A novel concealed object detection algorithm for entry control and security check applications using passive terahertz technology

Abstract Passive Terahertz (THz) imaging is rapidly being adopted in entry control and security applications, such as concealed object detection under clothing. A passive THz imaging system, which operates as a stand-off type sensor, can scan both indoor and outdoor environments. Therefore, an efficient and intelligent frameworks are necessary for the automatic detection of concealed objects. This paper proposes a novel framework based on binary segmentation of THz images and a visual attention mechanism for the automatic detection and localization of concealed objects. The core idea is to perform a comprehensive analysis of THz images using mixture models and multi-level segmentation to identify concealed objects as grayish blob-like structures in front of a bright human body. Thereafter, a computational model of visual saliency, motivated by biological vision, is used to accurately locate these structures and generate the concealed object map. Finally, the concealed object map is improved using the Superpixels module for precise localization of concealed objects. Experimental results indicate that the proposed framework enables effective hidden object detection and shows superior performance compared to other recently proposed methods. The quantitative and qualitative experimental results on a real passive THz dataset indicate that the proposed framework enables effective concealed object detection and depicts superior performance compared to traditional approaches. It achieves a 100% detection rate (100% recall), and 14% false alarm rate (86% precision) on the test dataset, demonstrating the high efficacy of the proposed approach in identifying concealed objects.