光学遥感系统的建模仿真及图像处理技术研究

发布时间：2018-06-15 01:12

  本文选题：MTF + 刃边法　； 参考：《中国科学院研究生院（西安光学精密机械研究所）》2014年博士论文

【摘要】：光学遥感成像技术的飞速发展，要求空间光学遥感系统的空间分辨率更高、光谱分辨率更高、辐射分辨率更高，然而遥感器的轻小型化与高分辨率成像之间是矛盾的，这就使得空间光学遥感器的设计、研制难度越来越大。受观测条件的限制，发送各种条件下的遥感数据变得很不现实，利用实验室技术模拟仿真从传感器、卫星平台、深空环境、大气传输、地表景物到最终图像获取的整个遥感物理过程，能够保证在经济成本很低的条件下来获得客户所需要的大量图像。由此可见，光学遥感仿真技术在遥感任务预测、成像系统设计、图像质量评估、图像处理算法验证、图像解译训练等一系列领域都具有极其重要的应用价值，业已逐渐成为国内外相关研究机构竞相开展的热门研究课题。 在对光学遥感系统进行建模和仿真之后，我们会得到各种类型的仿真图像，，需要我们利用各种相应的光学遥感图像处理技术对其进行处理，从而提取出有用的信息，从而实现了各种遥感图像的多样性应用。本文的主要研究工作包括以下几个方面： 1)基于经验总结的MTF数学模型，编写了基于MTF模型的遥感成像链仿真平台V1.0，实现在全数字环境中动态模拟大气、光学系统、平台、CCD探测器、离焦、像移以及电子学系统等各个环节对最终遥感像质的影响，有助于综合评价遥感器的成像性能，从而实现遥感器的优化设计。 2)在基于刃边法计算遥感图像的MTF的算法中，提出一种新型的非线性二乘迭代算法拟合Fermi函数，摒弃了模拟退火算法计算函数系数时的缺点，在精度和速度方面都有较大改进。 3)基于对各种传统彩色插值算法的分析和比较，本文提出了一种新型的基于权重边缘-权重色差的彩色插值算法，实现对Bayer图像进行彩色复原的过程。该算法将色差法则、加权定律、自适应法则和迭代优化算法巧妙地结合在一起，增加了R、G、B三种颜色分量之间的相关性，而且插值方向选取合适，有效抑制了彩色摩尔条纹。 4)对天文星图进行预处理、星点粗定位、星点细定位之后，本文提出了一种新的特征匹配法以全等三角形的重心为特征点。然后基于这些三角形重心点集合，利用RANSAC算法计算配准转换模型，完成配准过程。整个过程是利用GPU对星图配准算法进行了并行程序设计，并利用Matlab中的nvmex编译GPU代码，使得Matlab可以直接调用生成的MEX文件，有效的加速了天文星图的配准速度，总时间加速比达到29.043倍。 5)利用自制的多波段偏振CCD地面实验装置，将白炽灯作为照明光源，获取目标的偏振光谱图像。本文实验的目的是辨别真花和假花、真叶和假叶，以DOLP为评价指标，发现与灰度图像相比，偏振度图像中的真花和假花的对比度数值更大，即两者更容易被区分。
[Abstract]:The rapid development of optical remote sensing imaging technology requires higher spatial resolution, higher spectral resolution and higher radiation resolution of space optical remote sensing system. However, there is a contradiction between the miniaturization of remote sensor and high resolution imaging. This makes the design and development of space optical remote sensor more and more difficult. Limited by observation conditions, it is very unrealistic to send remote sensing data under various conditions, using laboratory technology to simulate and simulate the transmission from sensors, satellite platforms, deep space environments, atmosphere, The whole physical process of remote sensing from the surface to the final image can ensure that the large number of images needed by the customer can be obtained under the condition of low economic cost. It can be seen that optical remote sensing simulation technology has very important application value in a series of fields such as remote sensing task prediction, imaging system design, image quality evaluation, image processing algorithm verification, image interpretation training and so on. It has gradually become a hot research topic that related research institutions at home and abroad are competing to carry out. After modeling and simulating the optical remote sensing system, we will get various kinds of simulation images, which need to be processed by various optical remote sensing image processing techniques, so as to extract useful information. Thus, the diversity of remote sensing images is realized. The main research work of this paper includes the following aspects: 1) based on the MTF mathematical model, a remote sensing imaging chain simulation platform V1.0 based on MTF model is developed to realize the dynamic simulation of atmosphere and optical system in the full digital environment. The effects of various aspects of the platform CCD detector, defocus, image shift and electronic system on the final remote sensing image quality are helpful for the comprehensive evaluation of the imaging performance of the remote sensor. 2) in the algorithm of calculating MTF of remote sensing image based on edge method, a new nonlinear two-multiplication iterative algorithm is proposed to fit Fermi function. In this paper, the shortcomings of simulated annealing algorithm in calculating function coefficients are abandoned, and the accuracy and speed are improved greatly. 3) based on the analysis and comparison of various traditional color interpolation algorithms, In this paper, a new color interpolation algorithm based on weighted edge and weight color difference is proposed to realize the color restoration of Bayer image. The algorithm combines color difference law, weighting law, adaptive rule and iterative optimization algorithm skillfully, which increases the correlation among the three color components, and the direction of interpolation is suitable. In this paper, we propose a new feature matching method, which is based on the center of gravity of a full triangle, after the pretreatment of star map, the coarse location of star and the fine location of star. Then, based on the set of triangle barycenter points, the registration transformation model is calculated by using RANSAC algorithm, and the registration process is completed. The whole process is to make use of GPU to design the parallel program of star map registration algorithm, and compile GPU code by using nvmex in Matlab, so that Matlab can directly call the generated MEX file, which effectively accelerates the registration speed of astronomical star map. The acceleration ratio of total time is 29.043 times. 5) using the self-made multi-band polarization CCD ground experimental device, the incandescent lamp is used as the illuminating light source to obtain the polarizing spectrum image of the target. The purpose of this experiment is to distinguish true flower from false flower, true leaf and false leaf. Using DOLP as the evaluation index, we find that the contrast value of true flower and false flower in polarization degree image is larger than that in gray image, that is to say, it is easier to distinguish the true flower from false flower.
【学位授予单位】：中国科学院研究生院（西安光学精密机械研究所）
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP751

