Abstract

Land cover classification using optical satellite images plays an important role in the change detection of materials on Earth. This data source has contributed significantly to monitoring and managing natural resources, as well as hazard analysis and risk assessment. Various presented studies have approached traditional algorithms such as K-Means, Iso Data, Random Forest, etc, to quickly obtain the classification output for further research. It is crucial to assess each classificaiton method's performance to select the most optimal algorithm for different classification projects. Applying different methods, such as K-means and Quick shift to VNREDSat-1 images, allows us to improve classification accuracy. From that, the study considers the combination of these classification approaches, along with spectral analysis developed in previous works (WiPE algorithm), to propose an automated land cover classification method taking advantages of all results applied to VNREDSat-1 images.