Stereoscopic Camera Calibration Using Optical Flow and Epipolar Lines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Classical stereoscopic vision systems with identical cameras on a single imaging plane face challenges in diversity of imaging devices and varying locations, leading to complex system installation and high computational loads during image analysis, particularly in applications like advanced driver-assistance systems (ADAS) where cameras have different characteristics and locations.
Innovation Solution
A hybrid stereoscopic system using multiple, identical and/or non-identical cameras at initially unknown locations, where a processor calibrates the cameras by identifying optical flow and applying epipolar geometric constraints to converge epipolar lines, reducing computational costs and improving detection capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If cameras are placed at fixed positions on a single imaging plane, then mathematical complexity for object matching is reduced, but system installation becomes complicated and mutual field of view is reduced
Solution Approach 1:
The patent transitions from static single-plane camera arrangements to dynamic multi-plane configurations. Cameras are allowed to be positioned at different locations and orientations (different imaging planes) while the system adaptively calculates epipolar geometry parameters to maintain matching accuracy, thus resolving the contradiction between installation flexibility and matching complexity.
Solution Approach 2:
The patent changes the geometric parameters of the stereoscopic system by allowing cameras to operate on different imaging planes with varying positions and orientations. By dynamically adjusting epipolar geometry parameters (fundamental matrix, epipolar lines) based on actual camera configurations, the system maintains mathematical tractability while gaining installation versatility.
2Reliability
If diverse cameras with different characteristics and locations are used, then detection capabilities are improved, but computational load during image analysis increases significantly
Solution Approach 1:
The patent performs preliminary calibration to establish epipolar geometry parameters (fundamental matrix, epipolar lines) before actual image matching. This pre-computation creates a geometric framework that constrains subsequent matching operations, reducing the search space and computational load during real-time analysis while maintaining accurate matching across diverse camera configurations.
Solution Approach 2:
The patent introduces epipolar geometry as an intermediary mathematical framework that mediates between diverse camera configurations and the image matching process. By projecting the fundamental geometric constraints onto epipolar lines, the system simplifies the matching problem from a 2D search to a 1D search along constrained lines, significantly reducing computational complexity.
3Measurement precision
If epipolar lines are made to converge to an origin point through calibration, then pixel matching accuracy is improved, but calibration complexity increases
Solution Approach 1:
The patent employs feedback mechanisms in the calibration process by iteratively adjusting camera position parameters and computing corresponding epipolar geometry transformations. The system uses detected feature points and their correspondences across images to refine calibration parameters, creating a closed-loop process that improves matching accuracy while automating the calibration complexity.
Data Source
AI summary
A system and method of stereoscopic image processing by at least one processor may include receiving, from a first imaging device, having a first field of view (FOV), and located at a first, initially unknown position, a first image of a scene; receiving, from a second imaging device, having a second, different FOV, and located at a second, initially unknown position, a second image of the scene; calculating a plurality of flow lines in the first image, wherein each flow line represents an optical flow between a pixel of the first image and a corresponding pixel of the second image; and calibrating the imaging devices by determining at least one parameter of relative position between the first imaging device and second imaging device, based on the calculated flow lines.