scholarly journals Weakly Supervised Detection of Marine Animals in High Resolution Aerial Images

2022 ◽  
Vol 14 (2) ◽  
pp. 339
Author(s):  
Paul Berg ◽  
Deise Santana Maia ◽  
Minh-Tan Pham ◽  
Sébastien Lefèvre

Human activities in the sea, such as intensive fishing and exploitation of offshore wind farms, may impact negatively on the marine mega fauna. As an attempt to control such impacts, surveying, and tracking of marine animals are often performed on the sites where those activities take place. Nowadays, thank to high resolution cameras and to the development of machine learning techniques, tracking of wild animals can be performed remotely and the analysis of the acquired images can be automatized using state-of-the-art object detection models. However, most state-of-the-art detection methods require lots of annotated data to provide satisfactory results. Since analyzing thousands of images acquired during a flight survey can be a cumbersome and time consuming task, we focus in this article on the weakly supervised detection of marine animals. We propose a modification of the patch distribution modeling method (PaDiM), which is currently one of the state-of-the-art approaches for anomaly detection and localization for visual industrial inspection. In order to show its effectiveness and suitability for marine animal detection, we conduct a comparative evaluation of the proposed method against the original version, as well as other state-of-the-art approaches on two high-resolution marine animal image datasets. On both tested datasets, the proposed method yielded better F1 and recall scores (75% recall/41% precision, and 57% recall/60% precision, respectively) when trained on images known to contain no object of interest. This shows a great potential of the proposed approach to speed up the marine animal discovery in new flight surveys. Additionally, such a method could be adopted for bounding box proposals to perform faster and cheaper annotation within a fully-supervised detection framework.

2021 ◽  
Author(s):  
Thomas Spangehl ◽  
Michael Borsche ◽  
Deborah Niermann ◽  
Frank Kaspar ◽  
Birger Tinz

<p>The exploitation of offshore wind energy is an essential part of the German energy transition (Energiewende). The planning of new offshore wind farms demands detailed information on wind conditions at turbine hub heights in the North Sea and Baltic Sea. High-resolution reanalyses which are based on state-of-the-art numerical weather prediction (NWP) models combined with data assimilation systems offer the required meteorological data which are suitable for climatological assessment.</p><p>The regional reanalysis COSMO-REA6 operated by Germany’s national meteorological service (Deutscher Wetterdienst, DWD) provides hourly data of 6 km horizontal resolution for 1995-2019/08 (Kaspar et al., 2020). Moreover, hourly data of 31 km horizontal resolution for 1950 to present are available from the global reanalysis ERA5 produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). DWD delivers reanalysis data and statistical evaluation results to Bundesamt für Seeschifffahrt und Hydrographie (BSH) in order to facilitate offshore site tenders. Data and a report were recently published as part of the tenders for 2021 (https://pinta.bsh.de/).</p><p>Here we present an evaluation of the 100 m wind speed and direction from COSMO-REA6 and ERA5 based on a comprehensive statistical analysis. On the reference side the FINO measurements (Research platforms in the North Sea and Baltic Sea, https://www.fino-offshore.de/en/index.html) from FINO1 and FINO2 are used. The FINO measurements are not used by the data assimilation schemes of the two reanalyses and therefore constitute independent reference data. The focus is on episodes prior to the installation of wind farms in the direct vicinity of the FINO platforms to avoid wake effects. The quality of the two reanalyses is compared to other state-of-the-art reanalyses and wind atlas data.</p><p>Reference:</p><p>Kaspar et al. (2020): Regional atmospheric reanalysis activities at Deutscher Wetterdienst: review of evaluation results and application examples with a focus on renewable energy, Adv. Sci. Res., 17, 115–128, https://doi.org/10.5194/asr-17-115-2020.</p>


2020 ◽  
Vol 54 (1) ◽  
pp. qjegh2020-033 ◽  
Author(s):  
J. Guinan ◽  
C. McKeon ◽  
E. O'Keeffe ◽  
X. Monteys ◽  
F. Sacchetti ◽  
...  

