Research On Key Technology In Perception Intelligent Transportation Systems Based On AMR

At present,the ITS(Intelligent Transportation System)has a rapid development.Especially,the traffic information collection plays an important role in the ITS.The traffic information collection is of great significance for the ITS.The existing traffic information monitoring module are based on the video image information,geomagnetic coil sensors and a three-axis geomagnetic sensor(AMR).Due to the small size of AMR which can be installed easily and the rich features of information it can collects,the AMR has been widely researched recently.This paper studies collecting traffic information by an AMR,including the existence of a vehicle,vehicle classification and traffic volume.In this paper,we analyzed and compared the existing traffic information collection methods.We figured the geomagnetic signal is a better way to collect traffic information because it is much easier to get information which contains more features than other sensors.As the traffic information collecting system is of strict real-time requirements,we have to simplify the complicated vehicle classification algorithm.In this paper,we choose an algorithm combining filter-filter model and wrapper model.First,we evaluate and sort the extracted features by RelieF algorithm.Then we calculating the Poisson correlation coefficient to decide whether we need the feature or not.At last we use the filter algorithm to filter out the best subset of features.For determining whether the vehicle exists,we choose a state machine.In the no vehicle state,we update the background magnetic.In the vehicle state,we record and analysis the geomagnetic signals.We use the Adaboost algorithm to promote the vehicle classification result.The algorithm use SVM as weak classifiers.The improved algorithm has a good generalization ability.Compared with SVM algorithm,accuracy increased by 8% to 10%.This paper aims to realize the traffic information collection and analysis on the underlying ARM through simplify the classification algorithm.The main work of this paper is the study of traffic information collection by geomagnetic signal,including collecting traffic geomagnetic signal,processing,feature extraction and analyzing the traffic information.Innovative work of this paper are:1)Feature extraction and selection: We use an algorithm combining filter-filter model and wrapper model.First,we evaluate and sort the extracted features by RelieF algorithm.Then we calculating the Poisson correlation coefficient to decide whether we need the feature or not.At last we use the filter algorithm to filter out the best subset of features.2)Vehicle classification algorithm: We use the Adaboost algorithm to promote the vehicle classification result.The algorithm use SVM as weak classifiers.The classification result is better and generalization ability is much better.