3-D Trajectory Planning Scheme for Shoe Upper Processing

Xiaojie Hou,Junfeng Li

DOI: https://doi.org/10.1109/tim.2024.3476594

IF: 5.6

2024-10-25

IEEE Transactions on Instrumentation and Measurement

Abstract:Obtaining the processing trajectory for shoe uppers in automated shoe production presents significant challenges. The labor-intensive and inefficient nature of manual teaching, compounded by the need for individualized instruction for each shoe style, severely undermines the efficiency of automated shoemaking processes. Consequently, the automation and intelligent planning of processing trajectories for different shoe types has emerged as a challenging and prominent research topic in the shoemaking industry; the article, building upon this, the introduces a method for planning shoe upper processing trajectories using an improved maximal cliques registration approach, comprised primarily of two stages: 3-D point cloud reconstruction and processing trajectory extraction. Initially, point clouds of the shoe uppers are captured from multiple angles via a vision platform and subsequently preprocessed to minimize noise interference during registration. Subsequently, features are extracted from the preprocessed point clouds, with FPFH descriptors employed to characterize the shoe uppers' local features, thereby establishing initial correspondences from various perspectives; ultimately, the initial correspondences are represented as a compatibility graph to identify consistent correspondences, estimate optimal registration parameters, and apply an enhanced iterative closest point method with direction vector constraints for precise registration between point clouds, culminating in the 3-D reconstruction of the shoe uppers. Furthermore, the article outlines a camera-robot calibration method using principal component analysis, details the extraction of the processing trajectory on the shoe upper model, and describes the generation of the processing path in robot space based on this trajectory. Extensive experimental results demonstrate that this method significantly outperforms manual teaching techniques in efficiently extracting shoe upper processing trajectories, achieving a final z-axis processing accuracy within 1 mm and ensuring both high efficiency and consistency in trajectory acquisition.

engineering, electrical & electronic,instruments & instrumentation