Detection of two-wheeled vehicles based on Gaussian mixture model and AdaBoost algorithm

The paper describes a video detection method for two-wheeled vehicles appearing especially in the small-medium cities. The multi Gaussian mixture model is used to build the background and foreground. The single threshold method is efficiently used in shadow removal. Then Gaussian smooth filter processing and morphological image processing are used to filter out the noises in the foreground. The target region is obtained by comparing the difference of physical characteristics computed by labeled connecting area between general vehicles (cars) and two-wheeled vehicles. In the target region, the offline classifier based trained with local binary pattern (LBP) feature and AdaBoost algorithm, is used to accurate object detection. Experimental results show that the proposed method has a good performance in real-time and accurate detection of two-wheeled vehicles.