Noise induced from the traffics at night has continuously destroyed the peaceful night, resulting in harmful effects on general health. This paper presents a new method for the prediction and evaluation of the urban traffic noise, based on the fine segmentation of acoustic functional areas at night, to produce more accurate assessment on the noise pollution resulted from the traffics. The contribution of different types of vehicles to noise can be analyzed based on the prediction results with the parameter inputs of light-, medium-, or heavy-duty vehicles. To demonstrate the effectiveness of the proposed method, a typical area with various points of interest in Beijing was taken as an example, the prediction accuracy has been verified by comparing the predicted level with the monitored one. The experimental results also imply that the inflow of heavy-duty vehicles at night can significantly promote the traffic noise pollution.