| The splicing of point cloud is a very important research direction in 3D reconstruction.After decades of research,it has achieved a lot of research results,but there are still problems in algorithm complexity,calculation time and precision.Based on this,a free-view 3D point cloud splicing algorithm with simple algorithm,strong applicability and good real-time performance is proposed.The main work and innovations of this paper are as follows:(1)A structured light hardware system was built and the high-precision calibration of the system was completed.A high-precision single-view dense point cloud is obtained by using a Gray code combined with line-shift.(2)Research and implement a multi-view splicing scheme that requires additional calibration objects such as checkerboard method and hinge method.Multi-view splicing experiments are carried out on the surface of different texture characteristics,and the results are compared and analyzed.The algorithm based on additional calibration objects has low complexity,but the stitching accuracy is greatly affected by the calibration objects and this method limits the measurement range of the object to be measured and has poor applicability.(3)Research and implement a multi-view splicing method based on classical ICP(Iterative Closet Point)and binocular stereo vision restoration space 3D matching point pairs without additional calibration.The ICP algorithm requires high initial position,takes a long time to iterate,and the method based on binocular stereo vision with low accuracy.By analyzing its advantages and disadvantages,a non-iterative fast stitching algorithm based on motion screw is proposed,and multi-view fast stitching is realized for different texture surfaces.(4)In order to solve the problem of point cloud splicing of regular geometric objects such as cylinders,a high-precision splicing method combining 2D robust detection features and structured light 3D scan data is proposed.Firstly,2D image features are used to assist in locating structured light 3D scanning.The re-projection feature corresponding to the point cloud is used to find a reliable corresponding point cloud in different point cloud data sets,and then the accurate estimation of the stitching matrix is realized based on the searched matching point cloud,and a reliable and robust multi-view point cloud stitching is realized.The experimental results show that the non-iterative splicing algorithm proposed in this paper has the characteristics of high efficiency,reliability and robustness,and the root mean square error of the stitching is below 0.4mm. |
PDF Full Text Request