基于轨迹聚类的船舶异常行为识别研究

发布时间：2018-10-11 17:09

【摘要】：海上交通监控对于在航船舶的航行安全具有重要意义。船舶AIS的强制安装及沿海VTS的建立给海事监管部门带来极大的便捷,但海事部门对海上交通监控的主要方式仍为人工监控,该方法费时费力,且缺乏针对性,尤其在一些繁忙港口仅仅依靠人工监控很难满足港口安全需求。为了对船舶行为进行实时监控并自主发现异常行为船舶,提出了一种基于轨迹聚类的船舶异常行为识别模型。针对传统Hausdorff距离在度量轨迹不等长及轨迹点丢失情况下存在度量距离加大的问题,优化Hausdorff距离度量方式,将传统度量最邻近点距离改进为度量轨迹点到最邻近两点所在直线的垂直距离,使船舶轨迹间距离度量更加准确;将新提出的密度峰聚类算法应用到航海领域并对船舶轨迹进行聚类,该聚类方法不需要人为设置参数,可以避免人为因素的干扰;设置扫描线对聚类后的每一簇船舶轨迹进行扫描,获得船舶典型轨迹模型;统计每条船舶轨迹与船舶典型轨迹模型的距离、航向和航速偏差,根据准确率和误报警率来确定最优偏差阈值,若待测船舶偏差超过设置的阈值则识别为异常行为,从而达到智能识别出船舶异常行为的效果。通过对厦门港VTS监控中心发现的船舶异常行为案例进行验证,实验结果表明提出的算法模型能自主、有效识别船舶异常行为,并与值班人员相比能更早发现船舶行为发生异常,可为值班人员及早发现船舶异常行为提供参考依据。
[Abstract]:Maritime traffic monitoring plays an important role in the safety of navigation. The mandatory installation of ship AIS and the establishment of coastal VTS bring great convenience to maritime supervision department, but the main way of maritime traffic monitoring is manual monitoring, which is time-consuming and laborious, and lacks pertinence. Especially in some busy ports only rely on manual monitoring is difficult to meet port security needs. In order to monitor the ship's behavior in real time and find out the abnormal behavior of the ship independently, a ship abnormal behavior identification model based on trajectory clustering is proposed. Aiming at the problem that the traditional Hausdorff distance has the problem of increasing the measurement distance under the condition of the unequal length of the measuring path and the loss of the locus point, the Hausdorff distance measurement method is optimized. The distance between the most adjacent points is improved to measure the vertical distance between the trajectory points and the nearest two points, which makes the distance measurement of ship trajectory more accurate. The new density peak clustering algorithm is applied to the navigation field and the ship trajectory is clustered. This clustering method does not need to set the parameters artificially, so it can avoid the interference of human factors. Scanning lines are set to scan each cluster of ship trajectories after clustering, and the typical ship trajectory model is obtained, and the distance, course and speed deviation between each ship trajectory and the typical ship trajectory model are calculated. According to the accuracy and false alarm rate, the optimal deviation threshold is determined. If the deviation of the ship under test exceeds the set threshold, it will be recognized as abnormal behavior, so as to achieve the effect of intelligently recognizing the abnormal behavior of the ship. The experimental results show that the proposed algorithm model can identify the abnormal behaviors of ships independently and effectively, and can detect the abnormal behaviors of ships earlier than those of the watchmen, through the verification of the cases of abnormal behaviors of ships discovered by the VTS Monitoring and Control Center of Xiamen Port, and the experimental results show that the proposed algorithm can identify the abnormal behaviors of ships. It can provide reference basis for early detection of abnormal behavior of ship by duty personnel.
【学位授予单位】：集美大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U698

【参考文献】