Research On Key Technology Of Vessel Behavior Identification Based On Trajectory Data

As a big maritime country,China has abundant marine fishery resources and ecological resources.Over the past decades,offshore fishery resources have been over-exploited by competition and have been over-exploited for a long time.How to achieve sustainable use of marine fisheries is a serious problem.This requires not only the rational planning of marine property rights policies,but also the timely and effective management of the marine regulatory law enforcement agencies.With the rapid development of marine informatization,the Vessel Monitoring System(VMS)has gradually become an effective means of monitoring fishing intensity.The fishing vessel monitoring system is able to obtain fishing vessel trajectory data at regular intervals.The use of trajectory data for data mining of vessel fishing behavior is not only applied to the regulation of fishing vessels,but also applied to fishery identification,fishing intensity assessment,fishery resource development and marine ecological protection.The traditional research on the fishing behavior of vessels mostly takes the trajectory point as the research object.Based on the relationship between the speed and fishing behavior of the vessel,the behavior threshold of the vessel trajectory is realized by setting the fishing speed threshold interval.This one-size-fits-all method of identification cannot balance the precision of identification with the versatility of the algorithm.In view of the above problems,in recent years,the academic community has proposed a solution to the behavior identification of vessels based on the trajectory segment.This solution constructs an unsupervised multi-step clustering model,firstly divides and re-clusters the trajectory,and uses the identification model to extract the feature of the clusters,and establishes a trajectory-based fast identification model of the behavior of the fishing vessel.However,the existing multi-step clustering algorithm based on trajectory segment has high parameter sensitivity,which leads to difficulty in parameter adjustment and poor versatility in batch vessel behavior identification.Aiming at the above problems,this thesis takes trajectory data as the research object,makes full use of the time series of vessel trajectory data,the behavior continuity of the vessel and the characteristics of fast switching.Two methods of identifying the behavior of fishing vessels based on trajectory segments are proposed:OPTICS-based Multi-Step Clustering Fishing Behavior Identification(OMSC-FBI)and Core-distance-based Multi-Step Clustering Fishing Behavior Identification(CMSC-FBI).The main work of this thesis can be summarized as follows:(1)Based on the vessel trajectory data,this thesis conducts relevant theoretical research.Firstly,the behavior types and characteristics of vessels are introduced.Then the theory of trajectory data mining is expounded.Then the algorithm basis of multi-step clustering algorithm is described.Finally,the implementation methods of common discriminant techniques are introduced in detail.(2)Aiming at high time-consuming bottleneck of the existing trajectory segment-based fishing behavior identification algorithm,this thesis proposes a Time Series Neighbor Computing Principle(TSNCP)to reduce the unnecessary calculation of the similarity distance matrix.This principle can reduce the time complexity of matrix calculation fromO(n~2)to O(kn),and effectively solve the problem of high time-consuming calculation of similarity distance matrix.(3)Aiming at the problems of low-versatility of existing algorithms,this thesis proposes a new vessel behavior identification algorithm called OMSC-FBI,based on the theories of trajectory clustering.OMSC-FBI adopts the strategy of dividing and re-clustering trajectory data to establish discrimination model.And it can effectively improves the versatility by using low-parameter sensitivity algorithm.Experiments proof that the feasibility of the algorithm are verified,and the discriminant accuracy and algorithm versatility can be effectively considered.It can be used for the discrimination of batch vessels.(4)In order to further solve the problem of high time-consuming in the trajectory segment-based vessel behavior identification algorithm,this thesis proposes a more efficient solution based on the characteristics of vessel trajectory data,which is called CMSC-FBI algorithm.CMSC-FBI also adopts multi-step clustering strategy.Firstly,the core-distance algorithm is used to realize the segmentation of the track sub-segment,and then the trajectory segment is re-clustered based on the average speed of the trajectory segment.Finally,the discriminant model is established to realize the rapid discrimination of the behavior of the vessel.Experiments show that the algorithm is efficient and feasible and can balance accuracy and versatility.