【参考文献】

中国期刊全文数据库 前10条

1 张冬英;易维宁;王先华;乔延利;;空基遥感图像的MTF试验研究[J];安徽大学学报(自然科学版);2007年06期

2 罗志灶;周赢武;郑忠楷;;二值图像连通域标记优化算法[J];安庆师范学院学报(自然科学版);2010年04期

3 范鹏飞;范秀英;梁勇;杨必军;;基于成像过程的航空相机成像仿真系统[J];兵工自动化;2010年11期

4 魏新国,张广军,江洁;星敏感器中星图图像的星体细分定位方法研究[J];北京航空航天大学学报;2003年09期

5 林跃春;郭金海;雷阳;;浅析偏振光遥感的应用[J];测绘与空间地理信息;2011年02期

6 贾志科,崔慧娟,唐昆;彩色数字照相机的设计[J];电视技术;2000年11期

7 宋斌;;一种新的图像连通域快速标号算法[J];电子测量技术;2009年09期

8 蒋刚毅;黄大江;王旭;郁梅;;图像质量评价方法研究进展[J];电子与信息学报;2010年01期

9 蒋龙聪;刘江平;;模拟退火算法及其改进[J];工程地球物理学报;2007年02期

10 王钲旋,李志林,庞云阶;一个二值图像连通成份标记的快速算法[J];工程图学学报;1998年03期

中国博士学位论文全文数据库 前6条

1 蓝朝桢;空间目标天基光学观测系统建模与探测能力分析[D];解放军信息工程大学;2009年

2 肖汉;基于CPU+GPU的影像匹配高效能异构并行计算研究[D];武汉大学;2011年

3 杜西亮;基于振幅分割的光偏振测量技术的研究[D];哈尔滨工业大学;2007年

4 单列;视觉注意机制的若干关键技术及应用研究[D];中国科学技术大学;2008年

5 付强;基于成像链分析的光谱成像系统设计方法研究[D];中国科学院研究生院(西安光学精密机械研究所);2012年

6 黄涛;基于GPU的多点地质统计逐点模拟并行算法的研究[D];中国科学技术大学;2013年

本文编号：2019854