The Global Navigation Satellites Systems (GNSS) system is the conventional approach to provide the localization service in the urban area. However, the performance of GNSS is degraded by non-line-of-sight (NLOS) and multipath effects from the surrounding buildings. To solve this issue, the 3D building model is applied with the GNSS measurement by various matching schemes, which is called 3DMA GNSS. Recently, the geometry of the surrounding building can be estimated directly from the GNSS measurement instead of the 3D building model as sky visibility estimation, but this algorithm is limited by the number of GNSS satellites. In this paper, we developed an innovative sky visibility estimation by the SVM regression and simulated the real-time LEO constellation for the sky visibility estimation. With such improvement, the mean elevation differences between the 3D building model and our estimation is less than 10 degrees on our experiment in Hong Kong urban environment. After that, we developed sky visibility matching with the 3D building model to get the position solution. For the ideal condition, the 2D position error can be less than 1 meter in the urban canyon, and the performance of matching our estimated sky visibility also shows the potentials.