Corner Camera Surround View Calibration for Accurate Bird's-Eye Stitching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle 360° surround view systems with multiple cameras face challenges in calibration, leading to misregistration artifacts and complexity in stitching images, which complicates installation and reduces the clarity of composite images for operators.
Innovation Solution
A vehicle 360° surround view system using two corner-placed cameras with overlapping fields of view, combined with a calibration method that adjusts transformation parameters to minimize registration errors and enhance image alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple cameras are used to provide 360° surround view, then the coverage area is improved, but the calibration complexity and image stitching difficulty increase
Solution Approach 1:
The patent merges the calibration processes of multiple cameras into a unified system-level calibration. Instead of calibrating each camera independently and then stitching images, the system performs global calibration that simultaneously optimizes all camera parameters and their spatial relationships, reducing overall calibration complexity while maintaining complete 360° coverage
Solution Approach 2:
The patent introduces a bird's eye view dimension by projecting all camera images onto a common top-down plane. This dimensional transformation allows the system to handle the complexity of multiple camera angles and positions by mapping them to a single standardized view, simplifying the calibration and stitching process while providing comprehensive coverage
2Measurement precision
If manual calibration techniques are used, then the image registration accuracy is improved, but the time consumption and operator effort increase
Solution Approach 1:
The system implements self-calibration capabilities where the calibration process automatically determines camera parameters and spatial relationships without requiring manual intervention. The system uses image processing algorithms to detect features, compute transformations, and optimize registration automatically, maintaining high accuracy while eliminating time-consuming manual operations
Solution Approach 2:
The patent incorporates feedback mechanisms where the system continuously evaluates image registration quality and automatically adjusts calibration parameters to optimize alignment. This closed-loop approach ensures high registration accuracy while minimizing calibration time by iteratively improving results based on measured performance
3Area of stationary object
If cameras are placed at various positions on the vehicle, then the surround view coverage is improved, but the image stitching and composite image quality deteriorate
Solution Approach 1:
The patent transforms images from multiple camera positions by projecting them onto a common bird's eye view plane. This dimensional transformation allows the system to integrate images from various vehicle positions while maintaining geometric consistency and high composite image quality, as all views are rendered from a standardized top-down perspective
Solution Approach 2:
The system dynamically adjusts transformation parameters including projection angles, scaling factors, and alignment offsets to optimize the composite image quality. By continuously optimizing these parameters based on the specific camera positions and vehicle geometry, the system maintains high image quality across all surround view coverage areas
Data Source
Figure 1a~1c
Figure 1d~1e
Figure 2a
AI summary
Cameras having wide fields of view are placed at each of the front left, front right, rear left, and rear right corners of a nominally rectangular-shaped vehicle, thereby providing a continuous region of overlapping fields of view completely surrounding the vehicle, and enabling complete 360 stereoscopic vision detection around the vehicle. In an embodiment the regions of overlapping fields of view completely surround the vehicle. The cameras are first individually calibrated, then collectively calibrated considering errors in overlapping viewing areas to develop one or more calibration corrections according to an optimization method that adjusts imaging parameters including homography values for each of the cameras, to reduce the overall error in the 360 surround view image of the continuous region surrounding the vehicle.