This paper considers an underwater acoustic sensor network with one mobile surface node to collect data from multiple underwater nodes,where the mobile destination requests retransmission from each underwater node ind...This paper considers an underwater acoustic sensor network with one mobile surface node to collect data from multiple underwater nodes,where the mobile destination requests retransmission from each underwater node individually employing traditional automatic-repeat-request(ARQ) protocol.We propose a practical node cooperation(NC) protocol to enhance the collection efficiency,utilizing the fact that underwater nodes can overhear the transmission of others.To reduce the source level of underwater nodes,the underwater data collection area is divided into several sub-zones,and in each sub-zone,the mobile surface node adopting the NC protocol could switch adaptively between selective relay cooperation(SRC) and dynamic network coded cooperation(DNC) .The difference of SRC and DNC lies in whether or not the selected relay node combines the local data and the data overheard from undecoded node(s) to form network coded packets in the retransmission phase.The NC protocol could also be applied across the sub-zones due to the wiretap property.In addition,we investigate the effects of different mobile collection paths,collection area division and cooperative zone design for energy saving.The numerical results showthat the proposed NC protocol can effectively save energy compared with the traditional ARQ scheme.展开更多
针对传统目标检测方法在水下识别任务中误检率较高的问题,基于一阶段全卷积检测器(FCOS)引入多尺度特征选择及中心边界特征选择,实现高精度水下目标检测。模型中的自适应加权融合特征金字塔通过设置可学习权重加权融合所有的特征层级,...针对传统目标检测方法在水下识别任务中误检率较高的问题,基于一阶段全卷积检测器(FCOS)引入多尺度特征选择及中心边界特征选择,实现高精度水下目标检测。模型中的自适应加权融合特征金字塔通过设置可学习权重加权融合所有的特征层级,实现多尺度空间特征选择。此外,为了处理检测中分类和回归任务之间的特征耦合问题,并分离不同任务之间的共享特征,设计了基于空间特征解耦的检测头,实现了中心和边界区域的特征选择。实验中,针对水下数据集URPC2018和UWD2021进行性能测试,并与先进的目标检测方法进行对比。大量的实验结果表明,基于空间特征选择的FCOS模型在水下检测任务中展现出优异的性能,在URPC2018和UWD2021上的类平均精度(mean Average Precision,mAP)分别为82.7%和83.3%。展开更多
With increasing urbanization and agricultural expansion, large tracts of wetlands have been either disturbed or converted to other uses. To protect wetlands, accurate distribution maps are needed. However, because of ...With increasing urbanization and agricultural expansion, large tracts of wetlands have been either disturbed or converted to other uses. To protect wetlands, accurate distribution maps are needed. However, because of the dramatic diversity of wetlands and difficulties in field work, wetland mapping on a large spatial scale is very difficult to do. Until recently there were only a few high resolution global wetland distribution datasets developed for wetland protection and restoration. In this paper, we used hydrologic and climatic variables in combination with Compound Topographic Index （CTI） data in modeling the average annual water table depth at 30 arc-second grids over the continental areas of the world except for Antarctica. The water table depth data were modeled without considering influences of anthropogenic activities. We adopted a relationship between poten- tial wetland distribution and water table depth to develop the global wetland suitability distribution dataset. The modeling re- suits showed that the total area of global wetland reached 3.316× 10^7 km^2. Remote-sensing-based validation based on a compi- lation of wetland areas from multiple sources indicates that the overall accuracy of our product is 83.7%. This result can be used as the basis for mapping the actual global wetland distribution. Because the modeling process did not account for the im- pact of anthropogenic water management such as irrigation and reservoir construction over suitable wetland areas, our result represents the upper bound of wetland areas when compared with some other global wetland datasets. Our method requires relatively fewer datasets and has a higher accuracy than a recently developed global wetland dataset.展开更多
基金supported in part by National Key Research and Development Program of China under Grants No.2016YFC1400200 and 2016YFC1400204National Natural Science Foundation of China under Grants No.41476026,41676024 and 41376040Fundamental Research Funds for the Central Universities of China under Grant No.220720140506
文摘This paper considers an underwater acoustic sensor network with one mobile surface node to collect data from multiple underwater nodes,where the mobile destination requests retransmission from each underwater node individually employing traditional automatic-repeat-request(ARQ) protocol.We propose a practical node cooperation(NC) protocol to enhance the collection efficiency,utilizing the fact that underwater nodes can overhear the transmission of others.To reduce the source level of underwater nodes,the underwater data collection area is divided into several sub-zones,and in each sub-zone,the mobile surface node adopting the NC protocol could switch adaptively between selective relay cooperation(SRC) and dynamic network coded cooperation(DNC) .The difference of SRC and DNC lies in whether or not the selected relay node combines the local data and the data overheard from undecoded node(s) to form network coded packets in the retransmission phase.The NC protocol could also be applied across the sub-zones due to the wiretap property.In addition,we investigate the effects of different mobile collection paths,collection area division and cooperative zone design for energy saving.The numerical results showthat the proposed NC protocol can effectively save energy compared with the traditional ARQ scheme.
文摘针对传统目标检测方法在水下识别任务中误检率较高的问题,基于一阶段全卷积检测器(FCOS)引入多尺度特征选择及中心边界特征选择,实现高精度水下目标检测。模型中的自适应加权融合特征金字塔通过设置可学习权重加权融合所有的特征层级,实现多尺度空间特征选择。此外,为了处理检测中分类和回归任务之间的特征耦合问题,并分离不同任务之间的共享特征,设计了基于空间特征解耦的检测头,实现了中心和边界区域的特征选择。实验中,针对水下数据集URPC2018和UWD2021进行性能测试,并与先进的目标检测方法进行对比。大量的实验结果表明,基于空间特征选择的FCOS模型在水下检测任务中展现出优异的性能,在URPC2018和UWD2021上的类平均精度(mean Average Precision,mAP)分别为82.7%和83.3%。
基金supported by National High-tech R&D Program of China (Grant No. 2009AA12200101)
文摘With increasing urbanization and agricultural expansion, large tracts of wetlands have been either disturbed or converted to other uses. To protect wetlands, accurate distribution maps are needed. However, because of the dramatic diversity of wetlands and difficulties in field work, wetland mapping on a large spatial scale is very difficult to do. Until recently there were only a few high resolution global wetland distribution datasets developed for wetland protection and restoration. In this paper, we used hydrologic and climatic variables in combination with Compound Topographic Index （CTI） data in modeling the average annual water table depth at 30 arc-second grids over the continental areas of the world except for Antarctica. The water table depth data were modeled without considering influences of anthropogenic activities. We adopted a relationship between poten- tial wetland distribution and water table depth to develop the global wetland suitability distribution dataset. The modeling re- suits showed that the total area of global wetland reached 3.316× 10^7 km^2. Remote-sensing-based validation based on a compi- lation of wetland areas from multiple sources indicates that the overall accuracy of our product is 83.7%. This result can be used as the basis for mapping the actual global wetland distribution. Because the modeling process did not account for the im- pact of anthropogenic water management such as irrigation and reservoir construction over suitable wetland areas, our result represents the upper bound of wetland areas when compared with some other global wetland datasets. Our method requires relatively fewer datasets and has a higher accuracy than a recently developed global wetland dataset.
