Urban environments represent one of the most challenging areas for remote sensing analyses due to the great diversity of land cover materials found in their surface. The fusion of high spatial and high spectral resolution images arise as an alternative for urban applications, for the combination of these two characteristics allows better detection and discrimination of urban targets. The present work has a twofold objective: i) evaluate two datasets for the fine classification of urban targets at two levels of legend (with 11 and 38 land cover classes): one of them exclusively consisting of an orbital multispectral image (WV-2) and another one exclusively comprising an airborne hyperspectral image (SpecTIR), ii) as well as assess the performance of two different image classification methods, C4.5 Decision Tree and Random Forest, for both levels of legend. Eight classification experiments were executed to meet such objectives of investigating the efficacy of sensors and classification methods for the concerned two levels of detailing. The obtained classifications attained high accuracy. For all adopted levels of legend and methods, it was demonstrated that the classifications using SpecTIR data presented results significantly superior to those obtained with the WV-2 data
Ambientes urbanos representam uma das áreas mais desafiadoras do sensoriamento remoto devido à grande diversidade encontrada nos materiais presentes na sua superfície. O uso de imagens com alta resolução espacial e alta resolução espectral surge como uma alternativa para aplicações urbanas, pois a combinação destas duas características permite uma melhor detecção e discriminação de alvos. O presente trabalho tem um duplo objetivo: i) avaliar dois conjuntos de dados na classificação fina de alvos urbanos para dois níveis de legenda (com 11 e 38 classes de cobertura do solo): um deles composto exclusivamente por uma imagem orbital multiespectral (WV-2) e o outro conjunto composto exclusivamente por uma imagem aerotransportada hiperespectral (SpecTIR), ii) bem como testar o desempenho de dois métodos diferentes de classificação de imagens, Árvore de Decisão C4.5 e Floresta Randômica (Random Forest), para ambos os níveis de legenda. Oito experimentos de classificação foram realizados para atender a tais objetivos de investigar a eficácia dos sensores e dos métodos em dois níveis de detalhamento. Foram obtidas classificações de elevada acurácia. Demonstrou-se para todos os níveis de detalhamento e métodos que as classificações obtidas com dados do sensor SpecTIR apresentaram resultados significantemente superiores aos das classificações com dados do sensor WV-2.
The aim of data mining is to find useful knowledge inout of databases. In order to extract such knowledge, several methods can be used, among them machine learning (ML) algorithms. In this work we focus on ML algorithms that express the extracted knowledge in a symbolic form, such as rules. This representation may allow us to ''explain'' the data. Rule learning algorithms are mainly designed to induce classification rules that can predict new cases with high accuracy. However, these sorts of rules generally express common sense knowledge, resulting in many interesting and useful rules not being discovered. Furthermore, the domain independent biases, especially those related to the language used to express the induced knowledge, could induce rules that are difficult to understand. Exceptions might be used in order to overcome these drawbacks. Exceptions are defined as rules that contradict common believebeliefs. This kind of rules can play an important role in the process of understanding the underlying data as well as in making critical decisions. By contradicting the user's common beliefves, exceptions are bound to be interesting. This work proposes a method to find exceptions. In order to illustrate the potential of our approach, we apply the method in a real world data set to discover rules and exceptions in the HIV virus protein cleavage process. A good understanding of the process that generates this data plays an important role oin the research of cleavage inhibitors. We consider believe that the proposed approach may help the domain expert to further understand this process.