The techniques of digital image analysis, features detection, and Hidden Markov Model are employed to develop a vehicle speed prediction system, which used to find the trend of the speed changed. The proposed vehicle speed prediction model is used to set up a vehicle speeding early warning model. The proposed vehicle speed early warning system includes the vehicle speed computation and prediction model. All the data source of this study is obtained by the special design vehicular digital video recorder device that includes well defined driving data format. The data of digital video recorder, which represented the driving state data of the vehicle the speed is included, are analyzed to set up the speed computation and prediction model. The proposed approaches can closely match the vehicle speed and its concurrent video frame. The results of this study can also provide other vehicular computer vision techniques to reduce the processing time along with vehicle's speed.