Min Peng

Min Peng

Auto-generated profile, based on publicly available information

Min Peng

School of Information Technology & Engineering, Guangzhou College of Commerce, Guangzhou, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 February 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
MPMin Peng
Min Peng
SHShiqi Huang
Shiqi Huang
AKAsad Khan
Asad Khan
MBMAURICIO BARRIOS
MAURICIO BARRIOS
KMKhudoynazarov Egambergan Madrakhimovich
Khudoynazarov Egambergan Madrakhimovich
ATAhmad A. Telba
Ahmad A. Telba
The vital ecosystem services of coastal areas support biodiversity while storing carbon, protecting coasts, and conserving habitat for coastal species. Accurate mapping and monitoring of coastal ecosystems are essential for conservation and sustainable management, as these ecosystems face growing threats from human activities, sea-level rise, and climate change. A supervised Swin Transformer-based deep learning method using different hyperspectral datasets serves as the proposed algorithm for coastal cover mapping. The data requires preprocessing procedures that combine feature learning with normalization and dimensionality reduction to improve both spectral and spatial feature extraction. The Swin Transformer model extracts hierarchical features through its shifted window attention mechanisms, which combine local and global information. Through spectral-spatial fusion, the model utilizes the specific characteristics of each data source to enhance feature representation, enabling better discrimination of coastal area, ship detection, and large-scale coastal mapping. The integration of high-resolution spatial data with broader spectral information through multisource data methods supports robust classification and object detection. The algorithm achieves 92.4% overall classification accuracy through cross-validation and hyperparameter optimization while minimizing overfitting. It specifically enhances coastal area identification (>91%) and ship object detection (>90%). The analysis demonstrates that combining deep learning methods with diverse remote sensing data sources enables effective and precise mapping of coastal ecosystems.