Traffic Information Feature Analysis And Traffic Mode Detection

Nowadays with the rapid development of information technology,Intelligent Transportation System has become the main trend of future transportation development.Two problems in this field deserve attention:(1)Traffic congestion and(2)Traffic mode detection.Traffic congestions and accidents have a huge impact on the economy and the environment.As a result,it is necessary to analyze the causes that affect road congestion to guide the design and construction of urban traffic infrastructure and reduce the incidence of road congestion and traffic accidents.Traffic mode detection is another important problem.At the micro level,it helps mobile devices to serve users better.At the macro level,it helps to analyze the habits of urban users,which plays an important role in urban traffic operations,traffic planning and facility design.This paper first introduces the basic principle of deep learning.Secondly in order to solve the problem of the cause of traffic congestion on region level,this paper proposes a method for analyzing traffic congestion features based on neural networks.In this paper,the regional traffic networks are represented as images.By designing a convolutional neural network,the images are classified into two classes,i.e.congested and non-congested.In addition,the CAM(Class Activation Maps)method is used to extract and visualize factors that cause congested.Next in order to solve the problem that the artificially extracted features in traffic mode detection are not comprehensive enough,this paper proposes a method based on neural networks,including algorithm based on trajectory features and collection point features,and algorithm based on hierarchy taxonomy.By combining trajectory features with collection point features and using the taxonomy information,the designed neural netwoks get higher accuracy.In the experiments of traffic congestion,this paper concludes that the number of intersections in the region and the curvature of road segments in the region are two important features that affect regional congestion.In the experiment of traffic mode detection,this paper concludes that the proposed method can get higher accuracy in traffic mode detection compared to existing methods.