The invention relates to a vehicle positioning method in a tunnel based on vision and a millimeter wave radar map feature, which combines a sensing feature point data set with high-precision map layerdata, so that accurate vehicle body position is derived according to the data extracted and combined from the map side and the GPS data when an unpiloted vehicle runs in an environment with dark light and weak positioning signals. Compared with the prior art, the vehicle positioning method in a tunnel based on vision and a millimeter wave radar map feature in the invention has the following advantages: improving the accuracy of detecting static obstacles in a special environment, deducing the precise position of the unpiloted vehicle in a special scene and adapting to various scenes affectedby brightness and darkness of the light rays and the strength of the positioning signal.