Road Surface Condition Recognition Method Based on Color Models

A computer vision technique has been applied to analyze and research road surface meteorology. The original color data in HIS and RGB models has constituted feature vectors. Robust technique has been used to remove outliers before image process. And a BP neural network has been employed to identify the images collected from road surface in four kinds of states (namely, covered by dry asphalt, water, ice and snow). The result of experiments shows that the means is effective.