Visual-Inertial State Estimation With IMU Bias Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing state estimation methods for mobile objects suffer from significant estimation errors in position, speed, and attitude due to motion blur and peripheral object interference, particularly in visual inertial odometry systems.
Innovation Solution
A state estimation device and method that utilizes image data from a capture unit and inertial measurement data to extract feature points, track these points, calculate bias errors using bundle adjustment, and correct these errors to improve estimation accuracy by removing bias from inertial data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual inertial odometry is used to estimate position, speed, and attitude, then estimation can be performed using camera and IMU data, but significant estimation errors occur due to motion blur and peripheral object interference
Solution Approach 1:
The patent extracts and removes bias errors from the inertial measurement unit data through bundle adjustment and correction processes. The bias error calculation unit computes bias values based on feature point tracking and image data, then these biases are subtracted from the IMU measurements to produce corrected inertial data, thereby eliminating the harmful interference of motion blur and peripheral objects on the estimation accuracy
Solution Approach 2:
The patent implements a feedback mechanism where the estimated bias errors are continuously fed back to correct the inertial measurement data. The correction data generation unit uses the calculated bias values to adjust subsequent IMU readings, creating a closed-loop system that continuously improves estimation accuracy by using recent estimation results to correct past measurements
2Measurement precision
If bundle adjustment is performed to calculate bias error, then estimation accuracy can be improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary feature point extraction and tracking before the main bundle adjustment computation. By pre-processing the image data to identify and track feature points across frames, the system prepares the necessary input data for bias error calculation, thereby organizing the computational tasks in a way that reduces overall complexity while maintaining accuracy
Data Source
AI summary
By a state estimation device or a state estimation method, image data is read, a feature point included in the image data is extracted, the feature point is tracked, a position, a velocity, or an attitude of a mobile object is calculated based on inertia data, and a bias error of an inertial measurement unit is calculated. Correction data is calculated by removing the bias error from the inertia data, and a state including at least one of the position, the velocity, or the attitude of the mobile object is estimated based on the correction data.