Abstract:For indoor close-range high-accuracy attitude measurement using inertial and vision sensors, the major challenge is to achieve a reliable attitude in various scenarios, specifically in the absence of vision information. To address this challenge, we present a fusion strategy for reliable attitude measurement using a micro electro mechanical system (MEMS) gyroscope and camera, particularly during discontinuous vision observations. The proposed algorithm consecutively executes update and prediction modes. In the update mode, when both inertial and visual data are available, the complete Kalman Filter (KF) conducts the entire filtering process and updates residual and state errors. In the absence of visual data, the partial KF operates only when the time update and filtering state vectors are compensated by the transfer residual, which is estimated through self-propagating state error. The parameters of least-squares support vector machine (LSSVM-NARX) continuously update until the next discontinuous vision observation occurs, with the desired outputs and inner inputs being the error angles and the exogenous inputs being the gyroscope values and time. In the prediction mode, without visual data, LSSVM-NARX can provide estimated angle errors between the complete and partial KFs, which are used to eliminate errors from partial KF output vectors to obtain a reliable and accurate attitude solution. Comparisons and extensive semi-physical simulation experiments under various motion trajectories were performed to validate the effectiveness and superiority of the proposed scheme in an accurate and reliable attitude measurement capability associated with discontinuous observations.

Relative Pose Estimation Algorithm with Gyroscope Sensor.

Gyro-aided Camera-Odometer Online Calibration and Localization

Pose Estimation Of A Rigid Body And Its Supporting Moving Platform Using Two Gyroscopes And Relative Complementary Measurements

Stereo-inertial Pose Estimation and Online Sensors Extrinsic Calibration.

A Homography-Based Visual Inertial Fusion Method for Robust Sensing of a Micro Aerial Vehicle

Globally Optimal Solution to the Generalized Relative Pose Estimation Problem Using Affine Correspondences

Whole-body Pose Estimation in Physical Rider-Bicycle Interactions with a Monocular Camera and a Set of Wearable Gyroscopes

Pose Estimation in Physical Human-Machine Interactions with Application to Bicycle Riding

Globally Optimal Consensus Maximization for Relative Pose Estimation With Known Gravity Direction

LiDAR and Inertial Fusion for Pose Estimation by Non-linear Optimization

Developing Rigid Constraint for the Estimation of Pose and Structure from a Single Image

Globally Optimal Relative Pose and Scale Estimation from Only Image Correspondences with Known Vertical Direction

An Efficient Method to Recover Relative Pose for Vehicle-Mounted Cameras under Planar Motion.

An Efficient Solution to the Relative Pose Estimation with a Common Direction.

Tightly Coupled Visual and Inertial Measurements for Motion Estimation

Globally Optimal Relative Pose Estimation Using Affine Correspondences with Known Vertical Direction

Fast vision-based pose estimation iterative algorithm

Minimal Solutions to Relative Pose Estimation from Two Views Sharing a Common Direction with Unknown Focal Length.

Adaptive Sensor Data Fusion in Motion Capture.

A fusion strategy for reliable attitude measurement using MEMS gyroscope and camera during discontinuous vision observations

Pose Estimation for Omni-directional Cameras using Sinusoid Fitting