Search

The majority of Earth's surface lies deep in the oceans, where no surface light reaches. Robots diving down to great depths must bring light sources that create moving illumination patterns in the darkness, such that the same 3D point appears with different color in each image. On top, scattering and attenuation of light in the water makes images appear foggy and typically blueish, the degradation depending on each pixel's distance to its observed seafloor patch, on the local composition of the water and the relative poses and cones of the light sources. Consequently, visual mapping, including image matching and surface albedo estimation, severely suffers from the effects that co-moving light sources produce, and larger mosaic maps from photos are often dominated by lighting effects that obscure the actual seafloor structure. In this contribution a practical approach to estimating and compensating these lighting effects on predominantly homogeneous, flat seafloor regions, as can be found in the Abyssal plains of our oceans, is presented. The method is essentially parameter-free and intended as a preprocessing step to facilitate visual mapping, but already produces convincing lighting artefact compensation up to a global white balance factor. It does not require to be trained beforehand on huge sets of annotated images, which are not available for the deep sea. Rather, we motivate our work by physical models of light propagation, perform robust statistics-based estimates of additive and multiplicative nuisances that avoid explicit parameters for light, camera, water or scene, discuss the breakdown point of the algorithms and show results on imagery captured by robots in several kilometer water depth.

The calculation of the anharmonic modes of small to medium sized molecules for assigning experimentally measured frequencies to the corresponding type of molecular motions is computationally challenging at sufficiently high levels of quantum chemical theory. Here, a practical and affordable way to calculate coupled-cluster quality anharmonic frequencies using second order vibrational perturbation theory (VPT2) from machine-learned models is presented. The approach, referred to as "NN + VPT2", uses a high-dimensional neural network (PhysNet) to learn potential energy surfaces (PESs) at different levels of theory from which harmonic and VPT2 frequencies can be efficiently determined. The NN + VPT2 approach is applied to eight small to medium sized molecules (H$_2$CO, trans-HONO, HCOOH, CH$_3$OH, CH$_3$CHO, CH$_3$NO$_2$, CH$_3$COOH and CH$_3$CONH$_2$) and frequencies are reported from NN-learned models at the MP2/aug-cc-pVTZ, CCSD(T)/aug-cc-pVTZ and CCSD(T)-F12/aug-cc-pVTZ-F12 levels of theory. For the largest molecules and at the highest levels of theory, transfer learning (TL) is used to determine the necessary full-dimensional, near-equilibrium PESs. Overall, NN + VPT2 yields anharmonic frequencies to within 20 cm$^{-1}$ of experimentally determined frequencies for close to 90 % of the modes for the highest quality PES available and to within 10 cm$^{-1}$ for more than 60 % of the modes. For the MP2 PESs only around 60 % of the NN + VPT2 frequencies were within 20~cm$^{-1}$ of the experiment, with outliers up to 150 cm$^{-1}$ compared with experiment. It is also demonstrated that the approach allows to provide correct assignments for strongly interacting modes such as the OH bending and the OH torsional modes in formic acid monomer and the CO-stretch and OH-bend mode in acetic acid., Results unchanged - Minor rephrasing in Introduction and Methods - Added units on p. 20/21 - Added TL H2CO/CCSD(T)-F12 MAE in Table 1 - Added Acknowledgments and Data availability statement

Machine Learning (ML) has become a promising tool for improving the quality of atomistic simulations. Using formaldehyde as a benchmark system for intramolecular interactions, a comparative assessment of ML models based on state-of-the-art variants of deep neural networks (NN), reproducing kernel Hilbert space (RKHS+F), and kernel ridge regression (KRR) is presented. Learning curves for energies and atomic forces indicate rapid convergence towards excellent predictions for B3LYP, MP2, and CCSD(T)-F12 reference results for modestly sized (in the hundreds) training sets. Typically, learning curve off-sets decay as one goes from NN (PhysNet) to RKHS+F to KRR (FCHL). Conversely, the predictive power for extrapolation of energies towards new geometries increases in the same order with RKHS+F and FCHL performing almost equally. For harmonic vibrational frequencies, the picture is less clear, with PhysNet and FCHL yielding respectively flat learning at $\sim$ 1 and $\sim$ 0.2 cm$^{-1}$ no matter which reference method, while RKHS+F models level off for B3LYP, and exhibit continued improvements for MP2 and CCSD(T)-F12. Finite-temperature molecular dynamics (MD) simulations with the same initial conditions yield indistinguishable infrared spectra with good performance compared with experiment except for the high-frequency modes involving hydrogen stretch motion which is a known limitation of MD for vibrational spectroscopy. For sufficiently large training set sizes all three models can detect insufficient convergence (``noise'') of the reference electronic structure calculations in that the learning curves level off. Transfer learning (TL) from B3LYP to CCSD(T)-F12 with PhysNet indicates that additional improvements in data efficiency can be achieved.

