پیوندهای مفید
اخبار و اعلانات
آمار
| تعداد نشریات | 161 |
| تعداد شماره‌ها | 6,573 |
| تعداد مقالات | 71,037 |
| تعداد مشاهده مقاله | 125,517,264 |
| تعداد دریافت فایل اصل مقاله | 98,777,760 |
Identification of Coastal and Non-Coastal Vessels in SAR Radar Images Using YOLOv8 Deep Learning Network | ||
| Earth Observation and Geomatics Engineering | ||
| دوره 7، شماره 2، اسفند 2023 اصل مقاله (923.89 K) | ||
| نوع مقاله: Original Article | ||
| شناسه دیجیتال (DOI): 10.22059/eoge.2024.381462.1160 | ||
| نویسندگان | ||
| Farnaz Ashtari1؛ Tayabe Managhebi* 2؛ Farhad Samadzadegan1 | ||
| 1School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran | ||
| 2School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran - | ||
| چکیده | ||
| Unlike optical satellite imagery, Synthetic Aperture Radar (SAR) satellites can capture images at any time of day and in any weather, making them highly effective for ocean monitoring. Automatic ship detection in SAR images is crucial for both military and civilian applications. Traditional SAR-based methods detect ships by analyzing differences in backscatter between ships and the sea surface. However, challenges such as wave interference and the proximity of ships to the coastline often reduce the accuracy of these methods. To overcome these challenges, deep learning-based approaches have been developed, offering advanced data processing and automatic feature extraction. However, many existing deep learning models are complex and computationally heavy, limiting their effectiveness in real-time applications. To address this, a lightweight and efficient network named YOLOv8 is proposed for ship detection in SAR images. YOLOv8’s optimized architecture, featuring multiple convolutional layers, enhances the extraction of high-level semantic features, improving accuracy, speed, and robustness in ship detection. The effectiveness of this approach was evaluated using the SSDD ship dataset, which includes a wide variety of vessels in different backgrounds and distances from the coastline. This dataset, with various SAR image polarizations and resolutions, provides comprehensive coverage of SAR data types. The proposed YOLOv8 model achieved impressive results, with a detection accuracy of 99%, precision of 96%, mean Average Precision (mAP) of 98%, F1 score of 97%, and recall of 95%. These results highlight the model's efficiency and accuracy in detecting ships in SAR images. The proposed network was also able to detect small and large ships with small distance to each other in a variety of coastal images with a crowded background, showing the capability and acceptable performance of the network. | ||
| کلیدواژه‌ها | ||
| Ship detection؛ SAR radar images؛ SSDD dataset؛ Deep learning network؛ YOLOv8 network | ||
|
آمار تعداد مشاهده مقاله: 54 تعداد دریافت فایل اصل مقاله: 27 |
||
سامانه مدیریت نشریات علمی. قدرت گرفته از سیناوب