Camera Extrinsic Calibration via Vehicle Pose Reuse
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle systems face high computational loads and inefficiencies in estimating camera extrinsic parameters, particularly due to duplicative efforts in using inertial and GPS sensor measurements for localization, which are costly and inconvenient.
Innovation Solution
A system that determines calibrated camera extrinsic parameters by using a localization algorithm to estimate vehicle pose, identifying feature points between sequential image frames, and applying a non-linear optimization algorithm to minimize reprojection error, thereby reducing computational complexity and reusing vehicle pose information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If online calibration approaches are used to estimate camera extrinsic parameters, then calibration accuracy is improved, but computational load increases and system resources are wasted due to duplicative use of inertial and GPS sensor measurements
Solution Approach 1:
The patent merges the localization and calibration processes by having the calibration module reuse vehicle pose information already determined by the localization module. This eliminates duplicative processing of inertial and GPS sensor measurements while maintaining calibration accuracy, as the same sensor data is utilized efficiently across both functions without redundant computation.
Solution Approach 2:
The vehicle pose information determined by the localization module serves multiple functions: it is used both for localization (determining vehicle position and orientation) and for calibration (estimating camera extrinsic parameters). This multi-functional use of the same computed data eliminates redundant processing and reduces overall computational load while maintaining both localization and calibration accuracy.
2Device complexity
If offline manufacturing calibration is used to determine extrinsic parameters, then computational load is reduced, but calibration accuracy may be compromised and user convenience decreases
Solution Approach 1:
The system performs preliminary localization to determine vehicle pose information before conducting calibration. By having the localization module pre-compute vehicle position and orientation using inertial and GPS sensors, the calibration module receives ready-to-use pose data, eliminating the need for recalculating these values and reducing computational load while maintaining accuracy.
Solution Approach 2:
The vehicle pose information acts as an intermediary between the localization module and the calibration module. Instead of the calibration module directly processing raw sensor data (which would duplicate localization computations), it receives pre-processed pose information as an intermediary product, reducing computational complexity while preserving calibration accuracy.
3Measurement precision
If duplicative processing of inertial and GPS sensor measurements is performed for both localization and calibration, then calibration accuracy is maintained, but execution time increases
Solution Approach 1:
The patent merges the computational workflows of localization and calibration by having the calibration module utilize vehicle pose information already computed by the localization module. This eliminates duplicative processing of inertial and GPS sensor measurements, reducing execution time while maintaining calibration accuracy through shared use of the same sensor data.
Data Source
AI summary
A system for determining calibrated camera extrinsic parameters for an autonomous vehicle includes a camera mounted to the autonomous vehicle collecting image data including a plurality of image frames. The system also includes one or more automated driving controllers in electronic communication with the camera that executes instructions to determine a vehicle pose estimate based on position and movement of the autonomous vehicle by a localization algorithm. The one or more automated driving controllers determine the calibrated camera extrinsic parameters based on three dimensional coordinates for specific feature points of interests corresponding to two sequential image frames, the specific feature points of interests corresponding to the two sequential image frames, and the camera pose corresponding to the two sequential image frames.