Nowadays underwater vision systems are being widely applied in ocean research. However, the largest portion of the ocean - the deep sea - still remains mostly unexplored. Only relatively few image sets have been taken from the deep sea due to the physical limitations caused by technical challenges and enormous costs. Deep sea images are very different from the images taken in shallow waters and this area did not get much attention from the community. The shortage of deep sea images and the corresponding ground truth data for evaluation and training is becoming a bottleneck for the development of underwater computer vision methods. Thus, this paper presents a physical model-based image simulation solution, which uses an in-air texture and depth information as inputs, to generate underwater image sequences taken by robots in deep ocean scenarios. Different from shallow water conditions, artificial illumination plays a vital role in deep sea image formation as it strongly affects the scene appearance. Our radiometric image formation model considers both attenuation and scattering effects with co-moving spotlights in the dark. By detailed analysis and evaluation of the underwater image formation model, we propose a 3D lookup table structure in combination with a novel rendering strategy to improve simulation performance. This enables us to integrate an interactive deep sea robotic vision simulation in the Unmanned Underwater Vehicles simulator. To inspire further deep sea vision research by the community, we will release the source code of our deep sea image converter to the public., Accepted to ICPR 2021 Workshop: CVAUI 2020

The in silico exploration of chemical, physical and biological systems requires accurate and efficient energy functions to follow their nuclear dynamics at a molecular and atomistic level. Recently, machine learning tools gained a lot of attention in the field of molecular sciences and simulations and are increasingly used to investigate the dynamics of such systems. Among the various approaches, artificial neural networks (NNs) are one promising tool to learn a representation of potential energy surfaces. This is done by formulating the problem as a mapping from a set of atomic positions x and nuclear charges Zi to a potential energy V(x). Here, a fully-dimensional, reactive neural network representation for malonaldehyde (MA), acetoacetaldehyde (AAA) and acetylacetone (AcAc) is learned. It is used to run finite-temperature molecular dynamics simulations, and to determine the infrared spectra and the hydrogen transfer rates for the three molecules. The finite-temperature infrared spectrum for MA based on the NN learned on MP2 reference data provides a realistic representation of the low-frequency modes and the H-transfer band whereas the CH vibrations are somewhat too high in frequency. For AAA it is demonstrated that the IR spectroscopy is sensitive to the position of the transferring hydrogen at either the OCH- or OCCH3 end of the molecule. For the hydrogen transfer rates it is demonstrated that the O-O vibration is a gating mode and largely determines the rate at which the hydrogen is transferred between the donor and acceptor. Finally, possibilities to further improve such NN-based potential energy surfaces are explored. They include the transferability of an NN-learned energy function across chemical species (here methylation) and transfer learning from a lower level of reference data (MP2) to a higher level of theory (pair natural orbital-LCCSD(T)).

