浅析遥感图像的几何校正原理及方法.doc

《浅析遥感图像的几何校正原理及方法.doc》由会员分享，可在线阅读，更多相关《浅析遥感图像的几何校正原理及方法.doc（13页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、浅析遥感图像的几何校正原理及方法摘要：几何校正，就是清除遥感图像中的几何变形，是遥感影像应用的一项重要的前期处理工作。本文简单分析了几何校正的原理和基本方法，并以ERDAS软件为例，对青海海东地区遥感影像进行了几何校正，从而直观地表述了遥感图像几何校正的完整过程。结果表明，几何校正的精度受多方面因素影响，最主要的是控制点GCP的选取数量和选取位置。本次校正精度小于0.5个像元,符合要求。 关键词：遥感、ERDAS、几何校正、GCP引言：遥感20世纪60年代发展起来的对地观测综合性技术。狭义遥感指从远距离、高空，以至外层空间的平台上，利用可见光、红外、微波等遥感器, 通过摄影、扫描等各种方式，接

2、收来自地球表层各类地物的电磁波信息，并对这些信息进行加工处理，从而识别地面物质的性质和运动状态的综合技术。 遥感已然成为地理数据获取的重要工具。但是遥感技术的成图规律决定了遥感图像不能直接被应用，因为遥感图像在成像时 , 由于成像投影方式 、传感器外方位元素变化 、传感介质的不均匀、地球曲率 、地形起伏 、地球旋转等因素的影响 , 使得遥感图像存在一定的几何变形2 , 即图像上的像元在图像坐标系中的坐标与其在地图坐标系等参考坐标系统中的坐标之间存在差异 , 其主要表现为位移 、旋转 、缩放 、仿射 、弯曲和更高阶的歪曲3 。而且随着当今遥感技术的飞速发展，人们对遥感数据的需求也多源化,它们可以

3、是来自不同的波段 , 不同的传感器 , 不同的时间 。这些多源数据在使用时 , 必须具有较高的空间配准精度 。这就需要对原始影像进行高精度的几何校正 。因此 , 几何校正是遥感影像应用的一项重要的前期处理工作。ERDAS IMAGINE 是美国 ERDAS 公司开发的遥感图像处理系统，它以先进的图像处理技术友好灵活的用户界面和操作方式、面向广阔应用领域的产品模块、服务于不同层次用户的模型开发工具以及高度 RS/GIS 集成功能为遥感及相关应用领域的用户提供内容丰富且功能强大的图像处理工具，代表了遥感图像处理系统未来的发展趋势 5。基于此软件强大的功能性和灵活的操作性，本文采用erdas软件对海

4、东地区影像图进行几何纠正。2 研究区概况与研究方法 海东地区位于青海省东北部,海东以位于青海湖东而得名。地处祁连山支脉大板山南麓和昆仑山系余脉日月山东坡,属于黄土高原向青藏高原过渡镶嵌地带,海拔在16502835米之间。境内山峦起伏,沟整纵横,气候属于高原气候,高寒、干旱、日照时间长,太阳辐射强,昼夜温差大。年平均气温6.9,年均降水量为323.6 毫米,总蒸发量为1644毫米。本文采用校正过的2004年的海东地区参考影像对2009年对应影像进行校正。3 几何校正的原理与方法 遥感图像几何校正包括光学校正和数字纠正。本文主要介绍数字纠正。 数字纠正是通过计算机对图像每个像元逐个地解析纠正处理完

5、成的,其包括两方面,一是像元坐标变换，二是像元灰度值重新计算（重采样）。 (三) 数字图像灰度值的重采样 校正前后图像的分辨率变化、像元点位置相对变化引起输出图像阵列中的同名点灰度值变化,如图3所示重采样：P的灰度值取决于周围列阵点上像元的灰度值对其所作的贡献，这就是灰度值重采样. 1、 最近邻法 用距离投影点最近像元灰度值代替输出像元灰度值。 优点：保留大量原始灰度值，没有经过平滑处理，易于区分线性地物；简易省时：适用于专题文件。缺点：锯齿状，不平滑；对线性地物，可能出现不连续。2、 双线性内插法优点：平滑，没有锯齿状；与最近邻法相比，空间信息更准确；常用于改变向原大小，如数据融合。缺点：像

