An Adaptive Step Size Point Cloud Alignment Method Applied to Automatic Alignment of Parts

Abstract

An adaptive step-size point cloud alignment method applied to the automatic alignment of parts. A mathematical model is established for the 3D point cloud of the part CAD, the actual 3D point cloud, the target pose and the actual pose, and the machine learning algorithm is used to make the part more accurate. Accurate part pose transformation; the learning rule δ is adopted, and the step size ΔW is adaptively adjusted to avoid the target output from falling into the local optimal solution, so that the result is stable and the global optimal solution is obtained. The invention has the advantages of modifying the weight coefficient along the negative gradient direction to reduce the error, controlling the modification amount of each step with the size of the current optimal distance value, achieving the purpose of adjusting the learning rate in real time and accelerating the convergence of the objective function solution. It solves the problems of slow speed and poor system robustness of traditional machine learning algorithms.

Description

technical field [0001] The invention belongs to the technical field of computers and machinery, and relates to a self-adaptive step point cloud alignment method applied to automatic alignment of parts. Background technique [0002] At present, in the field of machining, for parts with complex spatial structure and surface shape, such as aero-engine turbine blades, the traditional six-point matching method is mostly used for automatic alignment. This alignment method is inefficient and difficult to realize automated parts processing. Machine Learning Algorithms: Traditional machine learning algorithms have shortcomings such as slow learning speed, easy to fall into local optimal solutions, and weak robustness, so that they cannot meet the accuracy requirements for automatic alignment of parts. SUMMARY OF THE INVENTION [0003] The present invention solves the technical problems existing in the prior art, thereby providing an adaptive step point cloud alignment method applie...

AI Technical Summary

Problems solved by technology

[0002] At present, in the field of mechanical processing, for parts with complex spatial structure and surface shape, such as the machining of aeroengine turbine blades, the traditional six-point matching method is often used for automatic alignment. This alignment method is inefficient and difficult to realize automatic part processing

Machine Learning Algorithms: Traditional machine learning algorithms have shortcomings such as slow learning speed, easy to fall into local optimal solutions, and weak robustness, so that they cannot meet the accuracy requirements for automatic alignment of parts

Images