Comparison of CNNs and ViTs Based Hybrid Models Using Gradient Profile Loss for Classification of Oil Spills in SAR Images

Oil spillage over a sea or ocean’s surface is a threat to marine and coastal ecosystems. Spaceborne synthetic aperture radar (SAR) data has been used efficiently for the detection of oil spills due to its operational capability in all-day all-weather conditions. The problem is often modeled as a semantic segmentation task. The images need to be segmented into multiple regions of interest such as sea surface, oil spill, look-alikes, ships and land. Training of a classifier for this task is particularly challenging since there is an inherent class imbalance. In this work, we train a convolutional neural network (CNN) with multiple feature extractors for pixel-wise classification; and introduce to use a new loss function, namely ‘gradient profile’ (GP) loss, which is in fact the constituent of the more generic Spatial Profile loss proposed for image translation problems. For the purpose of training, testing and performance evaluation, we use a publicly available dataset with selected oil spill events verified by the European Maritime Safety Agency (EMSA). The results obtained show that the proposed CNN trained with a combination of GP, Jaccard and focal loss functions can detect oil spills with an intersection over union (IoU) value of 63.95%. The IoU value for sea surface, look-alikes, ships and land class is 96.00%, 60.87%, 74.61% and 96.80%, respectively. The mean intersection over union (mIoU) value for all the classes is 78.45%, which accounts for a 13% improvement over the state of the art for this dataset. Moreover, we provide extensive ablation on different Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) based hybrid models to demonstrate the effectiveness of adding GP loss as an additional loss function for training. Results show that GP loss significantly improves the mIoU and F1 scores for CNNs as well as ViTs based hybrid models. GP loss turns out to be a promising loss function in the context of deep learning with SAR images.

Download Full-text

Revisit of a Case Study of Spilled Oil Slicks Caused by the Sanchi Accident (2018) in the East China Sea

Journal of Marine Science and Engineering ◽

10.3390/jmse9030279 ◽

2021 ◽

Vol 9 (3) ◽

pp. 279

Author(s):

Zhehao Yang ◽

Weizeng Shao ◽

Yuyi Hu ◽

Qiyan Ji ◽

Huan Li ◽

...

Keyword(s):

East China Sea ◽

Oil Spill ◽

Oil Spills ◽

Ocean Model ◽

Sea Surface ◽

East China ◽

The East China Sea ◽

Sar Images ◽

China Sea ◽

Spatial Coverage

Marine oil spills occur suddenly and pose a serious threat to ecosystems in coastal waters. Oil spills continuously affect the ocean environment for years. In this study, the oil spill caused by the accident of the Sanchi ship (2018) in the East China Sea was hindcast simulated using the oil particle-tracing method. Sea-surface winds from the European Centre for Medium-Range Weather Forecasts (ECMWF), currents simulated from the Finite-Volume Community Ocean Model (FVCOM), and waves simulated from the Simulating WAves Nearshore (SWAN) were employed as background marine dynamics fields. In particular, the oil spill simulation was compared with the detection from Chinese Gaofen-3 (GF-3) synthetic aperture radar (SAR) images. The validation of the SWAN-simulated significant wave height (SWH) against measurements from the Jason-2 altimeter showed a 0.58 m root mean square error (RMSE) with a 0.93 correlation (COR). Further, the sea-surface current was compared with that from the National Centers for Environmental Prediction (NCEP) Climate Forecast System Version 2 (CFSv2), yielding a 0.08 m/s RMSE and a 0.71 COR. Under these circumstances, we think the model-simulated sea-surface currents and waves are reliable for this work. A hindcast simulation of the tracks of oil slicks spilled from the Sanchi shipwreck was conducted during the period of 14–17 January 2018. It was found that the general track of the simulated oil slicks was consistent with the observations from the collected GF-3 SAR images. However, the details from the GF-3 SAR images were more obvious. The spatial coverage of oil slicks between the SAR-detected and simulated results was about 1 km2. In summary, we conclude that combining numerical simulation and SAR remote sensing is a promising technique for real-time oil spill monitoring and the prediction of oil spreading.

