针对无人机航拍图像目标检测中视野变化大、时空信息复杂等问题,文中基于YOLOv5(You Only Look Once Version5)架构,提出基于图像低维特征融合的航拍小目标检测模型.引入CA(Coordinate Attention),改进MobileNetV3的反转残差块,增加图...针对无人机航拍图像目标检测中视野变化大、时空信息复杂等问题,文中基于YOLOv5(You Only Look Once Version5)架构,提出基于图像低维特征融合的航拍小目标检测模型.引入CA(Coordinate Attention),改进MobileNetV3的反转残差块,增加图像空间维度信息的同时降低模型参数量.改进YOLOv5特征金字塔网络结构,融合浅层网络中的特征图,增加模型对图像低维有效信息的表达能力,进而提升小目标检测精度.同时为了降低航拍图像中复杂背景带来的干扰,引入无参平均注意力模块,同时关注图像的空间注意力与通道注意力;引入VariFocal Loss,降低负样本在训练过程中的权重占比.在VisDrone数据集上的实验验证文中模型的有效性,该模型在有效提升检测精度的同时明显降低复杂度.展开更多
Drone or unmanned aerial vehicle(UAV)technology has undergone significant changes.The technology allows UAV to carry out a wide range of tasks with an increasing level of sophistication,since drones can cover a large ...Drone or unmanned aerial vehicle(UAV)technology has undergone significant changes.The technology allows UAV to carry out a wide range of tasks with an increasing level of sophistication,since drones can cover a large area with cameras.Meanwhile,the increasing number of computer vision applications utilizing deep learning provides a unique insight into such applications.The primary target in UAV-based detection applications is humans,yet aerial recordings are not included in the massive datasets used to train object detectors,which makes it necessary to gather the model data from such platforms.You only look once(YOLO)version 4,RetinaNet,faster region-based convolutional neural network(R-CNN),and cascade R-CNN are several well-known detectors that have been studied in the past using a variety of datasets to replicate rescue scenes.Here,we used the search and rescue(SAR)dataset to train the you only look once version 5(YOLOv5)algorithm to validate its speed,accuracy,and low false detection rate.In comparison to YOLOv4 and R-CNN,the highest mean average accuracy of 96.9%is obtained by YOLOv5.For comparison,experimental findings utilizing the SAR and the human rescue imaging database on land(HERIDAL)datasets are presented.The results show that the YOLOv5-based approach is the most successful human detection model for SAR missions.展开更多
在自然场景中,天气情况、光照强度、背景干扰等问题影响火焰检测的准确性.为了实现复杂场景下实时准确的火焰检测,在目标检测网络YOLOv5的基础上,结合Focal Loss焦点损失函数、CIoU(Complete Intersection over Union)损失函数与多特征...在自然场景中,天气情况、光照强度、背景干扰等问题影响火焰检测的准确性.为了实现复杂场景下实时准确的火焰检测,在目标检测网络YOLOv5的基础上,结合Focal Loss焦点损失函数、CIoU(Complete Intersection over Union)损失函数与多特征融合,提出实时高效的火焰检测方法.为了缓解正负样本不均衡问题,并充分利用困难样本的信息,引入焦点损失函数,同时结合火焰静态特征和动态特征,设计多特征融合方法,达到剔除误报火焰的目的.针对国内外缺乏火焰数据集的问题,构建大规模、高质量的十万量级火焰数据集(http://www.yongxu.org/data bases.html).实验表明,文中方法在准确率、速度、精度和泛化能力等方面均有明显提升,同时降低误报率.展开更多
The simulation and prediction of the climatology and interannual variability of the East Asia winter monsoon(EAWM),as well as the associated atmospheric circulation,was investigated using the hindcast data from Global...The simulation and prediction of the climatology and interannual variability of the East Asia winter monsoon(EAWM),as well as the associated atmospheric circulation,was investigated using the hindcast data from Global Seasonal Forecast System version 5(GloSea5),with a focus on the evolution of model bias among different forecast lead times.While GloSea5 reproduces the climatological means of large-scale circulation systems related to the EAWM well,systematic biases exist,including a cold bias for most of China’s mainland,especially for North and Northeast China.GloSea5 shows robust skill in predicting the EAWM intensity index two months ahead,which can be attributed to the performance in representing the leading modes of surface air temperature and associated background circulation.GloSea5 realistically reproduces the synergistic effect of El Niño–Southern Oscillation(ENSO)and the Arctic Oscillation(AO)on the EAWM,especially for the western North Pacific anticyclone(WNPAC).Compared with the North Pacific and North America,the representation of circulation anomalies over Eurasia is poor,especially for sea level pressure(SLP),which limits the prediction skill for surface air temperature over East Asia.The representation of SLP anomalies might be associated with the model performance in simulating the interaction between atmospheric circulations and underlying surface conditions.展开更多
