An Integrated MEMS IMU/Camera System for Pedestrian Navigation in GPS-denied Environments

Abstract

Pedestrian navigation still remains an unsolved difficulty in the navigation community. Unlike land-vehicle, maritime and airborne navigation, users of pedestrian navigation device often require navigation solution in GNSS-denied environments, such as indoor, urban canyon, and thick foliage where GNSS signal is severely degraded by noise, blockage and multipath. Currently, the MEMS inertial sensors are extensively used in the GNSS-denied environments with advantages of being low-cost and small size. Many mobile devices such as smart phones and PDAs are also embedded with MEMS inertial sensors. But since the MEMS inertial sensors suffer from significant bias, drifts and noise, the navigation solution will diverge to hundreds of meters within several minutes if using the MEMS IMU alone. Therefore, external calibration from other sensors is necessary. A camera is currently integrated in most mobile devices, and it can provide the navigation information by tracking the locations of optical features in successive images. Previous research has shown the coupling of inertial sensors and camera to improve navigation performance in applications such as robotic navigation, unmanned aerial vehicle (UAV), and indoor positioning. Different from them, this paper focuses on monocular vision position aiding based MEMS IMU/Camera system to support pedestrian navigation in GPS-denied environments, especially in indoor conditions. Field test results demonstrate that the proposed vision-aided system can reduce the position error from more than several kilometres in INS-only solutions to less than 3 m.