A Multi-Channel Convolutional Neural Network Based Target Detection Approach For Coastal Surveillance Radars

Date

2024-01-01

Author

Aybar, Baba
Güvensen, Gökhan Muzaffer
Yilmaz, A. Ozgur

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

28
views

0
downloads

Cite This

Marine target detection plays an important role in coastal security, sea surveillance, navigation and vessel traffic management. Traditional methods for target detection in maritime environments often face challenges in high sea states when the target is small. In recent years, the application of convolutional neural networks (CNNs), specifically deep learning architectures, has demonstrated significant promise in addressing these challenges. Methods like YOLO (You-Only-Look-Once) and region-based CNNs are often applied to object detection and target classification problems. This paper presents a multi-channel deep learning object detection approach based on YOLOv4 network for marine target detection. The proposed method creates a false-alarm controllable target detection framework by applying multi-channel multi-scan YOLO directly on the raw radar data instead of radar images and creating a novel alternative for widely used scan-to-scan integration methods. In this paper, a mechanically scanning mono-static coastal surveillance radar is used as the radar framework. The strengths and limitations of the proposed model are evaluated in terms of detection and false alarm probabilities and compared with the traditional methods.

URI

https://hdl.handle.net/11511/116711

Collections

Department of Electrical and Electronics Engineering, Conference / Seminar