Performance analysis of vision-based terrain referenced navigation

Abstract

This paper studies a vision-based terrain referenced navigation. Generally, in terrain referenced navigation, a radar altimeter is utilized as a primary sensor. However, the single sensor information is not enough to compensate for the degradation of navigation performance caused by the increased sensor noise when the aircraft is operated at high altitude. To cope with that performance degradation, this paper adopts a vision sensor as a supplement sensor to expand the amount of measurement information. The measurement model is formed by the features' estimated positions based on stereopsis between sequential images. Monte Carlo simulation is conducted using 200 samples considering evenly distributed aircraft initial positions and initial position errors. This is to test the navigation performance of a highly nonlinear system characterized by the terrain altitude variation. The simulation result verifies that navigation error is reduced with the aid of the vision sensor when compared to the conventional filter in which the radar altimeter is used solely.