Intelligent Traffic Monitoring Using Fine-grained Wireless Signal

With the popularity of private cars and the increase in traffic volume,intelligent traffic monitoring technology has received more and more attention.Intelligent traffic monitoring can support a series of applications such as traffic jam guidance,intelligent traffic management and automatic driving assistance.For some specific scenarios that require low investment and a wide deployment range,such as rural road monitoring,traffic surveys at specific time and section,how to achieve portable,effective,easily scalable,low-cost,and easy-to-deploy traffic monitoring is especially important.The measurement of intelligent traffic parameters by measuring the changes of fine-grained wireless signal channel state information mainly includes four modules: data preprocessing,vehicle detection,vehicle classification and speed estimation.The method has the advantages of easy deployment,easy expansion,easy portability,and low cost.In order to remove the influence of noise,the low-pass filtering and data normalization methods are used to preprocess the raw channel state information data,and then the moving variance method is used to implement vehicle detection.The method of multi-branch convolutional neural network is used to realize the classification of vehicle types.The vehicle types are mainly divided into three categories: two-wheeled vehicles,small passenger cars and large trucks.For vehicles of the same category,the boundary detection method is used to determine the fluctuation time of the wireless signal,and then the polynomial fitting method is applied to complete the vehicle speed estimation.In order to verify the effect of the proposed method,a total of nearly 700 real data were collected on the roads around the campus for experiments.By analyzing the experimental results,we can get the conclusion that the method of moving variance can effectively achieve vehicle detection.When classifying vehicles,the accuracy rate of convolutional neural network is better than the traditional machine learning method.For the speed estimation,the accuracy of the binomial curve fitting is better.The experimental results show that the intelligent traffic monitoring based on the channel state information can effectively measure the traffic parameters,which is of great significance to the intelligent traffic monitoring in specific scenarios.