Search results for "Pattern recognition"
showing 10 items of 2301 documents
Applying Artificial Intelligence Methods to Detect and Classify Fish Calls from the Northern Gulf of Mexico
2021
Passive acoustic monitoring is a method that is commonly used to collect long-term data on soniferous animal presence and abundance. However, these large datasets require substantial effort for manual analysis
Superresolved imaging of remote moving targets.
2006
We present a superresolving approach that allows one to exceed the diffraction limit and recover highly resolved contours of moving targets from a sequence of low-resolution images. The presented approach is suitable for remote sensing applications. The resolution decoding algorithm that is used to recover the high-resolution features of the target can be run partially via optical means and that way can be used to reduce the required computational complexity.
The ATHENA X-ray Integral Field Unit (X-IFU)
2018
Event: SPIE Astronomical Telescopes + Instrumentation, 2018, Austin, Texas, United States.
Left-handed metamaterial coatings for subwavelength-resolution imaging
2012
We report on a procedure to improve the resolution of far-field imaging by using a neighboring high-index medium that is coated with a left-handed metamaterial. The resulting plot can also exhibit an enhanced transmission by considering proper conditions to retract backscattering. Based on negative refraction, geometrical aberrations are considered in detail since they may cause a great impact in this sort of diffraction-unlimited imaging by reducing its resolution power. We employ a standard aberration analysis to refine the asymmetric configuration of metamaterial superlenses. We demonstrate that low-order centrosymmetric aberrations can be fully corrected for a given object plane. For su…
Optoelectronic morphological image processor.
2009
A morphological optoelectronic image processor based on the threshold decomposition concept is described and demonstrated. Binary slices of a gray-scale input image are optically convolved with a binary structuring element of arbitrary size and shape in a noncoherent convolver. The slices are displayed on a liquid-crystal spatial light modulator of 320 × 264 pixels. The kernels are implemented as modifications of the system impulse response. The processor’s convolution patterns are recorded with a CCD camera and fed into a PC by a frame grabber. Subsequent elementary morphological operations are looped. Examples of processing an input image of 256 × 256 pixels and 16 gray levels with kernel…
Variational multiframe restoration of images degraded by noisy (stochastic) blur kernels
2013
This article introduces and explores a class of degradation models in which an image is blurred by a noisy (stochastic) point spread function (PSF). The aim is to restore a sharper and cleaner image from the degraded one. Due to the highly ill-posed nature of the problem, we propose to recover the image given a sequence of several observed degraded images or multiframes. Thus we adopt the idea of the multiframe approach introduced for image super-resolution, which reduces distortions appearing in the degraded images. Moreover, we formulate variational minimization problems with the robust (local or nonlocal) L^1 edge-preserving regularizing energy functionals, unlike prior works dealing wit…
Axial apodization in 4Pi-confocal microscopy by annular binary filters
2002
We present a novel technique for considerably decreasing the sidelobe height of the axial point-spread function of one-photon 4Pi-confocal microscopes. By means of a numerical example, in which the ratio between the excitation and the fluorescence wavelengths was set to epsilon = lambdaexc/lambdadet = 0.8, we show that simply inserting a pair of properly designed two-ring binary masks in the illumination set allows the height of the axial sidelobes to be reduced from 20% to 5% of the height of the central peak. This allows one to receive the full benefit of the strong narrowness of the central lobe provided by the 4Pi-confocal technique.
Combining Real-Time Segmentation and Classification of Rehabilitation Exercises with LSTM Networks and Pointwise Boosting
2020
Autonomous biofeedback tools in support of rehabilitation patients are commonly built as multi-tier pipelines, where a segmentation algorithm is first responsible for isolating motion primitives, and then classification can be performed on each primitive. In this paper, we present a novel segmentation technique that integrates on-the-fly qualitative classification of physical movements in the process. We adopt Long Short-Term Memory (LSTM) networks to model the temporal patterns of a streaming multivariate time series, obtained by sampling acceleration and angular velocity of the limb in motion, and then we aggregate the pointwise predictions of each isolated movement using different boosti…
Drawback of ICA Procedure on EEG: Polarity Indeterminacy at Local Optimization
2008
Independent component analysis (ICA) has been extensively applied to reject artifacts in electroencephalography (EEG) signal processing. The first step is to extract the independent component activations from the electrode records, and then project the desired components back to the electrodes. After the composition of the projected component is analyzed in details under ICA procedure, this study shows that since ICA may extract some source components at the local optimization in high-dimensional EEG signal space, the artificial polarity indeterminacy may happen on the projected component at some electrodes. By numerical simulations, this issue also exhibits that this polarity ambiguity occ…
Invariant pattern recognition based on 1-D Wavelet functions and the polynomial decomposition
1997
Abstract A new filter, consisting of 1-D Wavelet functions is suggested for achieving optical invariant pattern recognition. The formed filter is actually a real function, hence, it is theoretically possible to be implemented under both spatially coherent and spatially incoherent illuminations. The filter is based on the polynomial expansion, and is constructed out of a scaled bank of filters multiplied by 1-D Wavelet weight functions. The obtained output is shown to be invariant to 2-D scaling even when different scaling factors are applied on the different axes. The computer simulations and the experimental results demonstrate the potential hidden in this technique.