111 resultados para GPU acceleration
Resumo:
In this paper, we propose a multi-camera application capable of processing high resolution images and extracting features based on colors patterns over graphic processing units (GPU). The goal is to work in real time under the uncontrolled environment of a sport event like a football match. Since football players are composed for diverse and complex color patterns, a Gaussian Mixture Models (GMM) is applied as segmentation paradigm, in order to analyze sport live images and video. Optimization techniques have also been applied over the C++ implementation using profiling tools focused on high performance. Time consuming tasks were implemented over NVIDIA's CUDA platform, and later restructured and enhanced, speeding up the whole process significantly. Our resulting code is around 4-11 times faster on a low cost GPU than a highly optimized C++ version on a central processing unit (CPU) over the same data. Real time has been obtained processing until 64 frames per second. An important conclusion derived from our study is the scalability of the application to the number of cores on the GPU. © 2011 Springer-Verlag.
Resumo:
The Richardson-Lucy algorithm is one of the most important algorithms in the image deconvolution area. However, one of its drawbacks is slow convergence. A very significant acceleration is obtained by the technique proposed by Biggs and Andrews (BA), which is implemented in the deconvlucy function of the Image Processing MATLAB toolbox. The BA method was developed heuristically with no proof of convergence. In this paper, we introduce the Heavy-Ball (H-B) method for Poisson data optimization and extend it to a scaled H-B method, which includes the BA method as a special case. The method has proof of the convergence rate of O(k-2), where k is the number of iterations. We demonstrate the superior convergence performance of the scaled H-B method on both synthetic and real 3D images.
Resumo:
The R-matrix method when applied to the study of intermediate energy electron scattering by the hydrogen atom gives rise to a large number of two electron integrals between numerical basis functions. Each integral is evaluated independently of the others, thereby rendering this a prime candidate for a parallel implementation. In this paper, we present a parallel implementation of this routine which uses a Graphical Processing Unit as a co-processor, giving a speedup of approximately 20 times when compared with a sequential version. We briefly consider properties of this calculation which make a GPU implementation appropriate with a view to identifying other calculations which might similarly benet.
Resumo:
In recent years there have been a growing number of publications on procedures for damage detection in beams from analysing their dynamic response to the passage of a moving force. Most of this research demonstrates their effectiveness by showing that a singularity that did not appear in the healthy structure is present in the response of the damaged structure. This paper elucidates from first principles how the acceleration response can be assumed to consist of ‘static’ and ‘dynamic’ components, and where the beam has experienced a localised loss in stiffness, an additional ‘damage’ component. The combination of these components establishes how the damage singularity will appear in the total response. For a given damage severity, the amplitude of the ‘damage’ component will depend on how close the damage location is to the sensor, and its frequency content will increase with higher velocities of the moving force. The latter has implications for damage detection because if the frequency content of the ‘damage’ component includes bridge and/or vehicle frequencies, it becomes more difficult to identify damage. The paper illustrates how a thorough understanding of the relationship between the ‘static‘ and ‘damage’ components contributes to establish if damage has occurred and to provide an estimation of its location and severity. The findings are corroborated using accelerations from a planar finite element simulation model where the effects of force velocity and bridge span are examined.
Resumo:
Galactic cosmic-ray (CR) acceleration to the knee in the spectrum at a few PeV is only possible if the magnetic field ahead of a supernova remnant (SNR) shock is strongly amplified by CRs escaping the SNR. A model formulated in terms of the electric charge carried by escaping CRs predicts the maximum CR energy and the energy spectrum of CRs released into the surrounding medium. We find that historical SNRs such as Cas A, Tycho and Kepler may be expanding too slowly to accelerate CRs to the knee at the present time.
Resumo:
The self-consistent interaction between energetic particles and self-generated hydromagnetic waves in a cosmic ray pressure dominated plasma is considered. Using a three-dimensional hybrid magnetohydrodynamics (MHD)-kinetic code, which utilizes a spherical harmonic expansion of the Vlasov-Fokker-Planck equation, high-resolution simulations of the magnetic field growth including feedback on the cosmic rays are carried out. It is found that for shocks with high cosmic ray acceleration efficiency, the magnetic fields become highly disorganized, resulting in near isotropic diffusion, independent of the initial orientation of the ambient magnetic field. The possibility of sub-Bohm diffusion is demonstrated for parallel shocks, while the diffusion coefficient approaches the Bohm limit from below for oblique shocks. This universal behaviour suggests that Bohm diffusion in the root-mean-squared field inferred from observation may provide a realistic estimate for the maximum energy acceleration time-scale in young supernova remnants. Although disordered, the magnetic field is not self-similar suggesting a non-uniform energy-dependent behaviour of the energetic particle transport in the precursor. Possible indirect radiative signatures of cosmic ray driven magnetic field amplification are discussed.
