A Study on Oil Spill Detection in Ocean with Look a likes

Authors

  • J. Senthil Murugan, S. Durga Devi, M. Sakthivel, W. T. Chembian, G. Anushiya, S. Kavi priya, Infanta Mendez

DOI:

https://doi.org/10.17762/msea.v72i1.1684

Abstract

One of the biggest dangers to marine and coastal areas is oil spill. Effective oil slick monitoring and early detection are essential for the relevant authorities to respond quickly, control environmental pollution, and prevent additional harm. Because of their reliability and efficiency across a wide range of environmental and illuminative conditions, synthetic aperture radar (SAR) sensors are commonly used for this purpose. SAR sensors can clearly detect black patches that are likely connected to oil spills, but differentiating them from other objects that seem similar is a difficult task. Many alternative approaches have been put forth to automatically find and categorise these dark patches. The vast majority of them offer incomparable results due to the usage of different datasets. As an added complication, SAR images are generally labeled with a single label that applies to the whole picture, making it difficult to adjust nuanced parameters or extract relevant data. Deep convolutional neural networks (DCNNs) and the Random Forest Classifier are suggested as an effective method to get over these restrictions. A publicly accessible SAR picture collection is also introduced with the intention of serving as a standard for future oil spill detection technologies. The performance of well-known DCNN segmentation models and the Random Forest method in the given job is evaluated using the dataset that is being presented.Random Forest performed most efficiently in terms of test accuracy and inference time. Furthermore, using the provided dataset, it is explored and demonstrated how complex the given challenge is, particularly in light of the difficult task of differentiating actual oil spills and their imitations. Results suggest that effective oil spill detectors can be implemented using DCNN segmentation models with Random forest classifier, trained and assessed on the presented dataset. Future research on oil spill identification and SAR image processing is anticipated to benefit considerably from the current work.

Downloads

Published

2023-01-17

How to Cite

J. Senthil Murugan, S. Durga Devi, M. Sakthivel, W. T. Chembian, G. Anushiya, S. Kavi priya, Infanta Mendez. (2023). A Study on Oil Spill Detection in Ocean with Look a likes. Mathematical Statistician and Engineering Applications, 72(1), 90–97. https://doi.org/10.17762/msea.v72i1.1684

Issue

Section

Articles