A5071037237- Infrastructure Maintenance and Monitoring
- Advanced Image Processing Techniques
- Identification and Quantification in Food
- Coral and Marine Ecosystems Studies
- Video Coding and Compression Technologies
- Advanced Vision and Imaging
- Underwater Acoustics Research
- Image Retrieval and Classification Techniques
- Ocean Acidification Effects and Responses
- CO2 Sequestration and Geologic Interactions
- Underwater Vehicles and Communication Systems
- 3D Surveying and Cultural Heritage
- Evaluation Methods in Various Fields
- Genomics and Phylogenetic Studies
- Maritime Navigation and Safety
- Species Distribution and Climate Change
- Environmental DNA in Biodiversity Studies
- Ichthyology and Marine Biology
- Advanced Neural Network Applications
- Marine and coastal ecosystems
- Remote Sensing and Land Use
- Geophysical Methods and Applications
- Image and Object Detection Techniques
- Oceanographic and Atmospheric Processes
- Marine and fisheries research
National Oceanography Centre
2018-2022
GEOMAR Helmholtz Centre for Ocean Research Kiel
2022
ARM (United Kingdom)
2021
University of Southampton
2018-2019
University of Bristol
2015-2016
Carbon capture and storage (CCS) is a key technology to reduce carbon dioxide (CO2) emissions from industrial processes in feasible, substantial, timely manner. For geological CO2 be safe, reliable, accepted by society, robust strategies for leakage detection, quantification management are crucial. The STEMM-CCS (Strategies Environmental Monitoring of Marine Capture Storage) project aimed provide techniques understanding enable inform cost-effective monitoring CCS sites the marine...
Pothole detection is one of the most important tasks for road maintenance. Computer vision approaches are generally based on either 2D image analysis or 3D surface modeling. However, these two categories always used independently. Furthermore, pothole accuracy still far from satisfactory. Therefore, in this paper, we present a robust algorithm that both accurate and computationally efficient. A dense disparity map first transformed to better distinguish between damaged undamaged areas. To...
Digital imaging has become one of the most important techniques in environmental monitoring and exploration. In case marine environment, mobile platforms such as autonomous underwater vehicles (AUVs) are now equipped with high-resolution cameras to capture huge collections images from seabed. However, timely evaluation all these presents a bottleneck problem tens thousands or more can be collected during single dive. This makes computational support for image analysis essential....
Video and image data are regularly used in the field of benthic ecology to document biodiversity. However, their use is subject a number challenges, principally identification taxa within images without associated physical specimens. The challenge applying traditional taxonomic keys fauna from has led development personal, group, or institution level reference catalogues operational units (OTUs) morphospecies. Lack standardisation among these problems with observer bias inability combine...
Machine learning is rapidly developing as a tool for gathering data from imagery and may be useful in identifying (classifying) visible specimens large numbers of seabed photographs. Application an automated classification workflow requires manually identified to supplied training validating the model. These validation datasets are generally generated by partitioning available manual specimens; typical ratios dataset sizes 75:25 or 80:20. However, this approach does not facilitate desired...
This paper presents a novel pothole detection approach based on single-modal semantic segmentation. It first extracts visual features from input images using convolutional neural network. A channel attention module then reweighs the to enhance consistency of different feature maps. Subsequently, we employ an atrous spatial pyramid pooling (comprising convolutions in series, with progressive rates dilation) integrate context information. helps better distinguish between potholes and undamaged...
The evaluation of large amounts digital image data is growing importance for biology, including the exploration and monitoring marine habitats. However, only a tiny percentage collected evaluated by biologists who manually interpret annotate contents, which can be slow laborious. In order to overcome bottleneck in annotation, two strategies are increasingly proposed: “citizen science” “machine learning”. this study, we investigated how combination citizen science, detect objects, machine...
Carbon capture and storage is a key mitigation strategy proposed for keeping the global temperature rise below 1.5 °C. Offshore can provide up to 13% of CO2 reduction required achieve Intergovernmental Panel on Climate Change goals. The public must be assured that potential leakages from reservoirs detected therefore safely contained. We conducted controlled release 675 kg within sediments at 120 m water depth, simulate leak test novel detection, quantification attribution approaches. show...
In situ imaging of particles in the ocean are rapidly establishing themselves as powerful tools to investigate carbon cycle, including role sinking for sequestration via biological pump. A big challenge when analysing camera images is determining size particle, which required calculate content, velocity and flux. key image processing decision algorithm used decide part forms particle background. However, this critical analysis step often unmentioned its effect rarely explored. Here we show...
The current maturity of autonomous underwater vehicles (AUVs) has made their deployment practical and cost-effective, such that many scientific, industrial military applications now include AUV operations. However, the logistical difficulties high costs operating at sea are still critical limiting factors in further technology development, benchmarking new techniques reproducibility research results. To overcome this problem, paper presents a freely available dataset suitable to test...
Previous work has shown that spatial resampling can improve rate-distortion performance by providing a higher and more consistent level of video quality at low bitrates. Rate control aims to regulate the bitrate in accordance bit budget. While this is well studied problem single resolution case, very little progress been made on adaptive case. In paper we present an enhanced method rate for intra coding allows algorithm learn from previously coded frames make accurate predictions, resulting...
3-D road surface modeling has become an essential part of modern algorithms for pothole detection when point clouds are available. This paper introduces a scale-adaptive and tracking framework. It first fits quadratic to the cloud, generated using GPT-SGM, state-of-the-art disparity estimation algorithm. The process also incorporates normal vector information, obtained by three-filters-to-normal (3F2N), ultra-fast accurate estimator. By comparing actual modeled clouds, can be extracted....
Abstract Video and image data are regularly used in the field of benthic ecology to document biodiversity. However, their use is subject a number challenges, principally identification taxa within images without associated physical specimens. The challenge applying traditional taxonomic keys fauna from has led development personal, group, or institution level reference catalogues operational units (OTUs) morphospecies. Lack standardisation among these problems with observer bias inability...
Allocating low bit-budgets to intra-coded pictures can often lead high quantisation and loss of important frequency information. Coding at lower resolutions alleviate some this distortion when utilising effective resampling techniques filters that minimise the additional introduced as a result spatial resampling. However, it is not always possible reconstruct each portion picture same or similar level quality; within natural scenes content tends vary spatially therefore fixed scale factor...
As the demand for higher quality and resolution video increases, many applications fail to meet this due low bandwidth restrictions. One factor contributing problem is high bitrate requirement of intra-coded Instantaneous Decoding Refresh (IDR) frames featuring in all coding standards. Frequent IDR essential error resilience order prevent occurrence propagation. However, as each one consumes a huge portion available bitrate, future coded hindered by levels compression. This work presents new...
Unfortunately, we made two minor mistakes in the above paper. First of all, first graph on row (c) <xref ref-type="fig" rid="fig1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Fig. 11</xref> was same as third . Secondly, “precision” and “recall” ref-type="table" rid="table1" xmlns:xlink="http://www.w3.org/1999/xlink">Table III</xref> need to be switched. The correct figure table have no influence discussion conclusions paper, they are given here.