算法中,对于高效、智能、准确地进行图像解译具有重要意义。因此,本文基于人类认知机制,结合Pol SAR图像成像机理,建立“视觉认知-逻辑认知-心理认知”的高分辨率Pol SAR图像层次认知模型,并且先验知识全程参与,完成Pol SAR图像目标地物的快速、准确、智能地解译与识别。论文第一部分详细介绍了认知科学、Pol SAR图像处理等研究的国内外研究现状,如图像分割技术、图像目标识别技术。第二部分对认知领域及Pol SAR信息提取与目标解译技术的基本理论与模型进行了系统地阐述与说明。第三部分基于已有的理论知识,建立了基于先验知识的“视觉-逻辑-心理”认知的Pol SAR图像层次认知模型,并借助多层次图像分割、模糊逻辑、神经网络、上下文语义特征等理论,完成了相应的数学构建及编程实现。第四部分,利用三组不同成像条件下的Pol SAR图像所获得的实验结果进行算法验证。并且定量的结果分析表明,本文所提算法具有很好地图像解译效果,且具有一定的普适性,对Pol SAR图像解译研究具有重要的应用价值和意义。
[Abstract]:With the increasing air pollution and the frequent weather in haze, the use of traditional remote sensing system with optical sensor imaging is restricted by many restrictions, while the weather conditions such as weather conditions do not hinder the polarimetric synthetic Aperture Radar (Polarimetric Synthetic Aperture Radar,), which is referred to as Pol SAR), all day long. It is widely used in military and civilian research fields to obtain ground object information in all-weather, all-polarization mode. The development of .Pol SAR system can easily obtain ground object information by obtaining scattering information of target in different polarization state. Such as the physical characteristics of the target, spatial distribution, etc. However, in high-resolution Pol SAR images, the scattering characteristics of targets are more complex, and different parts of the same object may exhibit different scattering characteristics. The interpretation techniques at low resolution do not meet the requirements of current Pol SAR system image processing. Based on human cognitive mechanism and existing cognitive models, a target interpretation algorithm for high resolution Pol SAR images is established. Human cognition of image is the highest image interpretation mechanism at present, which is mainly attributed to three characteristics: one is that the visual cognitive process is from whole to local, the other is that the cognitive process is stratified to ensure the high efficiency and high accuracy of the operation process. Third, knowledge and experience is a prerequisite for the smooth development of cognition. It is of great significance for efficient, intelligent and accurate image interpretation to deeply study human image cognitive mechanism and introduce it into Pol SAR image target interpretation algorithm. Therefore, based on the human cognitive mechanism and the imaging mechanism of Pol SAR images, a high-level cognitive model of "visual cognitive-logical cognitive-psychological cognition" for high-resolution Pol SAR images is established, and prior knowledge is involved in the whole process. Fast, accurate and intelligent interpretation and recognition of target objects in Pol SAR images are completed. The first part of this paper introduces the research status of cognitive science Pol SAR image processing at home and abroad, such as image segmentation technology, image target recognition technology. In the second part, the basic theories and models of cognitive domain and Pol SAR information extraction and target interpretation technology are systematically expounded and explained. In the third part, based on the existing theoretical knowledge, the Pol SAR image hierarchical cognitive model based on transcendental knowledge is established, and with the help of multi-level image segmentation, fuzzy logic, neural network. The theory of contextual semantic features has completed the corresponding mathematical construction and programming implementation. In the fourth part, the experimental results of three groups of Pol SAR images under different imaging conditions are used to verify the algorithm. The quantitative results show that the proposed algorithm has a good map image interpretation effect, and has a certain universality, which has important application value and significance to the research of Pol SAR image interpretation.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN957.52
本文编号：2142741
[Abstract]:With the increasing air pollution and the frequent weather in haze, the use of traditional remote sensing system with optical sensor imaging is restricted by many restrictions, while the weather conditions such as weather conditions do not hinder the polarimetric synthetic Aperture Radar (Polarimetric Synthetic Aperture Radar,), which is referred to as Pol SAR), all day long. It is widely used in military and civilian research fields to obtain ground object information in all-weather, all-polarization mode. The development of .Pol SAR system can easily obtain ground object information by obtaining scattering information of target in different polarization state. Such as the physical characteristics of the target, spatial distribution, etc. However, in high-resolution Pol SAR images, the scattering characteristics of targets are more complex, and different parts of the same object may exhibit different scattering characteristics. The interpretation techniques at low resolution do not meet the requirements of current Pol SAR system image processing. Based on human cognitive mechanism and existing cognitive models, a target interpretation algorithm for high resolution Pol SAR images is established. Human cognition of image is the highest image interpretation mechanism at present, which is mainly attributed to three characteristics: one is that the visual cognitive process is from whole to local, the other is that the cognitive process is stratified to ensure the high efficiency and high accuracy of the operation process. Third, knowledge and experience is a prerequisite for the smooth development of cognition. It is of great significance for efficient, intelligent and accurate image interpretation to deeply study human image cognitive mechanism and introduce it into Pol SAR image target interpretation algorithm. Therefore, based on the human cognitive mechanism and the imaging mechanism of Pol SAR images, a high-level cognitive model of "visual cognitive-logical cognitive-psychological cognition" for high-resolution Pol SAR images is established, and prior knowledge is involved in the whole process. Fast, accurate and intelligent interpretation and recognition of target objects in Pol SAR images are completed. The first part of this paper introduces the research status of cognitive science Pol SAR image processing at home and abroad, such as image segmentation technology, image target recognition technology. In the second part, the basic theories and models of cognitive domain and Pol SAR information extraction and target interpretation technology are systematically expounded and explained. In the third part, based on the existing theoretical knowledge, the Pol SAR image hierarchical cognitive model based on transcendental knowledge is established, and with the help of multi-level image segmentation, fuzzy logic, neural network. The theory of contextual semantic features has completed the corresponding mathematical construction and programming implementation. In the fourth part, the experimental results of three groups of Pol SAR images under different imaging conditions are used to verify the algorithm. The quantitative results show that the proposed algorithm has a good map image interpretation effect, and has a certain universality, which has important application value and significance to the research of Pol SAR image interpretation.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN957.52
【参考文献】
相关期刊论文 前1条
1 肖鹏;徐军;陈少冲;;纹理特征提取方法[J];电子科技;2010年06期
相关博士学位论文 前1条
1 张腊梅;极化SAR图像人造目标特征提取与检测方法研究[D];哈尔滨工业大学;2010年
相关硕士学位论文 前1条
1 高贵;高分辨率SAR图像目标特征提取研究[D];国防科学技术大学;2003年
，本文编号：2142741
本文链接：https://www.wllwen.com/kejilunwen/wltx/2142741.html
