The Study Of Classification Of Passenger Vehicle Types Based On Artificial Neural Network

Electronic Toll Collection is an important part of Intelligent Transport System. Collectingtoll depends on the rapid speed of Passenger Vehicle and correct classification, which are alsothe key procedures to realize Electronic Toll Collection. The classification of Artificial NeuralNetwork has got satisfied results in laboratory and also solved many problems in practice.Extracting the geometrical features of vehicles and selecting the features will eliminatenon-relevant or redundant features, reduce the number of features, improve the accuracy ofthe model and reduce the time of operation. The advantage of Genetic Algorithm for featureselection is that it has relatively greater ability to search and good ability of robustness. TheGenetic Algorithm doesn’t adopt deterministic rules and emphasizes on making use of changeof probability to guide the process of search. Support vector machine is a kind of machinelearning with perceptron as its source, which has many special advantages in dealing withsmall sample, nolinear and high dimensional pattern recognition.This paper made use of BP neural network and Support Vector Machine to classify4 types of vehicles selecting from320samples. When Support Vector Machine doesn’t carry outparameter optimization, it doesn’t have a classification accuracy rate as BP neural network.When getting the same classification accuracy rate, BP neural network has fewer features butSupport Vector Machine use less time. BP neural network doesn’t have higher requirement forthe deflection of data. After parameter optimization of Genetic Algorithm, Support VectorMachine can have a higher classification accuracy rate than BP neural network.