Novel remote sensing image classification for LULC mapping in cloud-prone areas

Prof. WENG Qihao, Associate Director of the Research Institute for Land and Space (RILS), and his team have developed an integrated time series mapping method to enhance land use and land cover (LULC) mapping accuracy and frequency in cloud-prone areas. Other key PolyU members on the research team include Prof. DING Xiaoli, Director of RILS, and Dr LI Zhiwei, Research Assistant Professor in the Department of Land Surveying and Geo-Informatics.

The method incorporates spectral-indices-fused deep learning models and time series reconstruction techniques. When the team applied the method to the cloud- and rain-prone Pearl River Delta (i.e., Guangdong–Hong Kong–Macao Greater Bay Area, GBA), it yielded an overall mapping accuracy of up to 87.01%, outperforming existing LULC products.

This method has the potential to generate seamless and near-real-time maps for different regions of the world by using deep learning models trained on datasets collected globally. It can provide high-quality LULC data sets at different time intervals for various land and water dynamics in cloud- and rain-prone regions.