Download Full-text

Sacrificial amphiphiles: Eco-friendly chemical herders as oil spill mitigation chemicals

Science Advances ◽

10.1126/sciadv.1400265 ◽

2015 ◽

Vol 1 (5) ◽

pp. e1400265 ◽

Cited By ~ 27

Author(s):

Deeksha Gupta ◽

Bivas Sarker ◽

Keith Thadikaran ◽

Vijay John ◽

Charles Maldarelli ◽

...

Keyword(s):

Crude Oil ◽

Oil Spill ◽

Oil Spills ◽

Cold Water ◽

Marine Ecosystem ◽

Hot Water ◽

Sea Surface ◽

Marine Biota ◽

Polar Group ◽

Branched Chain

Crude oil spills are a major threat to marine biota and the environment. When light crude oil spills on water, it forms a thin layer that is difficult to clean by any methods of oil spill response. Under these circumstances, a special type of amphiphile termed as “chemical herder” is sprayed onto the water surrounding the spilled oil. The amphiphile forms a monomolecular layer on the water surface, reducing the air–sea surface tension and causing the oil slick to retract into a thick mass that can be burnt in situ. The current best-known chemical herders are chemically stable and nonbiodegradable, and hence remain in the marine ecosystem for years. We architect an eco-friendly, sacrificial, and effective green herder derived from the plant-based small-molecule phytol, which is abundant in the marine environment, as an alternative to the current chemical herders. Phytol consists of a regularly branched chain of isoprene units that form the hydrophobe of the amphiphile; the chain is esterified to cationic groups to form the polar group. The ester linkage is proximal to an allyl bond in phytol, which facilitates the hydrolysis of the amphiphile after adsorption to the sea surface into the phytol hydrophobic tail, which along with the unhydrolyzed herder, remains on the surface to maintain herding action, and the cationic group, which dissolves into the water column. Eventual degradation of the phytol tail and dilution of the cation make these sacrificial amphiphiles eco-friendly. The herding behavior of phytol-based amphiphiles is evaluated as a function of time, temperature, and water salinity to examine their versatility under different conditions, ranging from ice-cold water to hot water. The green chemical herder retracted oil slicks by up to ~500, 700, and 2500% at 5°, 20°, and 35°C, respectively, during the first 10 min of the experiment, which is on a par with the current best chemical herders in practice.

Download Full-text

Feature Merged Network for Oil Spill Detection using SAR Images

Remote Sensing ◽

10.3390/rs13163174 ◽

2021 ◽

Vol 13 (16) ◽

pp. 3174

Author(s):

Yonglei Fan ◽

Xiaoping Rui ◽

Guangyuan Zhang ◽

Tian Yu ◽

Xijie Xu ◽

...

Keyword(s):

Oil Spill ◽

Oil Spills ◽

Recognition Accuracy ◽

Original Model ◽

High Dimensional ◽

Sar Image ◽

Sar Images ◽

Marine Oil ◽

Marine Oil Spills ◽

Overfitting Problem

The frequency of marine oil spills has increased in recent years. The growing exploitation of marine oil and continuous increase in marine crude oil transportation has caused tremendous damage to the marine ecological environment. Using synthetic aperture radar (SAR) images to monitor marine oil spills can help control the spread of oil spill pollution over time and reduce the economic losses and environmental pollution caused by such spills. However, it is a significant challenge to distinguish between oil-spilled areas and oil-spill-like in SAR images. Semantic segmentation models based on deep learning have been used in this field to address this issue. In addition, this study is dedicated to improving the accuracy of the U-Shape Network (UNet) model in identifying oil spill areas and oil-spill-like areas and alleviating the overfitting problem of the model; a feature merge network (FMNet) is proposed for image segmentation. The global features of SAR image, which are high-frequency component in the frequency domain and represents the boundary between categories, are obtained by a threshold segmentation method. This can weaken the impact of spot noise in SAR image. Then high-dimensional features are extracted from the threshold segmentation results using convolution operation. These features are superimposed with to the down sampling and combined with the high-dimensional features of original image. The proposed model obtains more features, which allows the model to make more accurate decisions. The overall accuracy of the proposed method increased by 1.82% and reached 61.90% compared with the UNet. The recognition accuracy of oil spill areas and oil-spill-like areas increased by approximately 3% and reached 56.33%. The method proposed in this paper not only improves the recognition accuracy of the original model, but also alleviates the overfitting problem of the original model and provides a more effective monitoring method for marine oil spill monitoring. More importantly, the proposed method provides a design principle that opens up new development ideas for the optimization of other deep learning network models.

