Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time
Satellite images serve a crucial role in weather prediction, particularly in assessing the strength of tropical storms such as cyclones. Tropical cyclones are commonly observed in regions of open oceans where conventional meteorological stations are scarce or absent, posing difficulties in accuratel...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-02-01
|
| Series: | Alexandria Engineering Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1110016824012377 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1825206983895023616 |
|---|---|
| author | Priyanka Nandal Prerna Mann Navdeep Bohra Ghadah Aldehim Asma Abbas Hassan Elnour Randa Allafi |
| author_facet | Priyanka Nandal Prerna Mann Navdeep Bohra Ghadah Aldehim Asma Abbas Hassan Elnour Randa Allafi |
| author_sort | Priyanka Nandal |
| collection | DOAJ |
| description | Satellite images serve a crucial role in weather prediction, particularly in assessing the strength of tropical storms such as cyclones. Tropical cyclones are commonly observed in regions of open oceans where conventional meteorological stations are scarce or absent, posing difficulties in accurately measuring their precise intensity in those areas. Meteorologists utilise high-resolution satellite imagery to address this challenge by accurately perceiving the complexities of storm genesis and structure. This research presents a novel method for forecasting the strength of tropical storms using the You Only Look Once-Neural Architecture Search (YOLO-NAS) algorithm, leveraging real-time satellite imagery from the INSAT-3D platform. YOLO-NAS, a state-of-the-art object detection model, is used to detect and analyse key features in multispectral infrared satellite images, which are vital for assessing cyclone strength. The model is trained on an extensive dataset covering cyclone events in the Indian Ocean region. The experimental findings indicate an accuracy rate of 81 %, a retrieval rate of 83 %, mean Average Precision (mAP) of 85.19 %, and a Mean Square Error (MSE) of 0.10. The study demonstrates that accurate estimation of tropical cyclone strength can be obtained, leading to improved real-time capabilities. Extensive experimentation illustrates that the proposed model achieves superior performance on various metrics such as accuracy, precision, recall, F1-Score, mAP rates, MSE and inference time as compared to other similar cutting-edge algorithms. The meteorological significance of this work lies in its ability to offer a reliable, real-time tool for cyclone intensity prediction, which could be integrated into operational weather forecasting systems. The finding shows that there is potential for practical use of this method. |
| format | Article |
| id | doaj-art-db41a2910d064993ba647a33b9f7b05f |
| institution | Kabale University |
| issn | 1110-0168 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Alexandria Engineering Journal |
| spelling | doaj-art-db41a2910d064993ba647a33b9f7b05f2025-02-07T04:46:55ZengElsevierAlexandria Engineering Journal1110-01682025-02-01113227241Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real timePriyanka Nandal0Prerna Mann1Navdeep Bohra2Ghadah Aldehim3Asma Abbas Hassan Elnour4Randa Allafi5Department of Computer Science and Engineering, Maharaja Surajmal Institute of Technology, Delhi, IndiaDepartment of Computer Science and Engineering, Maharaja Agrasen Institute of Technology, Delhi, IndiaDepartment of Computer Science and Engineering, Maharaja Surajmal Institute of Technology, Delhi, IndiaDepartment of Information Systems, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, Riyadh, Saudi ArabiaDepartment of Computer Science, Applied College, Girls Section at Muhayel, King Khalid University, Muhayel Aseer, Saudi ArabiaDepartment of Computers and Information Technology, College of Sciences and Arts, Northern Border University, Arar, Saudi Arabia; Corresponding author.Satellite images serve a crucial role in weather prediction, particularly in assessing the strength of tropical storms such as cyclones. Tropical cyclones are commonly observed in regions of open oceans where conventional meteorological stations are scarce or absent, posing difficulties in accurately measuring their precise intensity in those areas. Meteorologists utilise high-resolution satellite imagery to address this challenge by accurately perceiving the complexities of storm genesis and structure. This research presents a novel method for forecasting the strength of tropical storms using the You Only Look Once-Neural Architecture Search (YOLO-NAS) algorithm, leveraging real-time satellite imagery from the INSAT-3D platform. YOLO-NAS, a state-of-the-art object detection model, is used to detect and analyse key features in multispectral infrared satellite images, which are vital for assessing cyclone strength. The model is trained on an extensive dataset covering cyclone events in the Indian Ocean region. The experimental findings indicate an accuracy rate of 81 %, a retrieval rate of 83 %, mean Average Precision (mAP) of 85.19 %, and a Mean Square Error (MSE) of 0.10. The study demonstrates that accurate estimation of tropical cyclone strength can be obtained, leading to improved real-time capabilities. Extensive experimentation illustrates that the proposed model achieves superior performance on various metrics such as accuracy, precision, recall, F1-Score, mAP rates, MSE and inference time as compared to other similar cutting-edge algorithms. The meteorological significance of this work lies in its ability to offer a reliable, real-time tool for cyclone intensity prediction, which could be integrated into operational weather forecasting systems. The finding shows that there is potential for practical use of this method.http://www.sciencedirect.com/science/article/pii/S1110016824012377Deep learningSatellite imagesTropical cyclone intensity detectionYOLOYOLO-NAS |
| spellingShingle | Priyanka Nandal Prerna Mann Navdeep Bohra Ghadah Aldehim Asma Abbas Hassan Elnour Randa Allafi Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time Alexandria Engineering Journal Deep learning Satellite images Tropical cyclone intensity detection YOLO YOLO-NAS |
| title | Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time |
| title_full | Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time |
| title_fullStr | Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time |
| title_full_unstemmed | Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time |
| title_short | Tropical cyclone intensity estimation based on YOLO-NAS using satellite images in real time |
| title_sort | tropical cyclone intensity estimation based on yolo nas using satellite images in real time |
| topic | Deep learning Satellite images Tropical cyclone intensity detection YOLO YOLO-NAS |
| url | http://www.sciencedirect.com/science/article/pii/S1110016824012377 |
| work_keys_str_mv | AT priyankanandal tropicalcycloneintensityestimationbasedonyolonasusingsatelliteimagesinrealtime AT prernamann tropicalcycloneintensityestimationbasedonyolonasusingsatelliteimagesinrealtime AT navdeepbohra tropicalcycloneintensityestimationbasedonyolonasusingsatelliteimagesinrealtime AT ghadahaldehim tropicalcycloneintensityestimationbasedonyolonasusingsatelliteimagesinrealtime AT asmaabbashassanelnour tropicalcycloneintensityestimationbasedonyolonasusingsatelliteimagesinrealtime AT randaallafi tropicalcycloneintensityestimationbasedonyolonasusingsatelliteimagesinrealtime |