Carlos Antonio Oliveira Vieira

Superfícies de resposta espectro-temporal de imagens do sensor MODIS para classificação de área de soja no Estado do Rio Grande do Sul

This paper was aimed at evaluating the potential and the limitations of MODIS images for soybean classification and area estimation through a Spectral-Temporal Response Surface (STRS) method. A soybean thematic map from Rio Grande do Sul State, Brazil, derived from Landsat images was used as reference data to assist both sample training and results comparison. Six 16-day composite MODIS images were classified through a supervised maximum likelihood algorithm (MAXVER) adapted to the STRS method. The results were evaluated using the Kappa coefficient for the entire study area and for one region dominated by large farms and another by small ones. The STRS method underestimated the soybean area by 6.6%, for the entire study area, with a Kappa coefficient of 0.503. For regions with large and small farms the soybean area was overestimated by 8% (Kappa=0.424) and underestimated by 43.4% (Kappa=0.358), respectively. Eventually, MODIS images, through the STRS method, demonstrated good potential to classify and estimate soybean area, mainly in regions with large farms. For regions with small farms the correct identification and classification of soybean areas showed to be less efficient due to the low spatial resolution of MODIS images.

STARS: A New Method for Multitemporal Remote Sensing

There is great potential for the development of remote sensing methods that integrate and exploit both multispectral and multitemporal information. This paper presents a new image processing method: Spectral-Temporal Analysis by Response Surface (STARS), which synthesizes the full information content of a multitemporal-multispectral remote sensing image data set to represent the spectral variation over time of features on the Earths surface. Depending on the application, STARS can be effectively implemented using a range of different models [e.g., polynomial trend surface (PTS) and collocation surface (CS)], exploiting data from different sensors, with varying spectral wavebands and acquiring data at irregular time intervals. A case study was used to test STARS, evaluating its potential to characterize sugarcane harvest practices in Brazil, specifically with and without preharvest straw burning. Although the CS model presented sharper and more defined spectral-temporal surfaces, abrupt changes related to the sugarcane harvest event were also well characterized with the PTS model when a suitable degree was set. Orthonormal coefficients were tested for both the PTS and CS models and performed more accurately than regular coefficients when used as input for three evaluated classifiers: instance based, decision tree, and neural network. Results show that STARS holds considerable potential for representing the spectral changes over time of features on the Earths surface, thus becoming an effective image processing method, which is useful not only for classification purposes but also for other applications such as understanding land-cover change. The STARS algorithm can be found at www.dsr.inpe.br/~mello.

Atributos topográficos e dados do Landsat7 no mapeamento digital de solos com uso de redes neurais

The objective of this study was to evaluate discriminant variables in digital soil mapping using artificial neural networks. The topographic attributes elevation, slope, aspect, plan curvature and topographic index, derived from a digital elevation model, and the indexes of clay minerals, iron oxide and normalized difference vegetation, derived from a Landsat7 image, were combined and evaluated for their ability to discriminate soils of an area at the northwest of Rio de Janeiro State. The Java neural simulator and the backpropagation learning algorithm were used. The maps generated by each of the six tested sets of variables were compared with reference points for determining the rating accuracy. This comparison showed that the map produced with the use of all the variables reached a performance (73.81% of agreement) superior to maps produced by other sets of variables. Possible sources of error in the use of this approach are mainly related to the great lithological heterogeneity of the area, which hindered the establishment of a more realistic model of environmental correlation. The approach can help make the soil survey in Brazil faster and less subjective.

Digital soilscape mapping of tropical hillslope areas by neural networks

Geomorphometric variables are applied in digital soil mapping because of their strong correlation with the disposition and distribution of pedological components of the landscapes. In this research, the relationship between environmental components of tropical hillslope areas in the Rio de Janeiro State, Brazil, artificial neural networks (ANN), and maximum likelihood algorithm (MaxLike) were evaluated with the aid of geoprocessing techniques. ANN and MaxLike were applied to soilscape mapping and the results were compared to the original map. The ANN architectures with seven and five neurons in the hidden layer produced the best classifications when using samples obtained systematically. When random samples were applied, the best neural net architectures were within 22 and 16 neurons in the hidden layer. In conclusion, the ANN can contribute to soilscape surveys, making map delineation faster and less expensive. The digital elevation model (DEM) and its derived attributes can contribute to the understanding of the soil-landscape relationship of tropical hillslope areas; the use of artificial neural networks and MaxLike is feasible for digital soilscape mapping. The systematic sampling method provided a global accuracy of 70 % and 65.9 % for the ANN and the MaxLike, respectively. When the random sampling method was applied, the ANN had a global accuracy of 69.6 %, and the MaxLike had an accuracy of 62.1 %, considering the total study area in relation to the reference map.

