Research on Visual-Inertial Measurement Unit Fusion Simultaneous Localization and Mapping Algorithm for Complex...

Area of Science:

• Robotics
• Computer Vision
• Geospatial Technology

Background:

• Intelligent robots in open-pit mining demand precise localization and digital mapping.
• Traditional visual SLAM struggles with open-pit mine terrain due to pitch variations and low texture.

Purpose of the Study:

• To develop an improved SLAM technique integrating visual and Inertial Measurement Unit (IMU) data.
• To enhance localization accuracy and system robustness for mining robots.

Main Methods:

• A point-line feature fusion matching strategy for robust line feature extraction.
• Integration of an enhanced Line Segment Detection (LSD) algorithm.
• Tightly coupled visual-inertial framework with IMU pre-integration and sliding window optimization.

Main Results:

• Improved RMSE accuracy by 36.62% (MH) and 26.88% (VR) on the EuRoC dataset compared to ORB-SLAM3.
• Significant reduction in trajectory drift in simulated open-pit mining tests.
• Localization accuracy improved by 40.62% and 61.32% in simulated tests.

Conclusions:

• The proposed visual-inertial SLAM method significantly enhances localization precision and robustness.
• This technique addresses the challenges of feature extraction and positioning in complex mining environments.
• The improved system demonstrates significant potential for intelligent mining applications.