Inertial Sensor Bias Error Estimation Using Rolling and Global Shutter Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for correcting bias errors in inertial sensors, particularly in vehicles equipped with optical sensors and external cameras, face challenges in accurately estimating and compensating for temperature-induced errors due to differences in shutter modes between sensors, leading to potential inaccuracies in motion estimation.
Innovation Solution
An error estimation device is implemented, comprising an error prediction unit, correction unit, motion compensation unit, difference extraction unit, and determination unit, which predicts bias errors in inertial sensors by comparing reflection distance images from optical sensors in rolling shutter mode with outside light images from external cameras in global shutter mode, allowing for accurate correction and compensation of vehicle movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bias error correction is performed using inertial sensor data alone, then the correction process is simple, but the accuracy of motion estimation deteriorates due to temperature-induced drift
Solution Approach 1:
The patent introduces image data from optical sensors and external cameras as an intermediary reference to correct inertial sensor bias errors. The system uses visual odometry calculations from image sequences as a mediator to detect and correct drift in inertial sensor measurements, thereby improving motion estimation accuracy without requiring complex hardware modifications
Solution Approach 2:
The system implements a feedback mechanism where the difference between motion estimated from inertial sensors and motion estimated from image data is continuously calculated. This feedback signal is used to adjust and correct the bias error in the inertial sensor, creating a closed-loop system that automatically compensates for temperature-induced drift over time
2Speed
If rolling shutter mode is used for optical sensor, then the image acquisition speed is high, but motion distortion occurs during vehicle movement
Solution Approach 1:
The patent applies motion compensation based on inertial sensor data before comparing images from rolling shutter and global shutter modes. By predicting vehicle movement using accelerometer and gyro data, the system pre-adjusts the rolling shutter image to compensate for expected motion, enabling accurate alignment without sacrificing acquisition speed
Solution Approach 2:
The system replaces physical mechanical synchronization mechanisms with computational motion compensation. Instead of mechanically synchronizing the rolling shutter with the global shutter, the patent uses software-based image warping and coordinate transformation algorithms to align images after the fact, maintaining high acquisition speeds while achieving precise alignment
3Measurement precision
If global shutter mode is used for external camera, then motion distortion is minimized, but the shutter synchronization with rolling shutter optical sensor becomes complex
Solution Approach 1:
The patent uses inertial sensor data as an intermediary to bridge the timing difference between global shutter and rolling shutter modes. By calculating vehicle motion during the exposure period using accelerometer and gyro data, the system creates a virtual synchronization reference that eliminates the need for complex hardware synchronization circuits
Solution Approach 2:
The system transforms the synchronization problem from the time domain to the spatial domain. Instead of synchronizing shutter timing, the patent applies geometric transformations to the image coordinates based on vehicle pose changes, aligning images from different shutter modes through spatial coordinate adjustment rather than temporal synchronization
4Measurement precision
If bias error is frequently repredicted to improve accuracy, then the estimation precision improves, but the processing time and computational load increase
Solution Approach 1:
The patent implements selective reprediction of bias error based on detected drift thresholds. Instead of continuously repredicting at every time step, the system monitors the divergence between inertial-based and vision-based motion estimates and only triggers reprediction when the error exceeds a predefined threshold, achieving high precision while minimizing unnecessary processing
Solution Approach 2:
The system employs periodic validation of bias error predictions at regular intervals using image data, rather than continuous correction. This periodic checking approach maintains estimation precision by regularly updating the bias model while reducing computational load compared to continuous real-time correction
Data Source
AI summary
An error estimation device includes an error prediction unit and a determination unit. The error prediction unit is configured to predict a bias error occurring in an inertial sensor mounted in a vehicle. The determination unit is configured to determine whether the bias error needs to be repredicted by the error prediction unit based on a reflection distance image acquired by an optical sensor in a rolling shutter mode and an outside light image acquired by an external camera in a global shutter mode.