跳到主要导航 跳到搜索 跳到主要内容

Adaptive Recognition and AR-Guided Assembly for Aviation Cable Connectors

  • Xi'an Jiaotong University

科研成果: 书/报告/会议事项章节 › 会议稿件 › 同行评审

摘要

Aviation cable assembly demands precise connector recognition under challenging industrial conditions, yet conventional methods suffer from inefficiency and high cognitive load. This study introduces an adaptive recognition system integrated with augmented reality (AR) guidance to address these limitations. The proposed framework combines a dual-constrained ROI extraction mechanism leveraging depth-based foreground segmentation and Hough transform for robust localization, a CBAM-enhanced EfficientNetV2M model optimized for cable connector recognition, a dynamic threshold stabilizer and a real-time minimized AR visual guidance mechanism. Evaluated on 12,800 connector images dataset, the system achieved 99.3% recognition accuracy. A user study demonstrated a 93.1% efficiency improvement over conventional methods, with AR-guided adaptive recognition significantly outperforming basic AR-guidance and conventional guidance in System Usability Scale (SUS: 91.00), cognitive load (Subjective Mental Effort Questionnaire, SMEQ: 5.73), and workload reduction (NASA-TLX: 12.83). This work bridges critical gaps in aviation assembly automation, offering a scalable solution for error-prone manual workflows.

源语言英语
主期刊名Fourth International Conference on New Materials, Machinery, and Vehicle Engineering, NMMVE 2025
编辑Jinyang Xu, J. Paulo Davim
出版商SPIE
ISBN(电子版)9781510698475
DOI
出版状态已出版 - 1 11月 2025
活动4th International Conference on New Materials, Machinery, and Vehicle Engineering, NMMVE 2025 - Changsha, 中国
期限: 11 7月 2025 → 13 7月 2025

丛书

姓名Proceedings of SPIE - The International Society for Optical Engineering
卷13953
ISSN(印刷版)0277-786X
ISSN(电子版)1996-756X

会议

会议4th International Conference on New Materials, Machinery, and Vehicle Engineering, NMMVE 2025
国家/地区中国
市Changsha
时期11/07/25 → 13/07/25

学术指纹

探究 'Adaptive Recognition and AR-Guided Assembly for Aviation Cable Connectors' 的科研主题。它们共同构成独一无二的学术指纹。

引用此