利用几何校正法进行不同影像间空间配准

在影像数据融合、动态变化监测等遥感影像集成分析和应用中，将来自不同传感器、不同时相获取的影像高精度快速配准是其中的关键技术之一。介绍利用几何校正法进行不同影像间空间配准的过程，在不同影像上选取同名点作为控制点，用最近邻、双线性内插或三次卷积内插运算法对分辨率较小的影像进行重采样，完成配准。并采用Landsat TM影像和SPOT pan影像进行实验，结果表明配准精度在1个像素之内，为进一步的遥感影像融合．分类作好数据准备。     

图像重采样对调制传递函数估算的影响

选取理想刃边图像和真实遥感图像作为原始图像,对其进行多种辐射和几何重采样操作,得到了一系列退化图像;用这些图像按照刃边法对成像系统的调制传递函数进行了估算。试验表明,采用刃边法估算成像系统的调制传递函数时,为了视觉效果,可以对图像进行适当的辐射重采样操作,基本不会影响成像系统的调制传递函数;而几何重采样则会严重影响成像系统的调制传递函数。     

基于FFT稀疏压缩感知域内遥感图像融合

压缩感知理论因其远低于乃奎斯特采样率的特性,减少了大量的采样数据。基于这一特性,提出一种在压缩感知域内进行遥感图像融合的方法。该方法首先对图像作快速傅里叶变换(FFT);然后进行测量采样获取压缩感知域数据;再采用权重法对数据进行融合;最后通过重构算法得到融合图像。通过实验得出压缩感知域内遥感图像融合具有数据量少,融合效果好等特点。     

海量遥感影像的存储与快速调度显示方法

随着计算机技术的发展,计算机的存储和处理的功能越来越强大,但是仍然无法满足日益增加的海量信息处理的需要。因此研究如何利用有限的计算机内存,处理无限大小的数据,成为研究的重点。本文根据海量遥感影像数据的特点以及对海量遥感图像进行处理的需要,研究了实现海量图像数据快速显示的方法以及其数据库存储结构。主要包括图像重采样、影像金字塔生成技术、图像快速显示技术等。     

遥感影像不同校正模型对反射率的影响分析

采用飞马D200四旋翼无人机携带多光谱相机,获取了可见光波段的数据,地面同步测量了地表反射率和大气参数.结合实验数据,采用基于POS数据的严格成像模型和多项式模型对图像进行了几何校正,对比了2种模型的校正精度和不同重采样方式产生的差异.在几何校正的基础上进行了大气校正,获得地表反射率,选取典型地物水体、植被、裸土的实测结果进行验证.结果表明低空无人机遥感可以得到精度较高的地表反射率结果,但不同的几何校正模型对地表反射率影响较小,可以忽略.通过开展该实验证明了利用无人机开展定量化遥感研究的可行性.     

一种遥感图像的坐标转换方法

在遥感应用中,要将遥感影像的空间信息转换到其他坐标系中,一般通过几何校正来实现。这种方法工作量大而且不可避免会引入新的误差。本文从坐标转换的角度出发,通过事先计算两个坐标系之间的转换参数,结合遥感影像中储存空间信息的方式,推导出遥感图像坐标转换的计算公式,并由此推导出根据坐标转换参数计算重采样像元坐标变换函数。根据本文提出方法,可以很方便地实现遥感图像的坐标转换,将遥感图像的空间信息转换到不同坐标系下。     

利用自动匹配与三角剖分进行遥感图像几何精校正

在研究传统几何精校正方法的基础上,提出了一种高精度的基于自动同名点匹配和三角剖分技术的几何精校正方法,该方法是通过基准底图对待校正图像进行几何精校正的。首先利用FAST (Features from Accelerated Segment Test)算子在基准底图上快速提取均匀分布的候选特征点,通过图像自身携带的地理定位信息确定初始同名点对;经平移误差消除、互相关双向匹配、RANSAC(Random Sample Consensus)粗差剔除、二元三点插值等步骤获取稳定可靠的亚像元级同名点对;最后根据亚像元级同名点对构建Delaunay三角网进行图像变换和重采样处理。以Landsat卫星ETM为基准底图对环境卫星CCD数据进行几何精校正试验,本算法几何精校正精度较传统的方法得到了很大提高。     

面向汛旱情监测的遥感影像GPU并行处理算法

针对面向汛旱情监测应用中遥感影像处理耗时过长的问题,包括辐射校正、几何纠正、遥感指数计算等过程,对其业务化工作流程进行了分解分析。结合统一计算架构(compute unified device architecture,CUDA)的存储结构和程序设计模型,将数据处理过程划分为数据读取、直方图统计、栅格分割、波段计算、重采样和数据输出等模块,对波段计算及重采样等模块设计了并行处理方案,并通过实验确定了栅格划分的最佳尺度,基于栅格数组图形处理器(graphics processing unit,GPU)映射方法加速了数据传输效率,最终提出了基于CUDA架构CPU-GPU协同的并行处理算法。实验结果表明,辐射校正及遥感指数计算的波段计算模块可节约58.9%的时间;几何纠正效果最为显著,最邻近像元重采样和双线性内插重采样模块的最终加速比分别能够达到9倍和7倍以上。     

