Output details
11 - Computer Science and Informatics
King's College London
A Kalman Filter-Integrated Optical Flow Method for Velocity Sensing of Mobile Robots
<02>This paper presents a novel optical flow algorithm to measure the velocity and location of mobile robots using a downward-looking camera. Differential optical flow methods require large image overlap for accurate velocity estimation, significantly limiting the maximum measurable velocities. To overcome the problem, a Kalman Filter is used to predict the image transformations, thus reducing the feature search area, resulting in rapid convergence of the optical flow algorithm. The proposed algorithm is validated under laboratory and outdoor environments, showing very good potential for mobile robot velocity and position sensing, demonstrating promising potential for application in GPS denied environments.