Comparison between artificial neural networks and maximum likelihood classification in digital soil mapping

O levantamento de solos e a principal fonte de informacao espacial sobre solos para diferentes usos, principalmente o uso agricola. No entanto, a continuidade dessa atividade tem sido grandemente comprometida, principalmente pela escassez de recursos financeiros. O objetivo deste estudo foi avaliar a eficiencia da utilizacao de dois classificadores distintos (redes neurais artificiais - RNAs e o algoritmo da maxima verossimilhanca - Maxver) na predicao de classes de solos em uma area na regiao noroeste do Estado do Rio de Janeiro. As variaveis discriminantes usadas incluem atributos do terreno, como elevacao, declividade, aspecto, plano de curvatura e indice topografico combinado (CTI) e indices clay minerals, iron oxide e de vegetacao NDVI, derivados de uma imagem do sensor ETM+ do LANDSAT 7. Para o treinamento e a validacao dos classificadores, foram utilizadas, respectivamente, 300 e 150 amostras por classe de solo, representativas das caracteristicas dessas classes, com relacao as variaveis discriminantes utilizadas. De acordo com os testes estatisticos realizados, o classificador com base na RNA produziu maior exatidao do que o classificador classico da maxima verossimilhanca (Maxver). A comparacao com 126 pontos de referencia coletados no campo evidenciou que o mapa produzido pela RNA teve desempenho superior (73,81 %) ao mapa produzido pelo algoritmo Maxver (57,94 %). As principais causas de erros detectadas na utilizacao desses classificadores foram: a heterogeneidade geologica da area aliada a problemas no mapa geologico utilizado; profundidade do contato litico e, ou, exposicao da rocha; e problemas com o modelo de correlacao ambiental utilizado em razao da natureza poligenetica dos solos. Os resultados obtidos permitem inferir que a utilizacao de atributos do terreno juntamente com dados de sensoriamento remoto em uma abordagem por RNAs pode contribuir para facilitar o mapeamento de solos no Brasil, principalmente por causa da disponibilidade de dados de sensores remotos a custos mais baixos e da facilidade de obtencao dos atributos do terreno.

Bayesian Networks for Raster Data (BayNeRD): Plausible Reasoning from Observations

This paper describes the basis functioning and implementation of a computer-aided Bayesian Network (BN) method that is able to incorporate experts’ knowledge for the benefit of remote sensing applications and other raster data analyses: Bayesian Network for Raster Data (BayNeRD). Using a case study of soybean mapping in Mato Grosso State, Brazil, BayNeRD was tested to evaluate its capability to support the understanding of a complex phenomenon through plausible reasoning based on data observation. Observations made upon Crop Enhanced Index (CEI) values for the current and previous crop years, soil type, terrain slope, and distance to the nearest road and water body were used to calculate the probability of soybean presence for the entire Mato Grosso State, showing strong adherence to the official data. CEI values were the most influencial variables in the calculated probability of soybean presence, stating the potential of remote sensing as a source of data. Moreover, the overall accuracy of over 91% confirmed the high accuracy of the thematic map derived from the calculated probability values. BayNeRD allows the expert to model the relationship among several observed variables, outputs variable importance information, handles incomplete and disparate forms of data, and offers a basis for plausible reasoning from observations. The BayNeRD algorithm has been implemented in R software and can be found on the internet. \ and

Geomorphometric Attributes Applied to Soil-Landscapes Supervised Classification of Mountainous Tropical Areas in Brazil: A Case Study

The present study aimed to improve the recognition of patterns of soils organization in mountainous tropical landscapes, hence helping soil surveys. The study area is located in the northwest Rio de Janeiro State, with a total area of approximately 16.470,ha. In this concern, geomorphometric features that define the geomorphic signature of the soil-landscape, were used. Geomorphometric features includes: elevation, relative elevation, aspect, curvature, curvature plane, curvature profile, slope, flow direction, flow accumulation and drainage’s Euclidian distance, being all these features obtained by geoprocessing techniques. Almost all attributes were obtained from a digital elevation model and, therefore, the primary elevation data were obtained from the topographic maps. Through these geomorphometric attributes, a geomorphometric signature of the landscape was elaborated, and the particularities of each soil-landscape unit improved the supervised classification. The results showed the feasibility of using geomorphometric attributes to perform a supervised classification, using either neural networks or a maximum likelihood algorithm for soil-landscapes classification of mountainous tropical areas. In addition, we showed that geoprocessing techniques used to extract geomorphometrics attributes can subsidize soil surveys, making soil mapping faster and less biased by subjectivity.

Historic queries in geosensor networks

This paper addresses--to our knowledge, for the first time--the problem of querying a geosensor network--a sensor network of mobile, locationaware nodes--for historical data. It compares different network architectures and querying strategies with respect to their performance in reconstructing events or processes that happened in the past, trying to support the hypothesis that reconstruction is possible within the limited capacities of a geosensor network only. In a concrete case study, these queries are studied in a simulated peer-to-peer ride sharing system.