Atomic Interaction-Based Traffic Bayonet Visual Analysis System

With the rapid advancement of urbanization,the rapid development of the national economy and the rapid increase in the number of urban motor vehicles,how to solve the traffic problems in the city has also become the focus of current research.At the same time,traffic-related data has also been taken seriously,and a large number of traffic bayonet data are collected and analyzed by cameras and other equipment installed at road intersections.For these complex and multidimensional traffic bayonet data,traditional data analysis methods have not been applied.Visual analysis technology combines visualization,human-computer interaction and automatic analysis,and makes the data analysis process transparent to help people get more and more useful information from the bayonet data.Therefore,this paper proposes the use of visual analysis technology to analyze traffic bayonet data.This paper designs a visual analysis system based on the Web platform,which mainly analyzes the data of urban road monitoring bayonet,extracts and atomizes traffic information to meet the needs of various complex researchers,and adopts similar traffic density maps.Visual analysis tools such as traffic timing charts explore the traffic conditions in a city.This article also uses visual analysis technology to explore traffic flow prediction based on neural network model.The main work and achievements of this paper are as follows:(1)Vehicle data cleanup.The core of data visual analysis is data,which needs to be cleaned to achieve the best visual effect.Due to malfunction of the surveillance device,there may exist error data in the dataset,such as the vehicle not being identified,vehicle type and vehicle color identification error.We first perform a data cleaning process.The process deletes the records in which vehicle plate is not identified,and unifies the color and type in different record of same trajectory.(2)Decomposition of traffic problems.Traditional visual analysis systems often satisfy only the current requirements and have poor scalability.This paper proposes a method for decomposing traffic analysis tasks.The user requirements are abstracted into different element operations,i.e.,atom operations including categorical atoms such as vehicle plate atom,and numerical atoms such as speed atom.These atom operations could be freely assembled into various forms to generate an essential subdataset for existed and unanticipated requirements.(3)Neural network predicts visualization of traffic flow processes.This paper also proposes to visualize the traffic flow forecasting process of the neural network and visually analyze the training process and results of the neural network.This helps researchers to construct a more reasonable neural network structure for traffic flow prediction.(4)Visual analysis system under B/S architecture.The system adopts a browser and server model design method.This method has strong portability.At the same time,there are many visual analysis libraries supporting this method.Through the main operations performed by the server,researchers can reduce the hardware requirements of the visual analysis system.The system is based on the bayonet,uses Mongodb as a data storage platform,realizes the needs of complex researchers through atomic interaction visual analysis methods,and uses different visual analysis tools to explore urban traffic problems.