|
An innovative optical tomography system for noninvasive reconstruction of absorption coefficient distributions in scattering environments is presented. The project integrates a light source, a fiber optic multiplexer and demultiplexer, a measurement tank with the analysed sample, an FPGA-based control and signal processing unit supporting Fourier-transform–based analysis. Deep learning models (U-Net, Vision Transformer, Self-Attention CNN) are used to process data acquired from 8 or 64 measurement sequences. The results demonstrate a significant impact of the number of channels on reconstruction performance. The system features are compact, design and near-real-time operation, meeting the requirements of Industry 4.0 for control and diagnostic applications.
|
