信阳师范学院学报（自然科学版）2024，Vol.37Issue(2)：240-245,6.DOI:10.3969/j.issn.1003-0972.2024.02.017
基于迁移学习的飞机线束模板图纸识别模型研究
Research on Recognition Model of Aircraft Wiring Harness Template Drawings Based on Transfer Learning
摘要
Abstract
A transfer learning-based classification method for aircraft harness template drawing content was designed for the recognition and classification of Electrical Wiring Interconnection Systems(EWIS)harness template drawings during change review.The harness template drawing element dataset was established from the existing harness template drawings,the VGG-16 network was selected as the basic network to optimize the model structure and parameters.To solve the problem of small dataset size and to improve the recognition accuracy and training efficiency of the model,the pre-trained model trained in the large-scale dataset was migrated to the harness template drawing element dataset for training,and the effects of different migration learning methods and different network structures on the model performance were compared.The experimental results showed that the fine-tuning transfer learning model had a high classification accuracy of 98.89%.Moreover,compared with other networks,the fine-tuning transfer learning model had a shorter training time and higher recognition accuracy,which has better application prospects.关键词
航空电气线路互联系统(EWIS)/线束模板图纸/卷积神经网络/VGG-16网络/迁移学习Key words
Electrical Wiring Interconnection Systems(EWIS)/aircraft wiring harness drawings/convolutional neural network/VGG-16 network/transfer learning分类
信息技术与安全科学引用本文复制引用
周文举,朱一丁,杨庆华,朱喜,王长法..基于迁移学习的飞机线束模板图纸识别模型研究[J].信阳师范学院学报（自然科学版）,2024,37(2):240-245,6.基金项目
国家自然科学基金项目(62005155) （62005155）
上海市科技攻关"揭榜挂帅"项目(YDZX20223100004008) （YDZX20223100004008）
