Page Position Estimation via Spine Endpoint Tracking

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Augmented reality systems face challenges in reliably estimating the position and orientation of fiduciary markers, especially when they become distorted during the turning of book pages, leading to disruptions in image augmentation and user experience.

Innovation Solution

The system employs fiduciary markings on each page of a book, using a video camera and entertainment device like the PlayStation 3 to detect and track these markings, and applies edge detection and Hough transforms to estimate the angular position of turning pages, even in conditions affected by motion blur, by analyzing regions centered on the book's spine endpoints.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If fiduciary markers are used for position estimation in augmented reality systems, then the system can track object position and orientation, but the markers become distorted during page turning which reduces tracking reliability

Engineering Contradiction:
Improveposition estimation reliabilityVSAvoidmarker shape distortion
Core Design Contradiction:
ReliabilityVSShape

Solution Approach 1:

The patent divides the book page into multiple regions and uses multiple fiduciary markers distributed across different pages. When one marker is distorted during page turning, other markers on different pages remain visible and usable, allowing the system to maintain position estimation through the undistorted markers.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a virtual book model as an intermediary that represents the physical book's state. This virtual model tracks page positions and orientations through computational geometry rather than relying solely on direct visual detection of physical markers, allowing the system to infer marker positions even when they are distorted.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Duration of action of moving object

If the system tracks fiduciary markers during page turning, then continuous position estimation is possible, but motion blur affects the accuracy of marker detection

Engineering Contradiction:
Improvecontinuous tracking capabilityVSAvoidmarker position measurement accuracy
Core Design Contradiction:
Duration of action of moving objectVSMeasurement precision

Solution Approach 1:

The patent performs preliminary actions by pre-defining the expected positions and orientations of fiduciary markers on each page of the virtual book model. During tracking, the system compares actual marker positions against these pre-defined expectations, allowing it to compensate for motion blur and maintain accurate position estimation even when markers are not perfectly detected.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from multiple sources including the virtual book model's expected marker positions, the current camera image, and the previously estimated book state. This multi-source feedback allows the system to correct for detection errors caused by motion blur and maintain continuous accurate tracking throughout the page turning process.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2642446B1System and method of estimating page position
Publication Date: 2017.11.01 SONY COMP ENTERTAINMENT EURO LTD
  • EP2642446B1 patent drawingFigure 1~2A
  • EP2642446B1 patent drawingFigure 2B
  • EP2642446B1 patent drawingFigure 3

AI summary

A method of estimating the angular position of a turning leaf in an image of a book comprises the steps of capturing a video image frame of the book, estimating the position of at least a first endpoint of the spine of the book, applying an edge detection operation to the video image frame of the book to generate an edge image, defining a first region having its lower boundary centred upon the first endpoint of the spine of the book, applying a Hough transform to the first region of the edge image to obtain a plurality of line estimates, and rejecting line estimates that do not substantially intersect with an estimated endpoint of the spine of the book; and for line estimates that are not rejected, detecting one or more clusters of angles of said line estimates with respect to an estimated endpoint of the spine of the book, and generating a respective average angle from the or each cluster of angles; then selecting a respective average angle from the available average angles, and estimating the angular position of the turning leaf in the image of the book responsive to the currently selected average angle.