Source author record

Peter Wilkinson

Peter Wilkinson appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

astro-ph.CO astro-ph.IM Computer Vision Machine Learning

Catalog footprint

What is connected

6works

4topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2020arXiv

Automating the assessment of biofouling in images using expert agreement as a gold standard

Biofouling is the accumulation of organisms on surfaces immersed in water. It is of particular concern to the international shipping industry because it increases fuel costs and presents a biosecurity risk by providing a pathway for non-indigenous marine species to establish in new areas. There is growing interest within jurisdictions to strengthen biofouling risk-management regulations, but it is expensive to conduct in-water inspections and assess the collected data to determine the biofouling state of vessel hulls. Machine learning is well suited to tackle the latter challenge, and here we apply deep learning to automate the classification of images from in-water inspections to identify the presence and severity of fouling. We combined several datasets to obtain over 10,000 images collected from in-water surveys which were annotated by a group biofouling experts. We compared the annotations from three experts on a 120-sample subset of these images, and found that they showed 89% agreement (95% CI: 87-92%). Subsequent labelling of the whole dataset by one of these experts achieved similar levels of agreement with this group of experts, which we defined as performing at most 5% worse (p=0.009-0.054). Using these expert labels, we were able to train a deep learning model that also agreed similarly with the group of experts (p=0.001-0.014), demonstrating that automated analysis of biofouling in images is feasible and effective using this method.

preprint2016arXiv

Update on the BINGO 21cm intensity mapping experiment

21cm intensity mapping is a novel approach aimed at measuring the power spectrum of density fluctuations and deducing cosmological information, notably from the Baryonic Acoustic Oscillations (BAO). We give an update on the progress of BAO from Integrated Neutral Gas Observations (BINGO) which is a single dish intensity mapping project. First we explain the basic ideas behind intensity mapping concept before updating the instrument design for BINGO. We also outline the survey we plan to make and its projected science output including estimates of cosmological parameters.

preprint2015arXiv

Localising fast radio bursts and other transients using interferometric arrays

A new population of sources emitting fast and bright transient radio bursts has recently been identified. The observed large dispersion measure values of FRBs suggests an extragalactic origin and an accurate determination of their positions and distances will provide an unique opportunity to study the magneto-ionic properties of the IGM. So far, FRBs have all been found using large dishes equipped with multi-pixel arrays. While large single dishes are well-suited for the discovery of transient sources they are poor at providing accurate localisations. A 2D snapshot image of the sky, made with a correlation interferometer array, can provide an accurate localisation of many compact radio sources simultaneously. However, the required time resolution to detect FRBs and a desire to detect them in real time, makes this currently impractical. In a beamforming approach, where many narrow tied-array beams are produced, the advantages of single dishes and interferometers can be combined. We present a proof-of-concept analysis of a new non-imaging method that utilises the additional spectral and comparative spatial information obtained from multiple overlapping TABs to estimate a transient source location with up to arcsecond accuracy in almost real time. We show that this method can work for a variety of interferometric configurations, including for LOFAR and MeerKAT, and that the estimated angular position may be sufficient to identify a host galaxy without reference to other simultaneous or follow-up observations. With this method, many transient sources can be localised to small fractions of a HPBW of a TAB, in the case of MeerKAT, sufficient to localise a source to arcsecond accuracy. In cases where the position is less accurately determined we can still significantly reduce the area that need be searched for associated emission at other wavelengths and potential host galaxies.

preprint2015arXiv

The SKA and the Unknown Unknowns

As new scientists and engineers join the SKA project and as the pressures come on to maintain costs within a chosen envelope it is worth restating and updating the rationale for the 'Exploration of the Unknown' (EoU). Maintaining an EoU philosophy will prove a vital ingredient for realizing the SKA's discovery potential. Since people make the discoveries enabled by technology a further axis in capability parameter space, the'human bandwidth' is emphasised. Using the morphological approach pioneered by Zwicky, a currently unexploited region of observational parameter space can be identified viz: time variable spectral patterns on all spectral and angular scales, one interesting example would be 'spectral transients'. We should be prepared to build up to 10 percent less collecting area for a given overall budget in order to enhance the ways in which SKA1 can be flexibly utilized.

preprint2014arXiv

Simulations of cm-wavelength Sunyaev-Zel'dovich galaxy cluster and point source blind sky surveys and predictions for the RT32/OCRA-f and the Hevelius 100-m radio telescope

We investigate the effectiveness of blind surveys for radio sources and galaxy cluster thermal Sunyaev-Zel'dovich effects (TSZEs) using the four-pair, beam-switched OCRA-f radiometer on the 32-m radio telescope in Poland. The predictions are based on mock maps that include the cosmic microwave background, TSZEs from hydrodynamical simulations, and unresolved radio sources. We estimate the effects of source clustering towards galaxy clusters from NVSS source counts around Planck-selected cluster candidates, and include appropriate correlations in our mock maps. The study allows us to quantify the effects of halo line-of-sight alignments, source confusion, and telescope angular resolution on the detections of TSZEs. We perform a similar analysis for the planned 100-m Hevelius radio telescope (RTH) equipped with a 49-beam radio camera. We find that RT32/OCRA-f will be suitable for small-field blind radio source surveys, and will detect $33^{+17}_{-11}$ new radio sources brighter than 0.87 mJy at 30~GHz in a 1 deg$^2$ field at $>5σ$ CL during a one-year, non-continuous, observing campaign, taking account of Polish weather conditions. It is unlikely that any galaxy cluster will be detected at $3σ$ CL in such a survey. A $60$-deg$^2$ survey, with field coverage of $2^2$ beams per pixel, at 15 GHz with the RTH, would find <1.5 galaxy clusters per year brighter than 60 $μ$Jy (at $3σ$ CL), and would detect about $3.4 \times 10^4$ point sources brighter than 1 mJy at $5σ$ CL, with confusion causing flux density errors $\lesssim 2\%$ (20%) in 68% (95%) of the detected sources. A primary goal of the planned RTH will be a wide-area ($π$~sr) radio source survey at 15 GHz. This survey will detect nearly $3 \times 10^5$ radio sources at $5σ$ CL down to 1.3 mJy, and tens of galaxy clusters, in one year of operation with typical weather conditions. ABRIDGED

preprint2011arXiv

Sunyaev Zel'dovich observations of a statistically complete sample of galaxy clusters with OCRA-p

We present 30 GHz Sunyaev Zel'dovich observations of a statistically complete sample of galaxy clusters with OCRA-p. The clusters are the 18 most X-ray luminous clusters at z > 0.2 in the ROSAT Brightest Cluster Sample. We correct for contaminant radio sources via supplementary observations with the Green Bank Telescope, also at 30 GHz, and remove a cluster that is contaminated by an unresolved X-ray source. All 17 remaining clusters have central SZ effects with Comptonisation parameter y_0 exceeding 1.9x10^-4, and 13 are detected at significance > 3 sigma. We use our data to examine scalings between y_0 and X-ray temperature, X-ray luminosity, and the X-ray mass proxy Y_X, and find good agreement with predictions from self-similar models of cluster formation, with an intrinsic scatter in y_0 of about 25%. We also comment on the success of the observations in the face of the contaminant source population, and the implications for upcoming cm-wave surveys.