NLOS Satellite Detection Using Fish-Eye Camera and Semantic Segmentation for Improving GNSS Positioning Accuracy in Urban Area

抄録

In this paper, for the GNSS (Global Navigation Satellite System) positioning, we propose a method to distinguish non-line-of-sight (NLOS) signals from line-of-sight (LOS) signals by utilizing the satellite geometry and the image data of the fish-eye view of the zenith direction at the antenna position. The NLOS signal is the diffracted signal by buildings or structures, and it greatly degrades the positioning accuracy. By applying the proposed method, we can exclude the NLOS signal and can improve the positioning results in urban areas. In recent years, image recognition using deep learning has developed rapidly. We use the semantic segmentation method using deep learning for segmentation distinguishing sky area from obstacle area.