The polarization observables $T, E, P, H$, and $G$ in photoproduction of $��$ mesons off protons are measured for photon energies from threshold to $W=2400\,$MeV ($T$), 2280 MeV ($E$), 1620 MeV ($P, H$), or 1820 MeV ($G$), covering nearly the full solid angle. The data are compared to predictions from the SAID, MAID, J��Bo, and BnGa partial-wave analyses. A refit within the BnGa approach including further data yields precise branching ratios for the $N��$ decay of nucleon resonances. A $N��$-branching ratio of $0.33\pm 0.04$ for $N(1650)1/2^-$ is found, which reduces the large and controversially discussed $N��$-branching ratio difference of the two lowest mass $J^P=1/2^-$-resonances significantly., 10 pages, 11 figures

The beam-helicity asymmetry was measured, for the first time, in photoproduction of $��^{0}��$ pairs on carbon, aluminum, and lead, with the A2 experimental setup at MAMI. The results are compared to an earlier measurement on a free proton and to the corresponding theoretical calculations. The Mainz model is used to predict the beam-helicity asymmetry for the nuclear targets. The present results indicate that the photoproduction mechanism for $��^{0}��$ pairs on nuclei is similar to photoproduction on a free nucleon. This process is dominated by the $D_{33}$ partial wave with the $����(1232)$ intermediate state., Accepted for publication in Physics Letters B (2020)

We report a measurement of the spin polarisation of the recoiling neutron in deuterium photodisintegration, utilising a new large acceptance polarimeter within the Crystal Ball at MAMI. The measured photon energy range of 300~--~700~MeV provides the first measurement of recoil neutron polarisation at photon energies where the quark substructure of the deuteron plays a role, thereby providing important new constraints on photodisintegration mechanisms. A very high neutron polarisation in a narrow structure centred around $E_��\sim$~570~MeV is observed, which is inconsistent with current theoretical predictions employing nucleon resonance degrees of freedom. A Legendre polynomial decomposition suggests this behaviour could be related to the excitation of the $d^*(2380)$ hexaquark.

The largest, at the moment, statistics of 7x10^6 eta->3pi^0 decays, based on 6.2x10^7 eta mesons produced in the gamma p -> eta p reaction, has been accumulated by the A2 Collaboration at the Mainz Microtron, MAMI. It allowed a detailed study of the eta->3pi^0 dynamics beyond its conventional parametrization with just the quadratic slope parameter alpha and enabled, for the first time, a measurement of the second-order term and a better understanding of the cusp structure in the neutral decay. The present data are also compared to recent theoretical calculations that predict a nonlinear dependence along the quadratic distance from the Dalitz-plot center., 11 pages, 6 figures, 1 table

This work measured $d��/d��$ for neutral kaon photoproduction reactions from threshold up to a c.m.\ energy of 1855MeV, focussing specifically on the $��p\rightarrow K^0��^+$, $��n\rightarrow K^0��$, and $��n\rightarrow K^0 ��^0$ reactions. Our results for $��n\rightarrow K^0 ��^0$ are the first-ever measurements for that reaction. These data will provide insight into the properties of $N^*$ resonances and, in particular, will lead to an improved knowledge about those states that couple only weakly to the $��N$ channel. Integrated cross sections were extracted by fitting the differential cross sections for each reaction as a series of Legendre polynomials and our results are compared with prior experimental results and theoretical predictions., 16 pages, 13 figures, 7 tables

Photoproduction of mesons off quasi-free nucleons bound in the deuteron allows to study the electromagnetic excitation spectrum of the neutron and the isospin structure of the excitation of nucleon resonances. The database for such reactions is much more sparse than for free proton targets. Single $��^0$ photoproduction off quasi-free nucleons from the deuteron was experimentally studied. Nuclear effects were investigated by a comparison of the results for free protons and quasi-free protons and used as a correction for the quasi-free neutron data. The experiment was performed at the tagged photon beam of the Mainz MAMI accelerator for photon energies between 0.45~GeV and 1.4~GeV, using an almost $4��$ electromagnetic calorimeter composed of the Crystal Ball and TAPS detectors. A complete kinematic reconstruction of the final state removed the effects of Fermi motion. Reaction model predictions and PWA for $��n\rightarrow n��^{0}$, based on fits to data for the other isospin channels, disagreed between themselves and no model provided a good description of the new data. The results demonstrate clearly the importance of a measurement of the fully neutral final state for the isospin decomposition of the cross section. Model refits, for example from the Bonn-Gatchina analysis, show that the new and the previous data for the other three isospin channels can be simultaneously described when the contributions of several partial waves are modified. The results are also relevant for the suppression of the higher resonance bumps in total photoabsorption on nuclei, which are not well understood., excepted for publication in Phys. Rev. C

