Research On Abnormal Detection Of The Internal Transport Vehicle Operation Trajectory In Inland Port

The throughput of small and medium-sized inland ports is relatively smaller,and internal transport vehicles(it is refer to vehicles that carry out cargo transfer inside the port,ITV)are usually used for loading and unloading operations of ships on inland ports.After long-term observation by port supervisors,during the process of cargo transfer in ITV,due to the complicated structure of the port roads,the variety of types of transfer operations,the harsh environment of the yard,and the uncontrollable human factors,some abnormal operation can be happened in the process of cargo transferr,which is closely related to the driving intention of ITV.How to detect the abnormality of the trajectory of ITV during operation and find out its abnormal intentions,and provide the evidence chain basis for the subsequent accountability,thus enhancing the safety of the operation of the inland port loading and unloading ship has become an urgent problem to be solved.In view of the above problems,this thesis judged whether ITV’operation conforms to the operation specification by detecting the trajectory of ITV.The specific research content of this thesis is as follows:(1)A feature filter was designed.Due to cost reasons,the trajectory data of ITV operation of the inland ports are usually collected by the civilian GPS,and its trajectory accuracy has certain errors,which needs to be preprocessed.The filter included three parts(trajectory classification,trajectory secondary filtering and trajectory normalization output)to eliminate invalid data,compress trajectories and normalize output trajectory data.(2)A real state sequence extraction method of trajectory was proposed.The real state of the trajectory(the dwell area formed or the area that satisfies the operation links by the trajectory)represented the driving intention of ITV operation,and it is the key to abnormal detection of the ITV operation trajectory.This method included two parts: judgment algorithm for road segment to which the trajectory belongs(TBRSJ)and the real state extraction algorithm of trajectory based on rules and DBSCAN(RSE-RD).In the algorithm of TBRSJ,a heuristic radius threshold which combined the port network grid index,as well as the limit of the road segment,the width of the road segment,and the accuracy of the trajectory was firstly defined to search for candidate road segments of the trajectory point.The observation probability was then defined based on various weighting factors.The transition probability was defined based on the A* algorithm using the optimal path search.The backward joint probability was defined based on the hidden Markov model.Context-based state transition probability was defined by combining with the transition probability and the backward joint probability.The joint probability of candidate road segments was finally calculated by combining with the observation probability and the state transition probability,and the road segment with the highest probability was selected as the road segment to which the trajectory point belongs.In the algorithm of RSE-RD,the invalid temporary states of trajectory were first removed based on the operating rules of ITV.After that,the moving states in the operation link were added to the real state of trajectory.Then the DBSCAN clustering processing was performed on the temporary state of the trajectory outside the rule,and the real state of the trajectory is obtained through the clustering result,and finally the real state sequence of the trajectory of ITV is obtained.(3)An adaptive Finite State Machine(AFSM)was designed to detect the abnormality of the operation trajectory of ITV and to supervise the safety of port cargo.Based on FSM,AFSM joined design of operation rules and feedback data.The design of the operation rules combined the experience of the port domain experts,the data of operation of ITV,and the real-time state of the port traffic.The purpose of the operation rules were to filter some legal states in the real state sequence of trajectory of ITV to avoid these legal states affecting the results of AFSM.Based on the dynamic time warping method,by calculating the similarity of the trajectory of each operation link in the corresponding operation area under the same operation instruction the feedback data was designed.The purpose of the feedback data was to determine whether the specific operation area of the multiple operations under the same operation instruction is legal to improve the accuracy of AFSM for trajectory anomaly detection.Finally,the method and model proposed in this thesis were experimentally verified.Experiments show that the trajectory preprocessing based on feature filter can effectively remove low-quality and invalid trajectory points and greatly compress the trajectory data.By comparing different parameters and the traditional hidden markov map matching algorithm(HMMM),it was verified that the algorithm of TBRSJ can more accurately determine road segment to which the trajectory belongs in inland ports.By comparing different types of contrast,it was verified that RSE-RD can effectively extract the real state of the trajectory of ITV.Based on the above method,AFSM had an accuracy of 95%～96% for the abnormal detetion of the trajectory of ITV.In summary,the work of this thesis can effectively detect the abnormal intention of operations of ITV,enhance the safety of cargo transfer operations,and has important practical significance for the intelligent construction of inland ports.