GPS data of vehicle on the road can reflect the actual road status, its analysis can be help to urban road planning, but its multi-source, mass and high dimension features of GPS big data has restricted its application for road traffic.Aiming that, a GPS big data visual computation architecture has been design in this paper, speed attribute and two-pass corner detection has been introduce to improve map matching and clustering analysis methods.Vehicle GPS data of Zibo city has been selected as the case, the relevant result shows that the improved methods can be effective and visual, and can get a better effect than flow map method for vehicle GPS big data processing.