A Comparative Analysis of Pixel-Based and Object-Based Approaches Using Multitemporal PlanetScope Imagery for Land Cover Classification

Remote sensing plays a crucial role in monitoring and managing land cover change and provides valuable insights for various applications, including environmental monitoring, urban planning, and natural resource management. In recent years, advances in sensor technology have led to the availability of high-resolution satellite imagery, enabling fine analysis of land cover dynamics. The study uses a multitemporal approach, where PlanetScope imagery are acquired at different points in time to capture temporal variations in land cover characteristics. The eight spectral bands provide improved ways to distinguish between different land cover types, including vegetation, water bodies, urban areas, and agricultural fields. Two classification approaches are evaluated: pixel-based (PB) classification, which assigns a land cover class to each individual pixel based on its spectral characteristics, and object-based (OB) classification, which groups neighbouring pixels into objects or segments and assigns a class label to each object based on its spectral, spatial, and contextual attributes. The OB approach performed better than the PB approach with an overall accuracy of 85.43%, compared to 81.90%, respectively. Also, ‘salt-and-pepper effect’ was significantly reduced using the OB approach. The study also investigates the potential advantages and limitations of each approach in capturing subtle land cover changes, spatial heterogeneity, and spectral variability.