CLASSIFICAÇÃO DE IMAGENS DE VEÍCULOS AÉREOS NÃO TRIPULADOS NO SOFTWARE SPRING

Abstract

Aerial images with high spatial resolution allow the obtaining of a variety of information and a high level of detail giving subsidy for the accomplishment of several mappings. The objective of this work was to evaluate the performance of supervised pixel-by-pixel classifiers and by region (Segmentation) in orthomotics produced by RPA model Phanton 4 Pro using SPRING software version 5.5.4. The orthosaic object of analysis has spatial resolution of approximately 6cm, representing a cultivation area of ​​rizicultura located at the mouth of the river Longá in the municipality of Buriti dos Lopes, northern region of the State of Piauí. The maximum likelihood classification algorithms (Pixel by Pixel) and Bhattacharya (Region) were applied. Both algorithms were satisfactory but presented great confusion between the classes attributed to the spectral mixture resulting from the very high special resolution of the images.