【作者】 王琦；

【导师】 赵章焰；

【作者基本信息】 武汉理工大学 ， 机械工程， 2019， 博士

【摘要】 目前大尺寸装备在监测与结构优化方面迫切需要高效且快捷的测量技术。研究工作表明摄影测量方法具有非接触、高效率、低成本和高测量精度的特点。但面对尺寸大、表面纹理匮乏且无法贴标志点、摄站受限及现场测量时间少的大尺寸装备时,存在局限与不适用,难以广泛应用。本文总结大尺寸装备的特点与限制,研究了摄影测量方法应用于大尺寸装备测量过程中的关键技术,提出了适合的大尺寸装备的相机标定方法、位姿估计方法、几何尺寸测量方法以及特殊圆柱形部件中心计算方法,并将其应用于港口机械测量的工程实践。实验结果显示本文所提方法均能够满足工程需要。主要工作包括:(1)针对大尺寸装备野外作业难以建立标定场的问题,提出了两步标定法。首先通过棋盘法标定出相机径向畸变系数。随后针对大尺寸装备控制点像点定位误差加大问题,提出了基于光学畸变与误差概率的亚像素级的像点定位法,在测量现场对每张照片的有效焦距进行高精度标定计算。6个场景实验结果显示,本文所提有效焦距标定改进效果显著,距离门座式起重机63.21m时,该算法的测量精度相对于传统方法提升98.84%。(2)针对大尺寸装备常面临拍摄角度倾斜大、控制点少及相片深度变化大的问题,提出了基于深度信息加权与投影算子建立目标函数求解最优相机位姿的方法,同时通过几何关系解决了位姿估计中的多解问题。仿真计算结果显示该方法优于其他目前常用方法。以龙门式起重机为测量对象的实验结果显示,控制点共面时的相机位置与姿态角相对误差分别为6.77e-6、0.455%,非共面时相对误差分别为4.67e-6、0.373%,表明该方法在受限条件下能准确、稳定地完成相机位姿估计。(3)针对缺乏纹理的大尺寸装备难以匹配同名点的问题,提出了轮廓约束与图像边缘提取的几何尺寸测量方法。该方法有效利用大尺寸装备自身几何特征已知的条件,结合测量对象几何轮廓约束与图像边缘建立误差,通过迭代求解测量对象几何参数精确值。在此过程中根据机械装备轮廓特点改进LSD直线提取算法,解决了原LSD算法无法提取存在钝角的直线以及曲线的问题。以港口常用的抓斗式卸船机为测量对象的实验结果表明在112.49m的测量距离条件下,相对误差优于1/5000,并且可以获取测量对象连续几何尺寸。(4)针对大尺寸装备部分圆柱体部件无法直接测量中心,只能间接测量筋板的问题,提出了基于光学畸变与选点方式的复合加权的中心精确定位算法。该方法对筋板测量数据获取过程中的光学畸变进行加权拟合空间圆,然后对选点组合过程的不同数据拟合圆心存在的偏差进行加权,准确拟合圆柱的圆心与半径。与可靠测量值比较,该方法拟合半径结果精度误差为0.15%。以门座式起重机为测量对象的实验结果表明,与人工贴标志点得出的圆心与半径相比,该算法的相对误差在2.5‰以内。更多还原

【Abstract】 At present,there is an urgent need for efficient and fast measurement technology in monitoring and structural optimization of large-scale equipment.The research shows that photogrammetry has non-contact,high efficiency,low cost and high measurement accuracy,but it is difficult to be widely used in large-scale equipment.Because largescale equipment are with large size,lack of surface texture,unable to stick markers and without enough measurement time.This paper summarizes the characteristics and limitations of large-scale equipment,studies the key technologies of photogrammetric method applied in the measurement process of large-scale equipment,and puts forward the camera calibration method,pose estimation method,geometric dimension measurement method and the calculation method of the center of special cylindrical components for large-scale equipment.The experimental results show that it can meet the engineering needs.The main work includes:(1)Aiming at the difficulty of establishing calibration field for large-scale equipment in field operation,a two-step calibration method is proposed.Firstly,the radial distortion coefficient of the camera is calibrated by checkerboard method.Sub pixel image location method based on optical distortion and error probability is proposed to solve the problem of increasing image location error of large-scale equipment control points.The effective focal length of each photograph is calibrated and calculated accurately in the measurement field.The experimental results of six scenarios show that the improved effective focal length calibration method proposed in this paper has a significant effect.When the distance is 63.21 m from the portal crane,the measurement accuracy of the algorithm is improved by 98.84% compared with the traditional method.(2)Aiming at the problems of large-scale equipment,such as large tilt of shooting angle,few control points and large change of photo depth,a method of establishing objective function based on depth information weighting and projection operator to solve the optimal camera pose is proposed.At the same time,the multi-solution problem in pose estimation is solved through geometric relationship.The simulation results show that the method is superior to other commonly used methods.The experimental results of gantry crane show that the relative errors of camera position and attitude angle are 6.77e-6 and 0.455% respectively when the control points are coplanar,and 4.67e-6 and 0.373% when the control points are not coplanar,respectively.It shows that the method can work well under Limited conditions.Accurate and stable camera pose estimation.(3)Aiming at the problem that large-scale equipment lacking texture is difficult to match homonymous points,a measurement method of contour constraint and image edge extraction is proposed.This method effectively utilizes the known conditions of the geometric characteristics of large-scale equipment,establishes errors by combining the geometric contour constraints of the measured object with the edge of the image,and solves the precise values of the geometric parameters of the measured object by iteration.In this process,the LSD straight line extraction algorithm is improved according to the characteristics of mechanical equipment contour,which solves the problem that the original LSD algorithm can not extract straight lines and curves with obtuse angles.The experimental results of grab ship unloader used in port show that the relative error is better than 1/5000 under the measurement distance of 112.49 m,and the continuous geometric dimensions of the measured object can be obtained.(4)Aiming at the problem that some cylindrical parts of large-scale equipment can not directly measure the center,but can only indirectly measure the stiffeners,a composite weighted central precise location algorithm based on optical distortion and point selection is proposed.This method fits the spatial circle by weighting the optical distortion in the process of obtaining the measured data of stiffeners,and then fits the center and radius of the cylinder accurately by weighting the deviations of the fitting centers of different data in the process of selecting points and combining.Compared with reliable measurements,the accuracy error of the fitting radius is 0.15%.The experimental results of portal crane show that the relative error of the algorithm is less than 2.5,and the radius deviation is less than 5 mm,compared with the calculation of circle center and radius by manual labeling.更多还原