Singapore green coverage analysis

• Main Author: Mok, Ying Chong
• Other Authors: Lee Bu Sung, Francis
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/171975
• Institution: Nanyang Technological University

With the increasing importance of producing precise and up to date land use land class (LULC) maps, which are crucial for governmental agencies and private companies involved in monitoring large-scale changes in land resources. This report proposes a pipeline for the generation of LULC maps from satellite imagery using a lightweight CNN model for semantic segmentation of satellite images. The proposed pipeline automatically conducts pre-processing on the input data and performs prediction to classify the data into pre-defined classes. The presented network is a novel lightweight model and then fine-tuned through varying hyperparameters. Overall accuracy of 95.15% was observed, with mean F1-score of 55.84% and mean Intersection over Union of 49.85%. The proposed model achieved better results compared to Random Forest model and U-Net model.