Abstract: wing to the lack of experience and visual auxiliary maintenance instructions,the improper maintenance was easy to occur.To address these issues,an intelligent maintenance method based on edge-cloud collaboration and wearable Augmented Reality(AR)was proposed.To overcome the loss of attitude tracking in maintenance scene with rich texture,the adaptive information entropy and sharpening adjustment algorithm was fused with the traditional Oriented Brief-Simultaneous Localization and Mapping(ORB-SLAM2).Besides,augmented reality was applied process human-computer interaction information during maintenance,and an industrial AR cloud platform was constructed to run the improved ORB-SLAM2 algorithm to obtain accurate poses.Meanwhile,the high-precision feature extraction and 3D registration could be realized through the edge-cloud collaboration computing framework.On this basis,faulty parts which needed to be repaired could be highlighted and description of maintenance process could be guided through enhanced visualization.Additionally,a remote expert system was constructed to solve the failure of unknown root causes,and the remote maintenance was realized.The comparison experiment of real cases in the workshop showed that the proposed method could effectively improve the efficiency and reliability of maintenance personnel.

Key words: edge-cloud collaboration, augmented reality, intelligent maintenance, tracking registration, remote expert

摘要： 针对运维人员对复杂结构设备的经验技术存在不足以及缺少直观可视化的辅助维修表征手段,导致运维人员的维修效率低、可靠性差且易受车间嘈杂环境影响而引起不当维修的现象,提出一种基于边云协同和增强现实的车间现场智能维修方法。通过融合自适应信息熵与锐化调整算法,克服传统面向快速旋转的二元鲁棒独立基本特征的同时定位与建图方法(ORB-SLAM2)在具有丰富纹理的维修场景中运行时,无法获得足够稳定的匹配点对导致姿态跟踪丢失的缺点。通过边云协同计算框架,在云端运行改进的ORB-SLAM2算法,反馈给可穿戴增强现实设备进行高精度特征提取与三维注册跟踪,实现车间现场环境下的待维修部件和维修过程说明的增强可视化指引。此外,构建远程专家系统解决未知根源的故障,实现车间现场远程维修。通过车间现场真实维修案例对比实验表明,所提方法能够有效提高车间现场人员维修的效率和可靠度。

关键词: 边云协同, 增强现实, 智能维修, 跟踪注册, 远程专家