基于LOF-FCM算法的船舶航行数据识别

Integrated Detection Method for Ship Navigation Data Based on Machine Learning

针对传统船舶自动识别系统数据在清洗异常数据和提取停留数据时分别采用不同的识别方式、类型判断阈值需要人为设定、识别效率不佳的局限性,首次提出了一种船舶航行轨迹中停留及异常数据的一体化检测方法。通过分析航行路线的3种数据(停留、异常和航行)异常因子特征,提出基于LOF-FCM的船舶航行数据、停留数据和异常数据一体化检测算法。实验对3类数据进行了识别,模型识别准确率达到了92.69%,有效提高了异常、停留、航行数据的识别能力。结果表明所提方法可一次性实现AIS数据中3种数据的检测,能高效分离出正常船舶航行数据,具有良好的工程应用价值。

To address the limitations of traditional ship automatic identification system in using different recognition methods for cleaning abnormal data and extracting stay data,requiring manual setting of type judgment thresholds,and poor recognition efficiency,a comprehensive detection method for stay and abnormal data in ship navigation trajectories is proposed for the first time.By analyzing the abnormal characteristics of three types of data(stay,anomaly and navigation)on the navigation route,a ship navigation data,stay data,and anomaly data integrated detection algorithm based on LOF-FCM is proposed.The experiment identified three types of data,and the model recognition accuracy reached 92.69%,effectively improving the recognition ability of anomaly,stay,and navigation data.The results indicate that the proposed method can detect three types of data in automatic identification system(AIS)data at once,efficiently separate normal ship navigation data,and has good engineering application value.