We propose a fusion terahertz deep learning computed tomography framework designed to precisely reconstruct object 3D geometric information from THz temporal-spatio-spectral signals acquired through a terahertz time-domain spectroscopy system. This Unet-based fusion framework utilizes multi-scale branches for extracting spatio-spectral features, which undergo processing through an element-wise filter adaptive convolutional layer, resulting in high-quality restoration of THz 3D images. Furthermore, the proposed framework offers high scalability and adjustability, allowing users to choose their processing signal domains and seamlessly integrate their own modified fusion network.
We present a diffractive terahertz sensor using a single-pixel detector to rapidly sense hidden defects within a target sample volume. Leveraging multiple spatially-engineered diffractive layers optimized via deep learning, this diffractive sensor can all-optically process the sample scattered waves and generate an output spectrum encoding information for indicating the presence/absence of hidden defects. We experimentally validated this framework using a single-pixel terahertz time-domain spectroscopy set-up and 3D-printed diffractive layers, successfully detecting unknown hidden defects within silicon samples. By circumventing raster scanning and digital image formation/reconstruction, this framework holds vast potential for various applications requiring high-throughput, non-destructive defect detection.
We report a polarization-encoded diffractive network to perform multiple arbitrary complex-valued linear transforms within a single diffractive processor. An array of pre-selected linear polarizers is placed between the trainable isotropic diffractive layers, and distinct complex-valued linear transformations are individually assigned to different combinations of input/output polarization states. A polarization-encoded diffractive network performs the target linear transforms with negligible error when N ≥ P x I x O, where N is the number of trainable diffractive features/neurons, I and O denote the number of pixels at the input and output fields-of-view, respectively, and P represents the number of target linear transforms.
Terahertz imaging system has aroused great attentions in recent years due to its unique applications in security screening, industrial inspection and biomedical evaluation. Most conventional terahertz PCA imaging setup is based on raster scan method, hence, image acquisition time is severely limited by the speed of mechanical movement. Typically, image acquisition time of terahertz PCA tomography systems costs hours to days depending on the size of observed objects, which severely limits their practical feasibility to real-world applications. Here, we propose a terahertz least absolute shrinkage and selection operator (LASSO) compressed sensing (CS) tomography system to reduce more than one order of magnitude of data acquisition time. Our approach replaces slow-moving mechanical raster scanning method by the highspeed spatial modulation of terahertz radiation with designed patterns, resulting in a compact, fast and noise-reduced way to obtain terahertz 3D image dataset. Based on the measured dataset, modified Hadamard and LASSO algorithms are designed to reconstruct high-quality 3D image (voxel number: 128x128x128) in 120s at 10% compressed rate. The reconstructed LASSO 3D terahertz image offers a less than 0.002 mean squared error and 80% structural similarity index compared with the ground truth image. This paves the way toward real-time terahertz 3D imaging in near future, which opens the door for varies of exciting applications in non-destructive sensing, imaging and material inspection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.