A carregar...
Projeto de investigação
Sem título
Financiador
Autores
Publicações
Improving imbalanced land cover classification with k-means smote
Publication . Fonseca, Joao; Douzas, Georgios; Bacao, Fernando; Information Management Research Center (MagIC) - NOVA Information Management School; NOVA Information Management School (NOVA IMS); MDPI - Multidisciplinary Digital Publishing Institute
Land cover maps are a critical tool to support informed policy development, planning, and resource management decisions. With significant upsides, the automatic production of Land Use/Land Cover maps has been a topic of interest for the remote sensing community for several years, but it is still fraught with technical challenges. One such challenge is the imbalanced nature of most remotely sensed data. The asymmetric class distribution impacts negatively the performance of classifiers and adds a new source of error to the production of these maps. In this paper, we address the imbalanced learning problem, by using K-means and the Synthetic Minority Oversampling Technique (SMOTE) as an improved oversampling algorithm. K-means SMOTE improves the quality of newly created artificial data by addressing both the between-class imbalance, as traditional oversamplers do, but also the within-class imbalance, avoiding the generation of noisy data while effectively overcoming data imbalance. The performance of K-means SMOTE is compared to three popular oversampling methods (Random Oversampling, SMOTE and Borderline-SMOTE) using seven remote sensing benchmark datasets, three classifiers (Logistic Regression, K-Nearest Neighbors and Random Forest Classifier) and three evaluation metrics using a five-fold cross-validation approach with three different initialization seeds. The statistical analysis of the results show that the proposed method consistently outperforms the remaining oversamplers producing higher quality land cover classifications. These results suggest that LULC data can benefit significantly from the use of more sophisticated oversamplers as spectral signatures for the same class can vary according to geographical distribution.
Assessment of the introduction of spatial stratification and manual training in automatic supervised image classification
Publication . Moraes, Daniel; Benevides, Pedro; Costa, Hugo; Moreira, Francisco; Caetano, Mário; NOVA Information Management School (NOVA IMS); Information Management Research Center (MagIC) - NOVA Information Management School
The performance of supervised classification depends on the size and quality of the training data. Multiple studies have used reference datasets to extract training data automatically in an efficient way. However, automatic extraction might be inappropriate for some classes. Furthermore, classes can have distinct spectral characteristics across large areas. Thus, dividing the study area into subregions can be beneficial. This study proposes to assess the impact of the introduction of spatial stratification and manually collected training data on classification performance. Two classifications were conducted with the Random Forest classifier and multi-temporal Sentinel-2 data. The classifications’ performance was evaluated by accuracy metrics and visual inspection of the maps. The results indicate that introducing spatial stratification and manual training yielded a higher overall accuracy (66.7%) when compared to the accuracy of a benchmark classification (60.2%) conducted without stratification and with training data collected exclusively by automatic methods. Visual inspection of the maps also revealed some advantages of the novel approach, namely constraining some land cover classes to be present only within specific strata, which avoids commission errors of the class to spread freely across the map. Most of the classification improvements were observed in subregions with specific landscapes and spectral patterns, although these strata represent a small fraction of the study area, which might have contributed to the small increase in accuracy.
Exploring BFAST to detect forest changes in Portugal
Publication . Costa, Hugo; Giraldo, Anny; Caetano, Mário; NOVA Information Management School (NOVA IMS); Information Management Research Center (MagIC) - NOVA Information Management School
Landsat 8 data and Breaks For Additive Season and Trend (BFAST) were used in a region of central Portugal to detect forest clear-cuts and burnt areas. A total of 79 Landsat 8 images from 2013 to 2019 were downloaded for path/row 204/032, and the NDVI was calculated. The same data processing was done for path/row 203/032 to create a denser time series in the overlapping area, which increased to 124 images. The output of the analysis is a binary map of change (i.e., forest loss) and no-change. A probabilistic accuracy assessment based on random stratified sampling was implemented with 100 random points per stratum. Each point was interpreted as being either "no-change", "clear-cut"or "burnt area"based on reference data. Furthermore, the date of change (if any) was defined. Results show an overall accuracy of 0.85±0.02 for the binary classification with omission and commission errors of class "Change"of 0.30±0.02 and 0.19±0.02. Moreover, it is estimated that 32% of the forested area in path/row 204/032 went through at least one episode of clear-cut or fire in the period analyzed. The time lag between the date of change and detection was about 2.5 months on average, which decreased to 1.5 months in the regions of the denser time series. The results are promising but BFAST is somewhat slow and hence some concerns remain about its efficiency in operation use.
Collaborative Data Transmission in Wireless Sensor Networks
Publication . Berbakov, Lazar; Dimic, Goran; Beko, Marko; Vasiljevic, Jelena; Stojkovic, Zeljko; CTS - Centro de Tecnologia e Sistemas; UNINOVA-Instituto de Desenvolvimento de Novas Tecnologias; Institute of Electrical and Electronics Engineers (IEEE)
Collaborative beamforming (CBF) is a promising technique aimed at improving energy efficiency of communication in wireless sensor networks (WSNs) which has attracted considerable attention in the research community recently. It is based on a fact that beampattern with stable mainlobe can be formed, if multiple sensors synchronize their oscillators and jointly transmit a common message signal. In this paper, we consider application of CBF with one bit of feedback in different communication scenarios and analyze the impact of constraints imposed by simple sensor node hardware, on the resulting signal strength. First, we present a CBF scheme capable of reducing interference levels in the nearby WSN clusters by employing joint feedback from multiple base stations that surround the WSN of interest. Then, we present a collaborative power allocation and sensor selection algorithm, capable of achieving beamforming gains with transmitters that are not able to adjust their oscillators' signal phase. The performance of the algorithms is assessed by means of achieved beamforming gain which is given as a function of algorithm iterations. The presented results, which are based on numerical simulations and mathematical analysis, are compared with the ideal case without constraints and with negligible noise at the Base Station (BS).
Automatic detection of vegetation cover changes in urban-rural interface areas
Publication . Barbosa, Bruno; Rocha, Jorge; Costa, Hugo; Caetano, Mário; NOVA Information Management School (NOVA IMS); Information Management Research Center (MagIC) - NOVA Information Management School; Elsevier Science Publisher B.V.
The present work started from the need to streamline the process of monitoring changes in vegetation in the in urban-rural interface fuel management bands, defined by Portuguese legislation as areas where the existing biomass must be totally or partially removed. The model developed uses a time series of Sentinel 2 satellite images to search for changes in the vegetation cover in a 100 meters buffer around built-up areas. The use of satellite data allows analysing large areas and speeds up the task of identifying the places where fuel management took place and the places where there is a need to carry out such management. The objective of the proposed method is to give a script in Python language that can verify the cleanliness of vegetation in the fuel management ranges through multi-temporal analysis of satellite images.
Unidades organizacionais
Descrição
Palavras-chave
Contribuidores
Financiadores
Entidade financiadora
Fundação para a Ciência e a Tecnologia
Programa de financiamento
3599-PPCDT
Número da atribuição
PCIF/SSI/0102/2017