3D-GeoArcs for Multi-Modal Image Registration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image registration techniques face challenges in accurately registering multi-modal images with 3D representations, especially in areas with insufficient ground-level image coverage, varying viewpoints, and different sensor modalities, due to reliance on direct feature matching and complex geometric models.

Innovation Solution

The method employs 3D-GeoArcs to determine the sensor's viewpoint by identifying feature pairs in both the target image and a 3D reference model, generating geoarc surfaces that represent relationships between features, and refining these surfaces to create a composite 3D representation, allowing for flexible and accurate fusion of multi-modal data without requiring precise geometric models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If direct feature matching techniques are used to register multi-modal images with 3D representations, then registration accuracy can be improved in areas with sufficient ground-level image coverage, but the technique fails in areas with insufficient ground-level image coverage and cannot handle varying viewpoints effectively

Engineering Contradiction:
Improveregistration accuracyVSAvoidapplicability to areas with insufficient ground-level coverage
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent introduces 3D-GeoArcs as an intermediary geometric construct that mediates between 2D image features and 3D spatial representation. Instead of directly matching features between multi-modal images and 3D representations, the system projects 3D geoarc surfaces into 2D image space and uses their intersections to infer sensor viewpoints and perform registration, enabling operation in areas without sufficient ground-level coverage

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transitions from 2D feature matching to 3D geometric reasoning by introducing 3D-GeoArcs and 3D reference models. The system creates 3D geoarc surfaces from 2D image features, performs intersections in 3D space to determine sensor viewpoints, and then uses these viewpoints for registration, effectively solving the adaptability problem by operating in an additional dimension

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

2Reliability

If complex geometric models are used to handle varying viewpoints and multi-modal data, then registration robustness can be improved, but computational intensity increases significantly

Engineering Contradiction:
Improveregistration robustnessVSAvoidcomputational intensity
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the complex registration problem into distinct geometric operations: (1) creating 3D-GeoArcs from 2D image features, (2) projecting these arcs into 3D space as surfaces, (3) finding intersections of these surfaces to determine sensor viewpoints, and (4) using these viewpoints for registration. This segmentation breaks down the computationally intensive problem into more manageable geometric operations that can be performed efficiently

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent replaces complex iterative optimization and machine learning-based geometric modeling with direct geometric constructions and intersections. Instead of using heavy computational mechanisms like deep learning feature matching or iterative closest point algorithms, the system uses analytical geometric operations (line intersections, surface intersections) that have closed-form solutions and lower computational requirements

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

3Measurement precision

If appearance-based feature matching is used to correlate landmarks between images, then registration can be performed with expert input, but the process requires manual identification of landmark features and is time-consuming

Engineering Contradiction:
Improvelandmark correspondence accuracyVSAvoidtime for expert landmark identification
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent enables the system to automatically identify and match features without requiring expert input. The 3D-GeoArc methodology automatically detects features in images, creates corresponding 3D geoarc surfaces, and computes intersections to determine registration parameters, making the system self-sufficient and eliminating the time-consuming manual landmark identification process

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP2660777B1Image registration of multimodal data using 3D geoarcs
Publication Date: 2019.01.02 THE BOEING CO
  • EP2660777B1 patent drawingFigure 1
  • EP2660777B1 patent drawingFigure 2
  • EP2660777B1 patent drawingFigure 3A~3B

AI summary

An accurate, flexible and scalable technique for multi-modal image registration is described, a technique that does not need to rely on direct feature matching and does not need to rely on precise geometric models. The methods and/or systems described in this disclosure enable the registration (fusion) of multi-modal images of a scene (700) with a three dimensional (3D) representation of the same scene (700) using, among other information, viewpoint data from a sensor (706, 1214, 1306) that generated a target image (402), as well as 3D-GeoArcs. The registration techniques of the present disclosure may be comprised of three main steps, as shown in FIG. 1. The first main step includes forming a 3D reference model of a scene (700). The second main step includes estimating the 3D geospatial viewpoint of a sensor (706, 1214, 1306) that generated a target image (402) using 3D-GeoArcs. The third main step includes projecting the target image's data into a composite 3D scene representation.