The western North Pacific subtropical high(WNPSH) dominates the summer climate over East Asia. The intensity,position, and shape of WNPSH influence the spatiotemporal distributions of precipitation, temperature, and t...The western North Pacific subtropical high(WNPSH) dominates the summer climate over East Asia. The intensity,position, and shape of WNPSH influence the spatiotemporal distributions of precipitation, temperature, and tropical cyclone activities in this region. This paper intends to investigate the performance of the UK Met Office Global Seasonal forecast system version 5(GloSea5) in simulation/prediction of the WNPSH based on a hindcast dataset. Analyses of the hindcast data show a systematic bias in the mean circulation over West Pacific, with negative geopotential height anomalies over the western North Pacific(WNP) and cyclonic anomalies in the 850-hPa winds and water vapor transport, indicating a weakening and eastward shift of the WNPSH. Despite the model’s bias in the climatology, it well captured the interannual variability of the monthly and seasonal-mean intensity of the WNPSH and the position of its ridge line in boreal summer from 1993 to 2015. The seasonal hindcasts indicate that there is significant prediction skill at up to three-month lead time for both the intensity and position of the WNPSH ridge line. The relationship between the WNPSH and different phases of the El Nino–Southern Oscillation(ENSO) in both the observational data and GloSea5 hindcasts was then investigated. The model captured the summer WNPSH anomalies well during most of the ENSO phases, except in the La Nina decaying and neutral summers. The intensity of the anticyclone in the WNP is weak in the decaying phase of El Nino in the GloSea5 hindcasts compared with the reanalysis data. GloSea5 is capable of representing the lagged teleconnection between El Nino events in the previous winter and the intensity of the WNPSH in the following summer. Regression analysis reveals weakened negative sea surface temperature anomalies(SSTAs) over the WNP in GloSea5, which reduced the gradient between the tropical western Pacific and the tropical Indian Ocean, resulting in a weaker easterly anomaly and stronger westerly anomaly, contributing to the weak anomalous anticyclone over the WNP and the weakened WNPSH relative to the reanalysis data.展开更多
The sounding data of meteorological satellites provide not only the real time weather information about the distribution of both cloud and rainfall,but also some others about the movement and state of atmosphere.They ...The sounding data of meteorological satellites provide not only the real time weather information about the distribution of both cloud and rainfall,but also some others about the movement and state of atmosphere.They are important variables and parameters for NWP model used to simulate and predict atmospheric state.In order to introduce remote sensing information from satellites into NWP model,there is an efficient way of establishing an RT model by use of the atmosphere radiation sounding data of meteorological satellites to get the variables and parameters valuable to NWP model.In this paper,we set up profiles of air temperature and water vapor from the surface to upper (0.1 hPa) using the radiosounding data and the surface data from May to August 1998 atmosphere East Asia.A TOVS RT model (RTTOV5) is provided to compute the value of radiation value of HIRS channels in NOAA14.Then the radiation values of 19 HIRS channels are gotten.After matching these data computed by the RT model and the corresponding values coming from satellite sounding in time,the statistic distribution of bias between tile model output and the satellite sounding at each sounding channel can be gotten.At the same time.the distribution of RMS to every TOVS HIRS channel,the standard biases to different scanning angle to each channel are also obtained.展开更多
The track, landfall, dynamic and thermodynamic and cloud-rain physical mesoscale structures and their evolution of typhoon HERB 1996 in 36 h from 0000 UTC 31 July to 1200 UTC 1 August 1996 were simulated by using the ...The track, landfall, dynamic and thermodynamic and cloud-rain physical mesoscale structures and their evolution of typhoon HERB 1996 in 36 h from 0000 UTC 31 July to 1200 UTC 1 August 1996 were simulated by using the non-hydrostatic mesoscale model MM5. This period covered the process of typhoon HERB landfall at Taiwan and Fujian Provinces. Results show that the model successfully simulated the landfall process of typhoon HERB, revealed the most important characteristics of the mesoscale dynamic and thermodynamic and cloud-rain physical structure during its landfall. The simulated typhoon track was close to the observation. The center of cyclonic circulation simulated at 0000 UTC on 1 August 1996 (24 h integration) was located in shore