Terasense

Simulation/Modeling

 

1. Model of THz scenes using Blender (a free open-source 3-D content creation suit)to assess the performance of a MMW imaging system and determine if certain patterns will be distinguished for a given set of scene temperatures and image system parameters.

UAB_Sim_pic

Fig. 1. Simulated passive millimeter-wave images, temperature in K, resolution 500 x 300 pixels (a) Outdoor, (b) Indoor.

 

2. Study of different face recognition systems: a commercial one (VeriLook SDK), PCA-SVM system, and DCT-GMM system. This is done in different environments (at short, mid and long distance) and with images acquired at the visible band of the spectrum. Mid and long distance scenarios are the ones where the GHz images are usually acquired. Therefore this approach constitutes the base to compare the behaviour of systems that use GHz images.

3. Generation of a database, called BIOGIGA, composed of 1200 synthetic images at 94 GHz of the body of 50 individuals. The images are the result of simulations carried out on corporal models at two types of scenarios (outdoors, indoors) and with two kinds of imaging systems (passive and active). These corporal models were previously generated using MakeHuman free software, based on body measurements taken from the subjects. Blender software was used to simulate the images at 94 GHz from the corporal 3D models.

ReducedImagenesBioGiga

Fig. 2. Synthetic images of one user simulated at 94 GHz, in different scenarios (indoors and outdoors) and with different imaging systems (passive and active).

4. Distance-based feature extraction for biometric recognition of Millimeter Wave body images. The developed extractor is tested on the previously described database. The results prove that the use of a small number of distance-based features provide good class separation.

ReducedImagingSteps  

Fig. 3. Main steps followed to extract the features from the image.

2Dplot

Fig. 4. Two dimensional representation of the discrimination power of the extracted features: (a) 2nd PCA component vs 1st PCA component and (b) Waist width vs height. Each cluster formed by the same kind of symbols represents an user.

5. Development of a high performance FDT parallel simulator. It has been included in a synthetic environment to simulate the effect of high intensity radiated fields on modern aircrafts and rotorcrafts. It gives a reduction of computational times in the order of thousands without losses of accuracy

6. Software MONURBS for the analysis and design of structures in THz. MONURBS can solve the EFIE (Electric Field Integral Equation) and the CFIE (Combined Field Integral Equation). It can also deal with conductors, dielectric and magnetic materials. The program uses a highly parallelized version of the Multilevel Fast Multipole Method.

7. Software HEMCUVE (Hybrid Electromagnetic Code Universities of Vigo and Extremadura) extension: proposal and implementation of the high-scalability and high-efficiency MLFMA-FFT algorithm. Optimal performance on mixed-memory massively parallel supercomputers: high scalability O(NlogN) method.

  • Rigourous solution of the largest computational electromagnetic problems to date: 500, 620 and 1000 million unknowns. This software has been awarded with two prizes: PRACE Award 2009 and Itanium Innovation Alliance Award 2009, in computationally intensive applications.
  • Current World record in computational electromagnetics: solution of the NASA Almond target at 3 THz (1000 million unknowns) using 1024 parallel processors, 5TB RAM and 35 hours of wall-clock time in the Finis Terrae supercomputer, at CESGA supercomputing center (ICTS-2009-40). 
  • HEMCUVE extension: development of surface integral-equation method of moments (SIE-MoM) formulation for metamaterial and plasmonic composite objects for efficient and highly accurate analysis of nanostructures in THz and optical regimes.
nasa almond_rcs_3thz
Fig. 6. Computed Bistatic RCS of the NASA Almond at 3 THz. 1 billion unknowns in the Finis-Terrae supercomputer.
  • Applied to the design of metamaterial super-lenses, nano-optical antennas for optical microscopy and spectroscopy, highly directive nano-optical antennas for wireless optical links and quantum processing, broadband nano-optical antennas for fiber to plasmonic guide interfacing.
lhm and_opticalnanoantenna

Fig. 7. (left) Simulated perfect lens made of double negative (DNG) metamaterial; (right) simulated nano-optical  gold-made plasmonic Yagi-Uda antenna at 817 nm.

logp and_metallodielectric_plasmonic_nanoantennas  

Fig. 8. (left) Nano-optical  silver-made Log-periodic broadband antenna for 600-1800 nm; (right) plasmonic metallo-dielectric antenna at 550 nm.