Self-Encoded Marker for Robust 3D Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D motion tracking methods in MRI and other imaging modalities are limited by the need for full marker visibility, susceptibility to lens distortions, and large pose changes, leading to reduced precision and accuracy, especially when part of the marker is obstructed or out of focus.
Innovation Solution
A self-encoded marker design with imprinted feature locations and rotational-invariant codes allows for 3D pose determination even from a subset of the marker, using a look-up table to associate quad codes with positions, enabling robust pose estimation and motion tracking with reduced tracking errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional optical motion tracking methods are used with full marker visibility requirements, then pose estimation can be achieved under ideal conditions, but tracking fails when part of the marker is obstructed or out of focus
Solution Approach 1:
The marker is segmented into multiple independently encodable regions (quads), each containing a rotational-invariant code that identifies its position on the marker. This segmentation allows the system to reconstruct full marker pose information from any subset of visible quads, eliminating the requirement for full marker visibility and enabling reliable tracking under partial obstruction or out-of-focus conditions
Solution Approach 2:
Each quad on the marker is equipped with unique local encoding information (rotational-invariant code) that identifies its specific position and orientation on the marker. This local quality ensures that even when only part of the marker is visible, the encoded position information in each visible quad allows the system to determine the marker's overall pose without requiring the entire marker to be in focus or visible
2Area of stationary object
If large camera apertures and large FOV are used to capture the entire marker, then the marker can be fully visible, but lens distortion and focus issues impair precision and accuracy
Solution Approach 1:
The system requires only partial visibility of the marker (a subset of quads) rather than the entire marker. This partial action approach allows the use of smaller camera apertures and narrower FOV, thereby reducing lens distortion and focus issues while still enabling accurate pose estimation through the position self-encoding capability of the marker quads
Solution Approach 2:
The invention changes the fundamental parameter of marker visibility from requiring 100% visibility to requiring only a subset of quads. This parameter change allows the system to operate with smaller FOV and aperture settings, improving measurement precision by reducing geometric distortions while maintaining the ability to track pose accurately through the encoded position information in visible quads
3Loss of information
If the entire marker must be within the camera FOV to ensure full visibility, then complete marker information is available, but the FOV must be large which increases geometric distortions
Solution Approach 1:
The marker information is segmented into multiple independently encodable quads, each containing complete position identification information. This segmentation allows the system to reconstruct full marker pose information from any subset of visible quads, eliminating the need for large FOV and thereby reducing geometric distortions while maintaining complete marker information availability
Solution Approach 2:
Each quad on the marker contains a copy of position identification information (rotational-invariant code) that allows it to be independently recognized and localized. This copying of position information to each marker element enables the system to determine full marker pose from partial marker visibility, reducing the required FOV and minimizing geometric distortions
Data Source
AI summary
The tracking and compensation of patient motion during a magnetic resonance imaging (MRI) acquisition is an unsolved problem. A self-encoded marker where each feature on the pattern is augmented with a 2-D barcode is provided. Hence, the marker can be tracked even if it is not completely visible in the camera image. Furthermore, it offers considerable advantages over a simple checkerboard marker in terms of processing speed, since it makes the correspondence search of feature points and marker-model coordinates, which are required for the pose estimation, redundant. Significantly improved accuracy is obtained for both phantom experiments and in-vivo experiments with substantial patient motion. In an alternative aspect, a marker having non-coplanar features can be employed to provide improved motion tracking. Such a marker provides depth cues that can be exploited to improve motion tracking. The aspects of non-coplanar patterns and self-encoded patterns can be practiced independently or in combination.