Vehicle Camera Extrinsic Calibration Using Sky-Filtered Visual Odometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera calibration methods for vehicles are prone to errors due to factors like windscreen replacement and mechanical loosening, leading to inaccurate downstream functionalities, and current techniques fail to effectively exclude slowly moving objects like clouds, causing bias in visual odometry.
Innovation Solution
A method for dynamic calibration that involves obtaining images, identifying feature points, removing sky and moving object points, and adjusting extrinsic parameters based on camera and sensor motion trajectories to correct for camera movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional visual odometry is used for camera calibration, then calibration can be performed dynamically, but long-term bias occurs due to slowly moving objects like clouds being mistakenly treated as stationary
Solution Approach 1:
The patent extracts and removes feature points belonging to the sky from the set of feature points used in visual odometry. By identifying sky regions in images and eliminating feature points within these regions, the method prevents slowly moving objects like clouds from being mistakenly treated as stationary, thereby removing the source of long-term bias while preserving dynamic calibration capability
Solution Approach 2:
The patent performs preliminary classification of feature points by determining which ones belong to the sky before the visual odometry calculation. This preliminary action of identifying and marking sky-related feature points allows the system to exclude them in advance, preventing bias introduction during the calibration process while maintaining the ability to perform dynamic calibration
2Measurement precision
If camera pose is adjusted to compensate for movement, then calibration accuracy is maintained, but errors may be introduced in the adjustment process leading to undesired calibration changes
Solution Approach 1:
The patent uses feedback by comparing the camera-based motion estimation (visual odometry) with motion estimation from other motion sensors such as accelerometers, gyroscopes, and wheel encoders. This cross-validation feedback mechanism allows the system to detect and compensate for camera movement while verifying adjustments against independent sensor data, thereby maintaining calibration accuracy without introducing erroneous changes
Data Source
AI summary
A method for calibrating a set of extrinsic parameters of a camera mounted on a vehicle is disclosed. The method includes: obtaining a sequence of images, wherein each image depicts a portion of a surrounding environment of the vehicle determining a set of feature points in the sequence of images; determining an area representing a sky in the sequence of images; removing a subset of feature points belonging to the area representing the sky, thereby forming an updated set of feature points; determining a first motion trajectory of the vehicle based on the updated set of feature points; obtaining a second motion trajectory of the vehicle which is based on motion data obtained from other sensors of the vehicle; and calibrating the camera by adjusting the set of extrinsic parameters of the camera based on a difference between the first motion trajectory and the second motion trajectory.