Resumo:
The non-thermal particle spectra responsible for the emission from many astrophysical systems are thought to originate from shocks via a first order Fermi process otherwise known as diffusive shock acceleration. The same mechanism is also widely believed to be responsible for the production of high energy cosmic rays. With the growing interest in collisionless shock physics in laser produced plasmas, the possibility of reproducing and detecting shock acceleration in controlled laboratory experiments should be considered. The various experimental constraints that must be satisfied are reviewed. It is demonstrated that several currently operating laser facilities may fulfil the necessary criteria to confirm the occurrence of diffusive shock acceleration of electrons at laser produced shocks. Successful reproduction of Fermi acceleration in the laboratory could open a range of possibilities, providing insight into the complex plasma processes that occur near astrophysical sources of cosmic rays.
Resumo:
We show that the diffusion approximation breaks down for particle acceleration at oblique shocks with velocities typical of young supernova remnants. Higher order anisotropies flatten the spectral index at quasi-parallel shocks and steepen the spectral index at quasi-perpendicular shocks. We compare the theory with observed spectral indices.
Resumo:
We investigate the acceleration of particles by Alfven waves via the second-order Fermi process in the lobes of giant radio galaxies. Such sites are candidates for the accelerators of ultra-high-energy cosmic rays (UHECR). We focus on the nearby Fanaroff-Riley type I radio galaxy Centaurus A. This is motivated by the coincidence of its position with the arrival direction of several of the highest energy Auger events. The conditions necessary for consistency with the acceleration time-scales predicted by quasi-linear theory are reviewed. Test particle calculations are performed in fields which guarantee electric fields with no component parallel to the local magnetic field. The results of quasi-linear theory are, to an order of magnitude, found to be accurate at low turbulence levels for non-relativistic Alfven waves and at both low and high turbulence levels in the mildly relativistic case. We conclude that for pure stochastic acceleration via Alfven waves to be plausible as the generator of UHECR in Cen A, the baryon number density would need to be several orders of magnitude below currently held upper limits.
Resumo:
Linear acceleration emission occurs when a charged particle is accelerated parallel to its velocity. We evaluate the spectral and angular distribution of this radiation for several special cases, including constant acceleration (hyperbolic motion) of finite duration. Based on these results, we find the following general properties of the emission from an electron in a linear accelerator that can be characterized by an electric field E acting over a distance L: (1) the spectrum extends to a cutoff frequency (h) over bar omega(c)/mc(2) approximate to L(E/E(Schw))(2)/(lambda) over bar (C), where E(Schw) = 1.3 x 10(18) V m(-1) is the Schwinger critical field and (lambda) over bar (C) = (h) over bar /mc = 3.86 x 10(-13) m is the Compton wavelength of the electron, (2) the total energy emitted by a particle traversing the accelerator is 4/3 alpha(f)(h) over bar omega(c) in accordance with the standard Larmor formula where alpha(f) is the fine-structure constant, and (3) the low frequency spectrum is flat for hyperbolic trajectories, but in general depends on the details of the accelerator. We also show that linear acceleration emission complements curvature radiation in the strongly magnetized pair formation regions in pulsar magnetospheres. It dominates when the length L of the accelerator is less than the formation length rho/gamma of curvature photons, where rho is the radius of curvature of the magnetic field lines and gamma the Lorentz factor of the emitting particle. In standard static models of pair creating regions linear acceleration emission is negligible, but it is important in more realistic dynamical models in which the accelerating field fluctuates on a short length scale.
Resumo:
Stationary solutions to the equations of nonlinear diffusive shock acceleration play a fundamental role in the theory of cosmic-ray acceleration. Their existence usually requires that a fraction of the accelerated particles be allowed to escape from the system. Because the scattering mean free path is thought to be an increasing function of energy, this condition is conventionally implemented as an upper cutoff in energy space-particles are then permitted to escape from any part of the system, once their energy exceeds this limit. However, because accelerated particles are responsible for the substantial amplification of the ambient magnetic field in a region upstream of the shock front, we examine an alternative approach in which particles escape over a spatial boundary. We use a simple iterative scheme that constructs stationary numerical solutions to the coupled kinetic and hydrodynamic equations. For parameters appropriate for supernova remnants, we find stationary solutions with efficient acceleration when the escape boundary is placed at the point where growth and advection of strongly driven nonresonant waves are in balance. We also present the energy dependence of the distribution function close to the energy where it cuts off-a diagnostic that is in principle accessible to observation.
Resumo:
Ion acceleration driven by superintense laser pulses is attracting an impressive and steadily increasing effort. Motivations can be found in the applicative potential and in the perspective to investigate novel regimes as available laser intensities will be increasing. Experiments have demonstrated, over a wide range of laser and target parameters, the generation of multi-MeV proton and ion beams with unique properties such as ultrashort duration, high brilliance, and low emittance. An overview is given of the state of the art of ion acceleration by laser pulses as well as an outlook on its future development and perspectives. The main features observed in the experiments, the observed scaling with laser and plasma parameters, and the main models used both to interpret experimental data and to suggest new research directions are described.