Microwave radar and video sensor fusion for vehicle classification using a Bayesian network

Vehicle classification is an important part of intelligent transportation systems by enabling collection of valuable information for various applications, such as road surveillance and system planning. This paper presents a vehicle classification system using a Bayesian network that fuses microwave radar information and video sensors. The microwave sensor gives the vehicle height profile while the video sensor gives the vehicle planar contour. The vehicle features are extracted by a Gaussian mixture model (GMM) to match existing templates. The vehicle classification was the GMM which is based on a Bayesian network to integrate the different vehicle features into a data fusion system. Tests show that the system improves the vehicle classification accuracy rate from 79% using only a microwave radar to 87% using sensor fusion and reduces the classification error rate from 9% to 2% between small and medium sized vehicles and large vehicles.