Low-Complexity Threat Objects Extraction Method Using Density-Based Clustering in Sub-THz Based Imaging

The wavelength of the 300 GHz sub-THz wave is approximately 1 mm, which enables tomographic imaging at a desired distance with a high resolution of less than 1 cm in the near field, and is expected to be used in security applications, such as threat object detection. However, determining the depth of a threat object involves generating images focused at each depth, which presents a considerable computation burden. In this study, we propose a low-complexity threat object identification method to reduce the computational complexity of the depth location by roughly dividing the data into the elevation, azimuth, and distance directions and applying a clustering method based on the sensing data density called DBSCAN. The results of the analysis using the measured sub-THz data indicate that the proposed algorithm can reduce the computational overhead by more than 80% while accurately determining the areas of the threat objects.