In this study, marine casualties which are recorded Global Integrated Shipping Information System （GISIS） in 2007-2011, result in death, injury, economic loss and environmental pollution are discussed. The studied t...In this study, marine casualties which are recorded Global Integrated Shipping Information System （GISIS） in 2007-2011, result in death, injury, economic loss and environmental pollution are discussed. The studied types of ships are container, bulk dry, general cargo, roro, ropax, passenger and tankers. The information contained in GISIS system is textual format and it is difficult to systematically analyze this information. For this reason, by creating a new data base, which is Microsoft Excel-based, ship accidents are classified according to name, flag state and type of ship, type, size and coordinates of the accidents that are evaluated by entering the Geographical Information System（GIS）. In the study, all marine areas having been separated into ranges in ArcGIS 10 program, the marine areas with intensive marine accidents have been determined and marine accident chart has been created. As a result of the study, high risk marine areas are Strait of Dover and Hamburg in the North Europe, Belfast Shores in Ireland, the seas surrounding Great Belt, Kattegat and Copenhagen in Baltic Sea, In far east, Kanmon Strait, Urage Channel and Bungo Strait in Japan, Shanghai, Ningbo and Hong Kong in China.展开更多
船舶行为特征挖掘与预测是水上智能交通系统的重要研究内容,也是交通运输工程领域的关键科学问题。为系统研究基于船舶自动识别系统(Automatic Identification System,AIS)数据的船舶行为特征挖掘与预测的研究现状与发展趋势,本文首先针...船舶行为特征挖掘与预测是水上智能交通系统的重要研究内容,也是交通运输工程领域的关键科学问题。为系统研究基于船舶自动识别系统(Automatic Identification System,AIS)数据的船舶行为特征挖掘与预测的研究现状与发展趋势,本文首先针对Web of Science(WOS)和中国知网(China National Knowledge Infrastructure,CNKI)收录的文献,用知识图谱分析软件VOSviewer对文献关键词进行处理,从文献计量学的角度生成高频关键词的聚类图谱和趋势演化。然后对基于AIS数据的水上交通要素挖掘、船舶行为聚类和船舶行为预测3个主题的研究内容、方法、存在问题进行了系统分析和展望,研究结果表明:①在基于AIS的水上交通要素挖掘方面,主要集中在对AIS数据中表征船舶行为空间特征和交通流的时间特征单独挖掘分析,缺乏对AIS数据的时间、空间以及环境因素特征的关联挖掘,对于如何进行交通要素的关联融合挖掘研究还有待深入探索;②在船舶行为聚类方面,研究主要是运用无监督聚类方法研究船舶航迹点和航迹段聚类,得到船舶航行行为模式的时空分布和船舶操纵意图辨识模型,然而融合多维特征的船舶轨迹的相似性计算方法、聚类参数的自适应选取以及船舶行为的语义特征建模有待进一步研究;③在船舶行为预测方面,主要集中在基于动力学方程、传统智能算法和深度循环神经网络的船舶行为预测研究,考虑船舶行为的随机性、多样性和耦合性的特点,运用混合神经网络模型以及神经网络与向量机、注意力机制相结合的模型实现多维的船舶航行行为特征的实时预测将是新的研究方向。最后提出了基于语义模型的船舶行为特征挖掘、基于深度卷积神经网络的船舶行为的预测和基于知识图谱的船舶行为特征挖掘和预测结果可视化等有待进一步研究的方向。展开更多
文摘In this study, marine casualties which are recorded Global Integrated Shipping Information System （GISIS） in 2007-2011, result in death, injury, economic loss and environmental pollution are discussed. The studied types of ships are container, bulk dry, general cargo, roro, ropax, passenger and tankers. The information contained in GISIS system is textual format and it is difficult to systematically analyze this information. For this reason, by creating a new data base, which is Microsoft Excel-based, ship accidents are classified according to name, flag state and type of ship, type, size and coordinates of the accidents that are evaluated by entering the Geographical Information System（GIS）. In the study, all marine areas having been separated into ranges in ArcGIS 10 program, the marine areas with intensive marine accidents have been determined and marine accident chart has been created. As a result of the study, high risk marine areas are Strait of Dover and Hamburg in the North Europe, Belfast Shores in Ireland, the seas surrounding Great Belt, Kattegat and Copenhagen in Baltic Sea, In far east, Kanmon Strait, Urage Channel and Bungo Strait in Japan, Shanghai, Ningbo and Hong Kong in China.
文摘船舶行为特征挖掘与预测是水上智能交通系统的重要研究内容,也是交通运输工程领域的关键科学问题。为系统研究基于船舶自动识别系统(Automatic Identification System,AIS)数据的船舶行为特征挖掘与预测的研究现状与发展趋势,本文首先针对Web of Science(WOS)和中国知网(China National Knowledge Infrastructure,CNKI)收录的文献,用知识图谱分析软件VOSviewer对文献关键词进行处理,从文献计量学的角度生成高频关键词的聚类图谱和趋势演化。然后对基于AIS数据的水上交通要素挖掘、船舶行为聚类和船舶行为预测3个主题的研究内容、方法、存在问题进行了系统分析和展望,研究结果表明:①在基于AIS的水上交通要素挖掘方面,主要集中在对AIS数据中表征船舶行为空间特征和交通流的时间特征单独挖掘分析,缺乏对AIS数据的时间、空间以及环境因素特征的关联挖掘,对于如何进行交通要素的关联融合挖掘研究还有待深入探索;②在船舶行为聚类方面,研究主要是运用无监督聚类方法研究船舶航迹点和航迹段聚类,得到船舶航行行为模式的时空分布和船舶操纵意图辨识模型,然而融合多维特征的船舶轨迹的相似性计算方法、聚类参数的自适应选取以及船舶行为的语义特征建模有待进一步研究;③在船舶行为预测方面,主要集中在基于动力学方程、传统智能算法和深度循环神经网络的船舶行为预测研究,考虑船舶行为的随机性、多样性和耦合性的特点,运用混合神经网络模型以及神经网络与向量机、注意力机制相结合的模型实现多维的船舶航行行为特征的实时预测将是新的研究方向。最后提出了基于语义模型的船舶行为特征挖掘、基于深度卷积神经网络的船舶行为的预测和基于知识图谱的船舶行为特征挖掘和预测结果可视化等有待进一步研究的方向。
