Research On Vehicle Detection And Classification Technology Based On Blind Source Separation

Intelligent transportation system has developed rapidly in recent years, which depends on the detection and classification of moving vehicles. Traditional detection and classification techniques based on the pressure sensor exist to install, high maintenance costs and other shortcomings, but the techniques based on the video image overcomes these drawbacks.Existing video-based classification techniques can be divided into four categories: license plate recognition method, based on the geometric characteristics of the vehicle, based on the vehicle contour shape, PCA and LDA-based method of algebraic features. They are either difficult to use either the classification accuracy is not high.Vehicle classification system is a key of traffic monitoring system. Blind source separation algorithm is introduced to the vehicle’s classification system in this paper. ICA remove higher-order statistical correlation of signals, and ICA is used for vehicle image feature extraction. Through experiments, ICA model in extracting vehicle feature extraction is superior to the traditional PCA and LDA algorithms. At the same time,geometrical characteristics(length + width) is extracted of the vehicle and corrected according to the position of vehicle in the image. This feature is used to the pre-classification system, combined with algebraic features, build a secondary classification system. Secondary classification system is based on geometric features and characteristics of algebra, to further enhance the accuracy of vehicle classification.Moving vehicle detection is the foundation of traffic monitoring system.By analyzing existing foreground detection and background modeling algorithm, a new three frame differencing method was proposed for traffic surveillance video for vehicles detection. Improved three frame differencing method has better adaptability in vehicles detection problem and runs as fast as the traditional frame difference algorithm. At the same time, Using morphological erosion operator to separate adhesions vehicle.In this paper, a road monitoring system based on OpenCV and the VS2010 is designed and implemented. The system can effectively detect and classify vehicles.