| With the development of modern industrial parts to become more complex and largescale,higher requirements are put forward for their quality inspection.Surface structured light measurement technology is widely used in the measurement of three-dimensional shape of industrial parts due to its advantages of fast measurement speed,high accuracy,and noncontact.However,large-scale industrial parts are affected by the large overall size,and a complete three-dimensional shape cannot be obtained in a single measurement.To obtain complete three-dimensional shape data,it is usually necessary to complete the splicing of point cloud data under multiple viewing angles.This process often has a large loss of accuracy.To this end,this paper proposes a multi-view point cloud global splicing method based on photogrammetry.And conducts research on two aspects: defocus calibration of large field of view cameras and multi-view point cloud global stitching,and finally builds a threedimensional shape measurement system based on global point cloud splicing.The specific research content is as follows:Aiming at the difficulty and low efficiency of existing camera calibration methods in large-scale measurement scenarios,a rapid defocus calibration method based on phase shift coding circles is proposed.The design principle of the phase shift coding circle is studied,and a three-step phase shift coding circle pattern is generated as the defocus target.Acquire the phase shift coded circle image in the defocused state,perform phase principal value calculation and circle center extraction on the defocused image,and obtain the precise image coordinates of the feature points.According to the obtained two-dimensional image coordinates and the spatial position information of the target,construct the objective equation with the smallest back projection error,and optimize the solution of the camera parameters.This method adopts the idea of defocus calibration,does not require high-precision large targets and large-scale calibration scenes.And only needs to shoot three phase-shift coded circle images from a fixed angle,which can realize the rapid speed of large-scale vision system cameras calibration.Aiming at the problem of accumulated errors in the local splicing method of multi-view point clouds,a global point cloud splicing method based on photogrammetry is proposed.The principle of photogrammetry is analyzed,and the high-resolution camera is used to analyze the high-precision global three-dimensional coordinates of the landmark points in the measurement scene.Then according to the invariance of the spatial characteristics of the landmark points,the rough matching between the local landmark points and the global landmark points is completed,and the random sampling consensus(RANSAC)algorithm is used to eliminate mismatches.Finally,the coordinate conversion matrix is calculated according to the marked points of the matching,and the local point cloud data is converted to the global coordinate system to realize the global splicing of the point cloud data.The proposed method not only effectively reduces the overlapping area of splicing point clouds,but also solves the cumulative error problem caused by the local splicing method.And improves the overall measurement accuracy.Based on the above research,a visual three-dimensional profile measurement system was built to complete the three-dimensional profile measurement of large-size automobile sheet metal parts.Experiments have verified that for a standard ball with a center-to-center distance of 999.068 mm,the absolute measurement error based on the local splicing method is 0.206 mm,and the relative error is 0.021%.The measurement absolute error based on the global splicing method is 0.124 mm,and the relative error is 0.012%.The results show that the measurement accuracy of the global splicing method is significantly improved compared with the local splicing method,which can meet the high-precision measurement requirements of large-size workpieces. |
