The environment of long highway tunnels is relatively closed, and vehicles are prone to traffic accidents when the road surface is wet and slippery. Once an accident occurs, it is difficult to rescue, and it is easy to cause a secondary accident. At present, the evaluation method of tunnel road surface slippery state has low spatial resolution and lacks real-time performance. To achieve a fine perception of tunnel road surface slippery state and improve time and spatial resolution of surface slippery state evaluation, the author proposed a dynamic data collection method and a method of tunnel road slippery state evaluation unit division. The evaluation model of the tunnel road surface slippery state was built based on the K-means clustering algorithm. The author selected the tunnel experimental environment in the Proving Ground for Highway of Ministry of Transportation, and set up controlled experiments under different surface slippery states. The experiment shows that the evaluation model for long highway tunnels’ road surface slippery state proposed in this paper can effectively reflect the actual road surface slippery state.