| Complex structural parts in aerospace,automotive and other fields are stacked together on the production line,requiring manual sorting and assembly,which seriously reduces efficiency.Surface structured light 3D measurement technology has the advantages of high precision,dense data and strong robustness,and has become the preferred solution for assisting industrial robots to achieve rapid measurement,sorting or assembly.However,the parts stacked on each other have visual occlusion,which easily leads to sparse measurement data,which makes subsequent data identification very difficult.For this reason,it is proposed to study the 3D measurement technology of multi-eye surface structured light under the existing technical framework of binocular surface structured light,and reduce the blind area of vision caused by stacking parts through multi-eye vision,thereby improving the integrity of the measurement scene data.Provide complete and accurate 3D data for subsequent data identification and robot positioning.The specific research contents are as follows:(1)Calibration method of multi-faceted structured light 3D measurement system.Using the four-step phase shift and multi-frequency heterodyne phase calculation method,the corresponding relationship between the camera image and the projected image is constructed,and the digital optical projector is used as a reverse camera to calibrate,so as to form a multi-eye vision measurement system with the binocular camera,and the system is calibrated.The initial value of the parameters is established,and the error model is established based on the beam adjustment method,and the accurate solution of the system parameters is obtained by iterative optimization.According to the experimental test,the standard deviation of the back-projection error of multi-objective vision calibration is 0.029 pixel,which has good accuracy.(2)Image matching based on multi-view geometric tensors.Taking the trinocular vision system composed of a projector and a binocular camera as an example,the left and right camera images are firstly subjected to stereo correction and transformation,and a trifocal tensor is introduced to establish a point-line-point correspondence between the three views.Search range,improve search efficiency.The experimental results show that the matching efficiency of this method is about twice that of the existing binocular algorithm,and the number of reconstructed 3D points is increased by more than 25%.(3)Development of multi-faceted structured light 3D measurement system.In view of the application requirements of industrial robots for grasping stacked parts,two types of measurement prototypes,the 3D measurement of large-scale fixed bases and the 3D measurement of robot hand-eye system,are developed.C++ language is used in the integrated development environment of QT Creator,and special 3D measurement software is developed to realize It has functions such as camera preview,calibration,measurement,and communication,and can provide the image and point cloud data required by the host computer.In order to improve computing efficiency,the NVIDIA CUDA computing architecture is applied to accelerate the aforementioned algorithms.The experimental results show that the measurement efficiency is more than 9 times higher than that on the CPU platform,and for the four selected industrial parts,the recognition success rate is increased by 12.5～42.9%.At present,the two types of systems have begun to carry out application testing in industrial production lines. |
PDF Full Text Request