Material Identification of Scanned Objects Based on the Classification of the Laser Reflection Intensity Profile

Area of Science:

• Robotics
• Computer Vision
• Sensor Technology

Background:

• Autonomous mobile robots require accurate environmental perception for navigation.
• Material classification is crucial for object recognition and scene understanding.
• Existing methods often rely on multi-scan data or multiple sensors, limiting real-time applications.

Purpose of the Study:

• To develop a single-frame material classification method for LiDAR data.
• To enable robots to distinguish materials with varying reflective properties, including glass.
• To enhance autonomous navigation capabilities through improved object identification.

Main Methods:

• Utilizing single-beam echo data from a single LiDAR frame.
• Comparing measured reflection intensity profiles (distance, incidence angle) with reference profiles.
• Employing the gradient of intensity profiles to improve material discrimination.

Main Results:

• Demonstrated reliable material classification in indoor environments.
• Successfully identified materials with diverse reflective properties, including transparent glass.
• Achieved accurate classification using only single-scan LiDAR data.

Conclusions:

• The proposed method offers effective material classification for autonomous mobile robots.
• Single-frame LiDAR data is sufficient for robust material identification, including challenging surfaces like glass.
• Eliminates the need for multi-scan accumulation or multi-sensor fusion, simplifying robot systems.