| Autor: |
Retheesh, V. V., Ambili, M. P., Kalady, Saidalavi |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2023, Vol. 2546 Issue 1, p1-7, 7p |
| Abstrakt: |
Detection of floating vessels from space borne optical images is important for a wide range of applications spanning from vessel traffic control to maritime security including naval warfare. Reliable detection of floating vessels enable the military to have strategic leap over the enemy during the time of conflicts. The complex background, types of vessels with different poses, shapes and size make the task of detection very difficult and challenging. Optical images stay ahead in vessel detection compared to other remote sensing images due to its higher resolution and visual contents. However these images pose challenges in detection process due to the influence of environmental effects mainly clouds, waves, clutter and sunglint. Also the increased resolution of these images cause high data volume leading to increased processing requirements. The vessel detection consists of mainly three sequential steps viz. finding the vessel candidate, classification of the same and establishing the identity of the vessel. Though Synthetic Aperture Radar (SAR) is a leading technology for maritime monitoring, the disadvantages of this technology like limited number of sensors, low revisit time and low resolution, causes hindrance to effective detection and monitoring. With increased number of optical satellites and very high resolution images, the optical image based method can partly overcome the limitation of SAR based approach. In this work a comparison of classical machine learning algorithms with convolutional neural network (CNN) based approach has been made. The study has revealed that the CNN approach outweighs the conventional machine learning algorithms in classifying vessels from space borne optical images. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
