Automatic Threat Object Detection from THz Images using Artificial Intelligence Algorithms

Terahertz (THz) waves ranging from 0.1 THz to 10 THz, have the capability to penetrate most non-metallic and non-polar materials. This particular property of THz wave has opened up several application fields, particularly in the field of non-destrutive testing and security scanning. In baggage scanning applications, the THz wave penetrates through various baggage materials and can see-through the objects concealed within them. However, due to the limited THz emission power levels and limited THz detector sensitivities available, THz images acquired are of poor quality. These images have (i) high background variations, (ii) poor contrast, (iii) blurry images, and (iv) lack proper edge continuities. Due to these reasons, advanced image processing and artificial intelligence algorithms are utilized to enable automatic threat object detection and identification from THz images. In this work, authors have employed artificial intelligence-based algorithms for object detection, namely YOLOv5, for automatic threat identification, and evaluated its performance using various metrics such as precision, recall, and mean average precision (mAP) {0.5 and 0.5:0.95}