Integrated Modeling and Target Classification Based on mmWave SAR and CNN Approach.

Area of Science:

• Applied Physics
• Computer Science
• Electrical Engineering

Background:

• Security screening technologies require advanced methods for threat detection.
• Millimeter-wave (mm-Wave) Frequency-Modulated Continuous-Wave (FMCW) radar offers potential for high-resolution imaging.
• Convolutional Neural Networks (CNNs) excel at image classification tasks.

Purpose of the Study:

• To develop and validate a numerical modeling approach for reconstructing and classifying concealed weapons.
• To assess the effectiveness of mm-Wave FMCW radar combined with CNNs for security applications.

Main Methods:

• A dataset of weapon images was used to generate reflectivity-added samples in MATLAB.
• Monostatic 2D Synthetic Aperture Radar (SAR) imaging reconstructed weapon profiles.
• A 10-layer CNN in Python classified the reconstructed weapon images.

Main Results:

• The CNN model achieved high accuracy, with precision and recall exceeding 98% for most weapon categories.
• The numerical modeling approach successfully reconstructed high-resolution weapon profiles.
• The system demonstrated robustness and reliability in weapon classification.

Conclusions:

• The proposed mm-Wave radar and CNN-based approach shows significant promise for enhancing security screening.
• This technology can be applied to various security checkpoints for effective threat detection.
• Further development could lead to more sophisticated and reliable security systems.