The characterization of the seafloor is a fundamental first step in informing resource management, marine spatial planning, conservation, fisheries, industry and research. Integrated Mapping for the Sustainable Development of Ireland's Marine Resource (INFOMAR), Ireland's national seabed mapping programme, delivers freely available, high-resolution seabed imagery derived from multibeam echosounder data in the Irish Exclusive Economic Zone. The European Union established the European Marine Observation and Data Network (EMODnet) Geology data portal, which provides harmonized broad-scale seabed substrate information for all European seas and confidence assessments of the information that underpins the geological interpretations. A multi-scale product has been produced using INFOMAR's high-resolution seabed substrate information at the 1:50 000 scale. As part of the Supporting Implementation of Maritime Spatial Planning in the Celtic Seas project, the EMODnet Geology seabed substrate data portal assisted in addressing the challenges associated with the implementation of the European Union's Marine Spatial Planning Directive. The seabed substrate data in the EMODnet Geology data portal were identified as a valuable tool for guiding the selection of sites for offshore wind farms in the Irish Sea and their subsequent characterization. This paper outlines the approach to delivering a multi-scale seabed substrate dataset for the Irish offshore and its applicability to marine spatial planning and the development of offshore energy resources.Thematic collection: This article is part of the Mapping the Geology and Topography of the European Seas (EMODnet) collection available at: https://www.lyellcollection.org/cc/EMODnet


Author(s):  
A. C. Carrilho ◽  
M. Galo

<p><strong>Abstract.</strong> Recent advances in machine learning techniques for image classification have led to the development of robust approaches to both object detection and extraction. Traditional CNN architectures, such as LeNet, AlexNet and CaffeNet, usually use as input images of fixed sizes taken from objects and attempt to assign labels to those images. Another possible approach is the Fast Region-based CNN (or Fast R-CNN), which works by using two models: (i) a Region Proposal Network (RPN) which generates a set of potential Regions of Interest (RoI) in the image; and (ii) a traditional CNN which assigns labels to the proposed RoI. As an alternative, this study proposes an approach to automatic object extraction from aerial images similar to the Fast R-CNN architecture, the main difference being the use of the Simple Linear Iterative Clustering (SLIC) algorithm instead of an RPN to generate the RoI. The dataset used is composed of high-resolution aerial images and the following classes were considered: house, sport court, hangar, building, swimming pool, tree, and street/road. The proposed method can generate RoI with different sizes by running a multi-scale SLIC approach. The overall accuracy obtained for object detection was 89% and the major advantage is that the proposed method is capable of semantic segmentation by assigning a label to each selected RoI. Some of the problems encountered are related to object proximity, in which different instances appeared merged in the results.</p>


2021 ◽  
Author(s):  
Jana Fischereit ◽  
Kurt Schaldemose Hansen ◽  
Xiaoli Guo Larsén ◽  
Maarten Paul van der Laan ◽  
Pierre-Elouan Réthoré ◽  
...  

Abstract. Numerical wind resource modelling across scales from mesoscale to turbine scale is of increasing interest due to the expansion of offshore wind energy. Offshore, wind farm wakes can last several tens kilometres downstream and thus affect the wind resources of a large area. So far, scale-specific models have been developed and it remains unclear, how well the different model types can represent intra-farm wakes, farm-to-farm wakes as well as the wake recovery behind a farm. Thus, in the present analysis the simulation of a set of wind farm models of different complexity, fidelity, scale and computational costs are compared among each other and with SCADA data. In particular, two mesoscale wind farm parameterizations implemented in the mesoscale Weather Research and Forecasting model (WRF), the Explicit Wake Parameterization (EWP) and the Wind Farm Parameterization (FIT), two different high-resolution RANS simulations using PyWakeEllipSys equipped with an actuator disk model, and three rapid engineering wake models from the PyWake suite are selected. The models are applied to the Nysted and Rødsand II wind farms, which are located in the Fehmarn Belt in the Baltic Sea. Based on the performed simulations, we can conclude that average intra-farm variability can be captured reasonable well with WRF+FIT using a resolution of 2 km, a typical resolution of mesoscale models for wind energy applications, while WRF+EWP underestimates wind speed deficits. However, both parameterizations can be used to estimate median wind resource reduction caused by an upstream farm. All considered engineering wake models from the PyWake suite simulate intra-farm wakes comparable to the high fidelity RANS simulations. However, they considerably underestimate the farm wake effect of an upstream farm although with different magnitudes. Overall, the higher computational costs of PyWakeEllipSys and WRF compared to PyWake pay off in terms of accuracy for situations when farm-to-farm wakes are important.


