Camera Parameter Estimation Using Principal-Point Likelihood Heatmaps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera calibration methods fail to accurately estimate camera parameters based on the distribution of image principal points at each pixel, leading to inaccuracies in camera calibration.
Innovation Solution
A camera parameter calculation device that generates a heatmap representing the likelihood of image principal points at each pixel using a learning model trained by machine learning, allowing the estimated image principal points to approach their true values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional geometry-based methods or deep learning methods are used for camera calibration, then the calibration process can be performed, but the camera parameter estimation accuracy is insufficient
Solution Approach 1:
The patent segments the camera calibration problem into two distinct stages: (1) generating a heatmap representing the likelihood distribution of image principal points at each pixel using a learning model, and (2) calculating camera parameters based on this heatmap. This segmentation allows the system to explicitly model and utilize the likelihood distribution, thereby improving estimation accuracy while maintaining calibration reliability.
Solution Approach 2:
The patent introduces a heatmap as an intermediary representation between the input image and the final camera parameter estimation. This heatmap serves as a probability distribution map that highlights likely locations of image principal points, enabling the system to incorporate uncertainty information and likelihood distributions into the calibration process, thus resolving the contradiction between precision and reliability.
2Measurement precision
If a learning model is trained to estimate image principal points, then the estimation can be performed, but the accuracy is limited without utilizing likelihood distribution
Solution Approach 1:
The patent performs preliminary action by generating a heatmap that represents the likelihood distribution of image principal points before final camera parameter calculation. This preliminary heatmap generation captures and preserves the likelihood distribution information that would otherwise be lost, allowing subsequent parameter estimation to benefit from this probabilistic information and achieve higher accuracy.
Solution Approach 2:
The patent implements feedback by using the generated heatmap to guide the camera parameter calculation process. The heatmap provides feedback about the likelihood distribution of principal points, which is then incorporated into the parameter estimation. This feedback mechanism ensures that likelihood distribution information is not lost but actively utilized to improve estimation precision.
Data Source
AI summary
A camera parameter calculation device acquires an image taken by a camera, generates a heatmap representing a likelihood of image principal point at each pixel by inputting the acquired image to a learning model, and calculates a camera parameter of the camera on the basis of the generated heatmap, the learning model being trained by machine learning so as to cause an estimated image principal point indicated by a heatmap to approach a true value for the estimated image principal point.