The Study Of Three Dimension Shape Reconstruction Based On Multi-View Point Cloud Registration Technique

Three-dimensional shape measurement of objects in the automotive,shipbuilding,civil engineering,metallurgy,aerospace,reverse engineering and other industries have been widely used.With the progress of science and technology,the measured object is also moving in the direction of large size and refinement,which brings about the contradiction between the measurement area and the measurement accuracy.How to reconstruct large-area,high-precision and rapid measurement of the three-dimensional shape of objects is the urgent problem to be solved at present.To achieve high-precision measurement,we first need to obtain reliable point cloud data.At present,the three-dimensional measurement technology has been developed very maturely,and various measurement methods emerge one after another.Due to its high precision,non-contact,real-time measurement and other advantages,optical three-dimensional measurement technology is widely used in 3-D shape reconstruction.In the real world measurement,it is difficult to obtain a full-field 3-D shape in a single measurement due to limitations of the shooting angle of the camera and the projection angle of the projector.One way to overcome this limitation is to obtain a multi-view measurement;however,the point clouds obtained in each viewpoint are in different local coordinate systems.In order to obtain a full-field shape,the point clouds in the different local coordinate systems must be registered into a single unified coordinate system for purposes of matching.Therefore,in order to realize large-area high-precision three-dimensional shape measurement,it is essential to have a high-precision measurement method to accurately obtain the three-dimensional point cloud in different viewpoints,and then,the accuracy of the matching algorithm needs to be as high as possible.Only meet these two requirements at the same time can we complete the request.In this paper,optical three-dimensional measurement technology and multi-view point cloud registration technology were used to study the large area,high-precision shape measurement.The main work includes the following aspects:First,an arbitrarily arranged fringe projection profilometry system is used to obtain point cloud of the object in single viewpoint.The method has the advantages of large single measurement area and high measurement accuracy,and the reliable multi-view point cloud can be obtained by using the method.Second,a new point cloud registration method is proposed to realize the full-field shape measurement of objects.We have designed an objective function that contains rotational and translational parameters that not only takes into account the registration between feature points but also accounts for the registration between two point clouds.In this paper,the initial value of the objective function is obtained by using the feature points extracted from the surface of the object.By optimizing the objective function,more accurate point clouds registration results can be obtained.Thirdly,a new criterion is designed to deal with the point cloud in the overlapping area.By resampling the point clouds of the overlapping regions we solved the problem of overlapping data redundancy and achieved good visualization results.In addition,in order to verify the validity and practicability of the proposed approach,the article measured a flat plate that with artificial markers and a cartoon toy that had complex shapes.The experimental results show that the measurement method and the matching algorithm can effectively improve the measurement precision and achieve full-field shape reconstruction of the object.