Vehicle Recognition Based on Multi-scale of Gauss Background Modeling

In the intelligent traffic system, vehicle recognition is always a hot topic. However, because the environment is easily influenced by the light, the weather, the shadow and the noise. It’s difficult to get ideal vehicle segmentation effect. This paper studies the existing background modeling algorithm and proposes a vehicle recognition method based on multi-scale of Gauss background modeling. This algorithm divides traffic image into several parts and multi-scale analysis, and extracts multi-scale features of the image. Describe the multi-scale features by using the mixed Gauss model, thus realize background modeling of the complex traffic image. This algorithm can be well suited to sudden change in the image, and restrain discrete noise points. As the research shows, the effectiveness of this algorithm solves the problem of the false alarm pixels interference.