Conformal Detection Of Anomalous Behaviors Of Vessel

At present,the ship abnormal behavior detection algorithm does not take into account the influence of the single ship's maneuvering pattern and ships interaction,which results in the detection of the ship abnormal behavior type being limited.In order to solve this question,this thesis takes the ship behavior as the research object,which adopts vessel behavior analysis technology based on trajectory data and conformal anomaly detection method to study the detection problem of abnormal vessel behavior.The main work of this thesis is as follows:A method for identifying the ship motion pattern based on one-way distance and spectral clustering is proposed.This method uses the one-way distance to measure the similarity between the trajectories to obtain the trajectory similarity matrix.Then the spectrum clustering algorithm is used to learn the spatial distribution of the trajectory in an unsupervised way,so as to identify the ship motion pattern.In this thesis,taking the measured AIS data of the Qiongzhou Strait as a sample,four kinds of ship movement patterns entering Xiuying Port of Haikou were identified,and the distances in modes and between modes were calculated.The results show that this method is practical and effective,and it can be applied to the coastal ports,narrow waterways and areas with complex ship traffic.The proposed method is robust and can run automatically in an unsupervised manner.A method for automatically identifying the single-vessel operation mode that is focused on from the ship AIS trajectory data is proposed.This method utilizes three adjacent trajectory points in the trajectory.According to the rules of ship maneuvering and kinematic constraints,eight kinds of ship maneuvering mode models are constructed,such as residence,uniform motion,acceleration,deceleration,small margin left turn,small right turn,large left turn and large right turn for further detecting the abnormal behavior of the ship.The experimental results are consistent with the actual ship's intention,indicating that this method can effectively identify the maneuvering behavior.This method has the advantages of automatic identification and meticulous description of ship maneuvering pattern.Aiming at the problem that the abnormal ship maneuvering behavior detection is limited because of neglecting the influence the ship maneuvering pattern,conformal anomaly algorithm detection for single ship behavior is proposed.By introducing the entry speed and the exit speed in the characteristics of the ship's trajectory point,the similarity of the maneuvering behavior was measured from the altering course behavior and the changing speed behavior.In order to compare the ship behavior similarity more accurately,the similarity of ship's comprehensive behavior was defined.The nearest neighbors to the points of the track from the training trajectory to be measured are respectively found,and along with the track points to be measured to construct a conformal detection sequence.The randomness test values are used to predict whether the track points to be measured is abnormal.Taking the measured AIS data from the Qiongzhou Strait as normal data,abnormal trajectories and artificially controlled abnormal behaviors were randomly generated by computer simulations as abnormal data.Anomaly detection tests were conducted separately.The probability of abnormal trajectories randomly generated were detected is 100%,but some trajectories in the trajectory evaluation set were incorrectly divided into abnormal trajectories.In the case of specified confidence levels were 99.0%and 99.7%,the false alarm rate respectively was 0.6%and 0.2%,which lower than the significance levels of 0.01 and 0.003 respectively.The results show that using the conformal detection algorithm can effectively detect the abnormal trajectory,and can strictly control the detection false alarm rate by specifying the level of significance.The proposed algorithm has the advantages of controllable false alarm rate and detection of subtle abnormal behaviors(such as slow braking and small steering)in ship maneuvering behavior.A conformal anomaly detection method for multi-ship situation is proposed.Field theory is used to calculate the propagation collision risk,and it is used as a non-uniform measure function of the consistency detector to detect the abnormal behavior in multi-ship situation.The ship trajectory in the real AIS trajectory set data was used as a research sample to carry out detection test in multi-ship situation respectively.The results show that under the premise of ensuring that the false alarm rate is controllable,this method can not only identify the abnormal behavior of the ship in the multi-ship situation,but also perform collision warning.The proposed method is with some advantages:controllable false alarm rate,and ease to implementation in practice,and early warning of ship collision.