6、元值被平均化，可能导致某些边缘信息丢失。3、 双三次卷积法 获取与投影点邻近的16个像元灰度值计算输出像元灰度值。优点：与其他方法相比，均值和标准偏差与原始像元一致；可以锐化图像，平滑噪声。缺点：数据只可能改变；计算费时。数字图像几何校正方法有多项式纠正法和共线方程纠正法。现主要介绍多项式纠正法。 多项式纠正法的基本思想：回避成像的空间几何过程，而真接对图像变形的本身进行数学模拟。常用的二元齐次多项式纠正变换方程为式中，x,y为某像元的原始图像坐标；X,Y为纠正后同名点的地面（或地图）坐示；ai,bi为多项式系（i=0,1,2） 实际工作中,多项式系数求出后,根据上述公式可以求解原始图像任一像

7、元的坐标,并对图像灰度进行内插,获取某种投影的纠正图像。 一般选择最小控制点的数量为:(n+1)(n+2)/2, n为多项式次数。4 利用Erdas对海东地区遥感影像进行几何校正基本步骤第一步：显示图像文件首先在ERDAS图标面板中双击Viewer,打开两个视窗(Viewer1/Viewer2分别打开1320352009.img(待校正影像)和132035ring.img(参考影像)第二步：启动几何校正模块选择多项式几何校正模型：Polynomial,定义多项式次方(Polynomial Order)为2。第三步：启动控制点工具选择采点模式，启动控制点工具，进入控制点采点状态。第四步：采集地面

8、控制点GCP在图像几何校正过程中，采集控制点是一项非常重要和繁重的工作。由上可知：控制点的个数最小为(n+1)(n+2)/2, n为多项式次数。可以适当增加控制点的个数，但不能无限制的增加，根据操作表明，采用二次多项式模型进行校正时，控制点个数在14-18个时，误差可以控制在半个像元之内。第五步：采集地面检查点以上采集的 GCP的类型均为控制点，用于控制计算，建立转换模型及多项式方程，下面所要采集的GCP类型是检查点。第六步：计算转换模型在控制点采集过程中，一般是设置为自动转换计算模型。所以随着控制点采集过程的完成，转换模型就自动计算生成。第七步：图像重采样重采样过程就是依据未校正图像的像元值

9、，计算生成一幅校正图像的过程。原图像中所有删格数据层都要进行重采样。第八步：保存几何校正模式在Geo-Correction Tools对话框中点击Exit按钮，退出几何校正过程，按照系统提示，选择保存图像几何校正模式，并定义模式文件，以便下一次直接利用。第九步：检验校正结果基本方法：(1)同时在两个视窗中打开两幅图像，一幅是矫正以后的图像，一幅是当时的参考图像，通过视窗地理连接功能，及查询光标功能进行目视定性检验。(2)只建立一个视窗，在一个是窗中分别打开参考图像和校正以后的图像，通过“Utility”-“Swipe”,可以选择水平和垂直方向调节，来观察校正前后同一目标区域的位置形态变化。结论

10、：1、 为了便于识别地物，校正开始前应进行通道设置，使得遥感图像清晰可辨。 2、 地面控制点GCP应选在山脉或者道路等位置不易改变的地物交叉处，河流交叉处尽量不要选择，以保证校正精度。 3、 GCP的选取可以遵循“中点-正方形”选取法，即先选图像四边的中点为GCP,然后各中点组成正方形，选取正方形四边的中点为GCP，以此类推进行均匀选取。参考文献：1王建敏,黄旭东等. 遥感制图技术的现状与趋势探讨.矿山测量,2007(3).2徐仕琪,张晓帆,周可法,赵同阳.基于ERDASIMAGINE软件的CCD影像几何精校正方法初探_以哈密地区为例. 新疆地质,2007（6）. 3 张世利 , 余坤勇等.

