Research On Visual Analysis Of Abnormal Trajectory Detection Oriented To Network Freight Platform

China's logistics industry has become one of the pillar industries of the service industry,and the emergence and development of network freight platform to further improve and optimize the operation process of logistics enterprises,to provide customers with more efficient transportation services,but also to the logistics industry to bring a new service model and operational concept.There are a large number of abnormal phenomena caused by improper operations by carriers in the online freight platform.For example,the carrier deliberately cancels the waybill or repeats the pick-up,which causes the departure place to be inconsistent or the destination to be inconsistent,forming a large number of abnormal trajectories,increasing the management cost of the platform.Existing related research on abnormal trajectory detection mostly relies on the definition of "abnormal" and threshold division in specific application scenarios.However,the abnormal trajectory data of network freight platform involves a wide range of scales and various abnormal situations.It is difficult to accurately define "abnormality" based on existing indicators.",thus failing to meet the requirements for detection accuracy in the network freight scene.The research contents and results are as follows:1)Propose the method of abnormal trajectory detection based on key nodes with differential intersection set distance metricThe method fuses multi-source data such as carrier vehicle trajectory,POI(Point of Interest)and road network,builds key nodes,enhances the expression of trajectory,and defines anomaly scoring function on this basis to calculate the difference between different types of anomaly trajectory and normal trajectory,and completes the detection and classification of anomaly trajectory.Compared with the existing differential intersection set distance metric anomaly detection algorithm,this method is suitable for the detection of anomalous trajectories in a wide range of intercity scales,while retaining the semantic information of trajectories.2)Constructing a visual analysis model of abnormal trajectory detection for network freight platformThe model introduces visual coding methods such as spatial location,shape,color,length,etc.,and designs interactive analysis means such as selection,screening and association,realizing visual analysis of information such as key node sequence of abnormal trajectories,spatio-temporal data of abnormal locations and driving behavior of abnormal carriers,helping analysts to evaluate and compare carriers and carrier vehicles from the whole and specific details.3)Proposed link graph visualization method based on key nodesThe method is based on the node link graph visualization technology,fusing the temporal characteristics of spatio-temporal data,taking the key nodes passed by the carrier vehicle trajectory as the nodes of the link graph,realizing the sequence characteristics of the carrier vehicle display,completing the temporal characteristics analysis of the carrier vehicle abnormalities overview,and providing direction for the detail exploration analysis.The method can highlight the key abnormal information in a large amount of trajectory data and provide subsequent analysis ideas.4)Proposed carrier evaluation index visualization methodThe method integrates the loop chart visualization and radar chart visualization techniques,designs interactive methods such as information prompting,selection and highlighting,defines eight dimensions of carrier evaluation,compares the abnormal evaluation of carriers in different time periods,and realizes detailed comparison of multiple attributes of carriers.The method can realize the comparative evaluation analysis of multidimensional data in different dimensions and scales.Based on the above research,this paper designs and implements a prototype system for visual analysis of abnormal trajectory detection for the network freight platform.Based on the J network freight platform,the system takes the carrier vehicle trajectory data,POI data and road network data of Guangdong Province in 2020 as data sources,and assists the network freight platform managers to explore the comparison of carrier vehicles and carriers in different granularity and dimensions through selection,screening,association and other interactive means to complete abnormal trajectory detection and analysis.Finally,the effectiveness of the system is verified through case studies and user experiments.The system can not only help the network freight platform to identify the abnormal behavior from the trajectory of the carrier vehicle,carry out the evaluation exploration of the carrier and provide detection and analysis,but also provide reference for the relevant departments to detect abnormal vehicles.