Neural Network Camera Parameter Learning via Spherical Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating camera parameters, such as those in Non-Patent Literature 1 and 2, face challenges in ease of calculation and accuracy, particularly for cameras with large lens distortion like fisheye cameras, due to complexity and insufficient expression of lens distortion.
Innovation Solution
A method involving a neural network that learns network parameters by projecting three-dimensional coordinate points onto a unit spherical surface and a predetermined plane, using true and estimated camera parameters to minimize the distance between true and estimated two-dimensional coordinate points, enabling accurate and easy calculation of camera parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional camera calibration methods are used, then calculation can be performed, but calculation accuracy is insufficient especially for cameras with large lens distortion
Solution Approach 1:
The patent transforms the camera calibration problem from direct parameter estimation to an optimization problem where the objective function (projection error) is minimized by adjusting camera parameters. This changes the approach from solving equations directly to iteratively optimizing parameters, improving accuracy especially for fisheye cameras with large distortion
Solution Approach 2:
The patent introduces an intermediate optimization process between the raw image data and the final camera parameters. By using projection error minimization as an intermediary step, the method bridges the gap between simple calculation and high accuracy, achieving both computational efficiency and precision
2Ease of operation
If simple camera parameter calculation methods are used, then calculation is easy, but calculation accuracy is insufficient
Solution Approach 1:
The patent implements an automated optimization process that self-adjusts camera parameters to minimize projection error. The system serves itself by iteratively refining parameters without manual intervention, achieving high accuracy while maintaining ease of operation through automation
Solution Approach 2:
The patent introduces feedback through the projection error calculation. The estimated camera parameters are used to project 3D points, the projection error is computed and fed back to refine the parameters, creating a closed-loop system that automatically improves accuracy while remaining computationally efficient
Data Source
AI summary
Provided is a method for learning a network parameter of a neural network including, by an information processor: acquiring a learning image; acquiring a true camera parameter related to the learning image; calculating a true two-dimensional coordinate point by projecting a three-dimensional coordinate point on a unit spherical surface onto a predetermined plane by using the true camera parameter; calculating an estimated two-dimensional coordinate point by projecting the three-dimensional coordinate point onto the predetermined plane by using the estimated camera parameter estimated by the neural network; and learning the network parameter of the neural network based on a distance between the true two-dimensional coordinate point and the estimated two-dimensional coordinate point.