It is known that different relationships exist between the strength and displacement of the stratospheric polar vortex(SPV),and the surface air temperature(SAT)patterns in Eurasia and North America,but the mechanisms ...It is known that different relationships exist between the strength and displacement of the stratospheric polar vortex(SPV),and the surface air temperature(SAT)patterns in Eurasia and North America,but the mechanisms behind these relationships remain unclear,especially on an interannual timescale.Based on empirical orthogonal function(EOF)analysis using NCEP reanalysis data over 1958–2018,this study attempts to ascertain the relationship between the SPV intensity and displacement over the Arctic and the SATs in the midlatitudes of the Northern Hemisphere.Our results indicate that a strengthened SPV corresponds to an SAT increase in Eurasia and a decrease in eastern North America and Greenland.When the SPV is shifted towards Eurasia,however,a corresponding SAT increase occurs in both North America and Eurasia,with a larger increase in North America than in Eurasia.Specifically,a strengthened SPV tends to correspond to a positive North Atlantic Oscillation-like circulation in the troposphere with negative geopotential height(GH)anomalies in Greenland and eastern North American continent and positive GH anomalies to the north of 45°N in Eurasia,which corresponds to lower SATs in North America than in Eurasia.However,when the SPV shifted towards Eurasia,it was accompanied by a positive Pacific/North American-like pattern with a deepened Aleutian low,which corresponds to the increasing SATs in North America.These tropospheric circulation changes are related to the response of tropospheric planetary wave activity to the SPV.A strengthened SPV corresponds to the weakening of tropospheric planetary wave-1 waves,which is accompanied by a negative GH in North America but a positive GH in Eurasia.If the SPV shifted towards Eurasia,the tropospheric planetary wave-1(-2)waves strengthened(weakened),and the combined effects of the planetary wave-1 and wave-2 waves would cause positive GH anomalies in both Eurasia and North America.展开更多
目的 视频异常行为检测是当前智能监控技术的研究热点之一,在社会安防领域具有重要应用。如何通过有效地对视频空间维度信息和时间维度信息建模来提高异常检测的精度仍是目前研究的难点。由于结构优势,生成对抗网络目前广泛应用于视频...目的 视频异常行为检测是当前智能监控技术的研究热点之一,在社会安防领域具有重要应用。如何通过有效地对视频空间维度信息和时间维度信息建模来提高异常检测的精度仍是目前研究的难点。由于结构优势,生成对抗网络目前广泛应用于视频异常检测任务。针对传统生成对抗网络时空特征利用率低和检测效果差等问题,本文提出一种融合门控自注意力机制的生成对抗网络进行视频异常行为检测。方法 在生成对抗网络的生成网络U-net部分引入门控自注意力机制,逐层对采样过程中的特征图进行权重分配,融合U-net网络和门控自注意力机制的性能优势,抑制输入视频帧中与异常检测任务不相关背景区域的特征表达,突出任务中不同目标对象的相关特征表达,更有效地针对时空维度信息进行建模。采用LiteFlownet网络对视频流中的运动信息进行提取,以保证视频序列之间的连续性。同时,加入强度损失函数、梯度损失函数和运动损失函数加强模型检测的稳定性,以实现对视频异常行为的检测。结果 在CUHK(Chinese University of Hong Kong) Avenue、UCSD(University of California, San Diego) Ped1和UCSD Ped2等视频异常事件数据集上进行实验。在CUHK Avenue数据集中,本文方法的AUC(area under curve)为87.2%,比同类方法高2.3%;在UCSD Ped1和UCSD Ped2数据集中,本文方法的AUC值均高于同类其他方法。同时,设计了4个消融实验并对实验结果进行对比分析,本文方法具有更高的AUC值。结论 实验结果表明,本文方法更适合视频异常检测任务,有效提高了异常行为检测任务模型的稳定性和准确率,且采用视频序列帧间运动信息能够显著提升异常行为检测性能。展开更多
基金Supported by the National Natural Science Foundation of China(42175072)Strategic Priority Research Program of Chinese Academy of Sciences(XDA2010030804)。
文摘It is known that different relationships exist between the strength and displacement of the stratospheric polar vortex(SPV),and the surface air temperature(SAT)patterns in Eurasia and North America,but the mechanisms behind these relationships remain unclear,especially on an interannual timescale.Based on empirical orthogonal function(EOF)analysis using NCEP reanalysis data over 1958–2018,this study attempts to ascertain the relationship between the SPV intensity and displacement over the Arctic and the SATs in the midlatitudes of the Northern Hemisphere.Our results indicate that a strengthened SPV corresponds to an SAT increase in Eurasia and a decrease in eastern North America and Greenland.When the SPV is shifted towards Eurasia,however,a corresponding SAT increase occurs in both North America and Eurasia,with a larger increase in North America than in Eurasia.Specifically,a strengthened SPV tends to correspond to a positive North Atlantic Oscillation-like circulation in the troposphere with negative geopotential height(GH)anomalies in Greenland and eastern North American continent and positive GH anomalies to the north of 45°N in Eurasia,which corresponds to lower SATs in North America than in Eurasia.However,when the SPV shifted towards Eurasia,it was accompanied by a positive Pacific/North American-like pattern with a deepened Aleutian low,which corresponds to the increasing SATs in North America.These tropospheric circulation changes are related to the response of tropospheric planetary wave activity to the SPV.A strengthened SPV corresponds to the weakening of tropospheric planetary wave-1 waves,which is accompanied by a negative GH in North America but a positive GH in Eurasia.If the SPV shifted towards Eurasia,the tropospheric planetary wave-1(-2)waves strengthened(weakened),and the combined effects of the planetary wave-1 and wave-2 waves would cause positive GH anomalies in both Eurasia and North America.
文摘目的 视频异常行为检测是当前智能监控技术的研究热点之一,在社会安防领域具有重要应用。如何通过有效地对视频空间维度信息和时间维度信息建模来提高异常检测的精度仍是目前研究的难点。由于结构优势,生成对抗网络目前广泛应用于视频异常检测任务。针对传统生成对抗网络时空特征利用率低和检测效果差等问题,本文提出一种融合门控自注意力机制的生成对抗网络进行视频异常行为检测。方法 在生成对抗网络的生成网络U-net部分引入门控自注意力机制,逐层对采样过程中的特征图进行权重分配,融合U-net网络和门控自注意力机制的性能优势,抑制输入视频帧中与异常检测任务不相关背景区域的特征表达,突出任务中不同目标对象的相关特征表达,更有效地针对时空维度信息进行建模。采用LiteFlownet网络对视频流中的运动信息进行提取,以保证视频序列之间的连续性。同时,加入强度损失函数、梯度损失函数和运动损失函数加强模型检测的稳定性,以实现对视频异常行为的检测。结果 在CUHK(Chinese University of Hong Kong) Avenue、UCSD(University of California, San Diego) Ped1和UCSD Ped2等视频异常事件数据集上进行实验。在CUHK Avenue数据集中,本文方法的AUC(area under curve)为87.2%,比同类方法高2.3%;在UCSD Ped1和UCSD Ped2数据集中,本文方法的AUC值均高于同类其他方法。同时,设计了4个消融实验并对实验结果进行对比分析,本文方法具有更高的AUC值。结论 实验结果表明,本文方法更适合视频异常检测任务,有效提高了异常行为检测任务模型的稳定性和准确率,且采用视频序列帧间运动信息能够显著提升异常行为检测性能。
