3D Bubble-View Panorama Reconstruction via Multi-View Stereo

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current 2D panorama techniques lack 3D navigation and measurement capabilities essential for factory planning and re-planning, and existing 3D laser scanning methods are expensive and less portable.

Innovation Solution

The development of an image-based 3D panorama technique, referred to as 'bubble-view,' which uses a consumer point-and-shoot camera for image acquisition, reconstructs 3D geometry through structure from motion and multi-view stereo with cylindrical surface sweeping, and registers multiple bubble-views in a common coordinate system for 3D navigation and visualization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If 2D panorama techniques are used, then the system is simple and cost-effective, but 3D navigation and measurement capabilities are lacking

Engineering Contradiction:
Improvesimplicity and cost-effectivenessVSAvoid3D navigation and measurement capabilities
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent transforms 2D panorama images into 3D representations by reconstructing depth information through multi-view stereo geometry. Multiple 2D images taken from different positions are processed to generate 3D point clouds and surface meshes, enabling 3D navigation and measurement while keeping the capture device simple (consumer camera).

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

Solution Approach 2:

The system creates virtual 3D copies of physical spaces using photographs taken with ordinary cameras. By reconstructing 3D geometry from 2D images through structure-from-motion and multi-view stereo algorithms, the patent produces digital twins that enable 3D interaction without requiring expensive laser scanners.

Inventive Principle:
Principle #26Copying

2Adaptability or versatility

If 3D laser scanning methods are used, then 3D navigation and measurement capabilities are achieved, but cost and portability are compromised

Engineering Contradiction:
Improve3D navigation and measurement capabilitiesVSAvoidcost and portability
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent replaces expensive, specialized 3D laser scanning equipment with inexpensive consumer-grade cameras. Multiple standard digital cameras or smartphone cameras can be used to capture the necessary images, dramatically reducing cost while achieving comparable 3D reconstruction results through computational photography methods.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Solution Approach 2:

The patent substitutes active mechanical 3D laser scanning systems with passive optical photography systems. Instead of using laser rangefinders and mechanical scanning mirrors, the system uses static or handheld cameras to capture images, then reconstructs 3D geometry computationally from the 2D photographs through multi-view stereo and structure-from-motion algorithms.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP2976748B1Image-based 3D panorama
Publication Date: 2017.08.02 SIEMENS INDUSTRY SOFTWARE INC
  • EP2976748B1 patent drawingFigure 1
  • EP2976748B1 patent drawingFigure 2~3
  • EP2976748B1 patent drawingFigure 4

AI summary

Various disclosed embodiments include methods, systems, and computer-readable media for generating a 3-dimensional (3D) panorama. A method includes receiving (1005) images of a 3D scene. The method includes reconstructing (1010) geometry of a plurality of 3D bubble-views from the images. Reconstructing includes using a structure from motion framework for camera localization (515, 520), generating a 3D surface mesh model of the scene using multi-view stereo via cylindrical surface sweeping for each bubble-view, and registering multiple 3D bubble-views in a common coordinate system. The method includes displaying (1020) the surface mesh model