| With the continuous optimization and upgrading of the social structure,the demand for intelligent management of transportation is also increasing.In order to realize intelligent management of traffic,it is necessary to monitor information about vehicles in real time.Nowadays,the application of image processing technology and artificial intelligence is in the ascendant.With these two technologies,the research on extracting road vehicle information has made great progress and has been widely used in reality.However,overcoming severe weather conditions such as heavy rain,blizzard,and fog is still a huge challenge for image recognition algorithms.In recent years,magnetic sensors have been favored by researchers because of their small size,low power consumption,and weather resistance.However,relevant research still needs to be improved in terms of equipment installation,miniaturization of vehicle detection nodes,vehicle feature extraction,and feature selection.Considering the above application background and technical problems,this paper mainly studies the vehicle detection and vehicle classification algorithm based on single magnetic sensor,and explores another alternative for vehicle information collection under the condition that image recognition cannot meet the demand.In this paper,a detector with single 3-axis magnetic sensor is designed to collect vehicle data,and an effective vehicle detection algorithm and vehicle speed estimation algorithm are implemented on the detector system.After obtaining the magnetic field disturbance signal,this paper deeply studies the feature extraction technology and the related feature selection algorithm,and proposes a feature selection algorithm FPE-Filter which is based on feature pairing elimination.Experimental results prove the efficiency of FPE-Filter.This paper also compares the performance of the four classification models:SVM,RF,KNN,and C4.5.The experimental results show that RF performs best and the classification accuracy reaches 97%. |
PDF Full Text Request