A method and system for point cloud splicing of complex components based on feature fusion

Abstract

The invention discloses a complex component point cloud splicing method and system based on feature fusion, belonging to the field of computer vision, including: collecting and positioning the three-dimensional point clouds of each part of the component; partial overlap between adjacent local three-dimensional point clouds ;According to the positioning results, convert each local 3D point cloud to the world coordinate system and extract key points after uniform downsampling to obtain the corresponding key point cloud; input the key point cloud to the feature fusion matching network to obtain any two adjacent key points The matching correspondence of point clouds; the feature fusion matching network takes two adjacent point clouds as input, and is used to extract multi-scale features or multiple features of points in each point cloud and fuse them into feature descriptors of corresponding points to calculate Matching correspondence of adjacent point clouds; use the matching correspondence of adjacent key point clouds for point cloud registration, and stitch each key point cloud into an overall 3D point cloud model based on the point cloud registration results. The invention can improve the splicing accuracy of point clouds of complex components.

Description

technical field [0001] The invention belongs to the field of computer vision, and more specifically relates to a method and system for splicing complex constructed point clouds based on feature fusion. Background technique [0002] Large and complex components, including aircraft wings, ship blades, wind blades, high-speed rail bodies, etc., are widely used in many major technical fields such as aerospace, energy and transportation. In order to achieve high-precision machining of the surface of large and complex components, it is necessary to develop an accurate and reliable three-dimensional measurement system to evaluate the surface accuracy of its various parts. Just like using a ruler to measure the scale of an object, if you want to get a sufficiently reliable measurement result, the accuracy of the ruler needs to be high enough. The same is true for the measurement of the surface accuracy of complex components. In order to obtain reliable measurement results, three-dim...

AI Technical Summary

Problems solved by technology

The six-degree-of-freedom transformation model estimated by traditional rigid body registration algorithms such as ICP and RANSAC cannot fit more complex non-rigid body transformations, so the relative pose transformation between local point clouds cannot be accurately estimated, resulting in the accuracy of point cloud stitching Decline

[0008] In summary, for the existing point cloud stitching method, it is not reliable in the context of the actual application of 3D measurement of large and complex components, and cannot meet the specific application requirements of high-precision 3D measurement

Images