AR Planar Shape Tracking via Projective Invariants
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality (AR) tracking methods face challenges in achieving high accuracy and robustness while being unobtrusive, especially in environments with low texture or complex shapes, and require scene engineering for fiducial-based tracking, which limits user interaction and natural experience.
Innovation Solution
The system employs natural feature tracking (NFT) methods to recognize and track planar shapes using projective invariant signatures and active contours, allowing for real-time camera pose estimation and augmentation of shapes without predefined fiducials, enabling flexible and accurate tracking in various AR applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fiducial-based tracking is used, then tracking accuracy and robustness are improved, but scene engineering is required which limits user interaction and natural experience
Solution Approach 1:
The patent uses projective invariant signatures as a mathematical copy or representation of the planar shape's geometric properties. Instead of requiring physical fiducials, the system creates and tracks a digital signature (set of projective invariants) that uniquely identifies the shape. This allows accurate tracking without scene engineering, as any planar shape can be used as long as its projective invariants can be computed.
Solution Approach 2:
The patent transforms the tracking problem from tracking physical fiducials to tracking projective invariant parameters. By changing the parameter space from physical coordinates to projective invariants (which are invariant under projective transformations), the system achieves fiducial-like accuracy for natural shapes. The camera pose is estimated by solving for the transformation that minimizes the difference between observed and expected projective invariants.
2Ease of operation
If natural feature tracking is used, then user interaction and natural experience are improved, but tracking accuracy and robustness deteriorate in environments with low texture or complex shapes
Solution Approach 1:
The patent extracts the essential geometric features needed for tracking by computing projective invariants from the planar shape's contour or internal features. Instead of relying on texture or complex surface details, the system extracts a minimal set of projective invariant parameters that capture the shape's geometric essence. This extraction process creates a robust representation that works reliably across different viewing conditions and environments.
Solution Approach 2:
The patent changes the parameter representation from raw image features (which are sensitive to texture and lighting) to projective invariants (which are mathematically guaranteed to be invariant under projective transformations). This parameter transformation ensures that tracking accuracy is maintained regardless of camera pose, lighting conditions, or texture variations, directly addressing the reliability issues of traditional NFT methods.
3Ease of operation
If NFT methods are used, then obtrusiveness is reduced and natural experience is improved, but computational complexity increases and accuracy is reduced
Solution Approach 1:
The patent extracts only the necessary projective invariant features from the planar shape, avoiding the need to process entire images or complex texture information. By focusing computation on extracting a small set of invariant parameters (such as distances between feature points, angles, or ratios that are projectively invariant), the system significantly reduces computational complexity compared to full NFT approaches while maintaining natural shape tracking.
Solution Approach 2:
The patent applies partial action by using only the minimal set of projective invariants needed for pose estimation, rather than analyzing all possible features in the image. The system computes projective invariants from a selected subset of feature points on the planar shape, which reduces computational load while providing sufficient information for accurate tracking. This selective approach balances natural experience with computational efficiency.
4Ease of operation
If NFT methods are used, then obtrusiveness is reduced and natural experience is improved, but tracking accuracy is reduced compared to fiducial-based methods
Solution Approach 1:
The patent changes the parameter space to projective invariants, which are mathematically guaranteed to be invariant under projective transformations. This parameter transformation ensures that the same geometric relationships are preserved regardless of camera pose, allowing for accurate pose estimation from natural shapes. The accuracy is improved by formulating the tracking as an optimization problem that minimizes the difference between observed and expected projective invariants, providing fiducial-level precision for natural shapes.
Solution Approach 2:
The patent creates a mathematical copy of the planar shape's geometric properties in the form of projective invariant signatures. By computing and tracking these invariant signatures, the system maintains a consistent representation of the shape's geometry across different views. This copying approach ensures that tracking accuracy is maintained by comparing the observed projective invariants with those computed from the known 3D model of the shape, enabling precise pose estimation without fiducials.
Data Source
AI summary
The present disclosure relates to systems and methods for tracking planar shapes for augmented-reality (AR) applications. Systems for real-time recognition and camera six degrees of freedom pose-estimation from planar shapes are disclosed. Recognizable shapes can be augmented with 3D content. Recognizable shapes can be in form of a predefined library being updated online using a network. Shapes can be added to the library when the user points to a shape and asks the system to start recognizing it. The systems perform shape recognition by analyzing contour structures and generating projective invariant signatures. Image features are further extracted for pose estimation and tracking. Sample points are matched by evolving an active contour in real time.