Land Use and Land Cover (LULC) information is an important data source for modeling environmental variables, so it is essential to develop high quality LULC maps. The hundreds of continuous spectral bands gathered with hyperspectral sensors provide high spectral detail and consequently confirm hyperspectral remote sensing as an appropriate option for many LULC applications. Despite increased spectral detail, issues like high dimensionality, huge volume of data and redundant information, mean that hyperspectral image classification is a complex task. It is therefore essential to develop classification approaches that deals with these issues. Since classification results are directly dependent on the dataset used, it is fundamental to compare and validate the classification approaches in public datasets. With this in mind, aiming to provide a baseline, four classification models in the relatively new hyperspectral HyRANK dataset were evaluated. The classification models were defined with three well-known classification algorithms: Spectral Angle Mapper (SAM), Support Vector Machine (SVM) and Random Forest (RF). A classification model with SAM and another with RF were defined with the 176 surface reflectance bands. A dimensionality reduction with principal component analysis was carried out and a classification model with SVM and another with RF were defined using 14 principal components as features. The results show that SVM and RF algorithms outperformed by far the SAM in terms of accuracy, and that the RF is slightly better than the SVM in this respect. It is also possible to see from the results that the use of principal components as features provided an improvement in the accuracy of the RF and an improvement of 28% in the time spent fitting the classification model.

Clouds are one of the major limitations to crop monitoring using optical satellite images. Despite all efforts to provide decision-makers with high-quality agricultural statistics, there is still a lack of techniques to optimally process satellite image time series in the presence of clouds. In this regard, in this article it was proposed to add a Multi-Layer Perceptron loss function to the pix2pix conditional Generative Adversarial Network (cGAN) objective function. The aim was to enforce the generative model to learn how to deliver synthetic pixels whose values were proxies for the spectral response improving further crop type mapping. Furthermore, it was evaluated the generalization capacity of the generative models in producing pixels with plausible values for images not used in the training. To assess the performance of the proposed approach it was compared real images with synthetic images generated with the proposed approach as well as with the original pix2pix cGAN. The comparative analysis was performed through visual analysis, pixel values analysis, semantic segmentation and similarity metrics. In general, the proposed approach provided slightly better synthetic pixels than the original pix2pix cGAN, removing more noise than the original pix2pix algorithm as well as providing better crop type semantic segmentation; the semantic segmentation of the synthetic image generated with the proposed approach achieved an F1-score of 44.2%, while the real image achieved 44.7%. Regarding the generalization, the models trained utilizing different regions of the same image provided better pixels than models trained using other images in the time series. Besides this, the experiments also showed that the models trained using a pair of images selected every three months along the time series also provided acceptable results on images that do not have cloud-free areas.

The regular acquisition of Earth Observations by remote sensing satellites provides long-term Satellite Image Time Series (SITS). Land surface spectral variability provides the capacity for the assessment of Land Use/Cover Change (LUCC) information through SITS. As the reduction of deforestation rates is a matter of global concern, we selected a test area in the Brazilian Amazon Rainforest to assess LUCC information through long-term SITS. Top of Atmosphere reflectance images acquired from Landsat satellites between 1984 and 2017 were downloaded. A filtering process was carried out through the analysis of cloud and shadow masks. A total of 279 images were used to build a long-term Normalized Difference Vegetation Index (NDVI) SITS for every pixel. Images from across 7 years were used to identify SITS for the classes No Change, Anthropic Change, and Natural Change in order to define reference SITS. The Fast- Dynamic Time Warping (FastDTW) algorithm was used to compute the similarity between the reference SITS and the SITS to be labeled. The K-Nearest Neighbor algorithm was applied to classify the SITS based on the similarity measurements. Two different values of the FastDTW radius parameter were used to build two LUCC maps. The overall accuracies of the LUCC maps were 58.06% and 55.02%, for the radius parameter of one and 20, respectively. It was observed that atmospheric effects, clouds, cloud shadows, smoke, among other noisy agents, can modify the real SITS shape. As a result, the use of raw SITS can lead to a reduction in the accuracy of LUCC maps. Furthermore, high cloud coverage in the Brazilian Amazon Rainforest results in high-frequency irregularity in the SITS, which further reduces the accuracy of the LUCC maps. However, the study showed that the long-term NDVI SITS can describe land cover types and the classes defined despite the constraints mentioned.

This work evaluated the performance of an Artificial Neural Network (ANN) in the temporal generalization of the Land Use and Land Cover (LULC) classes in the surroundings of the Salto Grande reservoir, located in a highly urbanized region of the São Paulo State, Brazil. Landsat-8 OLI (Operational Land Imager) multispectral images acquired in 2015, 2016 and 2017 were submitted to an ANN supervised classification. The ANN was trained with the image acquired in May 2015 to recognize five types of land cover (continental waters, forest, bare soil, agricultural area and urbanized area), using a learning rate of 0.1 and momentum of 0.5. As a classifier, the Multilayer Perceptron (MLP) ANN was used and the training algorithm was the backpropagation. To estimate classifications accuracies, checkpoints were randomly selected, and the error matrix was constructed for each date. The measures used in accuracy assessment were kappa, overall accuracy and the commission and omission errors per class. The results show that the image classification of 2015, the same year as the training data, resulted in a kappa index of 0.96, while the 2016 and 2017 classifications had kappa values of 0.72 and 0.74, respectively. Therefore, the experiments carried out to LULC classification from multitemporal scenes using a single-date trained ANN indicate the ANN's generalization capability and its potential in multitemporal analyzes. In addition, the 2016 classification, however, indicates the need to add non-spectral input data, which allows separate types of coverage of the body of water and landfill to be present when similar spectral responses.