Vision-Only Pose Relocalization via Coarse-Fine Matching

Solution Overview

Problem

Autonomous vehicles and robots face challenges in maintaining accurate position and orientation within environments without relying on external navigation systems, such as GPS or inertial sensors, due to the 'lost robot problem', where they become disoriented and lose their position.

Innovation Solution

A vision-only pose relocalization system that uses an image acquisition device to acquire query frames, compares them to a stored image database using coarse and fine matching algorithms, and performs a pose-solving algorithm to estimate the object's position in six degrees of freedom, relying solely on image data for localization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If vision-only pose relocalization is implemented without external navigation systems, then navigation autonomy and reliability in GPS-denied environments are improved, but system complexity and computational requirements increase

Engineering Contradiction:

Improvenavigation reliabilityVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The matching process is divided into two distinct stages: coarse matching using global descriptors to identify candidate frames, and fine matching using local descriptors to precisely determine pose. This segmentation reduces computational complexity by filtering out obviously mismatched frames in the coarse stage, allowing the computationally intensive fine matching to be applied only to a small subset of candidates.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces traditional mechanical navigation sensors (GPS, inertial measurement units) with a vision-based system that uses image processing and computer vision algorithms. This substitution eliminates dependency on external navigation systems while achieving comparable or superior reliability in GPS-denied environments through purely optical means.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If coarse and fine matching algorithms are used to identify matching images, then pose estimation accuracy is improved, but processing time and computational load increase

Engineering Contradiction:

Improvepose estimation accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The matching process is divided into two distinct stages: coarse matching using global descriptors to identify candidate frames, and fine matching using local descriptors to precisely determine pose. This segmentation reduces computational complexity by filtering out obviously mismatched frames in the coarse stage, allowing the computationally intensive fine matching to be applied only to a small subset of candidates.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The coarse matching algorithm performs preliminary filtering by identifying only those database frames that have a reasonable likelihood of matching the query frame. This preliminary action eliminates the need to perform fine matching on all database frames, significantly reducing processing time while maintaining accuracy by ensuring that the true match is not eliminated in the coarse stage.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If multiple algorithms (coarse-matching, fine-matching, pose-solving) are executed sequentially, then localization accuracy is improved, but computational resource consumption increases

Engineering Contradiction:

Improvelocalization accuracyVSAvoidcomputational energy consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The matching process is divided into two distinct stages: coarse matching using global descriptors to identify candidate frames, and fine matching using local descriptors to precisely determine pose. This segmentation reduces computational complexity by filtering out obviously mismatched frames in the coarse stage, allowing the computationally intensive fine matching to be applied only to a small subset of candidates.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The coarse matching algorithm performs preliminary filtering by identifying only those database frames that have a reasonable likelihood of matching the query frame. This preliminary action eliminates the need to perform fine matching on all database frames, significantly reducing processing time while maintaining accuracy by ensuring that the true match is not eliminated in the coarse stage.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentEP4481694A1Vision-only pose relocalization

Publication Date: 2024.12.25 HONEYWELL INTERNATIONAL INC

Figure 1
Figure 2
Figure 3

AI summary

Systems and methods for vision-only pose relocalization include an image acquisition device on an object that acquires a query frame of an environment containing the object; and a memory device that stores an image database. Further, the system can also include a processor that executes computer-readable instructions that direct the processor to perform a coarse-matching algorithm to identify a set of coarsely matched frames of data stored in the image database that coarsely match the query frame; perform a fine-matching algorithm to identify a candidate image in the set of coarsely matched frames that match the query frame; designate the candidate image as a matching image based on whether the candidate image satisfies a validity check; and perform a pose-solving algorithm based on the acquired query frame, the matching image, and parameters for the image acquisition device to estimate a pose of the object in six degrees of freedom.