Geolocation Detection Of Illegal Chemical Facilities Based On Spatiotemporal Data

With the advancement of sensor technology and the development of information infrastructure in modern cities,a large amount of spatiotemporal data generated in cities can be better sensed,stored,queried and managed.This enables many problems that were difficult to solve in the past to be solved by intelligent applications based on spatiotemporal data.Among them,spatiotemporal anomaly detection uses spatiotemporal data to detect abnormal events in cities in a timely manner,so as to reduce the possible harm caused by early action.This paper studies the problem of finding the geographic location of illegal chemical facilities based on spatiotemporal data such as the trajectory data of hazardous chemical transport vehicles.Chemicals are widely used in daily life and industrial production,but some chemicals are very dangerous,so they are strictly regulated by the government.However,driven by interests,there are many illegal small chemical facilities that smuggle and process hazardous chemicals.This brings great security risks to the surrounding residents.The past methods based on blanket searches and manual reporting are very inefficient and labor-intensive,making it difficult to solve this problem efficiently.This thesis innovatively proposes a framework for finding the geographic location of illegal chemical facilities mainly based on the trajectory data of hazardous chemical transport vehicles.The main goal of the framework is to predict from the trajectory data whether a location has ever had hazardous chemical loading and unloading behaviors.If a location is not in an industrial area and has been handling hazardous chemicals,it can be considered a potentially illegal small chemical location.First of all,identify the resident points of the cleaned trajectory data,and find the frequent resident points of all vehicles based on the spatial clustering algorithm,where illegal small chemicals are hidden.Next,use factory production information and POI data to preliminarily classify frequent residency locations,and combine the freight journeys contained in the trajectory data to construct a hazardous chemical freight heterogeneous graph.The heterogeneous graph includes hazardous chemical nodes and location nodes.For location nodes,this thesis proposes some spatiotemporal behavioral features based on trajectory data.In order to integrate spatiotemporal behavior features and freight context information to predict whether there has been hazardous chemical loading and unloading behavior at a location,this thesis continues to propose two location classification algorithms.The first algorithm is based on the heterogeneous graph structure representation learning algorithm of meta-path,learns the hidden distribution of hazardous chemicals related to the factory location,and based on this,constructs the fusion of freight upstream features and spatiotemporal behavior features for location classification.In the second algorithm,in order to fully disseminate and aggregate the behavior characteristics and freight information in the freight information of hazardous chemicals in the nodes of the graph network,so as to obtain a more effective high-dimensional representation,a spatiotemporal method on the heterogeneous graph neural network is used.Heterogeneous information dissemination aggregation mechanism,and the application of a variety of effective heterogeneous graph network structure optimization methods,has achieved better location classification results.In experiments,the data of real hazardous chemical transport vehicles are used to verify the effectiveness of the location classification algorithm,and the abnormal location detection effect of the algorithm is verified by combining the real illegal small chemical facilities data.