非常规航摄影像定位方法及精度评定

发布时间：2018-06-08 22:10

本文选题：非常规航摄影像 + 影像仿真　；参考：《武汉大学》2013年博士论文

【摘要】：航空摄影测量是获取及更新地理信息的重要手段,是基本比例尺地形图的主要成图方式,已经成为我国基础测绘的通用作业模式。随着科技的发展,摄影测量有了长足的发展。轻小型飞行平台已经应用于航空摄影测量,使其作业形式更加灵活,但也带来了影像姿态不稳定的新问题。本文围绕非常规航摄影像的定向和定位这一主线,针对其排列不规则、姿态变化大引起的新问题,以空中三角测量为基础,深入研究了适用于非常规航摄影像的定向方法,并对其定位精度进行了深入探讨。本文的主要研究内容和创新性成果如下： (1)改进了非常规航摄影像的模拟方法。为提高处理效率,改进了影像模拟方法,提出以面元为处理单位的反解法对航摄影像进行模拟。首先选取待模拟区域对应的DEM和DOM数据,根据航摄条件设置影像的内外方位元素；其次根据区域的地形特点将影像划分为面元；然后对面元四角的像素进行高程迭代以获取其对应地面点的平面位置,并建立面元与地面格网的近似仿射变换关系；最后基于仿射变换计算面元上所有像素对应的地面平面位置,并在DOM上获取灰度值。对平地和丘陵地的多种摄影条件下的航摄影像模拟结果表明,本文提出的影像模拟方法能够模拟各种摄影条件下的航摄影像,相比逐像素的影像模拟方法,处理效率提高了10倍左右。 (2)分析了航摄影像的匹配精度。通过对模拟的航摄影像进行高精度影像匹配,评价影像匹配的精度。试验结果表明,对影像间几何变形小的影像,匹配精度优于1/6像素；对影像间几何变形较大的影像,匹配精度一般不超过0.4像素。这一统计结果可作为影像自动量测的精度估值以及检验指标。 (3)针对影像间存在大旋角的立体像对,提出了以奇异值分解约束本质矩阵获得未知数初值的严格相对定向方法。首先构建影像间的本质矩阵,采用传统的8点法线性求解本质矩阵元素；其次顾及本质矩阵的秩亏特性,采用奇异值分解方法对本质矩阵进行降秩约束；然后从经约束的本质矩阵元素导出相对定向元素的初始值；最后将初始值带入到基于上下视差的严格相对定向模型进行迭代求解。对模拟影像、常规航摄影像、非常规航摄影像的相对定向试验表明,该方法适用于存在大旋角立体影像对的相对定向,能够满足实际生产重构立体模型的精度需求。 (4)针对存在大旋角、大倾角的影像模型,提出了基于平面旋转的绝对定向方法。方法的核心是根据拟合平面的法向量的旋转关系直接求解三维旋转参数。首先对模型点在像方坐标系和物方坐标系下各自拟合出最佳平面,并计算平面的法向量；然后通过平面法向量的空间旋转关系直接求解绝对定向角元素的初值；最后,将角元素初值连同由经验公式估计的比例系数初值一并代入绝对定向的严格模型进行绝对定向元素的精确求解。对模拟影像、航摄影像、近景影像的试验结果表明,本文方法适用于存在大旋转角的影像模型的绝对定向,定向精度满足实际生产需求。 (5)分析了解析空中三角测量各个环节的误差传播特性,统计了非常规影像定位定向的精度估值。通过对各种摄影条件下的模拟航摄影像区域网的平差处理,给出非常规航摄影像定位精度的估算公式。
[Abstract]:Aero photogrammetry is an important means of obtaining and updating geographic information. It is the main mapping method of basic scale topographic map. It has become the general operation mode of basic surveying and mapping in our country. With the development of science and technology, photogrammetry has made great progress. The light and small flying platform has been applied to aero photogrammetry to make its operation form more In this paper, based on the orientation and positioning of the unconventional aerial photography, this paper focuses on the orientation and positioning of the unconventional aerial photography, aiming at the new problems caused by the irregular arrangement and the large change of attitude. Based on the aerial triangulation, the orientation method suitable for the unconventional aerial photography is deeply studied and its positioning accuracy is advanced. The main research contents and innovative achievements of this paper are as follows:
(1) improved the simulation method of the unconventional aerial photography image. In order to improve the processing efficiency and improve the image simulation method, the inverse solution of the surface element as the processing unit is proposed to simulate the aerial photography. First, we select the DEM and DOM data corresponding to the simulated region, set the internal and external elements of the image according to the aerial conditions, and then according to the area of the area. The image is divided into a surface element, and then the pixel of the four corner of the opposite element is iterated to obtain the plane position of its corresponding ground point, and the approximate affine transformation relationship between the surface element and the ground grid is established. Finally, the ground plane position corresponding to all pixels on the surface element is calculated based on the affine transformation, and the gray value is obtained on the DOM. The simulation results of aerial photography under various photographing conditions in flat and hilly areas show that the proposed image simulation method can simulate the aerial photography under various photographing conditions. Compared with the pixel by pixel image simulation method, the processing efficiency is increased by about 10 times.
(2) the matching accuracy of aerial photography is analyzed. The accuracy of image matching is evaluated by matching the simulated aerial images with high precision images. The experimental results show that the matching precision is better than 1/6 pixels for small images with small geometric distortion, and the matching precision is generally not more than 0.4 pixels for the image with large geometric distortion between images. The results can be used as the accuracy estimation and inspection index for automatic image measurement.
(3) in view of the stereoscopic image pairs with large rotation angles between images, a strict relative orientation method is proposed to obtain the initial values of unknowns by the singular value decomposition constraint essential matrix. First, the essential matrix between the images is constructed, and the traditional 8 point method is used to solve the essential matrix elements linearly. Secondly, the rank deficiency characteristics of the essential matrix are taken into account, and the singular value decomposition square is adopted. The method has a reduced rank constraint on the essential matrix, and then derives the initial values of the relative orienting elements from the essential matrix elements that are constrained. Finally, the initial values are introduced into the strict relative orientation model based on the upper and lower parallax. The method is applicable to the relative orientation of stereo image pairs with large swirling angles, which can meet the accuracy requirement of reconstructing 3D models in actual production.
(4) an absolute orientation method based on plane rotation is proposed for the image model of large rotation angle and large dip angle. The core of the method is to solve the three dimensional rotation parameters directly according to the rotation relation of the normal vector of the fitting plane. First, the best plane of the model point in the image square coordinate system and the matter square coordinate system is fitted together, and the method of calculating the plane is calculated. The initial value of the absolute orientation angle element is solved directly by the space rotation relation of the plane normal vector. Finally, the exact solution of the absolute orientation element is solved with the initial values of the angular elements together with the initial values estimated by the empirical formula into the absolute orientation. The experimental results show that the method is applicable to the absolute orientation of the image model with large rotation angle, and the orientation accuracy meets the actual production demand.
(5) the error propagation characteristics of each link in the aerial triangulation are analyzed, and the accuracy estimation of the unconventional image location orientation is calculated. The estimation formula of the location precision of the unconventional aerial photography image is given through the adjustment of the simulated aerial photography area network under various photographing conditions.
【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：P231

【参考文献】