Ship Abnormal Trajectory Recognition Based On Support Vector Machine

The development of abnormal trajectory detection technology has significant implications for ensuring safe navigation of ships,supporting intelligent regulation of maritime areas,and providing decision-making support.Support Vector Machines(SVM),with its small sample size,nonlinearity,and strong generalization ability,have been widely used in abnormal detection.However,the selection of the abnormal threshold and parameters is a crucial factor affecting the accuracy of identifying abnormal trajectory points.Existing abnormal detection algorithms often rely on prior knowledge to select abnormal thresholds.Therefore,this thesis proposes a ship abnormal trajectory recognition solution based on multiple SVMs to improve the accuracy of identifying abnormal trajectory points by selecting multiple abnormal thresholds.AIS data is used as a sample to complete abnormal trajectory recognition,combining with SVM.The main research contents include:Firstly,the research background and significance of ship abnormal trajectory detection were elucidated,and the application of different anomaly detection algorithms in abnormal trajectory recognition was analyzed.The current research status of anomaly detection algorithms at home and abroad was summarized,and the content and characteristics of AIS data were introduced.The principles of SVM and factors affecting its recognition accuracy were analyzed,providing a theoretical basis for the establishment of a multi-support vector machine abnormal trajectory recognition model and the selection of abnormal thresholds.Secondly,AIS data was analyzed and the trajectory data was divided and processed based on the introduction of AIS information content.The support vector machine algorithm was used to identify abnormal trajectories.Experimental verification showed that the support vector machine algorithm can be used for ship abnormal trajectory recognition with high accuracy.Then,addressing the issue of low accuracy when selecting a fixed single abnormal threshold,this paper further analyzed the support vector machine algorithm and proposed a multi-support vector machine-based abnormal trajectory recognition algorithm.This method establishes multiple segment trajectory support vector machine models and sets multiple abnormal thresholds to overcome the shortcomings of using a single abnormal threshold,thereby improving the accuracy of abnormal trajectory recognition.Finally,on the basis of the abnormal track identification of ships in space,the abnormal identification of ship space-time track is realized.This method also uses the multi-support vector machine algorithm,and the dynamic time information constitutes the time-space correlation track space-time model,and realizes the ship in spatiotemporal anomaly trajectory identification under domain.In summary,the research work of abnormal trajectory recognition methods presented in this thesis lays a good foundation for ships to achieve safe navigation and accurate identification of abnormal trajectory points.