【作者】 苏国中；

【导师】 张剑清； 张祖勋；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2005， 博士

【摘要】 伽利略早在300年前就曾经说过:“有必要测量一切可测量的事物,而且要尽力去使那些还不可测量的事物可测(It is necessary to measure everything that can be measured and to try making messurable what isn’t as yet.)。”这一古老名言禅释了测量技术和测量手段的重要性。事实上,对客观事物测量方法的改进和测量精度的提高是时代科技进步的重要标志。从古到今,人们为测量方法的改进和测量精度的提高所做出的努力,从来就没有停止过。 本文着眼于飞机测试过程中飞机在空中飞行姿态测量方法研究。利用数字摄影测量的严密理论和计算机视觉的一些前沿成果,对利用光电经纬仪序列影像确定飞行三姿态这一国内外尚无先例的课题所涉及的关键问题进行深入研究,实现高精度、自动化的运动目标姿态测量。 对于飞机姿态测量这一课题,本文的研究思路是,利用近景摄影测量的方法,建立被测飞机的三维几何模型,再利用该模型数据和计算机图形学方法建立光电经纬仪成像模拟系统,通过模拟系统提供的不同影像模型与实际摄制的飞机序列影像进行快速匹配,达到利用单目影像序列直接确定飞机飞行姿态的目的。 本文的主要研究内容和成果包括: (1) 本文系统回顾了运动物体姿态测量的发展历史;总结了当前国内外的研究现状;指出了运动物体姿态测量中目前还有待于进一步解决的关键问题;展望了今后的发展趋势和应用前景。 (2) 详细介绍了目前采用的运动物体姿态测量技术和方法,阐述了它们的原理,全面系统地比较了各种方法的优缺点。大致可以将这些方法分为两大类:一种方法是在物体(如飞机)内部装载姿态GPS接收机和惯性测量器(IMU)对物体的位置和姿态进行测量,这种方法叫运动目标姿态内测。另一种方法是获取被测运动物体的序列影像然后依据影像资料和摄像机的姿态确定运动目标在每一时刻的姿态,这种方法叫运动目标姿态外测。本文对用外测方法确定中远距离飞机运动姿态这一国内外尚无成功解决方案的难题,进行了深入研究,提出了一套完整的解决方案,针对此方案开发了一套飞机姿态测量系统,并得到了实际应用的检验。 (3) 详细阐述了相机成像的透视几何原理,对透视投影成像过程和机理进行了深入剖析,分析了摄影测量的成像过程与OPENGL中成像过程的异同之处,推导了OpenGL成像机理与摄影测量方位元素之间的关系,并推导出两种成像过程在一定条件下的转换公式,阐述了两种成像过程的内在一致性,并进行了试验证明。本文的研究为OpenGL这一计算机图形学界的伟大成果在摄影测量界的应用铺平了道路。摄影测量与计算机视觉以及计算机图形学等学科的研究内容在某些方面有很大的相似,但由于发展背景不同,研究方法上存在很大差异,以致于很难交流和沟通,本文深入探索了摄影测量与计算机视觉以及计算机图形学之间的区别及其内在联系,对于学科之间的交流和沟通也将起到重要的作用。 (4) 将物理意义上的点扩展到数学意义上的点,建立了广义点摄影测量理论,本文对这一最更多还原

【Abstract】 "It is necessary to measure everything that can be measured and to try making messurable what isn’t as yet.", said by Galilei over 300 years ago. Such logion explains the importance of measurement techniques and methods. Actually, the progression of measurement method and improvement of measurement accuracy are also important signs of advancement of science and technology. This dissertation focuses on the study of aircraft pose measurement during the process of aircraft test. Photoelectric theodolite is the main equipment for aircraft pose measurement. The measurement process is as following: Firstly, 3D geometry model of such aircraft is obtained through close-range photogrammetry method; then photoelectric theodolite imaging simulation system has been built by use of such model and computer graphics method; through matching of images produced by such simulation system and real image taken by photoelectric theodolite, the pose of aircraft has been determined. In this process, only monocular image sequence has been used.To measure the pose of moving objects with digital image sequence is a foreland task in the fields of computer vision and digital photogrammetry. This dissertation tried to research such problem in-depth and high accurate and automatic pose measurement is the final goal.The main research content and achievement include:(1) Pose measurement development history is reviewed. The background and literatures of pose measurement is summarized. The trends and prospect applications of pose measurement in the future are expected. The unsolved crucial problems in pose measurement are pointed out.(2) The principle and system structure of pose measurement are introduced in detail. The advantages and disadvantages with respect to various aspects between existing pose measurement methods are discussed in detail. Generally, existing pose measurement methods falls in two main kinds: the first is to set GPS/IMU in the moving objects to be measured and GPS/IMU information is used to calculate the pose of objects, this kind of methods are called " Internal measurement of moving object"; the other is to acquire image sequences of moving objects to be measured with camera outside the object and the image sequences are used to identify the pose of objects, this kind of methods are named "External measurement of moving object".(3) Protective geometry of camera imaging process are studied in detail and imaging mechanism are exploded deeply, two imaging process descriptions respectively in photogrammetry and openGL are analyzed and transition equations between them are derived, their internal consistency are proved by real experiments. Such research impulse OpenGL, such big achievement in the field of computer graphics, tobe absorbed, assimilated and applied by the field of programmetry.(4) Generalized point photogrammetry theory has been established on the basis of the extension from physical point to mathmatics point. Traditional photogrammetry theory is based on collinear of image point, object point and center of camera. Such theory requires that the point must be able to be observed. Such requirement is unnecessary for generalized point photogrammetry theory. The point can be point in infinite far and strict correpondence between points is unnecessary and so its application is flexible and convenient. Because more points have been used, accuracy has been improved greatly. The proposal of such theory impulse the development of photogrammetry theory and can be used urban modeling and industrial photogrammetry and reconstruction of curve and so on.(5 ) Combining generalized point photogrammetry theory and profile matching, a novel algorithm, Generalized point Feedback Iteration, has been presented at first time and used in airplane pose measurement. Experiment results show its high accuracy and reliability. This method makes it possible to automatically extract pose generalized meters of moving objects from image sequences.(6) For the condition that images of aircraft in near distance are clear, feature point trace method has been used to obtain the feature points in each frame and their correspondence between different frames. Then the pyramid method and improved back resection have been used to calculate the pose of aircraft. For images of mid-long distance where the back resection is unstable, improved pyramid method and RANSAC methed has been used. For long distance image where the feature points of aircraft are not clear, GPFI method has been used to determine the profile of aircraft and determine its pose.(7) High accurate aircraft profile extraction algorithm based on line diffusion function has been proposed. In the meanwhile, aircraft model profile generation algorithm based on spatial index has been put forward and made good preparation for matching between image and model profile.(8) Based on Kalman Filter, prediction and smoothing of airplane pose has been researched. The used of KALMAN filter method has greatly improved the speed of pose calculation and works well in accuracy improvement and blunder elimination.更多还原