Airborne Vision-Aided Navigation Using Road Intersection Features

摘要

Modern airborne navigation systems for manned and unmanned platforms usually rely on GPS measurements to constrain the inertial position estimate of the platform. This reliance on GPS can quickly cause the navigational estimates of system to become unreliable when the system is operating in GPS-limited or denied areas. This paper presents a vision-aided inertial navigation system that uses ground features (in this case road intersections) matched to a database to provide position measurements. An image processing algorithm is used to extract the shapes of the road intersections from visual imagery, this shape is then matched to a reference database to provide image to map road intersection correspondences. The correspondence information is fused with the inertial solution in an Extended Kalman Filter to constrain the complete attitude and position inertial navigation solution. The system is developed to operate at non-zero attitude angles removing the level flight limitations of past approaches. Flight test results of the system demonstrate that the system can successfully produce accurate navigation estimates that are comparable to the use of GPS without the same limitations of GPS when operating in a GPS-limited or denied area.