| Annotation |
The article discusses Earth remote sensing methods used to detect and map oil spills. Research is aimed at solving the problems, studied in the field of computer vision, in accordance with the general line in solving problems of aerospace monitoring of oil development sites and oil pollution in the shelf waters.
The object of the study are images or a sequence of images of the natural environment.
The subject of the research is the study and development of mathematical modeling and hardware and software for image processing and analysis, image recognition and classification, machine learning, which makes it possible to assess the ecological state of offshore oil development sites.
The purpose of the article was to perform research in the field of semi-automatic image analysis, which allows for the rapid detection of oil pollution disturbances in offshore areas using ultra-precise neural network algorithms (ResNet-10) with long-term and short-term memory (LSTM) networks when processing materials from several sources of information that requires spatial correspondence between images.
Visual data analysis based on the formation of stable scene-forming features using deep neural networks, for the task of recognition of objects on the ground has been carried out. It is proposed to use multidimensional heterogeneous (several images) remote sensing data in a common depository as a basis for grouping or comparing the corresponding information.
The objectives of the study which realize the purpose of the article are the following:
– to form a set of exercise images containing scenes in a changing environment (weather conditions, illumination, seasonality and viewing angle);
– train deep convolutional neural networks using their own databases, which allow to identify stable features of images;
– to develop methods and algorithms for processing and analyzing images for the formation of stable scene-forming features from trained neural networks;
– to conduct computational experiments for comparative analysis and evaluation of the results of classification and categorization using a model of complex networks to manage complex information.
For this purpose, automatic registration of geometric deformations (translation, rotation and scaling) was performed using bilinear interpolation, testing was performed for a possible variation of the statistical model within a non-uniform sliding window, based on a semi-automatic approach, offshore areas of Oil Rocks (Caspian Sea).
Standard single-layer 2D LSTM networks [1] solve the problem of texture segmentation using texture perpixel classification.
The network accurately estimates texture regions and automatically adapts the different scale, orientation, and shape of texture regions in an image.
The article shows a simple and direct way to apply LSTM networks for texture segmentation, compared the efficiency (accuracy) using research-based classification quality measurement using a new similarity measure based on a statistical model (three versions of the nearest neighbor rule and the maximum likelihood method) [ 2].
The results of the performed studies generally confirmed the possibility and effectiveness of the proposed model. Allows you to simulate the competitive multi-temporal behavior of the model.
The second front of research, the analysis of object recognition, showed that contextual information is an important function that must be taken into account in object recognition systems.
In addition, evidence was found that the natural formation of clusters indicates that contexts have arisen that seem to were fundamental to performance outcomes.
However, it is important emphasize, that these experiments are empirical in nature and are carried out on a certain image base, however, well known in the academic scientific community
|
| References |
1. Bondur V. G. Issledovaniye Zemli iz kosmosa (Exploration of the Earth from space), 2000, no. 5, p. 28. Available at: https://www.aerocosmos.info/pdf/2006/Bon_model.pdf (date of access: 12/25/2021). Text: electronic.
2. Bondur V. G., Savin A. I. Issledovaniye Zemli iz kosmosa (Exploration of the Earth from space), 1995, no. 4, рр. 24‒33.
3. Bondur V. G., Arzhenenko N. I., Linnik V. N., Titova I. L. Issledovaniye Zemli iz kosmosa (Exploration of the Earth from space), 2003, no. 2, рр. 3‒17.
4. Bondur V. G., Starchenkov S. A. Izvestiya vysshikh uchebnykh zavedeniy: geodeziya i aerofotos\"yemka (News of higher educational institutions: geodesy and aerial photography), 2000, no. 3, рр. 118–143.
5. Vasil\'yev K. K. Spektr (Spektr). Ulyanovsk: Ul PI, 1992, рр. 3‒19.
6. Guliyev A. SH., Khlebnikova T. A. Interekspo GEO-Sibir\'. KHVII Mezhdunar. nauch. kongress (g. Novosibirsk, 19‒21 maya 2021 g.) (Interexpo GEO-Siberia. XVII Intern. scientific congress (Novosibirsk, May 19‒21, 2021): Sat. materials in 8 volumes. V. 4. Remote sensing methods of the Earth and photogrammetry, environmental monitoring, geoecology. Novosibirsk: SGUG), 2021, рр. 11‒16. Available at: https://doi.org/10.33764/2618-981X-2021-4-11-16.
7. Dzhayn A. K. TIIER (TIER), 1981, vol. 69, no. 5, рр. 9–39.
8. Alpers W., Holt B., Zeng K. Challenges and pitfalls. Remote Sens. Environ (Challenges and pitfalls. Remote Sens. Environ), ¬ 2017, рр. 133‒147.
9. Anil K., Karu J. IEEE Transactions on (IEEE Transactions on), ¬ 1996, vol. 18. ¬ no. 2, рр. 195‒205.
10. Brekke C., Solberg A.H.S. Remote Sens. Environ (Remote Sens. Environ), 2005, no. 95, рр. 1‒13. Available at: https://doi.org/10.3390/s8106642.
11. Efros A. IEEE International Conference on Computer Vision: Corfu, Greece (IEEE International Conference on Computer Vision: Corfu, Greece), -1999.
12. ESA Copernicus Open Access Hub (ESA Copernicus Open Access Hub), 2019. Available at: https://scihub.copernicus.eu (accessed: 25.12.2021). Text: electronic.
13. Fingas M., Brown C. E. In Oil Spill Science and Technology (In Oil Spill Science and Technology). Elsevier: Amsterdam, The Netherlands, 2011, рр. 111‒169.
14. Fingas M., Brown C. E. Standard Handbook Oil Spill Environmental Forensics (Standard Handbook Oil Spill Environmental Forensics). Elsevier: Amsterdam, The Netherlands, ¬ 2016, рр. 961‒981.
15. Fingas M., Brown C. Marine Pollution Bulletin (Marine Pollution Bulletin), 2014, no. 83, рр. 9‒23. Available at: https://doi.org/10.3390/s18010091.
16. Fingas M., Brown C. ¬ A Review of oil spill remote sensing (A Review of oil spill remote sensing), 2017, no. 18, рр. 91.
17. Graves A., Schmidhuber J. Advances in Neural Information Processing Systems (Advances in Neural Information Processing Systems), ¬2008, рр. 545–552.
18. Hochreiter S., Schmidhuber J. Neural Computation (Neural Computation), 1997, no. 9, рр. 1735‒1780. Available at: https://www.bioinf.jku.at/publications/older/2604.pdf (accessed: 25.12.2021). Text: electronic.
19. Kim K., Jung K., Park S. IEEE Transactions on (IEEE Transactions on), 2000, vol. 24, no. 11, рр. 1542–1550. Available at: https://core.ac.uk/download/pdf/42413589.pdf (accessed: 25.12.2021). Text: electronic.
20. Melendez J., Girones X. Image Processing (ICIP) 18th IEEE International Conference (Image Processing (ICIP) 18th IEEE International Conference), 2011,¬ рр. 2869–2872. Available at: https://10.1109/ICIP.2011.6116147 (date accessed: 25.12.2021). Text: electronic.
21. Tivive F., Bouzerdoum A. TENCON, IEEE Region 10 Conference (TENCON, IEEE Region 10 Conference), ¬ 2006, ¬ рр. 1–4.
|
