An inclusive classification optimization model for land use and land cover classification

The inclusive classification optimization model (ICOM) is an advanced fusion model specifically designed for large-scale land use and land cover classification. It automatically derives training samples from the intersection areas of five available top-tier classification products, and validation samples from their union areas. The reconciliation index is designed to refine the samples and aid in selecting the most suitable composite approach of the Landsat images for classification. ICOM further employs six classifiers from Google Earth Engine to comprehensively explore classification details of the customized Landsat images, and utilizes an accuracy-weighted plurality voting approach to integrate all the 11 classification results into a final optimized outcome. These accuracy indices comprises traditional metrics including producer's accuracy and user's accuracy, alongside a newly designed index of matching accuracy, which specifically assesses the capability of each land-cover class in accurately reflecting real-world conditions. Results demonstrate that that ICOM effectively capitalizes on the strengths of the incorporated high-quality classification products and classifiers, with the integrated classification output significantly enhancing both overall accuracy and localized classification performance.