Multi-sensor datasets of ultrasonic and mmWave reflected signals for material classification

Area of Science:

• Robotics and Sensor Technology
• Materials Science and Engineering

Background:

• Contactless material identification is crucial for various industrial applications.
• Existing datasets may lack comprehensive data from diverse sensing modalities.
• Standardized datasets are needed for training and validating material classification algorithms.

Purpose of the Study:

• To create and release a publicly accessible dataset of reflected signals from ultrasonic and millimetre-wave (mmWave) sensors.
• To facilitate research in contactless material classification using multiple sensor types.
• To provide a benchmark for developing and evaluating material identification algorithms.

Main Methods:

• Collected data using URM09 ultrasonic (40 kHz) and DFRobot mmWave C4001 (24 GHz) sensors.
• Tested six common materials (wood, plastic, metal, glass, cardboard, asbestos) at various thicknesses.
• Processed raw signals using Savitzky-Golay filtering and sliding window techniques.
• Extracted time-domain and frequency-domain features from sensor data.

Main Results:

• Generated a dataset with 1789 refined data points for ultrasonic sensing and 1504 for mmWave sensing.
• Included records for material-thickness combinations and material-only classification.
• The dataset covers diverse materials and sensor characteristics for robust analysis.

Conclusions:

• The released dataset provides valuable data for contactless material identification research.
• It supports advancements in robotics, industrial inspection, and non-destructive testing.
• This resource is expected to accelerate the development of sensor fusion and material classification techniques.