Accurate 2D and 3D maps information becomes essential in modern autonomous driving. This study focuses on the improvement of GNSS positioning using 3D maps. In ENC 2015, we presented a particle filter based GNSS positioning method using 3D maps, called 3D-GNSS. This paper investigates the improvement of 3D-GNSS applying differential GNSS (DGNSS) correction. DGNSS correction is consisted of not only the atmospheric and satellite clock/orbit corrections but also the system time offset between the GPS and other constellations. It enables the GNSS measurements can be used together to calculate the consistency based on the receiver bias. Furthermore, we include the idea of horizontal dilution of precision (HDOP) in the estimation of likelihood of the particles. Finally, the 3D-GNSS applying DGNSS correction achieves less than 1.5 meter in lateral positioning error.