Monocular SLAM Pose Estimation via Epipolar Constraint Filtering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Monocular visual SLAM technologies face inaccuracies in localization due to feature point mismatches, incomplete matching, and invalid matching during feature point selection, leading to inaccurate position and posture determination of a camera in unknown environments.

Innovation Solution

A method that involves obtaining rotation angular velocities and accelerations of a camera, identifying feature point pairs in two frames of images, constructing epipolar constraint equations, calculating rotation and displacement matrices, and filtering these matrices using reference values from the camera's angular velocities and accelerations to accurately determine the camera's pose.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If feature point matching is performed in monocular visual SLAM to determine camera position and posture, then localization functionality is achieved, but matching errors (mismatches, incomplete matching, invalid matching) occur leading to inaccurate localization

Engineering Contradiction:

Improvelocalization accuracyVSAvoidfeature point matching accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary verification mechanism using epipolar constraint equations as a mathematical mediator to validate feature point matches. The epipolar geometry serves as an intermediate check between raw feature matching and final pose estimation, filtering out incorrect matches before they corrupt the localization result.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent implements feedback by using the calculated rotation and displacement matrices to verify the consistency of feature point matches. The system continuously refines the matching process by comparing predicted feature positions (based on current pose estimates) with actual observed positions, creating a closed-loop verification system that improves localization accuracy.

Inventive Principle:

Principle #23Feedback

2Adaptability or versatility

If multiple rotation and displacement matrices are calculated from feature point pairs, then complete pose estimation is achieved, but incorrect solutions are included reducing accuracy

Engineering Contradiction:

Improvepose estimation completenessVSAvoidpose estimation accuracy

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent extracts and separates the valid pose solution from invalid ones by applying epipolar constraint verification. Among multiple mathematically possible rotation and displacement matrices, the system identifies and extracts only those that satisfy the geometric constraints of the epipolar model, discarding incorrect solutions.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent converts the potential harm of having multiple incorrect pose solutions into a benefit by using them as test cases for validation. The presence of multiple solutions allows the system to apply verification tests (epipolar constraints) that distinguish correct from incorrect solutions, turning the problem of solution multiplicity into an opportunity for robust validation.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

3Productivity

If feature point matching is performed without verification, then processing speed is maintained, but localization accuracy deteriorates due to matching errors

Engineering Contradiction:

Improveprocessing speedVSAvoidlocalization precision

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary verification of feature point matches using epipolar constraint equations before final pose estimation. This preliminary action filters out obviously incorrect matches early in the processing pipeline, preventing them from affecting the final result while minimizing additional computational overhead.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11468599B1Monocular visual simultaneous localization and mapping data processing method apparatus, terminal, and readable storage medium

Publication Date: 2022.10.11 SHENZHEN BAICHUAN SECURITY TECH CO LTD

AI summary

A monocular visual simultaneous localization and mapping (SLAM) data processing method. The SLAM data processing method comprises: obtaining rotation angular velocities and accelerations of a camera cyclically; obtaining a plurality of feature point pairs in two frames of images acquired by the camera, and obtaining pixel coordinate values of feature points in the feature point pairs, where each of the feature point pairs includes two feature points that correspond to a same feature of a same object and that are respectively in the two frames of images; obtaining to-be-selected rotation matrices and to-be-selected displacement matrices according to the pixel coordinate values; obtaining a reference rotation matrix of the camera according to the rotation angular velocities, and obtaining a reference displacement matrix of the camera according to the accelerations; and filtering the to-be-selected rotation matrices and the to-be-selected displacement matrices according to the reference rotation matrix and the reference displacement matrix.