near Fuqing, Fujian Province at which the typhoon was reported to landfall two hours later. It shows that strong upward motion formed by low level convergence existed in the eye-wall and subsidence at the eye. The wind field shows clear asymmetrical structure near the typhoon center. The cloud and rainband was screw-typed distributed around typhoon center, and consisted of meso-β scale rain cores. During the period of typhoon HERB staying near and passing over Taiwan, the lower cloud was developed in the eye region so that the previous clear typhoon eye on the satellite pictures became fuzzy. Observation shows that the typhoon center was 'warm', but the model simulations with higher space resolution show that in the mid-troposphere the region of eye-wall with stronger upward motion and more cloud-and rain- water was warmer than the eye. During the period of typhoon passing over Taiwan and its following landfall at Fujian, the track of model typhoon deviated about 30 km northward (i. e., rightward) because of the orographic effects of Taiwan Island, but the strength of the typhoon was not affected remarkably. The amount of rainfall on Taiwan in the 36 h simulations was enhanced more than six times by the orographic lifting of Taiwan Mountain.展开更多
Diagnosis is performed of a thunderstorm rainstorm event occurring in the summer of 1996 at Nanjing and numerical simulation undertaken in the context of hydrostatic equilibrium framework of MM5 as the fifth version o...Diagnosis is performed of a thunderstorm rainstorm event occurring in the summer of 1996 at Nanjing and numerical simulation undertaken in the context of hydrostatic equilibrium framework of MM5 as the fifth version of the PSU/NCAR mesoscale model.Analyses show that the rainstorm-associated thermal condition was the accumulation of unstable potential energy and the dynamic condition was vigorous convergence updrafts.And the simulation within the hydrostatic framework indicates the significant role of latent heat release in the rainstorm occurrence:that even for a 30 km grid spacing horizontally of great importance to the successful modeling of the meso-β event was a convection parameterization scheme that led to less rainfall in our run based only on its explicit version but to the prediction in closer agreement with the observed when its implicit version was used in combination:for the thunderstorm-accompanied torrential rain.the Grell scheme was superior to the version of Kuo and the improved Arakawa-Schubert parameterization scheme(Grell 1993:Anthese and Kuo 1987:Arakawa and Scherbt 1974:Grell et al,1991).Moreover,better results came from the simulation in the context of hydrostatic framework of the MM5 compared to those from the run within the nonhydrostatic equilibrium framework,a problem that awaits further efforts.展开更多
文摘针对无人机航拍图像目标检测中视野变化大、时空信息复杂等问题,文中基于YOLOv5(You Only Look Once Version5)架构,提出基于图像低维特征融合的航拍小目标检测模型.引入CA(Coordinate Attention),改进MobileNetV3的反转残差块,增加图像空间维度信息的同时降低模型参数量.改进YOLOv5特征金字塔网络结构,融合浅层网络中的特征图,增加模型对图像低维有效信息的表达能力,进而提升小目标检测精度.同时为了降低航拍图像中复杂背景带来的干扰,引入无参平均注意力模块,同时关注图像的空间注意力与通道注意力;引入VariFocal Loss,降低负样本在训练过程中的权重占比.在VisDrone数据集上的实验验证文中模型的有效性,该模型在有效提升检测精度的同时明显降低复杂度.
文摘Drone or unmanned aerial vehicle(UAV)technology has undergone significant changes.The technology allows UAV to carry out a wide range of tasks with an increasing level of sophistication,since drones can cover a large area with cameras.Meanwhile,the increasing number of computer vision applications utilizing deep learning provides a unique insight into such applications.The primary target in UAV-based detection applications is humans,yet aerial recordings are not included in the massive datasets used to train object detectors,which makes it necessary to gather the model data from such platforms.You only look once(YOLO)version 4,RetinaNet,faster region-based convolutional neural network(R-CNN),and cascade R-CNN are several well-known detectors that have been studied in the past using a variety of datasets to replicate rescue scenes.Here,we used the search and rescue(SAR)dataset to train the you only look once version 5(YOLOv5)algorithm to validate its speed,accuracy,and low false detection rate.In comparison to YOLOv4 and R-CNN,the highest mean average accuracy of 96.9%is obtained by YOLOv5.For comparison,experimental findings utilizing the SAR and the human rescue imaging database on land(HERIDAL)datasets are presented.The results show that the YOLOv5-based approach is the most successful human detection model for SAR missions.
