Multi-scale Fiducials for Distance-adaptive Target Tracking

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing fiducial recognition systems face challenges in varying distances, as scale-invariant algorithms are not always feasible due to processing power limitations, and traditional fiducials do not account for changes in size and position effectively.

Innovation Solution

The use of multi-scale fiducials, comprising a parent, child, and grandchild fiducials of varying sizes and positions, where each scale can independently determine relative positions and encode information, allowing for scale-variant recognition algorithms to compensate for changes in distance and visibility.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional single-scale fiducials are used, then the system is simple to implement, but the recognition accuracy deteriorates when distance varies causing scale changes

Engineering Contradiction:
Improvefiducial recognition accuracyVSAvoidfiducial structure complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements multi-scale fiducials where smaller fiducial instances are nested within or associated with larger ones, creating a hierarchical structure. Each scale level contains fiducial markers that can be independently detected, allowing the system to recognize the fiducial at appropriate scales regardless of distance, thereby maintaining recognition accuracy without requiring complex scale-invariant algorithms.

Inventive Principle:
Principle #7Nested doll (Nesting)

Solution Approach 2:

The patent extends the fiducial structure from a single scale (one dimension of size) to multiple scales (adding a dimensional aspect of size variation). By incorporating fiducial markers at different size levels within the same fiducial structure, the system gains the ability to handle scale changes due to distance variations, improving measurement precision across varying distances.

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

2Measurement precision

If scale-invariant algorithms are used to handle distance variations, then recognition accuracy across distances is improved, but processing power requirements increase

Engineering Contradiction:
Improvefiducial recognition accuracyVSAvoidprocessing power consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent pre-structures the fiducial with multiple scale levels before deployment. This preliminary design ensures that fiducial markers of appropriate sizes are already present in the scene, eliminating the need for complex real-time scale-invariant detection algorithms. The system can directly detect and use the pre-configured multi-scale markers, significantly reducing processing power requirements while maintaining recognition accuracy across distances.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If multi-scale fiducials are implemented, then recognition capability across distances is improved, but the fiducial structure complexity increases

Engineering Contradiction:
Improvedistance adaptabilityVSAvoidfiducial structure complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent employs a nested hierarchical structure where smaller fiducial instances are contained within or associated with larger ones. This nesting approach organizes the multi-scale complexity in a systematic manner, making the structure more manageable and easier to implement compared to completely independent multi-scale elements, thereby balancing adaptability with structural complexity.

Inventive Principle:
Principle #7Nested doll (Nesting)

Data Source

PatentEP3213262B1Multi-scale fiducials
Publication Date: 2020.05.06 AMAZON TECH INC
  • EP3213262B1 patent drawingFigure 1
  • EP3213262B1 patent drawingFigure 2
  • EP3213262B1 patent drawingFigure 3A

AI summary

Disclosed are various embodiments of a multi-scale fiducial. A multi-scale fiducial may have three or more scales, where the child fiducials are nested or otherwise linked by a relative position to the parent fiducials. Multi-scale fiducials may facilitate target identification and tracking at varying distances, potentially without the aid of a scale-invariant recognition algorithm. One application of multi-scale fiducials may involve target identification for autonomously controlled aerial vehicles.