【作者】 冯其强；

【导师】 李广云；

【作者基本信息】 解放军信息工程大学 ， 测绘科学与技术， 2010， 博士

【摘要】 论文根据我国制造业的发展特点和对高精度摄影测量技术的现实需求,针对数字工业摄影测量系统的软硬件组成、测量方式和精度要求,总结了国内外典型测量系统的特点,分析了数字工业摄影测量技术的未来发展趋势。论文的研究工作围绕高精度数字工业摄影测量系统涉及的相机、测量标志等硬件设备以及相机检校、标志点图像中心坐标提取、测量数据处理等关键技术展开,提出了一系列创新性技术和算法;并结合天线测量项目介绍了数字工业摄影测量技术在实际工程中的应用情况。论文主要工作总结如下:(1)分析了编码标志的设计原则,介绍了常用的同心圆环型编码标志和点分布型编码标志;设计制作了一种由5个模版点和3个编码点组成的点分布型编码标志,编码容量为496个;研究并实现了基于交比不变量的编码标志图像自动识别算法。(2)设计制作了由1个圆环标志和5个圆标志组成的定向靶;研究并实现了其图像自动识别算法。(3)介绍了CCD、CMOS及科学级CCD等影像传感器的工作原理及其成像性能差异;分析了数码相机中影响摄影测量精度的因素,包括图像分辨率、成像噪声、畸变差、结构稳定性及彩色成像机理等;提出了用于高精度数字工业摄影测量的数码相机选型原则。(4)介绍了3种常用的相机畸变模型:一般多项式模型、10参数模型和有限元模型,通过实验验证了不同模型的检校效果;提出了基于有限元模型和10参数模型的数码相机组合检校方法,有效提高了量测型相机的测量精度。(5)分析了圆形回光反射标志图像的灰度分布规律,介绍了常用的标志图像识别算法和中心坐标提取算法;提出了基于边界搜索的标志图像中心提取算法,像点坐标定位精度优于0.022像素。(6)利用均匀散焦退化模型生成了圆形回光反射标志的仿真图像;利用仿真图像验证了基于Canny算子的椭圆拟合算法及基于边界搜索的像点坐标定位算法的精度,分析了标志中心位置、噪声、标志图像半径等因素对定位精度的影响。(7)分析了圆形测量标志成像为椭圆时的椭圆偏心差分布规律;利用仿真图像定量分析了平面、球内表面、球外表面及圆柱面测量目标上的标志图像椭圆偏心差分布规律,给出了不同测量目标上的标志半径选择和摄站布设原则。(8)提出了基于4个非共线控制点的单张像片空间后方交会算法,以及基于定向靶和编码标志的多张像片自动概略定向算法;通过实验验证了像片定向算法的有效性。(9)介绍了基于核线约束的像点自动匹配算法,分析了核线约束的不足;提出了基于已知点和核面约束的像点匹配算法,并通过实验验证了算法的有效性。(10)推导了基于逐点法化消元法的自检校光束法平差快速计算方法;研究了利用多根基准尺确定测量坐标系长度基准的方法;提出了利用像点坐标残差和多余观测数检测粗差的方法。(11)介绍了在论文研究内容基础上集成开发的MetroIn_DPM数字工业摄影测量系统组成及软件主要功能模块;实验验证了系统测量精度;通过工程实例介绍了数字工业摄影测量技术在指导天线安装调整和天线工作状态下的变形测量方面的应用。更多还原

【Abstract】 Domestic manufacturing industry has been increasingly depending on the precise photogrammetric measurement during its development. The dissertation summarized characters and development trends of typical photogrammetry systems at home and abroad. The research included both hardware and algorithms. A series of innovative technologies and their applications in engineering projects were presented. Primary coverage of the dissertation was as follows:(1) Two kinds of typical coded targets and their design principles were introduced. A new set of dot-distributed coded target with code capacity of 496 was designed, which can be decoded automatically with the invariant of cross ratio.(2) A new device for orientation composed of a ring target and five circle ones were designed. Its decoded algorithms were also researched.(3) Both principles and performances of image sensors CCD, CMOS and scientific CCD were introduced. Influence factors of measurement accuracy in digital cameras such as resolution, imaging noises and distortions were analyzed. The principles for choosing digital cameras were presented.(4) Camera calibration experiments were carried out with three distortion models commonly used, which are polynomial model, ten-parameter model and finite element model. A new combined calibration method based on ten-parameter model and finite element model was presented. Calibration experiments showed that it can improve the precision of metric cameras dramatically.(5) Distribution regularities of gray scales in images of circle retro-reflective targets were analyzed. Several recognition and centring algorithms for target images were introduced. A new centring algorithm based on the edge searching and the intensity-weighted centroiding was presented, whose positioning error proved to be less than 0.02 pixels.(6) Images of circle retro-reflective targets were simulated with the image defocusing model. Experiments were carried out on the accuracies of different methods with simulation images, which showed that the intensity-weighted centroiding algorithm is much more precise than the ellipse fitting algorithm.(7) Ellipse image eccentric errors of targets on different objects were researched quantitatively with simulation images. Target radiuses and station parameters were calculated for the plane, spherical and cylindrical objects.(8) Elements of exterior orientation for one single image were calculated with space resection based on four non-collinear control points, while those for multiple images were calculated with the device for orientation and coded targets. Experiments proved the orientation method is effective.(9) The epipolar line matching method of corresponding image points was introduced. A new matching method based on known points and the epipolar line constraint was presented to improve the errors of epipolar line constraint, which proved effective.(10) Simplified solution of normal equation was deduced for the self-calibrated bundle adjustment. The length standard of the measurement coordinates was defined with several reference rulers. Blunders were detected with the residual errors of image point coordinates and the numbers of redundant observations.(11) The digital industrial photogrammetry system named MetroIn_DPM developed base on the research of the dissertation was introduced, whose measurement accuracy proved as high as V-STARS. The MetroIn_DPM system was applied in the measurements of one antenna during its assembly process and the other in several working positions.更多还原