Design of Dust-Filtering Algorithms for LiDAR Sensors Using Intensity and Range Information in Off-Road Vehicles.

Area of Science:

• Robotics and Sensor Technology
• Computer Vision
• Environmental Sensing

Background:

• LiDAR sensors offer high-resolution point cloud data crucial for robotics.
• Dust environments significantly degrade LiDAR sensor performance, leading to perception failures.
• Existing dust removal filters often compromise original data integrity.

Purpose of the Study:

• To design and evaluate an effective intensity-based filter for removing dust particles from LiDAR data.
• To improve the reliability of LiDAR-based perception systems in dusty conditions.
• To compare the proposed filter's performance against conventional dust removal methods.

Main Methods:

• Developed a two-step intensity-based filtering approach for LiDAR data.
• Step 1: Identified potential dust points using LiDAR intensity information.
• Step 2: Analyzed point density around candidate points, removing those below a threshold.

Main Results:

• Experimental datasets were collected and manually labeled in dusty environments.
• The proposed filter demonstrated superior performance compared to conventional filters.
• Achieved the highest F1 score, indicating effective dust removal and data preservation.

Conclusions:

• The novel intensity-based filter successfully removes dust from LiDAR data.
• The filter enhances robotic perception reliability in challenging dusty environments.
• Outperforms existing methods by effectively removing dust without sacrificing surrounding data quality.