A drone-mounted perception stack must estimate the vehicle's 6-DoF pose while navigating under the forest canopy. The hardware supplies an IMU streaming inertial data at 200 Hz and a stereo RGB camera providing visual odometry at 30 FPS. To maintain an accurate, real-time state estimate between camera frames and to correct long-term drift, which sensor-fusion strategy is most appropriate for this scenario?
Perform homography-based image stitching to align the inertial and visual measurements.
Fuse the two pose estimates by averaging their Euclidean distances at each timestamp.
Use an Extended Kalman Filter that predicts motion with the IMU and applies visual-odometry updates when camera frames arrive.
Apply cross-modal non-maximum suppression to merge bounding boxes from IMU and camera outputs.
An Extended Kalman Filter (EKF) is designed for continuous, nonlinear state estimation in which a high-rate process model (the IMU) predicts the state and a lower-rate measurement model (visual odometry from the cameras) provides periodic corrections. The IMU prediction keeps the pose updated at 200 Hz, while camera updates bound drift and bias accumulation. Non-maximum suppression is a post-processing technique for object-detection boxes, homography stitching fuses only image planes, and naively averaging Euclidean distances ignores sensor noise models and their covariance, so none of these alternatives can provide consistent, high-frequency pose estimates.
Ask Bash
Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.
What is an Extended Kalman Filter and how does it work?
Open an interactive chat with Bash
What is 6-DoF and why is it important for drones?
Open an interactive chat with Bash
How does visual odometry complement IMU data in state estimation?