The safety technology for preventing an accident in the automobile field in recent years attracts attention and many technologies are also implemented. However, many accidents based on a driver's wrong decision of a road condition still remain. To support drivers, it must be useful that the vehicles recognize a road condition in advance, and this technology might reduce the number of traffic accidents. In this paper, the method classifying the road condition in front of the vehicle into four classes (dry, wet, compacted snow and icy) is proposed. The proposal method is realized by assembling several classifiers, e.g. such as a dry-wet classifier, a dry-icy classifier and so on. This paper describes the architecture of classification and its performance by some experiments.