Robust laser-vision-inertial fusion slam method

By employing a laser-vision-inertial fusion SLAM method, the performance degradation of laser SLAM and visual SLAM algorithms in degraded scenes is addressed, achieving high robustness and high accuracy in real-time state estimation and mapping.

CN117782050BActive Publication Date: 2026-07-10FUDAN UNIVERSITY
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
FUDAN UNIVERSITY
Filing Date
2022-09-22
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing laser SLAM and visual SLAM algorithms perform poorly in degraded scenes, lacking geometric structure and optical texture information, leading to degraded localization performance and mapping failure.

Method used

A robust laser-vision-inertial fusion SLAM method is adopted. Through joint initialization, degradation detection and processing, multi-sensor feature layer fusion and back-end optimization, combined with observation data from lidar, monocular camera and IMU sensor, a highly robust and accurate state estimation and mapping is achieved.

Benefits of technology

The system significantly improves stability and accuracy in degradation scenarios, enhances the robustness of feature extraction, achieves high-confidence, low-latency degradation detection and processing, and outputs a tightly coupled, highly robust odometer.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117782050B_ABST
    Figure CN117782050B_ABST
Patent Text Reader

Abstract

The application provides a robust laser-vision-inertial fusion SLAM method, which improves the speed and robustness of system initialization in a degenerative scene by jointly initializing a radar inertial system and a vision inertial system, realizes feature enhancement in the degenerative scene by correlatively fusing laser depth features and vision texture features, realizes degenerative detection and degenerative robust enhancement of the front-end program by a CUSUM control chart based on a sequential probability ratio test, realizes degenerative detection and processing with high confidence, low time delay and high robustness, and finally globally optimizes a multi-sensor constraint output laser-vision-inertial odometer. Through the method of the application, the performance degradation problem of the existing SLAM algorithm without degenerative robustness enhancement in various degenerative scenes can be well overcome, and real-time state estimation and mapping with high robustness and high precision in the degenerative scene are realized.
Need to check novelty before this filing date? Find Prior Art