无线电设施原始资料附件特征的智能提取  

Intelligent Features Extraction for Original Attachment of Radio Facility

作　　者：陈华群[1] 张进 董世华 Chen Huaqun;Zhang Jin;Dong Shihua(College of Air Traffic Management,Civil Aviation Flight University of China,Guanghan 618307,Sichuan,China)

机构地区：[1]中国民用航空飞行学院空中交通管理学院,四川广汉618307

出　　处：《科技通报》2025年第3期15-20,共6页Bulletin of Science and Technology

基　　金：西藏自治区科技计划项目(XZ202403ZY0014);民航局安全能力建设项目(2022237);四川省大学生创新训练计划项目(S2024106240、S2024106075)。

摘　　要：为了提高航空原始资料的审核效率和降低人为差错,本文利用卷积神经网络模型和YOLOv5图像识别技术,研究无线电原始资料附件特征的智能提取。基于审核附件文件的模块分区属性,建立文件的完整性、准确性和合法性的评价函数;利用模块的信息变动属性和图像像素差异,构建卷积神经网络模型,实现分区模块的特征提取;将YOLOv5的Backbone中的FOCUS结构改进为ORT结构,提升图像检测的效率。以新增无线电导航设施的附件的特征提取为例进行实例验证。结果表明:分区模块式的卷积神经网络模型,降低了重复检测的冗余度;改进的YOLOv5的ORT结构,提升了审核的精确度和效率。To improve the auditing efficiency of aviation original data and reduce human error,Convolutional Neural Network model and YOLOv5 image recognition technology are used to study the intelligent extraction of radio original data attachment features.Based on the module partition attribute of the attachment file,the evaluation function of the integrity,accuracy and legitimacy of the file is established;Based on the information change attribute of the module and the difference of image pixels,the Convolutional Neural Network model is constructed to realize the feature extraction of the partition module;FOCUS structure in Backbone of YOLOv5 is improved to ORT structure to improve the efficiency of image detection.The feature extraction of the attachment of the newly increased radio navigation facility is used for example verification.The results show that:the Convolutional Neural Network model based on partition module reduces the redundancy of repeated detection;The ORT results of YOLOv5 improve the accuracy and efficiency of the audit.

关 键 词：航空原始资料 无线电附件 特征提取 卷积神经网络模型 YOLOv5图像检测 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]