High precision positioning has become the point of discussion for many revolutionary applications. Like unmanned autonomous system (UAS), requires a highly precise positioning with centimeter-level accuracies. Real-time Kinematic (RTK) is one of the most precise positioning technologies and provide a centimeter-level positioning in opensky or sub-urban environment. However, in the urban environment, severe non-line-of-sight (NLOS) and multipath effects degrade the RTK GNSS positioning. Ambiguity resolution (AR) is the key for RTK GNSS, the carrierphase measurement with integer ambiguity resolved can provide a centimeter accuracy for positioning. The NLOS reception and multipath effect mentioned will result in a noisy measurement for AR and result in low fixing rate in the urban area for RTK GNSS. Therefore, removal on the bad measurements and remain those good-condition signal become essential for RTK GNSS. We believed that NLOS satellite exclusion by 3D building model dynamically is better comparing to with fixed elevation angle. Based on this idea, the study proposes using the 3D building model with position hypothesis to filter out unhealthy satellite from AR and RTK positioning, namely 3DMA GNSS RTK. The designed experiment in Hong Kong urban environment with geodetic-grade receivers will be used to evaluate the performance of our proposed algorithm. The experiment results show that 3DMA GNSS RTK can provide positioning accuracy with 10cm averagely.