Fechar

@InProceedings{SautterRosa:2017:ImGrPa,
               author = "Sautter, Rubens Andreas and Rosa, Reinaldo Roberto",
          affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto 
                         Nacional de Pesquisas Espaciais (INPE)}",
                title = "Improvement to Gradient Pattern Analysis and applications",
            booktitle = "Anais...",
                 year = "2017",
         organization = "Workshop dos Cursos de Computa{\c{c}}{\~a}o Aplicada do INPE, 
                         17. (WORCAP)",
             keywords = "Sistemas Din{\^a}micos, Morfologia de gal{\'a}xias, GPU.",
             abstract = "The description of complex patterns, especially spatial patterns, 
                         is a challenge. Some features, such as classification features(in 
                         static cases) and regime (in dynamical systems), are difficult to 
                         detect. In this work we present the Gradient Pattern Analysis 
                         (GPA), an operator that describes spatially extended system by 
                         means of vectorial alignment and vectorial symmetry. This operator 
                         showed interesting results dynamical systems [1] and astronomical 
                         problems[2, 3]. We briefly review this operator and propose some 
                         improvements, with respect to the measured feature and the 
                         computation cost. To improve the feature extraction, we introduce 
                         an operator which analyses the second order gradient moment. To 
                         reduce the computation time, a version of this code is presented 
                         using OpenCL. In this work we also briefly showcase two 
                         applications to this operator: Coupled Map Lattices (CML) and 
                         galaxy morphological classification. As result we observed better 
                         results with the improved operator and a lower computational cost 
                         in the parallel version.",
  conference-location = "S{\~a}o Jos{\'e} dos Campos, SP",
      conference-year = "20-22 nov. 2017",
             language = "en",
           targetfile = "Sautter_improvement.pdf",
        urlaccessdate = "15 jun. 2024"
}


Fechar