Real-time moving vehicle recognition under snowy condition

The recognition of moving vehicle plays an important role in ITS (intelligent transport system). This paper describes a tracking-based recognition method for moving vehicles under snowy weather conditions when it is difficult to detect the vehicles because of environmental noise, caused by snowflakes, snow accumulation, reflections, and so on. First of all, the moving objects, including the obvious noise, are segmented from the traffic road scenes. Then a tracking strategy is used to estimate the objects trajectories. Last, on basis of the trajectory analysis, the vehicles are recognized. Instead of using the traditional gray images, we base our algorithm on the RGB color images to improve the accuracy of segmentation. The Intel streaming SIMD (single instruction multiple data) Extensions technology is also used in the algorithm implementation to achieve the real-time capability. The experiment results show that the proposed scheme has high recognition rate and enjoys satisfactory real-time performance.