The data available from the A2 Collaboration at MAMI were analyzed to select the $��p\to ��^0��p$ reaction on an event-by-event basis, which allows for partial-wave analyses of three-body final states to obtain more reliable results, compared to fits to measured distributions. These data provide the world's best statistical accuracy in the energy range from threshold to $E_��=1.45$ GeV, allowing a finer energy binning in the measurement of all observables needed for understanding the reaction dynamics. The results obtained for the measured observables are compared to existing models, and the impact from the new data is checked by the fit with the revised Mainz model., 15 pages, 18 figures

The double polarization observable $E$ and the helicity dependent cross sections $��_{1/2}$ and $��_{3/2}$ were measured for $��$ photoproduction from quasi-free protons and neutrons. The circularly polarized tagged photon beam of the A2 experiment at the Mainz MAMI accelerator was used in combination with a longitudinally polarized deuterated butanol target. The almost $4��$ detector setup of the Crystal Ball and TAPS is ideally suited to detect the recoil nucleons and the decay photons from $��\rightarrow 2��$ and $��\rightarrow 3��^0$. The results show that the narrow structure previously observed in $��$ photoproduction from the neutron is only apparent in $��_{1/2}$ and hence, most likely related to a spin-1/2 amplitude. Nucleon resonances that contribute to this partial wave in $��$ production are only $N1/2^-$ ($S_{11}$) and $N1/2^+$ ($P_{11}$). Furthermore, the extracted Legendre coefficients of the angular distributions for $��_{1/2}$ are in good agreement with recent reaction model predictions assuming a narrow resonance in the $P_{11}$ wave as the origin of this structure.

An experimental study of the $��'\to ��^0��^0��\to 6��$ decay has been conducted with the best up-to-date statistical accuracy, by measuring $��'$ mesons produced in the $��p \to ��' p$ reaction with the A2 tagged-photon facility at the Mainz Microtron, MAMI. The results obtained for the standard parametrization of the $��'\to ��^0��^0��$ matrix element are consistent with the most recent results for $��'\to������$ decays, but have smaller uncertainties. The available statistics and experimental resolution allowed, for the first time, an observation of a structure below the $��^+��^-$ mass threshold, the magnitude and sign of which, checked within the framework of the nonrelativistic effective-field theory, demonstrated good agreement with the cusp that was predicted based on the $����$ scattering length combination, $a_0-a_2$, extracted from $K \to 3��$ decays., 15 pages, 8 figures, 2 tables, accepted for publication in Phys. Rev. D

The scalar dipole polarizabilities, $��_{E1}$ and $��_{M1}$, are fundamental properties related to the internal dynamics of the nucleon. The currently accepted values of the proton polarizabilities were determined by fitting to unpolarized proton Compton scattering cross section data. The measurement of the beam asymmetry $��_{3}$ in a certain kinematical range provides an alternative approach to the extraction of the scalar polarizabilities. At the Mainz Microtron (MAMI) the beam asymmetry was measured for Compton scattering below pion photoproduction threshold for the first time. The results are compared with model calculations and the influence of the experimental data on the extraction of the scalar polarizabilities is determined., 6 pages, 5 figures

High statistics measurements of the photon asymmetry $\mathrm��$ for the $\overrightarrow��$p$\rightarrow��^{0}$p reaction have been made in the center of mass energy range W=1214-1450 MeV. The data were measured with the MAMI A2 real photon beam and Crystal Ball/TAPS detector systems in Mainz, Germany. The results significantly improve the existing world data and are shown to be in good agreement with previous measurements, and with the MAID, SAID, and Bonn-Gatchina predictions. We have also combined the photon asymmetry results with recent cross-section measurements from Mainz to calculate the profile functions, $\check{\mathrm��}$ (= $��_{0}\mathrm��$), and perform a moment analysis. Comparison with calculations from the Bonn-Gatchina model shows that the precision of the data is good enough to further constrain the higher partial waves, and there is an indication of interference between the very small $F$-waves and the $N(1520) 3/2^{-}$ and $N(1535) 1/2^{-}$ resonances.

