Research And Implementation Of Vehicle Type Recognition System In Video

This paper mainly researches the basic principle and key technology of vehicle recognition technology which based on video image in intelligent traffic system. Among them, the important branch of ITS systems and key technology is the automatic recognition models. Through the study of relevant literature. On the basis of summarizing and analyzing the existing vehicle recognition technology, combined with the basic theory of image processing, puts forward a new method, and through the relevant experiments prove the validity of this method, at the same time, a simple classification of vehicle recognition system was implemented.Adopt the method of image processing and pattern recognition to identify models generally contain moving target detection, feature extraction and recognition model, such as three steps. In this paper, image averaging method to sliding update the background image for vehicle detection, using the projection technology to locate the vehicle, then according to the geometry features and texture features of the vehicle, using the BP neural network classification method for sequence image recognition, And displays the test results with the MFC and Opencv.This paper first introduces some common methods of moving vehicle detection, finally compared with image average method to update the background image slide background subtraction image method, through the gray image, an adaptive threshold selection method of two value, then use Digital morphology processing method for vehicle region filling algorithms, regional elimination and edge extraction operation, at the same time, a flood fill algorithm to remove the body of big noise, finally, extract the contour, the contour image of complete vehicle.In the feature extraction of the vehicle, this paper chooses the moment invariant features as well as the texture feature graylevel co-occurrence matrix feature as a characteristic parameter. And extract the four graylevel co-occurrence matrix texture characteristics and seven moment invariants as a model of the characteristic feature extraction of characteristic parameters. Through the relevant experiment prove the validity of these two characteristics, and makes a detailed analysis.In terms of models of pattern recognition, the paper introduces the related knowledge of artificial neural network and BP neural network, and based on the analysis of the BP neural network, on the basis of theory and its classification, according to the characteristics of the model in this paper sample library a classifier was determined by the experiments of related parameters, the experimental results, first of all, using the same moment and texture features graylevel co-occurrence of moments using BP neural network to identify models respectively, and then through test seven moment invariant features and texture features different combination of the characteristics of the moments, get a set of high recognition rate characteristics of two kinds of combination.In the final stage of this paper, which introduces the hardware and software environment of the vehicle recognition system, this paper builds experimental platform with Opencv and visual studio2008, and designed a kind of vehicle recognition system, and has carried on the system test, achieved good effect.