2019 ◽  
Vol 11 (9) ◽  
pp. 1128 ◽  
Author(s):  
Maryam Rahnemoonfar ◽  
Dugan Dobbs ◽  
Masoud Yari ◽  
Michael J. Starek

Recent deep-learning counting techniques revolve around two distinct features of data—sparse data, which favors detection networks, or dense data where density map networks are used. Both techniques fail to address a third scenario, where dense objects are sparsely located. Raw aerial images represent sparse distributions of data in most situations. To address this issue, we propose a novel and exceedingly portable end-to-end model, DisCountNet, and an example dataset to test it on. DisCountNet is a two-stage network that uses theories from both detection and heat-map networks to provide a simple yet powerful design. The first stage, DiscNet, operates on the theory of coarse detection, but does so by converting a rich and high-resolution image into a sparse representation where only important information is encoded. Following this, CountNet operates on the dense regions of the sparse matrix to generate a density map, which provides fine locations and count predictions on densities of objects. Comparing the proposed network to current state-of-the-art networks, we find that we can maintain competitive performance while using a fraction of the computational complexity, resulting in a real-time solution.


Sensors ◽  
2021 ◽  
Vol 21 (4) ◽  
pp. 1033
Author(s):  
Qiaodi Wen ◽  
Ziqi Luo ◽  
Ruitao Chen ◽  
Yifan Yang ◽  
Guofa Li

By detecting the defect location in high-resolution insulator images collected by unmanned aerial vehicle (UAV) in various environments, the occurrence of power failure can be timely detected and the caused economic loss can be reduced. However, the accuracies of existing detection methods are greatly limited by the complex background interference and small target detection. To solve this problem, two deep learning methods based on Faster R-CNN (faster region-based convolutional neural network) are proposed in this paper, namely Exact R-CNN (exact region-based convolutional neural network) and CME-CNN (cascade the mask extraction and exact region-based convolutional neural network). Firstly, we proposed an Exact R-CNN based on a series of advanced techniques including FPN (feature pyramid network), cascade regression, and GIoU (generalized intersection over union). RoI Align (region of interest align) is introduced to replace RoI pooling (region of interest pooling) to address the misalignment problem, and the depthwise separable convolution and linear bottleneck are introduced to reduce the computational burden. Secondly, a new pipeline is innovatively proposed to improve the performance of insulator defect detection, namely CME-CNN. In our proposed CME-CNN, an insulator mask image is firstly generated to eliminate the complex background by using an encoder-decoder mask extraction network, and then the Exact R-CNN is used to detect the insulator defects. The experimental results show that our proposed method can effectively detect insulator defects, and its accuracy is better than the examined mainstream target detection algorithms.


Energies ◽  
2019 ◽  
Vol 12 (14) ◽  
pp. 2780
Author(s):  
Jared A. Lee ◽  
Paula Doubrawa ◽  
Lulin Xue ◽  
Andrew J. Newman ◽  
Caroline Draxl ◽  
...  

Offshore wind resource assessments for the conterminous U.S. and Hawai’i have been developed before, but Alaska’s offshore wind resource has never been rigorously assessed. Alaska, with its vast coastline, presents ample potential territory in which to build offshore wind farms, but significant challenges have thus far limited Alaska’s deployment of utility-scale wind energy capacity to a modest 62 MW (or approximately 2.7% of the state’s electric generation) as of this writing, all in land-based wind farms. This study provides an assessment of Alaska’s offshore wind resource, the first such assessment for Alaska, using a 14-year, high-resolution simulation from a numerical weather prediction and regional climate model. This is the longest-known high-resolution model data set to be used in a wind resource assessment. Widespread areas with relatively shallow ocean depth and high long-term average 100-m wind speeds and estimated net capacity factors over 50% were found, including a small area near Alaska’s population centers and the largest transmission grid that, if even partially developed, could provide the bulk of the state’s energy needs. The regional climate simulations were validated against available radiosonde and surface wind observations to provide the confidence of the model-based assessment. The model-simulated wind speed was found to be skillful and with near-zero average bias (−0.4–0.2 m s−1) when averaged over the domain. Small sample sizes made regional validation noisy, however.


Author(s):  
Shubair Abdullah

Detecting the denial of service attacks that solely target the router is a maximum security imperative in deploying IPv6 networks. The state-of-the-art Denial of Service detection methods aim at leveraging the advantages of flow statistical features and machine learning techniques. However, the detection performance is highly affected by the quality of the feature selector and the reliability of datasets of IPv6 flow information. This paper proposes a new neuro-fuzzy inference system to tackle the problem of classifying the packets in IPv6 networks in crucial situation of small-supervised training dataset. The proposed system is capable of classifying the IPv6 router alert option packets into denial of service and normal by utilizing the neuro-fuzzy strengths to boost the classification accuracy. A mathematical analysis from the fuzzy sets theory perspective is provided to express performance benefit of the proposed system. An empirical performance test is conducted on comprehensive dataset of IPv6 packets produced in a supervised environment. The result shows that the proposed system overcomes robustly some state-of-the-art systems.


Sign in / Sign up

Export Citation Format

Share Document