Download Full-text

The effect of vertical mixing on the horizontal drift of oil spills

Ocean Science ◽

10.5194/os-14-1581-2018 ◽

2018 ◽

Vol 14 (6) ◽

pp. 1581-1601 ◽

Cited By ~ 8

Author(s):

Johannes Röhrs ◽

Knut-Frode Dagestad ◽

Helene Asbjørnsen ◽

Tor Nordam ◽

Jørgen Skancke ◽

...

Keyword(s):

Water Column ◽

Oil Spill ◽

Oil Spills ◽

Vertical Mixing ◽

Weather Conditions ◽

Model Simulations ◽

Horizontal Transport ◽

Crucial Component ◽

Drift Behavior ◽

A Chain

Abstract. Vertical and horizontal transport mechanisms for marine oil spills are investigated using numerical model simulations. To realistically resolve the 3-D development of a spill on the ocean surface and in the water column, recently published parameterizations for the vertical mixing of oil spills are implemented in the open-source trajectory framework OpenDrift (https://doi.org/10.5281/zenodo.1300358, last access: 7 April 2018). The parameterizations include the wave entrainment of oil, two alternative formulations for the droplet size spectra, and turbulent mixing. The performance of the integrated oil spill model is evaluated by comparing model simulations with airborne observations of an oil slick. The results show that an accurate description of a chain of physical processes, in particular vertical mixing and oil weathering, is needed to represent the horizontal spreading of the oil spill. Using ensembles of simulations of hypothetic oil spills, the general drift behavior of an oil spill during the first 10 days after initial spillage is evaluated in relation to how vertical processes control the horizontal transport. Transport of oil between the surface slick and the water column is identified as a crucial component affecting the horizontal transport of oil spills. The vertical processes are shown to control differences in the drift of various types of oil and in various weather conditions.

Download Full-text

The effect of vertical mixing on the horizontal drift of oil spills

10.5194/os-2018-100 ◽

2018 ◽

Cited By ~ 1

Author(s):

Johannes Röhrs ◽

Knut-Frode Dagestad ◽

Helene Asbjørnsen ◽

Tor Nordam ◽

Jørgen Skancke ◽

...

Keyword(s):

Oil Spill ◽

Oil Spills ◽

Vertical Mixing ◽

Weather Conditions ◽

Size Spectra ◽

Model Simulations ◽

Horizontal Transport ◽

Crucial Component ◽

Drift Behavior ◽

A Chain

Abstract. Vertical and horizontal transport mechanisms of marine oil spills are investigated using numerical model simulations. To realistically resolve the 3D-development of a spill on the ocean surface and in the water column, recently published parameterizations for the vertical mixing of oil spills are implemented in the open source trajectory framework OpenDrift1. These encompass the wave-entrainment of oil, two alternative formulations for the droplet size spectra, and turbulent mixing. The performance of the integrated oil spill model is evaluated by comparing model simulations with airborne observations of an oil slick. The results show that an accurate description of a chain of physical processes, in particular vertical mixing and oil weathering, is needed to represent the horizontal spreading of the oil spill. Using ensembles of simulations of hypothetic oil spills, the general drift behavior of an oil spill during the first 10 days after initial spillage is evaluated in relation to how vertical processes control the horizontal transport. Vertical mixing of oil between the surface slick and entrained oil is identified as a crucial component affecting the horizontal transport of oil spills. The vertical processes are shown to control differences in the drift of various types of oil and in various weather conditions. 1 https://github.com/opendrift/opendrift

Download Full-text

International Cooperation in Oil Spill Response

International Oil Spill Conference Proceedings ◽

10.7901/2169-3358-1991-1-61 ◽

1991 ◽

Vol 1991 (1) ◽

pp. 61-64 ◽

Cited By ~ 1

Author(s):

J. A. Nichols ◽

T. H. Moller

Keyword(s):

