点云
弹道
计算机视觉
计算机科学
人工智能
过程(计算)
胶水
切片
运动规划
点(几何)
校准
特征(语言学)
旋转(数学)
姿势
GSM演进的增强数据速率
运动学
工程类
机器视觉
空白
路径(计算)
实体造型
迭代最近点
面子(社会学概念)
坐标系
有效载荷(计算)
特征提取
棱镜
三维重建
迭代重建
作者
Tianyu Wang,Wenjie Feng,Junfeng Li,Jinfeng Gao
标识
DOI:10.1109/tim.2025.3623792
摘要
Existing automated shoe glue spraying solutions face two major technical challenges: first, during the 3D reconstruction stage, traditional multi-view point cloud registration methods struggle to balance speed and accuracy when dealing with low-overlap point clouds; second, during the trajectory planning phase, strategies based on fixed path parameters are inadequate for adapting to the geometric feature variations of the complex shoe upper surface. To address these issues, this paper proposes an intelligent solution that combines coarse-to-fine registration with adaptive trajectory optimization. To address the 3D reconstruction challenge of the shoe upper, a multi-view 3D imaging system is constructed, where shoe upper point cloud data is acquired using structured light cameras and a rotating platform. A K4PCS coarse registration framework, assisted by a 3D calibration disk, is designed to quickly align multi-view point clouds. Furthermore, an improved TEAR algorithm is proposed, which strengthens the rotation weight optimization strategy to suppress symmetry mismatches and achieve fast and accurate alignment of low-overlap point clouds across views (root mean square error less than 0.5 mm, single-view registration time less than 1 s). For the trajectory planning problem, an improved slicing method based on local normal vectors is designed, and combined with the glue spraying process model, adaptive adjustments to the spraying posture along the edge trajectory are made to dynamically adapt to the curvature changes of the shoe upper, generating a high-precision glue spraying path. Experimental results show that compared to traditional methods, the proposed solution reduces the total time for the multi-view 3D imaging system of the shoe upper to less than 3 seconds, and the error in the height direction of the glue spraying trajectory is reduced to within 0.5 mm. This research provides an efficient and robust 3D vision technology solution for the intelligent upgrading of the footwear industry.
科研通智能强力驱动
Strongly Powered by AbleSci AI