DOAJ Open Access 2025

Hybrid deep learning and boosting for multi-source coastal scene classification using aerial imagery

Alireza Sharifi Bayan Alabdullah

Abstrak

This paper tackles a key challenge for protecting our coasts: quickly and accurately identifying different coastal landscapes from aerial photos. We present a smart AI system that combines deep learning with powerful boosting algorithms. Our method uses a pretrained neural network (ResNet18) to extract detailed visual features from high-resolution RGB images of Beaches, Rivers, and Ports, taken from the diverse AID dataset. These features are then classified by an XGBoost model, creating a robust fusion of techniques. Trained on globally sourced Google Earth imagery, the system proves highly effective across different sensors. It achieves an excellent 94.1% accuracy and F1-score, reliably distinguishing between visually similar scenes like beaches and rivers. This work demonstrates a practical and accurate tool for coastal monitoring, supporting better management of these vital ecosystems.

Topik & Kata Kunci

Penulis (2)

A

Alireza Sharifi

B

Bayan Alabdullah

Format Sitasi

Sharifi, A., Alabdullah, B. (2025). Hybrid deep learning and boosting for multi-source coastal scene classification using aerial imagery. https://doi.org/10.1080/10106049.2025.2596965

Akses Cepat

PDF tidak tersedia langsung

Cek di sumber asli →
Lihat di Sumber doi.org/10.1080/10106049.2025.2596965
Informasi Jurnal
Tahun Terbit
2025
Sumber Database
DOAJ
DOI
10.1080/10106049.2025.2596965
Akses
Open Access ✓