Research On Key Technologies Of Optical Three-dimensional Shape Measurement Based On Close-range Industry Photogrammetry

In recent years,the demand of optical 3D shape measurement has been increasing in many fields,such as industrial product testing,reverse engineering,heritage preservation,biomedical modeling,virtual reality and so on.As we all know,the traditional contact 3D measurement method can not meet the needs of practice.The high-precision and high-efficiency acquisition of three dimensional point coordinates,geometric dimensions and shape of objects have already become the research hotspots in measurement technology.At present,there are still some shortcomings and problems in measuring accuracy and efficiency,measuring range and evaluation criteria in optical 3D shape measurement.In this paper,the key techniques of 3D shape measurement are systematically and deeply studied by combining theoretical analysis with experimental research,which are based on the mechanical engineering,optics,electronics,photogrammetry,computer vision,digital image processing and other disciplines.The main contents and achievements are listed as follows:1.The principle and characteristics of various optical 3D shape measurement techniques are expounded,and the research status and development trend of optical3D shape measurement based on close-range industrial photogrammetry are systematically analyzed.Combining close-range industrial photogrammetry and computer vision,coordinate system transformation,collinear equation,coplanar equation,epipolar geometry and fundamental matrix,3d reconstruction method are studied.Aiming at the time-consuming problem of traditional bundle adjustment method in close-range industrial photogrammetry,a fast bundle adjustment algorithm is proposed,in which the step vector will be partitioned into camera and sturcture parameters,and based on the sparse nature of Jacobian matrix,the adjustment efficiency is greatly improved.The above contents provide the theoretical basis for the following research work.2.In order to improve the camera calibration accuracy in close-range industrial photogrammetry,a large size chessboard with high precision is designed,and the calibration algorithm based on homography matrix and nonlinear optimization are introduced.With the known intrinsic parameters in the camera,an iterative undistortion method for pixel coordinates is presented.Considering the influence of lens auto-focus on the camera parameters,a layered camera calibration method is proposed,in which the camera will be calibrated several times at different shooting distances,and multiple sets of intrinsic parameters are obtained.The experiment based on grating ruler shows that the layered calibration method can significantly improve the accuracy of optical 3D measurement.3.In order to solve the problem of multi-baseline and large-dip angle between camera stations in optical 3D shape measurement of polyhedral,rotational objects and large-scale surface,two orientation algorithms are proposed and two orientation devices are designed.One is the absolute orientation algorithm based on the coplanar points,in which the pose of the camera stations are determined by the homography matrix between the word plane of coplanar points and their image plane.The simulation experiment shows the absolute orientation algorithm is correct,and also evaluates the orientation accuracy at different pixel noise level.The second is the hybrid relative orientation algorithm,in which the initial values of the two pose parameters are obtained by the essential matrix decomposition,and then are further optimized by nonlinear optimization algorithm.The experiment has shown that the algorithm can achieve correct relative orientation when two stations are in various complex pose relationships.At the same time,two orientation devices,3D orientation target and mobile planar orientation target,are designed.They are suitable for measuring objects with polyhedral,rotational and large-size-surface structures.In 3D shape measurement,these two orientation devices are used for camera stations orienting and local point clouds stitching.4.In view of the shortage of the method which extracts the 3D points by sticking artificial markers in traditional close-range photogrammetry,such as time-consuming,labor-consuming and sparse point cloud data,two kinds of optical extraction techniques for spatial 3d points,namely dot-matrix structured light and optical probe,are proposed.Fistly,the dot-matrix structured light acquires dense 3d point clouds on the surface of an object,and its projection pattern is designed as a densely arranged luminous-dot matrix.A gray-scale exponential weighting method is proposed to accurately extract the pixel coordinates of the dot-center in the structured light image.Aiming at the characteristics of structured light pattern,a beads-string matching algorithm is proposed,which combines the epipolar-line-constraint matching algorithm between three images to accurately realize the row-by-column sorting of all the dot-centers.Secondly,optical probe is designed to extract single three-dimensional point in the occluded area,deep hole and groove in the object,and the precise 3D coordinates of the probe-head in the probe coordinate system are calibrated.The experiment shows that the relative accuracy of distance measurement is about7?10^-5in the view field of1800mm?1200mm when the camera is fixed.5.Based on the above theory and technology,two 3D shape measurement systems based on 3D orientation target and mobile planar orientation target,and one probe-based multi-view field 3D shape measurement system based on 3D orientation target are developed independently.For objects such as large-diameter satellite antenna,A380 aircraft model,polyhedral solid model,the industrial products with large-size surface or occluded areas,and so on,3d shape measurement experiments are carried out and the data are analyzed,and good results in engineering applications are obtained.