International Cooperation ◽

Oil Spill ◽

Oil Spills ◽

Oil Pollution ◽

Weather Conditions ◽

Federal Republic Of Germany ◽

Effective Response ◽

International Convention ◽

Specialized Equipment ◽

Favorable Weather

ABSTRACT Effective response to a major marine oil spill occasionally calls for specialized equipment, personnel, and expertise that is beyond the capability of the country or company concerned. In recognition of this fact, a new International Convention on International Cooperation in Oil Pollution Preparedness and Response has been developed under the auspices of the International Maritime Organization. There is already considerable potential for international cooperation through existing regional conventions and agreements, and other less formal arrangements. This cooperation involves governmental agencies, the oil and shipping industries, commercial companies, insurers, intergovernmental organizations, and international industry organizations. This will be illustrated by reference to two recent major oil spills in Europe where this international cooperation proved very successful. The first involved the cleanup of some 15,000 metric tons of heavy crude oil that impacted the holiday island of Porto Santo in the Madeiran archipelago. Cooperation among the Portuguese government, The International Tanker Owners Pollution Federation, the tanker's oil pollution insurer, the Commission of the European Communities, and the governments of France, Germany, the Netherlands, and the United Kingdom resulted in the rapid provision of specialized equipment and associated personnel to deal with the major shoreline contamination. The second incident, involving a spill of waste oil from a tanker in the Baltic Sea off the coast of Sweden, resulted in the rapid mobilization of cleanup resources from Sweden, Finland, Denmark, the Federal Republic of Germany, and the U.S.S.R. under the terms of the Helsinki Convention. During favorable weather conditions, the combined forces of the five countries were successful in recovering a high percentage of the oil at sea, with the result that the contamination of shorelines was minimal.

Download Full-text

Large-Scale Detection and Categorization of Oil Spills from SAR Images with Deep Learning

Remote Sensing ◽

10.3390/rs12142260 ◽

2020 ◽

Vol 12 (14) ◽

pp. 2260 ◽

Cited By ~ 2

Author(s):

Filippo Maria Bianchi ◽

Martine M. Espeseth ◽

Njål Borch

Keyword(s):

Deep Learning ◽

Oil Spill ◽

Large Scale ◽

Oil Spills ◽

Sar Images ◽

Oil Spill Detection ◽

Learning Framework ◽

Large Scale Data ◽

Human Operators ◽

Scale Detection

We propose a deep-learning framework to detect and categorize oil spills in synthetic aperture radar (SAR) images at a large scale. Through a carefully designed neural network model for image segmentation trained on an extensive dataset, we obtain state-of-the-art performance in oil spill detection, achieving results that are comparable to results produced by human operators. We also introduce a classification task, which is novel in the context of oil spill detection in SAR. Specifically, after being detected, each oil spill is also classified according to different categories of its shape and texture characteristics. The classification results provide valuable insights for improving the design of services for oil spill monitoring by world-leading providers. Finally, we present our operational pipeline and a visualization tool for large-scale data, which allows detection and analysis of the historical occurrence of oil spills worldwide.

Download Full-text

South Baltic Oil Spill Response Project (SBOIL)—Development and Implementation of Models of Drift and Fall Trajectories of Biogenic Oil Binders

Sustainability ◽

10.3390/su13179889 ◽

2021 ◽

Vol 13 (17) ◽

pp. 9889

Author(s):

Fokke Saathoff ◽

Marcus Siewert ◽

Marcin Przywarty ◽

Mateusz Bilewski ◽

Bartosz Muczyński ◽

...

Keyword(s):

Oil Spill ◽

Oil Recovery ◽

Oil Spills ◽

New Technology ◽

Weather Conditions ◽

Current Response ◽

Adverse Weather ◽

Recovery System ◽

Oil Spill Response

This paper presents the methodology, assumptions, and functionalities of an application developed during the realization of the project “South Baltic Oil Spill Response through Clean-up with Biogenic Oil Binders” (SBOIL). The SBOIL project is a continuation of the BioBind project, the primary goal of which was to develop and deploy an oil recovery system designed for use in coastal waters and adverse weather conditions. The goal of the SBOIL project was to use this new technology to improve the current response capabilities for cross-border oil spills. The developed application allows for the determination of the position of an aircraft at the time of dropping the oil binders, the determination of the oil binders’ position after falling in terms of a specific aircraft’s position, the determination of the position of oil binders after a certain time in order to plan the action of recovering it from the water surface, and the determination of the time when the binders will be in their assumed position.