11、基于ERDAS几何校正及在闽江流域影像处理中应用.福建林学院学报，2007（10）. 4 朱树龙,朱宝山,王红卫.遥感图像处理与应用 .科学出版社 ，2006. 5邢建军,王勇.浅谈基于ERDASIMAGINE软件的几何精纠正方法.测绘与空间地理信息，2007（4）.翻译原文Analysis of remote sensing images of the principles and methods of geometric correctionSummary：Geometric correction，Is the clear geometric distortion in remote se

12、nsing image，Is an important pre-processing of remote sensing image applications. This article briefly analyzes the principles and basic geometric correction method, and ERDAS software, geometric correction of remote sensing images of Haidong Prefecture in Qinghai intuitive representation of the comp

13、lete process of geometric correction of remote sensing image. The results show that the geometric correction accuracy is affected by various factors, the most important is to select the number and select the location of the control point GCP. The calibration accuracy of less than 0.5 pixel, to meet

14、the requirements.Keywords: remote sensing, ERDAS, geometric correction, GCPIntroduction: Remote Sensing (Remote Sensing) earth observation integrated technology developed in 1960s. The narrow sense remote sensing from long-range, high-altitude, and even outer space platform (plant form), visible lig

15、ht, infrared, microwave remote sensing (Remote Sensor) receives from all kinds of places of the earths surface through a variety of ways such as photography, scanning, the electromagnetic wave information of the objects, and for processing the information, thereby to identify the nature of the groun

16、d material and the motion state of integrated technology. Remote sensing has become an important tool for geographic data acquisition. But remote sensing technology mapping rule remote sensing images can not be directly applied because of the remote sensing image in the imaging, imaging projection m

17、ode, the sensor exterior orientation elements of change, the non-uniformity of the sensing medium, the curvature of the earth, undulating terrain, Earth rotation factors, making the remote sensing image there is a certain geometric distortion 2, that exists between the image coordinates of the pixel

18、 in the image coordinate system to its coordinates in the map coordinate system reference coordinate system differences, its main performance displacement, rotation, scaling, affine, bent and distorted 3. With the rapid development of todays remote sensing technology, the demand for remote sensing d

19、ata has multi-source, they can be from a different band, different sensors, different time. These multi-source data space, when used, must have a high registration accuracy. This high-precision geometric correction of the original image. Therefore, geometric correction is an important pre-processing

20、 of the application of remote sensing images.ERDAS IMAGINE ERDAS developed remote sensing image processing system, friendly state-of-the-art image processing technology and flexible user interface and operating methods for broad applications module to serve users at different levels of model develop

21、ment tools, as well as a high degree of RS/ GIS integration capabilities for users in the field of remote sensing and related applications to provide content-rich and powerful image processing tools, on behalf of the remote sensing image processing system future trends (sufer1) 5. Powerful functiona

22、lity and flexibility of this software-based interoperability, Haidong Region Image Geometric Correction erdas software.2 Study area and research methodsHaidong Prefecture in Qinghai Province, northeast of, Haidong in Qinghai Lake East, the name. Located in the Qilian, large plate Mountains, offshoot

23、 foothill, Kunlun Mountains Yumo Riyue Dongpo, belonging to the Qinghai-Tibet Plateau, the Loess Plateau transition inlaid zone, elevation of between 1650 to 2835 meters. Within the hills, ditch the whole aspect, the climate is highland climate, cold, drought, long hours of sunshine, solar radiation

24、, temperature difference between day and night. The annual average temperature of 6.9 C, average annual rainfall of 323.6 mm, the total evaporation of 1644 mm. The 2009 corresponding image correction using calibrated the 2004 Hai dong Prefecture reference image.3 geometric correction principles and

25、methods Remote sensing image geometric correction including optical correction and digital correction. This paper describes the digital corrective.Digital corrective corrected image each pixel basis to parse through computer processing is complete, which includes two aspects, one pixel coordinates t

26、ransform recalculate the pixel gray values (resampling).(3) digital image gray value resamplingCorrecting the image before and after changing resolution, like meta-point position relative gray value changes homonymous points caused by the change in the output image array, as shown in Figure 3Resampl

27、ing: P gray value depending on the surrounding array point its contribution made by the gradation value of the pixel, which is the gray scale value resampling.1 the nearest neighbor methodDistance projection point instead of pixel gray value output pixel gray value.Advantages: to retain a large numb

28、er of the original gray value, not been smooth, easy to distinguish between the linear feature; Occasional Paper: Suitable for easy and timesaving. Disadvantages: jagged, not smooth; linear feature may appear discontinuous.2 bilinear interpolationAdvantages: smooth, not jagged; compared with the nea

29、rest neighbor method, spatial information is more accurate; commonly used to change the original sizeSuch as data fusion. Disadvantages: pixel values are averaged, and the information loss may result in some of the edges.3 double cubic convolutionAcquires the projection point 16 adjacent output pixe

