船舶在避碰状态下开展异常轨迹点数据动态挖掘时，若不能及时了解避碰危险等级，会降低船舶异常轨迹点数据挖掘的挖掘性能。为提升异常轨迹点数据动态挖掘精度，提出面向船舶避碰的异常轨迹点数据动态挖掘方法。首先对船舶航行避碰危险程度展开具体分析，并根据分析结果采集船舶异常轨迹点数据，通过小波去噪算法完成船舶异常轨迹点数据的去噪处理。再进行数据去噪处理，提取轨迹点船舶的各项参数特征，结合长短期记忆网络构建船舶异常轨迹点的数据动态挖掘模型。最后将提取的特征向量实施变异赋值，将其赋值结果作为模型输入值输入模型中。根据模型输出实现船舶避碰情况下，异常轨迹点数据的动态挖掘。实验结果表明，使用该方法开展船舶异常轨迹点数据动态挖掘时，挖掘性能较高，挖掘效果好。
When the ship carries out the dynamic mining of abnormal trajectory point data in the collision avoidance state, if the collision avoidance risk level of the ship cannot be known in time, the mining performance of the abnormal trajectory point data mining of the ship will be reduced. In order to improve the dynamic mining accuracy of abnormal trajectory point data, a dynamic mining method of abnormal trajectory point data for ship collision avoidance is proposed. Firstly, the risk degree of collision avoidance under ship navigation is analyzed in detail, and according to the analysis results, the abnormal trajectory point data of the ship is collected, and the denoising process of the abnormal ship trajectory point data is completed by the wavelet denoising algorithm. Then, the data is denoised to extract the parameters of the ship's trajectory points, and the long-term and short-term memory network is combined to build a data dynamic mining model for abnormal ship trajectory points. Finally, the extracted feature vector is subjected to mutation assignment, and the assignment result is input into the model as the model input value. In the case of ship collision avoidance based on the model output, the dynamic mining of abnormal trajectory point data is completed. The experimental results show that when the method is used to carry out dynamic mining of ship abnormal trajectory point data, the mining performance is high and the mining effect is good.
2022,44(22): 136-139 收稿日期：2022-08-18
DOI：10.3404/j.issn.1672-7649.2022.22.026
分类号：TP311
基金项目：河南省省级项目（22B470007；2018GGJS298）；河南省教育规划课题（2020YB0592）
作者简介：耿瑞焕(1986-),女,硕士,讲师,研究方向为信息处理及数据挖掘
参考文献：
[1] 范爱龙, 李方轩. 基于实船监测的内河船舶能效数据特征挖掘及建模研究[J]. 武汉理工大学学报, 2020, 42(6): 26–34
[2] 周海, 陈姚节, 陈黎. 船舶轨迹聚类分析与应用[J]. 计算机仿真, 2020, 37(10): 113–118+199
[3] 蒋通, 崔良中, 周钢, 等. 多步骤船舶轨迹聚类方法研究与实现[J]. 舰船电子工程, 2021, 41(9): 53–57+92
[4] 李倍莹, 张新宇, 沈忱, 等. 基于改进K中心点聚类的船舶典型轨迹自适应挖掘算法[J]. 上海海事大学学报, 2021, 42(3): 15–22
