An Introduction to the Kalman Filter
An Introduction to the Kalman Filter
Vision-IMU integration using a slow-frame-rate monocular vision system in an actual roadway setting
IEEE Transactions on Intelligent Transportation Systems
| Hi-index | 0.00 |
Timely and precision indoor spatial awareness is critical for a variety of facility management and building emergency response scenarios, such as facility manager navigation for regular building system maintenance, and firefighter navigation. Since indoor environments do not have GPS coverage and many indoor spaces have similar appearances, people under stress tend to lose their spatial awareness and have difficulty in analyzing their surroundings for completion of their tasks or finding a safe-haven. State-of-art indoor navigation solutions collectively use the magnetic field of the Earth, or deploy a wireless sensor network serving as spatial reference framework to achieve high precision localization. The magnetic field based approaches have decreased performance in areas with objects having strong magnetic fields, while the wireless sensor-network based approaches involve substantial investments into the wireless infrastructure. Moreover, complex indoor environments usually cause complicated interactions between the wireless sensors and the indoor objects, which pose additional technical challenges. The research presented in this paper explores an indoor navigation solution composed of passive sensors and uniquely integrates data acquired from an Inertial Measurement Unit (IMU), a camera system (vision sensor), and a step sensor for precision tracking of the trajectory of a user. Contrary to other methods, the proposed system does not rely on magnetic field sensing and a wireless sensor networks. The developed algorithms exploit the IMU signals for localization, the vision sensors for heading estimation and the step sensor for detecting stationary phases of the foot based. An Extended Kalman Filter (EKF) integrates this information and achieves disclosure error of 5% of tracking accuracy on average. Preliminary experimental results show the potential of this approach in urban setting.