Application of Local-Feature-Based 3-D Point Cloud Stitching Method of Low-Overlap Point Cloud to Aero-Engine Blade Measurement

Blades are a core component of aero-engines. The accuracy of the blade profile of an aero-engine is crucial for its normal operation. Considering the limitations of low-overlap scanning point cloud data stitching in blade profile measurements, this paper proposes an improved feature fusion-trimmed iterative closest-point (TrICP) algorithm, realizing automatic stitching of three-dimensional (3D) point clouds scanned by laser measurements. In the stitching experiment of the Stanford 3D scan dataset Dragon-scan point cloud, the success rates of viewing angle differences of 24° and 48° were 100% and 66.7%, respectively, which were higher than those obtained using the TrICP algorithm, FPFH+SAC-IA, and ISS_BR+SHOT. The proposed algorithm exhibited high stitching success rates and efficiencies in the point cloud stitching experiment with large transformations. Moreover, the algorithm was employed as a prestitching tool. An automatic stitching method was further proposed by combining the point-to-plane iterative closest-point algorithm for performing precise stitching and the pose-map optimization algorithm for performing automatic stitching experiments on blade laser measurement data. The point cloud data measured using a coordinate-measuring machine further verified the stitching accuracy of our algorithm. The automatic stitching method exhibited good performance with regard to the scanning point cloud data of turbine rotor and guide blades (turbine rotor and guide blades have different shapes). The root-mean-square errors of the stitching experiments were 0.0354 and 0.0398 mm, meeting the error requirement of blade design and processing. Results show that the proposed algorithm is superior to traditional algorithms and shows promise for engineering applications.