The Dalitz decay pi^0 -> e^+e^-gamma has been measured in the gamma p -> pi^0 p reaction with the A2 tagged-photon facility at the Mainz Microtron, MAMI. The value obtained for the slope parameter of the pi^0 electromagnetic transition form factor, a_pi = 0.030+/-0.010_tot, is in agreement with existing measurements of this decay and with recent theoretical calculations. The uncertainty obtained in the value of a_pi is lower than in previous results based on the pi^0 -> e^+e^-gamma decay., 11 pages, 7 figures, 1 table

The Dalitz decays eta -> e^+e^-g and omega -> pi^0 e^+e^- have been measured in the g p -> eta p and g p -> omega p reactions, respectively, with the A2 tagged-photon facility at the Mainz Microtron, MAMI. The value obtained for the slope parameter of the electromagnetic transition form factor of eta, Lambda^{-2}_eta=(1.97+/-0.11_tot) GeV^{-2}, is in good agreement with previous measurements of the eta -> e^+e^-g and eta -> mu^+mu^-g decays. The uncertainty obtained in the value of Lambda^{-2}_eta is lower than in previous results based on the eta -> e^+e^-g decay. The value obtained for the omega slope parameter, Lambda^{-2}_omega_pi^0 = (1.99+/-0.21_tot) GeV^{-2}, is somewhat lower than previous measurements based on omega -> pi^0 mu^+mu^-, but the results for the omega transition form factor are in better agreement with theoretical calculations, compared to earlier experiments., 19 figures, 2 tables

Data on the polarization observables T, P, and H for the reaction $��p\to p��^0$ are reported. Compared to earlier data from other experiments, our data are more precise and extend the covered range in energy and angle substantially. The results were extracted from azimuthal asymmetries measured using a transversely polarized target and linearly polarized photons. The data were taken at the Bonn electron stretcher accelerator ELSA with the CBELSA/TAPS detector. Within the Bonn-Gatchina partial wave analysis, the new polarization data lead to a significant narrowing of the error band for the multipoles for neutral-pion photoproduction.

Total cross sections, angular distributions, and invariant-mass distributions have been measured for the photoproduction of $��^0��^0$ pairs off free protons and off nucleons bound in the deuteron. The experiments were performed at the MAMI accelerator facility in Mainz using the Glasgow photon tagging spectrometer and the Crystal Ball/TAPS detector. The accelerator delivered electron beams of 1508 and 1557~MeV, which produced bremsstrahlung in thin radiator foils. The tagged photon beam covered energies up to 1400~MeV. The data from the free proton target are in good agreement with previous measurements and were only used to test the analysis procedures. The results for differential cross sections (angular distributions and invariant-mass distributions) for free and quasi-free protons are almost identical in shape, but differ in absolute magnitude up to 15\%. Thus, moderate final-state interaction effects are present. The data for quasi-free neutrons are similar to the proton data in the second resonance region (final state invariant masses up to $\approx$1550~MeV), where both reactions are dominated by the $N(1520)3/2^-\rightarrow ��(1232)3/2^+��$ decay. At higher energies, angular and invariant-mass distributions are different. A simple analysis of the shapes of the invariant-mass distributions in the third resonance region is consistent with strong contributions of an $N^{\star}\rightarrow N��$ decay for the proton, while the reaction is dominated by a sequential decay via a $����$ intermediate state for the neutron. The data are compared to predictions from the Two-Pion-MAID model and the Bonn-Gatchina coupled channel analysis., accepted for publication in Eur. Phys. J. A

A new measurement of the rare, doubly radiative decay eta->pi^0 gamma gamma was conducted with the Crystal Ball and TAPS multiphoton spectrometers together with the photon tagging facility at the Mainz Microtron MAMI. New data on the dependence of the partial decay width, Gamma(eta->pi^0 gamma gamma), on the two-photon invariant mass squared, m^2(gamma gamma), as well as a new, more precise value for the decay width, Gamma(eta->pi^0 gamma gamma) = (0.33+/-0.03_tot) eV, are based on analysis of 1.2 x 10^3 eta->pi^0 gamma gamma decays from a total of 6 x 10^7 eta mesons produced in the gamma p -> eta p reaction. The present results for dGamma(eta->pi^0 gamma gamma)/dm^2(gamma gamma) are in good agreement with previous measurements and recent theoretical calculations for this dependence., 10 pages, 8 figures, 1 table