基金supported by the State Key Program of the National Natural Science of China(Grant No.41730964)the National Key Research and Development Program on Monitoring,Early Warning and Prevention of Major Natural Disaster(2018YFC1506000)+2 种基金the National Natural Science Foundation of China(Grant Nos.41975091 and 42175047)National Basic Research Program of China(2015CB453203)UK-China Research&Innovation Partnership Fund through the Met Office Climate Science for Service Partnership(CSSP)China as part of the Newton Fund.
文摘The simulation and prediction of the climatology and interannual variability of the East Asia winter monsoon(EAWM),as well as the associated atmospheric circulation,was investigated using the hindcast data from Global Seasonal Forecast System version 5(GloSea5),with a focus on the evolution of model bias among different forecast lead times.While GloSea5 reproduces the climatological means of large-scale circulation systems related to the EAWM well,systematic biases exist,including a cold bias for most of China’s mainland,especially for North and Northeast China.GloSea5 shows robust skill in predicting the EAWM intensity index two months ahead,which can be attributed to the performance in representing the leading modes of surface air temperature and associated background circulation.GloSea5 realistically reproduces the synergistic effect of El Niño–Southern Oscillation(ENSO)and the Arctic Oscillation(AO)on the EAWM,especially for the western North Pacific anticyclone(WNPAC).Compared with the North Pacific and North America,the representation of circulation anomalies over Eurasia is poor,especially for sea level pressure(SLP),which limits the prediction skill for surface air temperature over East Asia.The representation of SLP anomalies might be associated with the model performance in simulating the interaction between atmospheric circulations and underlying surface conditions.
基金Supported by the National Key Research and Development Program of China(2017YFC1502303)National Natural Science Fundation of China(41730964,41975091,and 41605078)UK–China Research and Innovation Partnership Fund through the Met Office Climate Science for Service Partnership(CSSP)China as part of the Newton Fund。
文摘The western North Pacific subtropical high(WNPSH) dominates the summer climate over East Asia. The intensity,position, and shape of WNPSH influence the spatiotemporal distributions of precipitation, temperature, and tropical cyclone activities in this region. This paper intends to investigate the performance of the UK Met Office Global Seasonal forecast system version 5(GloSea5) in simulation/prediction of the WNPSH based on a hindcast dataset. Analyses of the hindcast data show a systematic bias in the mean circulation over West Pacific, with negative geopotential height anomalies over the western North Pacific(WNP) and cyclonic anomalies in the 850-hPa winds and water vapor transport, indicating a weakening and eastward shift of the WNPSH. Despite the model’s bias in the climatology, it well captured the interannual variability of the monthly and seasonal-mean intensity of the WNPSH and the position of its ridge line in boreal summer from 1993 to 2015. The seasonal hindcasts indicate that there is significant prediction skill at up to three-month lead time for both the intensity and position of the WNPSH ridge line. The relationship between the WNPSH and different phases of the El Nino–Southern Oscillation(ENSO) in both the observational data and GloSea5 hindcasts was then investigated. The model captured the summer WNPSH anomalies well during most of the ENSO phases, except in the La Nina decaying and neutral summers. The intensity of the anticyclone in the WNP is weak in the decaying phase of El Nino in the GloSea5 hindcasts compared with the reanalysis data. GloSea5 is capable of representing the lagged teleconnection between El Nino events in the previous winter and the intensity of the WNPSH in the following summer. Regression analysis reveals weakened negative sea surface temperature anomalies(SSTAs) over the WNP in GloSea5, which reduced the gradient between the tropical western Pacific and the tropical Indian Ocean, resulting in a weaker easterly anomaly and stronger westerly anomaly, contributing to the weak anomalous anticyclone over the WNP and the weakened WNPSH relative to the reanalysis data.
