3D Vision Aided GNSS Real-Time Kinematic Positioning for Autonomous Systems in Urban Canyons

Abstract In this paper, a three-dimensional vision-aided method is proposed to improve global navigation satellite system (GNSS) real-time kinematic (RTK) positioning. To mitigate the impact of reflected non-line-of-sight (NLOS) reception, sky-pointing camera with deep neural network was employed exclude these measurements. However, NLOS exclusion results in distorted geometry. fill gap, complementarity between low-lying visual landmarks and healthy but high-elevation measurements explored geometric constraints. Specifically, inertial measurement units, captured by forward-looking camera, GNSS were tightly integrated via sliding window optimization estimate GNSS-RTK float solution. The integer ambiguities fixed solution then resolved. effectiveness verified using several challenging data sets collected urban canyons Hong Kong.