Download Full-text

DARK SPOT DETECTION USING INTENSITY AND THE DEGREE OF POLARIZATION IN FULLY POLARIMETRIC SAR IMAGES FOR OIL POLUTION MONITORING

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ◽

10.5194/isprsarchives-xl-1-w5-749-2015 ◽

2015 ◽

Vol XL-1-W5 ◽

pp. 749-753

Author(s):

F. Zakeri ◽

J. Amini

Keyword(s):

Oil Spill ◽

Oil Spills ◽

Degree Of Polarization ◽

Dark Spot ◽

Sar Images ◽

Simple Method ◽

The Past ◽

Spot Detection ◽

Otsu Method ◽

Image Pixels

Oil spill surveillance is of great environmental and economical interest, directly contributing to improve environmental protection. Monitoring of oil spills using synthetic aperture radar (SAR) has received a considerable attention over the past few years, notably because of SAR data abilities like all-weather and day-and-night capturing. The degree of polarization (DoP) is a less computationally complex quantity characterizing a partially polarized electromagnetic field. The key to the proposed approach is making use of DoP as polarimetric information besides intensity ones to improve dark patches detection as the first step of oil spill monitoring. In the proposed approach first simple intensity threshold segmentation like Otsu method is applied to the image. Pixels with intensities below the threshold are regarded as potential dark spot pixels while the others are potential background pixels. Second, the DoP of potential dark spot pixels is estimated. Pixels with DoP below a certain threshold are the real dark-spot pixels. Choosing the threshold is a critical and challenging step. In order to solve choosing the appropriate threshold, we introduce a novel but simple method based on DoP of potential dark spot pixels. Finally, an area threshold is used to eliminate any remaining false targets. The proposed approach is tested on L band NASA/JPL UAVSAR data, covering the Deepwater Horizon oil spill in the Gulf of Mexico. Comparing the obtained results from the new method with conventional approaches like Otsu, K-means and GrowCut shows better achievement of the proposed algorithm. For instance, mean square error (MSE) 65%, Overall Accuracy 20% and correlation 40% are improved.

Download Full-text

Oil Spill Identification from SAR Images for Low Power Embedded Systems Using CNN

Remote Sensing ◽

10.3390/rs13183606 ◽

2021 ◽

Vol 13 (18) ◽

pp. 3606

Author(s):

Lorenzo Diana ◽

Jia Xu ◽

Luca Fanucci

Keyword(s):

Embedded Systems ◽

Power Consumption ◽

Low Power ◽

Oil Spills ◽

Weather Conditions ◽

Hardware Accelerators ◽

Sar Images ◽

Satellite Systems ◽

High Resolution Images ◽

Dedicated Hardware

Oil spills represent one of the major threats to marine ecosystems. Satellite synthetic-aperture radar (SAR) sensors have been widely used to identify oil spills due to their ability to provide high resolution images during day and night under all weather conditions. In recent years, the use of artificial intelligence (AI) systems, especially convolutional neural networks (CNNs), have led to many important improvements in performing this task. However, most of the previous solutions to this problem have focused on obtaining the best performance under the assumption that there are no constraints on the amount of hardware resources being used. For this reason, the amounts of hardware resources such as memory and power consumption required by previous solutions make them unsuitable for remote embedded systems such as nano and micro-satellites, which usually have very limited hardware capability and very strict limits on power consumption. In this paper, we present a CNN architecture for semantically segmenting SAR images into multiple classes. The proposed CNN is specifically designed to run on remote embedded systems, which have very limited hardware capability and strict limits on power consumption. Even if the performance in terms of results accuracy does not represent a step forward compared with previous solutions, the presented CNN has the important advantage of being able to run on remote embedded systems with limited hardware resources while achieving good performance. The presented CNN is compatible with dedicated hardware accelerators available on the market due to its low memory footprint and small size. It also provides many additional very significant advantages, such as having shorter inference times, requiring shorter training times, and avoiding transmission of irrelevant data. Our goal is to allow embedded low power remote devices such as satellite systems for remote sensing to be able to directly run CNNs on board, so that the amount of data that needs to be transmitted to ground and processed on ground can be substantially reduced, which will be greatly beneficial in significantly reducing the amount of time needed for identification of oil spills from SAR images.

Download Full-text