Search

Neural networks have been applied to landmine detection from data generated by different kinds of sensors. Real-valued neural networks have been used for detecting landmines from scattering parameters measured by ground penetrating radar (GPR) after disregarding phase information. This paper presents results using complex-valued neural networks, capable of phase-sensitive detection followed by classification. A two-layer hybrid neural network structure incorporating both supervised and unsupervised learning is proposed to detect and then classify the types of landmines. Tests are also reported on a benchmark data. Index Terms--Hybrid artificial neural networks, landmine classification, landmine detection, scattering parameters.

Time-resolved Stokes vector components of light transmitted through filamentous tissues were measured with a view to improving the imaging quality of optical images in such tissues. Temporal profiles of the Stokes vectors and the time-resolved degree of polarization (DOP) were calibrated to produce higher image quality than that of images based on time gating, polarization discrimination, or both. A thin chicken bone inserted into chicken breast tissue with filament orientation in different directions with respect to the direction of input linear polarization was scanned to demonstrate images of higher spatial resolution and contrast based on the measurement of time-resolved DOP. OCIS codes: 170.6920, 170.3880.

A temperature-independent fiber-Bragg-grating strain-sensing system, based on a novel optical-power-detection scheme, is developed and analyzed. In this system a pair of fiber Bragg gratings with reflection spectra either partially or substantially overlapping is placed side by side to form a temperature-independent strain-sensor unit. Conventional wavelength-interrogation techniques are not used here, and instead an optical-power-detection scheme is proposed to directly calibrate the measurand, i.e., the strain. Unlike the conventional approach in a multiplexed sensing system, the presented power-detection-based interrogation method does not need the fiber-Bragg-grating sensors to be spectrally separate. The only requirement is that the spectra of the two fiber Bragg gratings of each sensor unit in a multiplexed system be identical or slightly separate (slightly overlapping spectra would also work in the sensing scheme) and the source's optical power be sufficient for sensitive measurement. Based on a three-sensor-unit system, we demonstrate simple strain measurements of high linearity ([+ or -] 0.4%), good sensitivity [2 microstrains ([micro]S)], high thermal stability ([+ or -] 0.8%), and zero cross talk. The effects of light source spectral flatness and fiber bending loss on measurement accuracy are also discussed. OCIS codes: 060.2310, 060.2370, 050.2770, 040.1880, 120.6810, 060.2410.