液晶面板喷墨打印表面缺陷检测中存在目标小、样本少、纹理背景干扰等问题,应用传统图像处理算法检测精度低、泛化性差,针对以上问题提出了一种改进RT-DETR(real-time detection transformer)的目标检测算法。改进RT-DETR算法通过将主...液晶面板喷墨打印表面缺陷检测中存在目标小、样本少、纹理背景干扰等问题,应用传统图像处理算法检测精度低、泛化性差,针对以上问题提出了一种改进RT-DETR(real-time detection transformer)的目标检测算法。改进RT-DETR算法通过将主干网络ResNet模型替换为特征提取性能更优的ConvNeXt模型,提高算法整体检测精度。设计了基于通道注意力的增强通道压缩模块,使算法更有效地消除背景干扰专注于定位缺陷目标,加快算法收敛,提高小目标检测精度。在构建的喷墨打印缺陷数据集训练实验上,改进RT-DETR算法检测平均精度mAP(mean average precision)为80.58%,较原始RT-DETR算法提升了2.89%,较原始DETR算法提升了15.88%,检测速度达到20 FPS(frames per second),改进RT-DETR算法的综合检测性能更优。改进RT-DETR算法在小目标检测数据集VisDrone训练实验上表现出良好的通用性,为其他工业场景下的表面小目标缺陷检测提供了参考价值。展开更多
The dominant error source of mobile terminal location in wireless sensor networks (WSNs) is the non-line-of-sight (NLOS) propagation error. Among the algorithms proposed to mitigate the influence of NLOS propagati...The dominant error source of mobile terminal location in wireless sensor networks (WSNs) is the non-line-of-sight (NLOS) propagation error. Among the algorithms proposed to mitigate the influence of NLOS propagation error, residual test (RT) is an efficient one, however with high computational complexity (CC). An improved algorithm that memorizes the light of sight (LOS) range measurements (RMs) identified memorize LOS range measurements identified residual test (MLSI-RT) is presented in this paper to address this problem. The MLSI-RT is based on the assumption that when all RMs are from LOS propagations, the normalized residual follows the central Chi-Square distribution while for NLOS cases it is non-central. This study can reduce the CC by more than 90%.展开更多
针对基于Linux和TCG软件栈(Trusted computing group Software Stack,TSS)的复杂性问题,提出一种轻量级的可信软件栈。分析了TSS的基本结构与TSS在嵌入式系统的局限,总结出基于嵌入式系统的可信软件栈设计需求,设计出软件栈命令调用的...针对基于Linux和TCG软件栈(Trusted computing group Software Stack,TSS)的复杂性问题,提出一种轻量级的可信软件栈。分析了TSS的基本结构与TSS在嵌入式系统的局限,总结出基于嵌入式系统的可信软件栈设计需求,设计出软件栈命令调用的机制和软件栈的结构。此外,分析了TSS密钥管理缓存算法,在flash中定义一块密钥槽空间,方便密钥管理中直接访问,阐述密钥生成的逻辑过程,实现面向嵌入式系统的可信软件系统。经实验验证,该软件栈可以结合RT-Thread实时系统实现基本的可信计算功能。展开更多
文摘液晶面板喷墨打印表面缺陷检测中存在目标小、样本少、纹理背景干扰等问题,应用传统图像处理算法检测精度低、泛化性差,针对以上问题提出了一种改进RT-DETR(real-time detection transformer)的目标检测算法。改进RT-DETR算法通过将主干网络ResNet模型替换为特征提取性能更优的ConvNeXt模型,提高算法整体检测精度。设计了基于通道注意力的增强通道压缩模块,使算法更有效地消除背景干扰专注于定位缺陷目标,加快算法收敛,提高小目标检测精度。在构建的喷墨打印缺陷数据集训练实验上,改进RT-DETR算法检测平均精度mAP(mean average precision)为80.58%,较原始RT-DETR算法提升了2.89%,较原始DETR算法提升了15.88%,检测速度达到20 FPS(frames per second),改进RT-DETR算法的综合检测性能更优。改进RT-DETR算法在小目标检测数据集VisDrone训练实验上表现出良好的通用性,为其他工业场景下的表面小目标缺陷检测提供了参考价值。
基金supported by the State Key Program of National Natural Science of China (Grant No.60532030)the New Century Excellent Talents in University (Grant No.NCET-08-0333)the Natural Science Foundation of Shandong Province (Grant No.Y2007G10)
文摘The dominant error source of mobile terminal location in wireless sensor networks (WSNs) is the non-line-of-sight (NLOS) propagation error. Among the algorithms proposed to mitigate the influence of NLOS propagation error, residual test (RT) is an efficient one, however with high computational complexity (CC). An improved algorithm that memorizes the light of sight (LOS) range measurements (RMs) identified memorize LOS range measurements identified residual test (MLSI-RT) is presented in this paper to address this problem. The MLSI-RT is based on the assumption that when all RMs are from LOS propagations, the normalized residual follows the central Chi-Square distribution while for NLOS cases it is non-central. This study can reduce the CC by more than 90%.
文摘针对基于Linux和TCG软件栈(Trusted computing group Software Stack,TSS)的复杂性问题,提出一种轻量级的可信软件栈。分析了TSS的基本结构与TSS在嵌入式系统的局限,总结出基于嵌入式系统的可信软件栈设计需求,设计出软件栈命令调用的机制和软件栈的结构。此外,分析了TSS密钥管理缓存算法,在flash中定义一块密钥槽空间,方便密钥管理中直接访问,阐述密钥生成的逻辑过程,实现面向嵌入式系统的可信软件系统。经实验验证,该软件栈可以结合RT-Thread实时系统实现基本的可信计算功能。