Research On Global Stitching Method Based On High-Resolution Multi-View Point Cloud

With the expansion of the application scope of 3D imaging technology,the application of 3D reconstruction in various fields has become more widespread,and the requirements for 3D reconstruction technology have also increased.Due to the blind spot of the camera’s perspective,it is not possible to capture the complete surface of the object at once.Each shot can only generate a local point cloud of the object.In order to obtain a complete high-resolution 3D point cloud model of the object,it is necessary to use stitching technology to restore the complete 3D morphology of the object after multiple shots.Therefore,in order to complete the complete three-dimensional morphology reconstruction of objects,this article first takes two point clouds from different perspectives as the object and achieves the stitching of two point clouds.Furthermore,a global stitching method is studied for high-resolution point clouds from multiple perspectives.The main work of this article is as follows:For two point clouds with overlapping areas in a multi view point cloud,this paper introduces the complete stitching process of transforming a point cloud from one perspective to another perspective point cloud coordinate system.After completing the preprocessing of the point cloud,the two high-resolution point clouds are roughly and precisely spliced.Due to the fact that point clouds require a point cloud with overlapping regions as a medium when transforming coordinate systems,and there may not be a point cloud with overlapping regions with other point clouds in multiple point clouds from different perspectives,the stitching method between point clouds from two perspectives cannot directly transform point clouds from each perspective into the same coordinate system.Therefore,this paper studies the stitching algorithm for multi perspective point clouds.After the multi view point clouds are spliced according to the shooting sequence,the global optimization ICP algorithm is adopted to solve the problem of point cloud dislocation caused by error accumulation,and the accumulated error is reasonably distributed among the point clouds,so that the global error between the spliced point clouds is minimized.In the process of stitching multi view point clouds,this article optimized the corresponding point pairs selected based on the nearest neighbor algorithm,eliminated the point pairs with low reliability,and introduced a weight matrix to improve the algorithm,achieving accuracy improvement in global stitching of multi view point clouds.Due to the need for coarse registration of multiple viewpoint point clouds in the global ICP algorithm to obtain a good positional relationship in order to complete the global stitching of multi viewpoint point clouds,it is more suitable to use a fast global stitching method with less initial position restrictions on point clouds for high-resolution point clouds with obvious appearance features.The point cloud’s own characteristics are used to extract the corresponding point set,and the corresponding point set is optimized through a triplet optimization algorithm,Use the GM penalty function to suppress erroneous point pairs and achieve fast global stitching between multi view point clouds.Considering that the GM penalty function suppresses incorrect corresponding point pairs through distance,the suppression ability weakens when the positional relationship between multi view point clouds is extreme.Therefore,this paper proposes a penalty function based on normal vectors to enhance the robustness of the algorithm through the normal vector relationship between point cloud pairs,breaking the limitation of the initial positional relationship in multi view point cloud stitching,and completing the stitching of multi view point clouds.