Research On Point Cloud Stitching Methods Of Robotic Mobile Measurement For Large Surfaces

Large surfaces such as aircraft skins,wind turbine blades,and high-speed train bodies are the core components of large aircraft,energy equipment,and high-speed trains,which are usually characterized by large dimensions and variable curvature.In order to meet the stringent aerodynamic performance,these surfaces have high requirements for profile accuracy,which requires 3D measurement technology to realize the perception and control of the manufacturing accuracy.The traditional Li DAR scanning and laser tracker measurement can hardly balance the harsh measurement requirements of high efficiency and high accuracy,and a measurement strategy via point cloud stitching that combines local multi-view scanning with global data fusion must be adopted.Mobile industrial robots have the advantages of a large working range and good kinematics dexterity and can realize the automatic and flexible measurement of large surfaces,which has gradually become a trend in large surface measurement.The mobile measurement via industrial robots can obtain the complete 3D data of the measured surface by multi-view scanning and point cloud stitching,but the stitching of multi-view point clouds faces some difficulties,such as the dynamic changes of point clouds’ reference,the lack of valid features for stitching,and the error accumulation due to successive stitching.Therefore,this dissertation researches from three aspects: the global measurement pose estimation of multi-station and multi-view point clouds,the high-accuracy registration of adjacent point clouds with weak features,the global consistency optimization of successively stitched point clouds,and proposes a robotic mobile measurement method based on multi-view scanning,local stitching and global optimization for large surfaces.The main research results include:For the problem of dynamic change of point clouds’ measurement reference at multiple stations and multiple views,a pose estimation model based on mobile stereo vision using fixed markers and a mobile binocular camera is established,and a pose estimation method based on iterative correspondence point registration is proposed,thus realizing the global measurement pose estimation with markers;A pose estimation strategy of single-station and multi-station hierarchical sensing positioning based on robot’s joint sensors and Li DAR sensors is established to achieve global measurement pose estimation without markers in complex environments.For the problem of the large surface point clouds lacking valid features and having low registration accuracy,the characteristic of multiple locally optimal solutions for weak feature surface point cloud registration is studied,and a two-layer optimization model for registration accuracy integrating global intelligent optimization and local iterative optimization is designed.A hybrid optimization registration algorithm based on the fruit fly optimization algorithm and improved iterative closest point algorithm is proposed,which solves the problem that the point cloud registration tends to fall into local optimal and achieves high-accuracy registration of local adjacent point clouds from surfaces with weak features.For the problem of error accumulation caused by multiple successive stitching for large range point clouds,an adjacency determination method of point clouds based on point cloud collision detection is proposed,and a loop pose constraint based on the point cloud adjacency matrix and a surface constraint based on the registration error information matrix are established.A global optimization method based on the improved pose graph that combines the pose consistency constraint and surface consistency constraint is proposed,which solves the accumulation error problem of successively stitched point clouds in a large range and improves the global stitching accuracy of point clouds from robotic mobile measurement.Based on the above key technologies,the robotic mobile measurement experiments are conducted,and the robotic mobile measurement system and software are developed.The3 D measurement and quality assessment are performed on a large wind turbine blade.The results prove that the proposed method can achieve high-accuracy automatic measurement of large surfaces,which affords a firm assurance for high-accuracy manufacturing of large surfaces.