Non-rigid monocular simultaneous localization and mapping method and device based on dense optical flow

CN122156302APending Publication Date: 2026-06-05INST OF AUTOMATION CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
INST OF AUTOMATION CHINESE ACAD OF SCI
Filing Date
2026-02-11
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing monocular deformable V-SLAM methods suffer from insufficient tracking stability and reconstruction accuracy in abrupt changes in illumination, weakly textured regions, and highly dynamic deformation scenes. Traditional sparse optical flow algorithms perform poorly under these conditions, and dynamic deformation map management strategies are prone to introducing incorrect or missing edges.

Method used

We employ the RAFT dense optical flow algorithm based on deep learning for short-term data association, combined with an adaptive threshold strategy to manage dynamic deformation maps, and introduce viscous and elastic regularization terms to jointly estimate camera pose and deformation, and perform lost point recovery and deformable mapping.

Benefits of technology

It improves the accuracy and tracking stability of 3D reconstruction, especially in complex dynamic environments, reducing tracking failures and reconstruction errors, and enhancing the robustness and accuracy of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122156302A_ABST
    Figure CN122156302A_ABST
Patent Text Reader

Abstract

The application belongs to the technical field of simultaneous localization and mapping, and discloses a non-rigid monocular simultaneous localization and mapping method and device based on dense optical flow; wherein the non-rigid monocular simultaneous localization and mapping method comprises the following steps: obtaining a candidate feature set based on a current frame image; calling a RAFT dense optical flow algorithm to obtain a dense optical flow vector based on the current frame image and a previous frame image; sequentially performing sparse processing and structure similarity index screening processing on the dense optical flow vector based on the candidate feature set to obtain an effective feature point association result; updating a dynamic deformation graph by adopting an adaptive threshold strategy based on the effective feature point association result; performing camera pose and deformation joint estimation based on the updated dynamic deformation graph, and performing missing point recovery and deformable mapping based on the joint estimation result. The technical scheme disclosed by the application can alleviate the tracking failure problem in a complex dynamic environment and improve the three-dimensional reconstruction accuracy.
Need to check novelty before this filing date? Find Prior Art