面向震害信息提取的多源遥感图像自动配准

为了从多源遥感图像中提取震害信息,研究了一种图像自动配准的高精度算法.首先,利用Moravec算子提取图像的特征点,并采用随机抽样一致性(random sample consensus,RANSAC)算法剔除误匹配点;然后,根据特征点对图像进行对数极坐标变换,求得基准图像和待配准图像间的旋转和缩放因子;进行旋转角度和缩放系数补偿后,采用与相位相关的亚像元级算法计算图像间的偏移量;最后,利用插值算法对待配准图像进行重采样,实现待配准图像与基准图像的几何配准.将配准后的震前和震后图像作为分类目标,分别进行基于特征的分类和震害信息提取,并对分类结果进行变化检测.实验结果表明,该算法流程具有较高的鲁棒性与配准精度,适用于多源遥感图像配准,所提取的震害信息能较好地应用于震害评估.     

以“不变特征点集”为控制数据集的遥感图像自动化处理框架

全球变化研究需要将多种传感器光谱波长接近的图像波段一起使用,以满足遥感应用对时间分辨率和区域覆盖的要求,这给遥感图像处理提出了新要求,涉及多传感器多时相数据几何定位一致性问题、辐射归一化问题及地类属性标识一致性问题以及高度自动化处理的问题。针对上述几方面问题,提出了一个基于"不变特征点集"IFPs(Invariant Feature Points set)作为控制数据集的区域级遥感图像自动化处理框架,将图像的几何空间、辐射值空间和类别属性值空间的时空对齐问题纳入到统一框架,提供了一种间接快速处理的手段和理念,并对构建IFPs的关键技术进行了综述。     

火星快车HRSC线阵影像投影轨迹法近似核线重采样

火星快车高分辫率立体相机(high resolution stereo camera, HRSC)线阵推扫式影像具有分辫率高、全球覆盖性好等特点,对火星快车HRSL影像进行摄影测量处理可获取火星全球高分辫率地形数据,而影像匹配是自动获取地形数据的关键利用核线几何约束可以将影像匹配搜索范围由二维降至一维本文在构建火星快车HRSC线阵影像严格几何模型的基础上,提出了一种近似核线重采样方法首先,利用投影轨迹法分析 HRSC线阵影像核线的直线性与平行性;然后,将原始HRSC影像进行微分纠正生成水平影像,分析投影轨迹法得出的核线在水平影像上的几何关系并确定近似核线方向,再沿核线方向进行核线重排生成近似核线影像实验结果表明,本文算法生成的近似核线影像同名点上下视差优于1个像素,且在水平影像上进行近似核线重采样减少了由外方位元素变化引起的几何畸变,更有利于影像匹配。     

数字摄影中的核线影像算法研究

针对数字摄影立体像对自动生成核线影像的问题,介绍了基于数字影像几何纠正的核线解析方法及实现过程。内容包括基于数字影像的几何纠正、双线性插值法进行核线重采样、算法的VC++实现等问题的研究。最后对生成的核线影像的精度进行了分析,证明了用这一方法生成核线影像的可行性。     

Vector-based approach for combining ascending and descending persistent scatterers interferometric point measurements

The combination of ascending and descending persistent scatterers interferometric (PSI) data by means of resampling and/or spatial interpolation, separately for each synthetic aperture radar geometry, is a commonly followed procedure, limited though by the reduced spatial coverage and the introduced uncertainties from multiple rasterization steps. Herein, an alternative approach is proposed for combining different PSI line-of-sight observables in the vector domain, based on the geographic proximity of PS point targets. In the proposed nearest neighbour vector approach all necessary analysis steps are performed by means of attributes transfer and calculations between features geodatabases, prior to any rasterization. By increasing the number of input point vectors during subsequent interpolation, the overall error budget coming from spatial interpolation is being reduced. The advantages of the proposed vector-based approach compared to the commonly used grid-based procedure are being demonstrated using real data.     

Multiresolution and Multispectral Data Fusion Using Discrete Wavelet Transform with IRS Images: Cartosat-1, IRS LISS III and LISS IV

Image fusion techniques integrate complimentary information from multiple image sensor data such that the new images are more suitable for the purpose of human visual perception and computer based processing tasks for extraction of detail information. As an important part of image fusion algorithms, pixel-level image fusion can combine spectral information of coarse resolution imagery with finer spatial resolution imagery. Ideally, the method used to merge data sets with high-spatial and highspectral resolution should not distort the spectral characteristics of the high-spectral resolution data. This paper describes the Discrete Wavelet Transform (DWT) algorithm for the fusion of two images using different spectral transform methods and nearest neighbor resampling techniques. This research paper investigates the performance of fused image with high spatial resolution Cartosat-1(PAN) with LISS IV and Cartosat-1(PAN) sensor images with the LISS III sensor image of Indian Remote Sensing satellites. The visual and statistical analysis of fused images has shown that the DWT method outperforms in terms of Geometric, Radiometric, and Spectral fidelity.     

