Multi-Camera Depth Measurement Using Noise Variance Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image capturing systems with fish-eye lenses face challenges in stereoscopic depth measurement due to the epipole issue, where the accuracy of depth measurement is compromised by noise variance, especially when cameras are positioned close to each other, leading to increased occlusion and reduced accuracy.
Innovation Solution
An image capturing apparatus with at least three cameras, each equipped with a fish-eye lens, acquires images and camera parameters, extracts image pairs, calculates three-dimensional position information, and performs weighted addition using effective baseline length or viewing angle to minimize depth measurement error by accounting for noise variance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If fish-eye lenses with wide angle of view are used to enable depth measurement in wide range, then measurement range is improved, but measurement precision deteriorates due to epipole issue and noise variance
Solution Approach 1:
The patent segments the depth measurement task by using multiple camera pairs from a three-camera system. Each camera pair measures depth in specific regions where epipole issues are avoided, and the results are integrated to achieve complete coverage with high precision throughout the wide field of view.
Solution Approach 2:
The patent merges depth measurement results from multiple camera pairs through weighted addition. By combining measurements from different camera pairs that observe the same target from different angles, the system achieves high-precision depth measurement across the entire wide range while compensating for individual camera pair limitations.
2Device complexity
If cameras are positioned close to each other to reduce occlusion, then device complexity is reduced, but measurement precision deteriorates due to increased noise variance
Solution Approach 1:
The patent applies feedback by using weighted addition where the weight for each camera pair's measurement is determined by the inverse of its noise variance. This feedback mechanism automatically adjusts the contribution of each measurement based on its quality, ensuring high precision even when cameras are positioned close together.
3Device complexity
If conventional weighted addition is performed without considering noise variance, then device complexity is reduced, but measurement precision deteriorates due to unaccounted noise distribution
Solution Approach 1:
The patent changes the weighting parameter from uniform or simple disparity-based weights to weights based on noise variance. By calculating the noise variance for each camera pair's measurement and using its inverse as the weight, the system optimizes measurement precision while maintaining computational feasibility.
Data Source
AI summary
An image capturing apparatus: calculates, for each of pairs of images, based on the pairs of images and camera parameters, position information including information specifying a three-dimensional position of a measurement point with respect to a pair of pixels respectively included in the pair of images, and information specifying a position of each of the pair of pixels; performs, based on the position information and the camera parameters, weighted addition on the information specifying the three-dimensional position of the measurement point in each of the pair of images, with information using an effective baseline length or a viewing angle, with respect to the three-dimensional position, of a pair of the cameras corresponding to the pair of images; and outputs position information of the measurement point to which a weighted addition value is applied.