A Novel Robust Polarization Skylight Navigation Algorithm Based on Obstacles Detection

Research on polarization skylight navigation has attracted much attention because of its strong autonomy and non-accumulating error. However, polarization information obtained directly from the skylight will inevitably contain much interference, because the views of polarization sensor will be occluded by many obstacles. The existing methods of polarized navigation usually use cloud detection or support vector machine (SVM) classifier for obstacles detection, which are overly dependent on the prior knowledge about the specific obstacles colors and are difficult to detect obstacles looking similar to the sky. In this paper, a novel angular feature based on polarization E-vector is proposed, which is highly sensitive to all obstacles regardless of their colors. Based on this, a multi-obstacles detector is designed, and the obstacles detection and navigation are closely combined into an optimization problem. Our algorithm is effective for obstacles detection and avoids the influence of false detection on navigation precision. Simulation results show that the multi-obstacles detector based on E-vector angular feature achieves a small rate of false detection and thus implements robust polarization navigation under inference of obstacles.