ABSTRACT
Remote sensing allows obtaining information on agriculture regularly with non-invasive measurement approaches. Field data is crucial for adequate agricultural monitoring by remote sensing. However, public available field data are scarce, mainly in tropical regions, where agriculture is highly dynamic. The present publication aims to support the reduction of this gap. The LEM+ dataset provides information monthly about 16 land use classes for 1854 fields from October 2019 to September 2020 (one Brazilian agricultural year) from Luís Eduardo Magalhães (LEM) and other municipalities in the west of Bahia state, Brazil. The reference data were collected in two fieldworks (March 2020 - first crop season, and August 2020 - second crop season). The boundaries of the fields visited in situ were delimited using Sentinel-2 false color compositions (near infrared - red - green) at 10â¯m spatial resolution. The land use classes were labeled monthly based on information collected in situ (agricultural land use and photographs) and by visual interpretation of Sentinel-2 false color composition (near infrared - shortwave infrared - red) and MODIS/Terra (Normalized Difference Vegetation Index) time series. The dataset can be useful for the development of new pattern recognition methods for agricultural land use mapping and monitoring, comparison of different classification methods, and optical and SAR remote sensing time series analysis. This dataset contributes to complement previous initiatives [1], [2] to make tropical agriculture field reference data publicly available.
ABSTRACT
O presente trabalho teve o objetivo de avaliar as imagens CCD/CBERS-2 quanto à possibilidade de discriminarem variedades de citros. A área de estudo localiza-se em Itirapina (SP) e, para este estudo, foram utilizadas imagens CCD de três datas (30/05/2004, 16/08/2004 e 11/09/2004). Um modelo que integra os elementos componentes da cena citrícola sensoriada é proposto com o objetivo de explicar a variabilidade das respostas das parcelas de citros em imagens orbitais do tipo CCD/CBERS-2. Foram feitas classificações pelos algoritmos Isoseg e Maxver e, de acordo com o índice kappa, concluiu-se que é possível obterem-se exatidões qualificadas como muito boas, sendo que as melhores classificações foram conseguidas com imagens da estação seca.
This paper was aimed at evaluating the possibility of discriminating citrus varieties in CCD imageries from CBERS-2 satellite ("China-Brazil Earth Resouces Satellite"). The study area is located in Itirapina, São Paulo State. For this study, three CCD images from 2004 were acquired (May 30, August 16, and September 11). In order to acquire a better understanding and for explaining the variability of the spectral behavior of the citrus areas in orbital images (like as the CCD/CBERS-2 images) a model that integrates the elements of the citrus scene is proposed and discussed. The images were classified by Isoseg and MaxVer classifiers. According to kappa index, it was possible to obtain classifications qualified as 'very good'. The best results were obtained with the images from the dry season.