30、l gray value calculation pixel grayscale value.Advantages: Compared with other methods, the mean and standard deviation of the original pixel consistent; can sharpen the image, smoothing the noise.Disadvantages: The data can only be changed; calculate time-consuming.Digital image geometric correctio

31、n method polynomial corrective law and common the line equation corrective method. Now introduces polynomial corrective method. The basic idea of the correct method of polynomial: to avoid the spatial geometry of the imaging process, and really take a mathematical simulation of image deformation its

32、elf. The dual homogeneous polynomial correct transformation equationsWherein, x, y of an original image pixel coordinates; X, Y to correct the same locations in the ground (or map) sit shown; ai, bi of the polynomial coefficients (i = 0,1,2 .)The actual work, the polynomial coefficients determined a

33、ccording to the above equation can be solved any of the original image a pixel coordinates, and the image gradation interpolation, obtain certain projection corrective images.Usually select the smallest control point number is: (n +1) (n +2) / 2, n polynomial.Four Erdas Haidong area of remote sensin

34、g image geometric correction of the basic stepsThe first step: the display of image filesFirst in ERDAS icon panel, double-click the Viewer, open two windows (Viewer1/Viewer2 open 1320352009.img (to be corrected image) and 132035ring.img (reference image)Step 2: Start the geometric correction module

35、Polynomial geometric correction model: Polynomial, defined polynomial power (Polynomial Order) 2.The third step: start control point toolsSelect mining point mode, start the control point tools into the state of the control points collected points.Step 4: collecting ground control points GCPIn the i

36、mage geometry correction process, the acquisition and control point is a very important and arduous work. Seen from the above: the number of control points is a minimum of (n +1) (n +2) / 2, n polynomial. May be appropriate to increase the number of control points, but not unlimited increase, accord

37、ing to the operation showed that quadratic polynomial model correction, the number of control points at 14-18, the error can be controlled in half pixel within.Step 5: collecting ground checkpointTypes are control points above the GCP collected for control calculation, and create a transformation mo

38、del, and the polynomial equation to be collected below the GCP type checkpoint.Step 6: Calculate the conversion modelIn the control points in the acquisition process, is set to automatic conversion calculation model. With the completion of the acquisition process of the control points, the conversio

39、n model automatically calculates the generation.Step 7: image resamplingThe resampling process is calculated based on the non-corrected image pixel values, the process of generating a corrected image. Deleted grid data layer should be carried out in the original image resampling.Step 8: Save the geo

40、metric correction modeGeo-Correction Tools dialog box, click the Exit button to exit the geometric correction process, follow the prompts, choose to save the image geometric correction mode, and define the schema file, so that the next time the direct use.Step 9: inspection and calibration resultsTh

41、e basic methods: (1) simultaneously in the two windows open two images, the a correction subsequent image, a reference image, visual qualitative test through the window geographical connection function, and the query cursor function. (2) create a window is a window, respectively to open the referenc

42、e image and the image after the correction, Utility - Swipe, the horizontal and vertical adjustment can be selected to observe the same target area before and after correction The position of morphological changes.Conclusion:1, in order to facilitate the identification feature, before the correction

43、 began to set the channel, making the remote sensing image legible.2, Ground Control Point GCP should be selected in the mountains or road position is not easy to change the feature at the intersection of the river at the intersection of try not to choose, in order to ensure the accuracy of the corr

44、ection.GCP selection can follow the midpoint - square selection method, that is, before the election four sides of the mid-point of the image for GCP, then the midpoint composed of square, select the midpoint of the four edges of the square for GCP, and so uniformly select.References:1 Wang Jianmin,

45、 THE INNOVATION. Remote sensing mapping techniques Status and Trends of Mine Surveying, 2007 (3).2 Xu Shiqi, Zhang Xiaofan Zhou method, Zhao Tong Yang. Based the CCD image software ERDASIMAGINE geometric precision correction methods Preliminary _ Hami region as an example. Xinjiang Geology, 2007 (6)

46、.3 Zhang Shili, Evolvement Based on ERDAS geometric correction and in the Minjiang River Basin image processing applications. Fujian College of Forestry, 2007 (10).4 Zhu Shulong, Baoshan, Wang Hongwei sensing image processing and application of Science Press, 2006.5 Xing Jianjun, Wang Yong. Talking Based Geometric Correction ERDASIMAGINE software. Geomatics & Spatial Information Technology, 2007 (4).