An on-line prediction method for ship navigation behavior

Abstract

The invention provides an on-line prediction method of ship navigation behavior, belonging to the ship navigation field. The method includes training learning stage and on-line prediction stage. The training learning stage sorts and divides the historical AIS big data first, and then trains and learns the intercepted data by using two-way long-short-time memory loop neural network. In the on-lineprediction stage, the AIS receiver is used to collect real-time AIS data, and then the sliding window algorithm is used to obtain the key feature points of ship trajectory, and then the future six consecutive ship trajectory points are predicted, and the final six consecutive future ship trajectory points are used as the predicted ship sailing behavior after multiple predictions. The mature modeltrained by big data has a wide application range and high versatility, which can be used in the later practical application to further enhance the prediction ability of the frequently predicted ships,and can be used in the ship intelligent collision avoidance decision-making and ship abnormal behavior detection, to provide services for the water traffic management department.

Description

technical field [0001] The invention belongs to the field of ship navigation and relates to an online prediction method for ship navigation behavior. Background technique [0002] In the study of ship intelligent collision avoidance, in order to generate more reliable and effective collision avoidance decisions, the types of information sources are not limited to the current position, course, speed and other factors of the two ships, but also consider the navigation intention of other ships. The following ship movement and the possible position of other ships in the future make the overall intelligent collision avoidance system a priori and predictable, which can make the final collision avoidance decision more reliable, effectively reduce the risk of collision, and avoid resulting in personal and property damage. [0003] At the same time, when the port supervision department monitors the safety of the managed area, it needs to dynamically identify the abnormal behavior of...

AI Technical Summary

Problems solved by technology

However, the selection of the kernel function of this model relies too much on experience and a large amount of prior knowledge, which leads to a very high error rate, high requirements for the quality of learning samples, and the algorithm cannot be updated online. The model after learning is relatively fixed and flexi

Images