Based on vision sensing and dedicated short range communication (DSRC) disseminating car’s GPS information via v2v communication, a fast collision warning response to an imminent crash has been successfully developed. An embedded dual-platform, DaVinci+XScale-NAV270 was employed to achieve line-marking identification, neighboring vehicles detection, and headway measurement. In order to tackle the problem of driving drowsiness in a vehicle, this paper has also considered the specific four factors: fatigue level, breath alcohol concentration, carbon monoxide concentration, and carbon dioxide concentration. To do so, this study applied a data fusion QT-BPNN/ANFIS to fuse heterogeneous data and then infer the precise collision warning signal. The experimental results show that the proposed approach outperforms two alternative well-known systems.