3D-2D Pose Estimation for Accurate Depth Orientation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for estimating the position and orientation of an object using 2D images have low accuracy in the camera's depth direction due to inadequate estimation of translation z, rotation ϕ, and rotation γ parameters.

Innovation Solution

A position and orientation estimation apparatus that optimizes translation in the X-axis and Y-axis directions and rotation about the Z-axis using a 2D image, while utilizing 3D data for accurate positioning, thereby improving estimation accuracy in the depth direction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If position/orientation estimation is performed using only 2D image data, then the estimation process is simple and fast, but the accuracy in the camera's depth direction (translation z, rotation φ, rotation γ) is low

Engineering Contradiction:
Improveposition/orientation estimation accuracy in depth directionVSAvoidestimation process complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the position/orientation estimation process into two separate stages: first estimating position/orientation using 2D image data, then refining the depth-direction parameters (translation z, rotation φ, rotation γ) using 3D data. This segmentation allows each stage to focus on specific parameters, improving overall accuracy while maintaining process efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent combines 2D image data and 3D data in a two-stage estimation process. The first stage uses 2D image data for initial estimation, and the second stage integrates 3D data to refine depth-direction parameters, achieving high accuracy without requiring complex simultaneous processing of all parameters.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If all six position/orientation parameters are optimized simultaneously using combined 3D and 2D data, then comprehensive accuracy is achieved, but the computational complexity and processing time increase significantly

Engineering Contradiction:
Improveoverall position/orientation estimation accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the optimization process into two distinct phases: Phase 1 optimizes parameters using 2D image data, and Phase 2 optimizes depth-direction parameters using 3D data. This segmentation reduces computational complexity at each stage compared to simultaneous optimization of all six parameters.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary estimation of position/orientation using 2D image data before refining with 3D data. This preliminary action provides a good initial estimate that reduces the computational burden of the subsequent refinement stage, decreasing overall processing time.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3501759B1Position and orientation estimation apparatus, position and orientation estimation method, and program
Publication Date: 2022.08.03 OMRON CORP
  • EP3501759B1 patent drawingFigure 1
  • EP3501759B1 patent drawingFigure 2
  • EP3501759B1 patent drawingFigure 3

AI summary

Provided is a position/orientation estimation apparatus capable of increasing the accuracy of estimating a position and an orientation of an object, compared with a conventional apparatus, even if both pieces of information on the 3D data and a 2D image are used, for example. A three-dimensional detailed position/orientation estimation apparatus (140) includes a first position/orientation estimation unit (145) and a second position/orientation estimation unit (146) that are configured to estimate three-dimensional position and orientation. The first position/orientation estimation unit (145) optimizes six parameters (translations x, y, and z, and rotations ϕ, γ, and θ) using 3D data, and the second position/orientation estimation unit (146) optimizes only three parameters (translations x and y, and rotation θ) that can be estimated with high accuracy using a 2D image, based on the result of the three-dimensional position/orientation estimation performed by the first position/orientation estimation unit (145) using the 3D data.