A topographic target light scattering-differential optical absorption spectroscopy （＇IbTaL-DOA~） system is de- veloped for measuring average concentrations along a known optical path and studying surface-near distr...A topographic target light scattering-differential optical absorption spectroscopy （＇IbTaL-DOA~） system is de- veloped for measuring average concentrations along a known optical path and studying surface-near distributions of atmospheric trace gases. The telescope of the ToTaL-DOAS system points to targets which are located at known dis- tances from the measurement device and illuminated by sunlight. Average concentrations with high spatial resolution can be retrieved by receiving sunlight reflected from the targets, A filed measurement of NO2 concentration is performed with the ToTaL-DOAS system in Shijiazhuang in the autumn of 2011. The measurement data are compared with con- centrations measured by the point monitoring technique at the same site. The results show that the ToTaL-DOAS system is sensitive to the variation of NO2 concentrations along the optical path.展开更多
针对空间非合作目标(航天器、空间碎片等)因光线变化、目标旋转等导致跟踪失败问题,提出以带有C2F模块的YOLO(You Only Look Once)v8模型用于目标检测,并将经微调的全尺度网络(Omin-Scale Network,OSNet)替换StrongSORT中ResNet50进行...针对空间非合作目标(航天器、空间碎片等)因光线变化、目标旋转等导致跟踪失败问题,提出以带有C2F模块的YOLO(You Only Look Once)v8模型用于目标检测,并将经微调的全尺度网络(Omin-Scale Network,OSNet)替换StrongSORT中ResNet50进行重识别,以改进YOLO-StrongSORT算法。该算法首先以YOLOv8从采集的空间影像中检测目标;随即将检测结果分别送入包含OSNet网络的外性特征检测器和以NSA卡尔曼滤波器为核心的运动信息检测器提取检测目标的外观特征和预测目标运动轨迹;然后采用Vanilla匹配机制匹配连续帧中同一目标,实现对目标的跟踪;最后,用SNCOVT数据集和自研的近红外相机采集影像验证。结果表明,改进YOLO-StrongSORT算法在SNCOVT数据集上平均跟踪精度和成功率分别为61.7%和60.7%,较传统算法更准确;在实验室采集影像上跟踪精度和成功率分别为60.0%和56.2%,可用于空间目标的跟踪。展开更多
提出了一种基于Vulkan架构的弹跳射线(shooting and bouncing ray,SBR)加速计算方法,用于电大复杂目标雷达散射截面的快速计算。设计了高效的Vulkan计算着色器,充分利用GPU硬件光追,显著提升了SBR法中光线求交的计算速度;引入了双命令...提出了一种基于Vulkan架构的弹跳射线(shooting and bouncing ray,SBR)加速计算方法,用于电大复杂目标雷达散射截面的快速计算。设计了高效的Vulkan计算着色器,充分利用GPU硬件光追,显著提升了SBR法中光线求交的计算速度;引入了双命令缓冲机制,使得CPU与GPU能够高效协同工作,从而加速多角度扫描任务的执行;在虚拟孔径面上划分互不干扰的子任务,进一步提升了多GPU并行的利用效率。实验结果表明:所提出方法在计算电大复杂目标雷达散射截面时相较于FEKO RL-GO方法实现了40倍以上的加速;双命令缓冲机制提升了约42%的多角度扫描速度;双GPU计算并行效率超过90%。展开更多
基金Project supported by the National High Technology Research and Development of China (Grant No.2009AA063006)the National Natural Science Foundation of China (Grant No. 40905010)the Special Project of Environmental Nonprofit Industry Research,China (Grant No. 201109007)
文摘A topographic target light scattering-differential optical absorption spectroscopy （＇IbTaL-DOA~） system is de- veloped for measuring average concentrations along a known optical path and studying surface-near distributions of atmospheric trace gases. The telescope of the ToTaL-DOAS system points to targets which are located at known dis- tances from the measurement device and illuminated by sunlight. Average concentrations with high spatial resolution can be retrieved by receiving sunlight reflected from the targets, A filed measurement of NO2 concentration is performed with the ToTaL-DOAS system in Shijiazhuang in the autumn of 2011. The measurement data are compared with con- centrations measured by the point monitoring technique at the same site. The results show that the ToTaL-DOAS system is sensitive to the variation of NO2 concentrations along the optical path.
文摘针对空间非合作目标(航天器、空间碎片等)因光线变化、目标旋转等导致跟踪失败问题,提出以带有C2F模块的YOLO(You Only Look Once)v8模型用于目标检测,并将经微调的全尺度网络(Omin-Scale Network,OSNet)替换StrongSORT中ResNet50进行重识别,以改进YOLO-StrongSORT算法。该算法首先以YOLOv8从采集的空间影像中检测目标;随即将检测结果分别送入包含OSNet网络的外性特征检测器和以NSA卡尔曼滤波器为核心的运动信息检测器提取检测目标的外观特征和预测目标运动轨迹;然后采用Vanilla匹配机制匹配连续帧中同一目标,实现对目标的跟踪;最后,用SNCOVT数据集和自研的近红外相机采集影像验证。结果表明,改进YOLO-StrongSORT算法在SNCOVT数据集上平均跟踪精度和成功率分别为61.7%和60.7%,较传统算法更准确;在实验室采集影像上跟踪精度和成功率分别为60.0%和56.2%,可用于空间目标的跟踪。
文摘提出了一种基于Vulkan架构的弹跳射线(shooting and bouncing ray,SBR)加速计算方法,用于电大复杂目标雷达散射截面的快速计算。设计了高效的Vulkan计算着色器,充分利用GPU硬件光追,显著提升了SBR法中光线求交的计算速度;引入了双命令缓冲机制,使得CPU与GPU能够高效协同工作,从而加速多角度扫描任务的执行;在虚拟孔径面上划分互不干扰的子任务,进一步提升了多GPU并行的利用效率。实验结果表明:所提出方法在计算电大复杂目标雷达散射截面时相较于FEKO RL-GO方法实现了40倍以上的加速;双命令缓冲机制提升了约42%的多角度扫描速度;双GPU计算并行效率超过90%。
