6D Object Pose Estimation Using Marker Point Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to precisely detect and determine the situation, including position and orientation, of objects with planar surfaces, such as small load carriers, in logistics facilities, especially under varying light conditions and with complex object geometries.
Innovation Solution
A method involving multiple steps: capturing photographs with depth information, detecting objects using machine learning algorithms, retrieving and aligning 3D models with marker points, performing a plausibility check, computing the 2D pose, selecting points from depth information, and aligning these points with the computed orientation to determine the 6-dimensional situation of the object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection methods are used, then the detection process is simple, but the measurement precision of object situation is insufficient
Solution Approach 1:
The detection process is segmented into multiple specialized steps: object detection to identify the load carrier, marker point detection to locate reference points, pose computation to calculate 2D orientation, and situation determination to compute 6D position and orientation. Each step processes specific features independently, improving overall measurement precision while managing complexity through modular organization.
Solution Approach 2:
The system transitions from 2D image plane detection to 6D situation determination by computing both position (x, y, z) and orientation (roll, pitch, yaw) coordinates. Depth information from the photograph is integrated with 2D pose computation to establish three-dimensional spatial relationships, enabling precise six-dimensional characterization of object situation.
2Measurement precision
If multiple detection steps are implemented, then the measurement precision improves, but the loss of time increases
Solution Approach 1:
A 3D model of the object is retrieved in advance, containing pre-defined marker points with known individual situations relative to the planar surface. This preliminary preparation enables rapid matching during detection, as the system only needs to identify marker point positions in the photograph and compute transformations, rather than establishing the complete 3D model during the detection process itself.
Solution Approach 2:
The system performs plausibility checks on detected marker points to verify detection accuracy. The computed 6D situation is validated against the known 3D model geometry and marker point configurations. This feedback mechanism ensures high measurement precision by detecting and correcting errors, while the automated validation process efficiently manages the additional time required through algorithmic optimization.
3Measurement precision
If marker points are used for pose computation, then the orientation accuracy improves, but the device complexity increases
Solution Approach 1:
A 3D model copy of the object is created and retrieved, containing virtual marker points that correspond to physical marker points on the object. This digital twin enables virtual-to-real matching, where the known geometric relationships in the 3D model are used to interpret 2D image measurements and compute accurate 6D situation without requiring complex physical measurement devices.
Solution Approach 2:
The system transforms the problem from direct 6D measurement to a two-stage parameter computation: first computing 2D pose parameters (position and orientation in image plane) from marker point coordinates, then transforming these to 6D situation parameters by integrating depth information and applying 3D transformation matrices. This parameter transformation approach simplifies the computational geometry while maintaining high orientation accuracy.
Data Source
AI summary
A method for determining a position relative to a capture device of an object having at least one planar surface includes capturing at least one recording of the object includes depth information and image information by a camera; recognizing the object in the image information; retrieving a model of the object includes at least one marker point, the inherent position of which with respect to the at least one planar surface is predefined; plausibilizing the at least one marker point on the object in the recording; calculating a 2-dimensional pose of the at least one planar surface on the basis of the at least one identified marker point in the recording; selecting a point set from the depth information of the recording; and calculating the position of the object by adapting the point set to the calculated pose.