一种基于 Contourlet 变换的遥感影像超分辨率重建改进方法

针对原有方法轮廓波(Contourlet)分解高频信息处理的不足,提出用零填充重采样代替双三次插值用于高频信息处理的新方法,该方法可使Contourlet分解后的高频信息重采样过程不引入噪声能量的同时获得理想的插值结果。改进方法利用Contourlet多分辨率分析的特征,以及零填充重采样方法较传统插值方法优越的性能,此两者相结合可改进原有方法的不足,提高单帧影像的超分辨率重建效果。经理论分析与试验验证,改进后方法的超分辨率重建结果无论在定性分析与定量对比上都要优于传统方法。     

基于SPOT5遥感影像的湿地资源调查方法研究

通过利用SPOT5遥感影像对湿地资源调查方法的实验研究,探讨了遥感影像的融合方法,选取适合湿地资源调查的数据融合和重采样方法,并解译融合后的遥感数据。从而获取湿地研究区的分类数据。     

Suitability of the DCS460c colour digital camera for quantitative remote sensing analysis of vegetation

The Kodak DCS460c is a high resolution, colour digital camera with a specific CCD array mosaic structure that requires a dedicated processing algorithm for production of a three-band image. Kodak's Active Interpolation (KAI) algorithm is compared with a simpler nearest neighbour interpolation method (NNI) with regard to applicability in scientific investigations of target features in imagery. Detailed quantitative analyses of flat-field imagery were undertaken in order to provide insight into artefacts and to correct for the off nadir reduction in brightness. Characteristics of the aerial photography were also studied and aspects of processing and analysis of the imagery for scientific investigations are discussed. We present a novel method for correction of brightness fall-off (due to the vignette effects and increasing view angle) which is generally applicable to remote sensing of rural scenes. Quantitative spectral comparisons between related imagery from the DCS460c (e.g. for interpretation of relative vegetation cover) are only scientifically reliable when the raw (i.e. non-interpolated) DN values for green and red are below about 67% of their maximum range. This is due to preferential vertical leakage or charge diffusion in the CCD. A combination of CCD characteristics leads to possibly poor spatial resolution for the infrared band. The imagery after KAI shows good spatial resolution and has a natural looking colour reproduction useful for discerning features by eye. However, when using KAI, reliable spectral measurements can probably only be acquired from portions of objects that are at least four pixels from the object's boundary. In general, quantitative spatial and spectral analyses can be carried out with the DCS460c but the KAI, the broad band response and the vertical leakage require specific solutions to produce quantitative results.     

自组织网络在遥感土地覆盖分类中应用研究

设计完成和比较了自组织网络的几种算法在遥感土地覆盖分类中的应用,结果表明非监督和监督学习结合方法进行遥感土地覆盖分类,各算法在分类性能上无显著差异,因此可采用算法和较简单的单竞争学习网络,根据最邻近原则进行非参数分类。     

Misleading information from direct interpretation of Geometrically Incorrect aerial photographs

Remote sensing represents a powerful method to study the change of several phenomena over time. However, useful input data (for example, aerial photos) can produce misleading information if an inadequate geometric correction is applied. Among mathematical models used for this type of correction, orthorectification (differential correction) seems to be the only one that guarantees accurate results. However, many authors are still basing their results on incorrectly transformed images. This phenomenon is especially due to the incidence of several user-friendly tools and interfaces for the rectification of images. This paper tests both polynomial rectification and orthorectification. Aerial photography of 1954 of the same area was acquired and polynomial rectification and orthorectification were applied, using a 1998 orthophoto as a base. An unsupervised classification was performed by utilising the nearest neighbour method for the resampling of images. The area occupied by each class was calculated. A multitemporal analysis was then performed, considering the overlap between the 1954 orthophoto and the 1998 orthophoto as the base of reference. An overestimate up to 21% of class area was found. When multitemporal analysis was carried out, an overestimate up to 100% of class area change was found.     

Epipolar resampling of linear pushbroom satellite imagery by a new epipolarity model

This paper presents a practical epipolarity model for high-resolution linear pushbroom satellite images acquired in either along-track or cross-track mode, based on the projection reference plane in object space. A new method for epipolar resampling of satellite stereo imagery based on this model is then developed. In this method, the pixel-to-pixel relationship between the original image and the generated epipolar image is established directly by the geometric sensor model. The approximate epipolar images are generated in a manner similar to digital image rectification. In addition, by arranging the approximate epipolar lines on the defined projection reference plane, a stereoscopic model with consistent ground sampling distance and parallel to the object space is thus available, which is more convenient for three-dimensional measurement and interpretation. The results obtained from SPOT5, IKONOS, IRS-P5, and QuickBird stereo images indicate that the generated epipolar images all achieve high accuracy. Moreover, the vertical parallaxes at check points are at sub-pixel level, thus proving the feasibility, correctness, and applicability of the method.