Terahertz Image Detection with the Improved Faster Region-Based Convolutional Neural Network.

Area of Science:

• Optics and Photonics
• Computer Science
• Security Technology

Background:

• Terahertz (THz) imaging offers non-ionizing, clothing-penetrating capabilities, making it ideal for security applications.
• Current security screening methods face limitations in detecting concealed items effectively and in real-time.
• Deep learning presents a promising approach for enhancing the performance and speed of THz imaging analysis.

Purpose of the Study:

• To develop and validate a deep learning-based system for detecting concealed weapons and contraband using terahertz imaging.
• To improve the accuracy and real-time detection speed of security screening processes.
• To address the challenges associated with analyzing the unique characteristics of terahertz images.

Main Methods:

• Collection and labeling of a comprehensive dataset of terahertz images.
• Development of a classification method leveraging transfer learning for terahertz image analysis.
• Implementation of an improved Faster Region-based Convolutional Neural Network (Faster R-CNN) with threshold segmentation for independent object detection.

Main Results:

• The proposed deep learning system demonstrated high effectiveness in detecting concealed items in terahertz images.
• Experimental results confirmed the efficiency and real-time detection capabilities of the developed method.
• The Faster R-CNN approach, enhanced with threshold segmentation, proved successful in distinguishing between human bodies and concealed objects.

Conclusions:

• The developed deep learning system offers a robust solution for concealed item detection in terahertz imaging.
• The proposed method significantly enhances the performance and speed of security screening applications.
• This research validates the potential of advanced deep learning techniques for advancing terahertz imaging security solutions.