CLASSFITO: SOFTWARE PARA A CLASSIFICAÇÃO DE FITOFISIONOMIAS POR IMAGENS

Abstract

The Cerrado biome is a hotspot composed of thousands of species of endemic plants and animals, with one of the largest river basins in South America, and has been constantly threatened by increasing deforestation, which is greater than the percentage of deforestation in the Amazon in recent years. In this sense, the identification of the remaining Cerrado vegetation classes, whether for conservation purposes or other applications, has become increasingly important. The works to identify vegetation classes, such as classification into formations and their phytophysiognomies by Ribeiro and Walter (1998), combined with remote sensing combined with supervised classification techniques, become a practical and agile option. The objective of this study was to articulate a geospatial data analysis process, through the development of the first version of the open access web software CLASSFITO. This software was developed in the python language, using the scikit learn library. It allows the training and application of previously trained machine learning models for vegetation classification by orbital or aerial images. The software proves to be efficient in training Random Forest and Support Vector Machine models, providing reports in raster format and information about class prediction. CLASSFITO is available at: https://classfito.app.br/, and is in the process of being registered with the Instituto Nacional de Propriedade Industrial - INPI.