地球信息科学学报 ›› 2022, Vol. 24 ›› Issue (9): 1676-1687.doi: 10.12082/dqxxkx.2022.210455

• 地理信息技术在海事领域的应用 • 上一篇    下一篇

基于厦门港的海上交通事故地理空间分布及风险预测研究

杨洋1(), 邵哲平1,2, 赵强1,2, 潘家财1,2, 胡雨4, 梅强1,2,3,*()   

  1. 1.集美大学航海学院,厦门 361021
    2.船舶辅助导航技术国家地方联合工程研究中心,厦门 361021
    3.上海海事大学商船学院,上海 201306
    4.中科(厦门)数据智能研究院,厦门 361021
  • 收稿日期:2021-08-04 修回日期:2021-12-14 出版日期:2022-09-25 发布日期:2022-11-25
  • 通讯作者: *梅 强(1987— ),男,安徽宿州人,助教,主要从事海事地理信息分析研究等。E-mail: meiqiang@jmu.edu.cn
  • 作者简介:杨 洋(1994— ),女,四川眉山人,硕士生,主要从事海事地理信息研究。E-mail: 1042280675@qq.com
  • 基金资助:
    国家自然科学基金项目(71804059);福建省自然科学基金项目(2021J01821);福建省教育厅面上项目(JAT200265);船舶辅助导航技术国家地方联合工程研究中心开放基金(JMCBZD202002)

Geographical Spatial Distribution and Risk Prediction of Maritime Traffic Accidents in Port of Xiamen

YANG Yang1(), SHAO Zheping1,2, ZHAO Qiang1,2, PAN Jiacai1,2, HU Yu4, MEI Qiang1,2,3,*()   

  1. 1. Navigation College, Jimei University, Xiamen 361021, China
    2. National-Local Joint Engineering Research Center for Aids to Navigation, Xiamen 361021, China
    3. Merchant Marine College, Shanghai Maritime University, Shanghai 201306, China
    4. Xiamen Data Intelligence Academy of Chinese Academy Sciences, ICT, Xiamen 361021, China
  • Received:2021-08-04 Revised:2021-12-14 Online:2022-09-25 Published:2022-11-25
  • Supported by:
    National Natural Science Foundation of China(71804059);Natural Science Foundation of Fujian Province(2021J01821);Science and Technology Project of Education Department of Fujian Province(JAT200265);Open Foundation of National-Local Joint Engineering Research Center for Aids to Navigation(JMCBZD202002)

摘要:

利用海上交通事故空间分布特征进行安全分析是海上交通安全管理的重要组成部分。本文使用厦门港2008—2020年的海上交通事故数据,经过事故数据空间分布特征提取、分析及预测等流程,最终得到厦门水域海上交通事故潜在危险区域。本文首先使用原始事故数据在GIS软件中进行空间定位,形成事故点的可视化空间分布图,然后使用核密度分析法鉴别海上交通事故多发区域,再利用空间自相关分析法,得到该区域事故空间的分布特征和具体的聚集点,最后使用该分布特征、对目标水域数据进行标准化网格切分,并利用机器学习算法对潜在事故风险区域进行预测。本文在核密度分析结果中发现:就事故频度而言,厦门湾和西海域交通事故频度较高。在空间自相关分析的结果中表明:就空间分布特征而言,厦门港的空间分布出现聚集特征且为空间正相关模式,且就事故具体的空间聚集点而言,厦门湾和西海域仍是事故高发的中心区域。而最后的厦门湾及周边水域风险预测模型显示:潜在事故风险区域多位于沿海和河口交汇区域。本文研究结果表明在基于地理空间数据分布特征提取和网格化分析的基础上,结合机器学习方法(随机森林),对于海上交通事故的预测具有良好的效果。

关键词: 海上交通事故, 空间分布特征, 核密度分析, 空间自相关分析, 事故预测, 随机森林, 厦门港, 风险

Abstract:

Safety analysis based on the spatial distribution of maritime traffic accidents is of great significance for maritime traffic safety management. In this paper, we obtained the potential risk areas of maritime traffic accidents in Xiamen waters through extraction, analysis, and prediction of spatial distribution of maritime traffic accident data in Xiamen Port from 2008 to 2020. The original accident data were firstly used for spatial orientation in GIS software to form a visualized spatial distribution map of accident points. Then, the areas where maritime traffic accidents frequently occur were identified using the kernel density analysis. In addition, the spatial autocorrelation analysis method was employed to obtain the spatial distribution characteristics and specific gathering points of accidents in this region. Finally, we performed standardized grid segmentation on the data in the target waters based on the spatial distribution characteristics and predicted potential risk areas of maritime traffic accidents using a machine learning algorithm (i.e., random forest). The kernel density analysis results showed that Xiamen Bay and Western Sea were high risk areas in terms of accident frequency. Moreover, the spatial autocorrelation analysis results indicated that in terms of spatial distribution characteristics, the spatial distribution of Xiamen Port showed aggregation characteristics and was positively correlated with time. For specific spatial gathering points of accidents, Xiamen Bay and Western Sea were still the centers with a high incidence of accidents. Furthermore, the risk prediction model of Xiamen Bay and its surrounding waters demonstrated that potential risk areas of maritime traffic accidents were mostly located at the intersection of coastal and estuary areas, because there were frequent crossings of vessels and fishing boats. The results of this study show that based on the distribution characteristics extracted and grid analysis of geospatial data, our proposed method has a good effect in predicting maritime traffic accidents by use of machine learning methods.

Key words: maritime traffic accidents, spatial distribution characteristics, kernel density analysis, spatial auto-correlation analysis, accident forecast, random forest, Xiamen port, risk