Differential and total cross sections for the quasifree reactions $��p\rightarrow��p$ and $��n\rightarrow��n$ have been determined at the MAMI-C electron accelerator using a liquid deuterium target. Photons were produced via bremsstrahlung from the 1.5 GeV incident electron beam and energy-tagged with the Glasgow photon tagger. Decay photons of the neutral decay modes $��\rightarrow 2��$ and $��\rightarrow 3��^0 \rightarrow 6��$ and coincident recoil nucleons were detected in a combined setup of the Crystal Ball and the TAPS calorimeters. The $��$-production cross sections were measured in coincidence with recoil protons, recoil neutrons, and in an inclusive mode without a condition on recoil nucleons, which allowed a check of the internal consistency of the data. The effects from nuclear Fermi motion were removed by a kinematic reconstruction of the final-state invariant mass and possible nuclear effects on the quasifree cross section were investigated by a comparison of free and quasifree proton data. The results, which represent a significant improvement in statistical quality compared to previous measurements, agree with the known neutron-to-proton cross-section ratio in the peak of the $S_{11}(1535)$ resonance and confirm a peak in the neutron cross section, which is absent for the proton, at a center-of-mass energy $W = (1670\pm 5)$ MeV with an intrinsic width of $��\approx 30$ MeV.

One of the main disadvantages of the Hough techniques is the requirement of heavy computation for the transform. To reduce the computation time, a simple sampling technique of the image pixels is suggested. The performance of this sampling Hough Transform is comparable to the standard Hough techniques but the computation time is significantly reduced. In the evaluation of the performance of this new sampling Hough technique, a number of real images have been studied. The present approach is capable of detecting lines and circles with any thickness, even for a line with one pixel width. >

In this paper, we suggest a novel concept of applying the self-organizing map (SOM) in the Hough domain for a significant reduction of the Hough space. By using the SOM as the output space of the generalized Hough transform, the conventional 4-D Hough domain is replaced by a 10/spl times/10 map, organized in a rectangular grid. Experimental results indicate high accuracy in voting is attainable despite its small memory requirement.

By making use of the Hough and the contour sequence matching technique, the authors suggest a novel approach for solving the problem of the large memory requirement of the Hough space. For the conventional Hough algorithm it is necessary to take a four-dimensional space for the recognition of an irregular pattern, while in the proposed algorithm the Hough space is replaced with a two-dimensional domain. The orientation of a possible object is determined by comparing the contour sequence of the object and that of the prototype. The reduction of the scaling and orientation parameters in the new Hough space also simplifies the peak searching process to verify the results. In order to enhance the accuracy of the recognition rate, a novel peak searching technique for eliminating false peaks in the Hough space is suggested. The overall approach was tested using practical images and was found to be very promising. >

The authors present a new approach to the use of the Hough transform for the detection of ellipses in a 2-D image. In the proposed algorithm, the conventional 5-D Hough voting space is replaced by four 2-D Hough planes which require only 90 kbytes of memory for a 384/spl times/256 image. One of the main differences between the proposed transform and other techniques is the way to extract feature points from the image under question. For the accumulation process in the Hough domain, an inherent property of the suggested algorithm is its capability to effect verification. Experimental results from the authors' work on real and synthetic images show that a significant improvement of the recognition is achieved as compared to other algorithms. Furthermore, the proposed algorithm is applicable to the detection of both circular and elliptical objects concurrently.

By using the combination of the Hough and the contour sequence matching techniques, the authors suggest a new algorithm of scale and rotation invariant for pattern recognition. In the proposed algorithm, the conventional four-dimensional Hough space is replaced by a two-dimensional one. Because of the significant reduction of memory requirement, the authors can provide a much faster and more efficient generalised Hough algorithm. In order to enhance the performance of the Hough transform, they also suggest a new peak searching technique on the Hough space to achieve an accurate location of the detected objects.