Study of object detection in linear terahertz imaging systems

Abstract

A neural network-based system designed for automated detection of concealed items using a postal terahertz scanner is presented, with system optimization provided. A dataset of objects scanned by a THz scanner is introduced. A convolutional neural network is trained on this dataset of terahertz images to classify and detect whether an image contains a prohibited item or not. The system is tested using realworld samples, achieving an accuracy of 95.5% mAP@0.5. The results demonstrate the effectiveness of employing a neural network in postal terahertz scanners and its potential for use in security and surveillance applications.

Concealed objects detection and recognition in active transmission sub-terahertz imaging scanner for automatic security screening

“…Exploiting this advantage, active transmission sub-THz imaging scanners, particularly operating at 100 GHz, have been recently proposed for screening postal items. 1,3,[8][9][10] Automatic concealed objects detection and recognition represents a current challenge in developing modern sub-THz scanners for security screening. The recognition process typically encompasses preprocessing, detection, localization, classification, and tagging tasks.…”

“…Exploiting this advantage, active transmission sub-THz imaging scanners, particularly operating at 100 GHz, have been recently proposed for screening postal items 1 , 3 , 8 – 10 …”

“…Recent advancements in artificial intelligence techniques have opened up new possibilities for the automatic detection and recognition of concealed objects in sub-THz imaging scanners 9 , 16 . Statistical machine learning mainly relies on classification methods based on feature extraction for object detection.…”

“…Most related research works focused on concealed weapon detection under clothing for personal screening. In active transmission sub-THz imaging, transfer learning based CNNs using pretrained models, such ass YOLOv3, 6 YOLOv5, 29 YOLOv7, 30 RetinaNet, 16 region-based CNN (R-CNN), 2 and others, 1 , 9 , 31 have been proposed for concealed item recognition, particularly in sub-THz imaging scanners for postal screening applications.…”