講演情報
[10p-A22-1][JSAP-Optica Joint Symposia Invited Talk] AI-Enhanced Single-Pixel Terahertz Camera for Hidden Object Detection
〇Adam Valles1 (1.Univ. Autonoma de Barcelona)
キーワード：
Single-pixel imaging
Terahertz (THz) radiation penetrates most non-conducting materials yet is strongly reflected by metals, making THz imaging a natural candidate for non-contact detection of hidden metallic objects. However, THz imaging has historically been constrained by the poor sensitivity of bolometer arrays and the high source powers they require. Single-pixel imaging (SPI) offers an alternative by concentrating all signal onto a single high-sensitivity detector. The AI-enhanced Single-Pixel Camera (AISPIC) implements this using a rotating metallic disk as a passive spatial modulator, paired with a pyroelectric detector and a convolutional neural network (CNN) reconstruction stage. Operating at 0.1 THz, AISPIC reconstructs 40x40-pixel images from just ~10% of the measurements a full raster scan would require. The CNN achieves a twofold improvement in peak signal-to-noise ratio over brute-force reconstruction at the same acquisition speed, with image reconstruction completed in under 200 ms, opening a realistic path toward real-time single-pixel THz imaging.