摘要
3D物体识别与位姿估计是AR辅助维修信息增强的基础,目前的3D物体识别技术在识别准确率、运行实时性和位姿估计准确度上仍然存在不足。以物体识别的实时性和位姿估计的准确性为要求,研究基于纹理点云的局部参考系估计技术和点云全局特征描述符生成与匹配技术,提出融合形状与纹理信息的全局正交对象描述符及其3D物体识别算法,基于蜗轮蜗杆减速器的拆装维修场景进行实例验证。验证结果表明,3D物体识别算法在AR辅助维修场景中具备实际可行性和真实有效性。
3D object recognition and pose estimation are the basis of AR-assisted maintenance information enhancement.At present,the recognition accuracy and real-time and pose estimation accuracy shortage also exists in 3D object recognition technology.To meet the requirements of the target recognition real-time and pose estimation accuracy of scene recognition,this paper researches on the local reference frame estimation and descriptor generation and matching based on global feature of point cloud and proposes a fusion shape and texture global orthographic object descriptor and the 3D object recognition algorithm.Then,the algorithm is verified based on the disassembly and maintenance of the worm gear reducer.The verification results show that the algorithm above has practical feasibility and real validity in the AR-assisted maintenance real scene.
作者
王维
张丹
胡曾一震
左敦稳
WANG Wei;ZHANG Dan;HU Zeng yizhen;ZUO Dunwen(College of Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)
出处
《机械制造与自动化》
2020年第3期155-158,共4页
Machine Building & Automation
基金
国家自然科学基金(51705246)。
关键词
AR辅助维修
局部参考系
点云描述符
3D物体识别
位姿估计
AR-assisted maintenance
local reference frame
point cloud descriptor
3D object recognition
pose estimation