基金This paper is supported by the National Key Project of Basic Theory Research"the Formation Mechanism and Prediction Theory of Severe Climatic and Synoptic Disasters in China" under Grant 199804096.
文摘The sounding data of meteorological satellites provide not only the real time weather information about the distribution of both cloud and rainfall,but also some others about the movement and state of atmosphere.They are important variables and parameters for NWP model used to simulate and predict atmospheric state.In order to introduce remote sensing information from satellites into NWP model,there is an efficient way of establishing an RT model by use of the atmosphere radiation sounding data of meteorological satellites to get the variables and parameters valuable to NWP model.In this paper,we set up profiles of air temperature and water vapor from the surface to upper (0.1 hPa) using the radiosounding data and the surface data from May to August 1998 atmosphere East Asia.A TOVS RT model (RTTOV5) is provided to compute the value of radiation value of HIRS channels in NOAA14.Then the radiation values of 19 HIRS channels are gotten.After matching these data computed by the RT model and the corresponding values coming from satellite sounding in time,the statistic distribution of bias between tile model output and the satellite sounding at each sounding channel can be gotten.At the same time.the distribution of RMS to every TOVS HIRS channel,the standard biases to different scanning angle to each channel are also obtained.
基金Supported by the Program of "Research on the Formation MechanismPrediction Theory of Severe Synoptic Disasters in China" (G1998040907).
文摘The track, landfall, dynamic and thermodynamic and cloud-rain physical mesoscale structures and their evolution of typhoon HERB 1996 in 36 h from 0000 UTC 31 July to 1200 UTC 1 August 1996 were simulated by using the non-hydrostatic mesoscale model MM5. This period covered the process of typhoon HERB landfall at Taiwan and Fujian Provinces. Results show that the model successfully simulated the landfall process of typhoon HERB, revealed the most important characteristics of the mesoscale dynamic and thermodynamic and cloud-rain physical structure during its landfall. The simulated typhoon track was close to the observation. The center of cyclonic circulation simulated at 0000 UTC on 1 August 1996 (24 h integration) was located in shore near Fuqing, Fujian Province at which the typhoon was reported to landfall two hours later. It shows that strong upward motion formed by low level convergence existed in the eye-wall and subsidence at the eye. The wind field shows clear asymmetrical structure near the typhoon center. The cloud and rainband was screw-typed distributed around typhoon center, and consisted of meso-β scale rain cores. During the period of typhoon HERB staying near and passing over Taiwan, the lower cloud was developed in the eye region so that the previous clear typhoon eye on the satellite pictures became fuzzy. Observation shows that the typhoon center was 'warm', but the model simulations with higher space resolution show that in the mid-troposphere the region of eye-wall with stronger upward motion and more cloud-and rain- water was warmer than the eye. During the period of typhoon passing over Taiwan and its following landfall at Fujian, the track of model typhoon deviated about 30 km northward (i. e., rightward) because of the orographic effects of Taiwan Island, but the strength of the typhoon was not affected remarkably. The amount of rainfall on Taiwan in the 36 h simulations was enhanced more than six times by the orographic lifting of Taiwan Mountain.
基金National Natural Science Foundation of China under Grants 49375246.
文摘Diagnosis is performed of a thunderstorm rainstorm event occurring in the summer of 1996 at Nanjing and numerical simulation undertaken in the context of hydrostatic equilibrium framework of MM5 as the fifth version of the PSU/NCAR mesoscale model.Analyses show that the rainstorm-associated thermal condition was the accumulation of unstable potential energy and the dynamic condition was vigorous convergence updrafts.And the simulation within the hydrostatic framework indicates the significant role of latent heat release in the rainstorm occurrence:that even for a 30 km grid spacing horizontally of great importance to the successful modeling of the meso-β event was a convection parameterization scheme that led to less rainfall in our run based only on its explicit version but to the prediction in closer agreement with the observed when its implicit version was used in combination:for the thunderstorm-accompanied torrential rain.the Grell scheme was superior to the version of Kuo and the improved Arakawa-Schubert parameterization scheme(Grell 1993:Anthese and Kuo 1987:Arakawa and Scherbt 1974:Grell et al,1991).Moreover,better results came from the simulation in the context of hydrostatic framework of the MM5 compared to those from the run within the nonhydrostatic equilibrium framework,a problem that awaits further efforts.
