Remote sensing data plays an important role in natural disaster management.However,with the increase of the variety and quantity of remote sensors,the problem of“knowledge barriers”arises when data users in disaster...Remote sensing data plays an important role in natural disaster management.However,with the increase of the variety and quantity of remote sensors,the problem of“knowledge barriers”arises when data users in disaster field retrieve remote sensing data.To improve this problem,this paper proposes an ontology and rule based retrieval(ORR)method to retrieve disaster remote sensing data,and this method introduces ontology technology to express earthquake disaster and remote sensing knowledge,on this basis,and realizes the task suitability reasoning of earthquake disaster remote sensing data,mining the semantic relationship between remote sensing metadata and disasters.The prototype system is built according to the ORR method,which is compared with the traditional method,using the ORR method to retrieve disaster remote sensing data can reduce the knowledge requirements of data users in the retrieval process and improve data retrieval efficiency.展开更多
针对消防设施选址问题,构建考虑时效性、市民等待救援的焦急心理和建设成本的三目标消防设施选址模型,以实现更科学的消防设施布局。鉴于该问题的NP难特性,提出基于算子学习的多目标深度强化学习模型(multi-objective deep reinforcemen...针对消防设施选址问题,构建考虑时效性、市民等待救援的焦急心理和建设成本的三目标消防设施选址模型,以实现更科学的消防设施布局。鉴于该问题的NP难特性,提出基于算子学习的多目标深度强化学习模型(multi-objective deep reinforcement learning,MDRL)。设计多种优化算子作为强化学习的动作空间,训练策略网络以选择最佳优化算子来改进解决方案。针对多目标问题,设计基于优势差异的方法(MDRL-AD)和基于支配性评估的方法(MDRL-DE)。采用四种规模的测试算例及实际案例进行数值实验,将MDRL和改进的NSGA-Ⅱ、MOPSO、L2I算法进行比较,并利用Hypervolume指标、Spacing指标、Ω指标、IGD指标对算法性能进行评估。实验结果表明,MDRL-AD方法更适用于求解小规模算例,MDRL-DE方法则在求解大规模和超大规模算例时相比其他算法优势明显。MDRL在非劣解集的收敛性和均匀性方面明显优于其他对比算法,为消防设施布局规划提供了一种有竞争力的解决方案。展开更多
基金supported by the National Key Research and Development Program of China(2020YFC1512304).
文摘Remote sensing data plays an important role in natural disaster management.However,with the increase of the variety and quantity of remote sensors,the problem of“knowledge barriers”arises when data users in disaster field retrieve remote sensing data.To improve this problem,this paper proposes an ontology and rule based retrieval(ORR)method to retrieve disaster remote sensing data,and this method introduces ontology technology to express earthquake disaster and remote sensing knowledge,on this basis,and realizes the task suitability reasoning of earthquake disaster remote sensing data,mining the semantic relationship between remote sensing metadata and disasters.The prototype system is built according to the ORR method,which is compared with the traditional method,using the ORR method to retrieve disaster remote sensing data can reduce the knowledge requirements of data users in the retrieval process and improve data retrieval efficiency.
文摘针对消防设施选址问题,构建考虑时效性、市民等待救援的焦急心理和建设成本的三目标消防设施选址模型,以实现更科学的消防设施布局。鉴于该问题的NP难特性,提出基于算子学习的多目标深度强化学习模型(multi-objective deep reinforcement learning,MDRL)。设计多种优化算子作为强化学习的动作空间,训练策略网络以选择最佳优化算子来改进解决方案。针对多目标问题,设计基于优势差异的方法(MDRL-AD)和基于支配性评估的方法(MDRL-DE)。采用四种规模的测试算例及实际案例进行数值实验,将MDRL和改进的NSGA-Ⅱ、MOPSO、L2I算法进行比较,并利用Hypervolume指标、Spacing指标、Ω指标、IGD指标对算法性能进行评估。实验结果表明,MDRL-AD方法更适用于求解小规模算例,MDRL-DE方法则在求解大规模和超大规模算例时相比其他算法优势明显。MDRL在非劣解集的收敛性和均匀性方面明显优于其他对比算法,为消防设施布局规划提供了一种有竞争力的解决方案。