Multi-Camera Extrinsic Calibration with Pose-Based Error Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing extrinsic calibration methods for multi-camera systems suffer from low precision in detected keypoints, leading to inaccurate calibration data, and are prone to errors due to camera displacement, requiring time-consuming manual recalibration.

Innovation Solution

A method for extrinsic calibration using detected poses of objects, particularly human individuals, employs a bundle adjustment technique with a weighted objective function that includes a reprojection error and pose quality functions to refine calibration data, minimizing the impact of observation errors and ensuring high accuracy and robustness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual measurements are used for extrinsic calibration, then calibration data can be obtained, but the process is time consuming and error prone

Engineering Contradiction:
Improvecalibration data accuracyVSAvoidcalibration time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system uses detected poses of objects in the scene as calibration patterns, allowing the cameras to self-calibrate using naturally occurring features rather than requiring external calibration equipment or manual measurements. The calibration process leverages the scene content itself to compute extrinsic parameters automatically.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The method minimizes an error function that combines reprojection error with constraints on object pose parameters, dynamically adjusting calibration parameters based on observed pose inconsistencies across multiple views and time points to achieve accurate calibration without manual intervention.

Inventive Principle:
Principle #35Parameter changes

2Extent of automation

If dedicated calibration equipment is used, then calibration data can be calculated automatically, but the system is sensitive to camera displacement and requires re-calibration

Engineering Contradiction:
Improvecalibration automationVSAvoidcalibration stability
Core Design Contradiction:
Extent of automationVSReliability

Solution Approach 1:

The calibration method works with any object that has detectable pose features, making it universally applicable without requiring specialized calibration equipment. The same calibration approach can be used regardless of camera displacement, as long as pose-detectable objects are present in the scene.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system continuously monitors pose detection quality and uses this feedback to determine when re-calibration is needed. By tracking the consistency of pose measurements across views and time, the system can automatically trigger recalibration only when necessary, improving reliability while maintaining automation.

Inventive Principle:
Principle #23Feedback

3Extent of automation

If keypoints are detected in images for calibration, then extrinsic calibration can be automated, but low precision in detected keypoints propagates into inaccurate calibration data

Engineering Contradiction:
Improvecalibration automationVSAvoidcalibration data precision
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The method introduces pose estimation as an intermediary step between keypoint detection and calibration computation. By first estimating object poses from keypoints and then using these pose estimates as calibration patterns, the system separates the low-precision keypoint detection from the final calibration computation, reducing error propagation.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The approach transitions from using only 2D image plane information to incorporating 3D pose information as an additional dimension. By minimizing an error function that includes both reprojection error and pose consistency constraints, the method leverages temporal and spatial dimensions to improve calibration precision despite imperfect keypoint detection.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

4Reliability

If existing calibration methods are used, then calibration can be performed, but the process is sensitive to camera displacement requiring re-calibration

Engineering Contradiction:
Improvecalibration robustnessVSAvoidre-calibration flexibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The calibration method is designed to be dynamically adaptable to camera displacement. By continuously monitoring pose detection quality and automatically triggering recalibration when pose inconsistencies exceed thresholds, the system maintains robustness while adapting to changes in camera positions without requiring manual intervention.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3882857B1Extrinsic calibration of multi-camera system
Publication Date: 2025.07.30 SONY GROUP CORP
  • EP3882857B1 patent drawingFigure 1A~1B
  • EP3882857B1 patent drawingFigure 1C~2A
  • EP3882857B1 patent drawingFigure 2B~3A

AI summary

A method for extrinsic calibration of cameras in a monitoring system, performed by a calibration device, comprises obtaining (301) preliminary calibration data of the cameras, obtaining (302) a preliminary pose for an object observed by the cameras, and determining (304) calibration data that results in an optimization of an objective function using the preliminary calibration data and the preliminary pose as starting values for the optimization. The method further comprises defining (303) the objective function with calibration data and pose as unknown variables and including a first function (303A) yielding aggregated reprojection error, and a second function (303B) yielding pose quality based on relative orientation and relative distance between 3D points of the pose. The second function enables improved accuracy of the calibration data and may be configured to yield pose quality for individual poses or motion quality for a time sequence of poses.