基于面部特征融合的飞行员专注度识别方法  

A Pilot Concentration Recognition Method Based on Objective Experiment

作　　者：包其建 顾人舒 顾宏斌[1] BAO Qi-jian;GU Ren-shu;GU Hong-bin(Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China;Hangzhou Dianzi University,Hangzhou 310000,China)

机构地区：[1]南京航空航天大学,江苏南京211000 [2]杭州电子科技大学,浙江杭州310000

出　　处：《航空计算技术》2025年第1期33-37,共5页Aeronautical Computing Technique

基　　金：国家自然科学基金项目资助(62202130);浙江省自然科学基金项目资助(LQ22F020026)。

摘　　要：飞行员的训练质量对于飞行安全至关重要,循证训练中要求评估飞行员包含一系列核心胜任力的全面能力。应用计算机视觉技术,非接触式识别飞行员模拟训练时的专注度等级,为循证训练提供更多信息。为了划分适用训练评估的专注度等级,设计了诱发注意力的实验,并以客观的指标对数据进行标注,接着提出一种面部特征融合的专注度识别网络模型,提取面部图像特征以及局部二值模式(LBP)图像特征,同时引入自适应参数调节两种特征的比重。模型在测试集上的精确率、召回率和F1值分别达到了98.6%、98.5%和98.5%,优于其他先进的分类算法。实验结果表明,该模型能有效识别飞行员的专注度等级,同时为循证训练提供更好数据基础。The quality of pilot training is critical to flight safety,and evidence-based training requires the assessment of a pilot′s overall competence,including a set of core competencies.From the perspective of computer vision,this paper identifies the concentration level of pilot simulation training without invasiveness,and provides data basis and evaluation method for evidence-based training.In order to classify the concentration level suitable for training evaluation,an attention-inducing experiment was designed and the data were labeled with objective indicators.Then,a concentration recognition network model based on facial feature fusion was proposed to extract facial image features and LBP image features,and adaptive parameters were introduced to adjust the proportion of the two features.The experimental results show that the accuracy rate,recall rate and F1 value of the model on the test set reach 98.6%,98.5%and 98.5%respectively,which is better than other advanced classification algorithms.The model can effectively identify the concentration level of pilot simulation training and provide methods and evaluation indexes for evidence-based training.

关 键 词：循证训练 专注度等级 LBP特征 特征融合 

分 类 号：V328[航空宇航科学与技术—人机与环境工程] TP317.4[自动化与计算机技术—计算机软件与理论]