针对传统检测方法中摄像头视角受限问题,提出了一种结合面部姿态矫正和改进ViViT的多视角下人脸疲倦检测方法。采用Mediapipe Face Mesh定位面部三维特征点并将其矫正为正面,利用提出的FGR-ViViT模型来捕捉矫正后的眼睛、眉毛、嘴巴线...针对传统检测方法中摄像头视角受限问题,提出了一种结合面部姿态矫正和改进ViViT的多视角下人脸疲倦检测方法。采用Mediapipe Face Mesh定位面部三维特征点并将其矫正为正面,利用提出的FGR-ViViT模型来捕捉矫正后的眼睛、眉毛、嘴巴线条图像帧序列变化。FGR-ViViT通过在ViViT的Temporal Transformer Encoder中添加部件选择模块来捕捉特征在时间维度中的细微差异,同时融合2次dropout和改进的对比损失函数来调整样本的相似性,降低模型过拟合风险并提高泛化能力。实验结果表明,提出的方法在YawDD和DROZY矫正后的线条图像帧的测试集上,F1-分数达到了94.5%和97.6%,相较于原始人脸图像帧分别提高了3.2%和10.4%,其FGR-ViViT相较于原始ViViT分别提高了6.1%和0.7%。所提方法适用于摄像头灵活摆放的多种应用场景,对解决多视角人脸睡意判断具有积极意义。展开更多
共因失效普遍存在于现实的工程系统中。如果忽略共因失效对系统的影响,按照传统假设其部件独立的方法进行可靠性分析会得出过于乐观的结论,并且作出认为系统符合可靠性需求的错误判断。在考虑系统由部件选择性失效传播引起的共因失效的...共因失效普遍存在于现实的工程系统中。如果忽略共因失效对系统的影响,按照传统假设其部件独立的方法进行可靠性分析会得出过于乐观的结论,并且作出认为系统符合可靠性需求的错误判断。在考虑系统由部件选择性失效传播引起的共因失效的同时,提出了一种面向复杂系统的基于蒙特卡罗模拟(Monte Carlo simulation,MCS)与元胞自动机(cellular automata,CA)集成的可靠性评估算法。传统方法只能解决可以转化为串、并联结构的简单系统的可靠性评估问题,而所提出的算法打破了这种限制,使得其应用范围更加广泛。展开更多
The influence of active metal components of catalyst, additives and catalyst preparation method on the reactivity of catalyst for selective hydrodesulfurization (HDS) of FCC naphtha was investigated, and the RSDS-21 c...The influence of active metal components of catalyst, additives and catalyst preparation method on the reactivity of catalyst for selective hydrodesulfurization (HDS) of FCC naphtha was investigated, and the RSDS-21 catalyst with high HDS performance and the RSDS-22 catalyst with high selectivity were developed by RIPP. The composite loading of a new series of catalysts for selective HDS of FCC gasoline has demonstrated excellent desulfurization activity and selectivity and can under conventional hydrotreating conditions manufacture clean gasoline product meeting the national IV emission standard and the Euro V emission standard with less loss in antiknock index. The finalized new series of FCC catalysts upon being adopted for selective HDS of FCC naphtha have good adaptability to different feedstocks along with good stability.展开更多
文摘针对传统检测方法中摄像头视角受限问题,提出了一种结合面部姿态矫正和改进ViViT的多视角下人脸疲倦检测方法。采用Mediapipe Face Mesh定位面部三维特征点并将其矫正为正面,利用提出的FGR-ViViT模型来捕捉矫正后的眼睛、眉毛、嘴巴线条图像帧序列变化。FGR-ViViT通过在ViViT的Temporal Transformer Encoder中添加部件选择模块来捕捉特征在时间维度中的细微差异,同时融合2次dropout和改进的对比损失函数来调整样本的相似性,降低模型过拟合风险并提高泛化能力。实验结果表明,提出的方法在YawDD和DROZY矫正后的线条图像帧的测试集上,F1-分数达到了94.5%和97.6%,相较于原始人脸图像帧分别提高了3.2%和10.4%,其FGR-ViViT相较于原始ViViT分别提高了6.1%和0.7%。所提方法适用于摄像头灵活摆放的多种应用场景,对解决多视角人脸睡意判断具有积极意义。
文摘共因失效普遍存在于现实的工程系统中。如果忽略共因失效对系统的影响,按照传统假设其部件独立的方法进行可靠性分析会得出过于乐观的结论,并且作出认为系统符合可靠性需求的错误判断。在考虑系统由部件选择性失效传播引起的共因失效的同时,提出了一种面向复杂系统的基于蒙特卡罗模拟(Monte Carlo simulation,MCS)与元胞自动机(cellular automata,CA)集成的可靠性评估算法。传统方法只能解决可以转化为串、并联结构的简单系统的可靠性评估问题,而所提出的算法打破了这种限制,使得其应用范围更加广泛。
文摘The influence of active metal components of catalyst, additives and catalyst preparation method on the reactivity of catalyst for selective hydrodesulfurization (HDS) of FCC naphtha was investigated, and the RSDS-21 catalyst with high HDS performance and the RSDS-22 catalyst with high selectivity were developed by RIPP. The composite loading of a new series of catalysts for selective HDS of FCC gasoline has demonstrated excellent desulfurization activity and selectivity and can under conventional hydrotreating conditions manufacture clean gasoline product meeting the national IV emission standard and the Euro V emission standard with less loss in antiknock index. The finalized new series of FCC catalysts upon being adopted for selective HDS of FCC naphtha have good adaptability to different feedstocks along with good stability.
