Emprego de segmentação multiresolucional no mapeamento digital de formas de relevo
Resumo
Resumo: A abordagem geomorfométrica tem sido cada vez mais utilizada no mapeamento de formas de relevo. Dentre a infinidade de técnicas que foram e são desenvolvidas e aplicadas neste campo, aquelas baseadas na análise orientada ao objeto e que mais precisamente empregam o algoritmo de segmentação multiresolucional veem apresentando importantes potencialidades. Assim, o objetivo desta pesquisa foi desenvolver e avaliar uma técnica semiautomatizada baseada no emprego de segmentação multiresolucional para o mapeamento de padrões de formas de relevo compatíveis ao 4o táxon da proposta de Ross (1992), na área da Carta Curitiba (leste paranaense). Para isto, a partir de um modelo digital de terreno (MDT) de resolução espacial de 20 metros, calculou-se os seguintes atributos topográficos: declividade e relevo sombreado (locais), e índice de posição topográfico (IPT), amplitude altimétrica, média da declividade, média da densidade de rupturas (regionais). Em seguida, foram identificados os padrões de formas de relevo da área de estudo, baseada no mapeamento prévio realizado no âmbito do projeto de mapeamento geomorfológico do Paraná na escala 1:100.000. Os atributos foram padronizados e em seguida segmentados em seis níveis, gerando assim objetos de imagem que foram classificados segundo regras adaptadas a partir do mapeamento prévio anteriormente citado. No primeiro nível de segmentação foram empregados o IPT e a declividade, mapeando as planícies fluviais. Nos outros cinco níveis, foram utilizados os demais atributos topográficos, sendo os atributos regionais calculados a partir de raios de vizinhança distintos para cada nível. Também foram utilizados parâmetros de segmentação distintos em cada nível, sendo adequados às respectivas formas de relevo mapeadas. O 2° e o 3° níveis classificaram as áreas segundo o grau de dissecação (baixa, média e alta), enquanto os demais níveis mapearam os padrões de formas de relevo pertencente a cada um dos graus de dissecação. Foram mapeados 12 padrões de formas de relevo: 1) planícies fluviais, 2) colinas, 3) colinas onduladas, 4) morrotes I, 5) morrotes II, 6) morros I, 7) morros II, 8) morros declivosos I, 9) morros declivosos II, 10) morros declivosos III, 11) morros elevados I e 12) morros elevados II. Os resultados apresentaram coerência quando conferidos em campo. O método apresentou potencialidades, visto que reduziu o efeito de borda e apresentou formas de relevo com tamanhos adequados à escala pretendida no mapeamento. Entretanto, limitações associadas a erros na segmentação ou classificação de algumas formas de relevo ocorreram, mas sem inviabilizar a aplicabilidade do método. Por fim, evidencia-se a viabilidade do emprego da segmentação multiresolucional no mapeamento do 4° táxon, com este algoritmo possuindo potencial para ser explorado no mapeamento em outras escalas. Porém, é necessário atenção na definição dos parâmetros de segmentação, principalmente o parâmetro de escala, pois apresentam subjetividade. Palavras-chave: Cartografia geomorfológica. Taxonomia do relevo. Geomorfometria. Análise digital do relevo. Análise orientada ao objeto. Abstract: The geomorphometric approach has been increasingly used in the mapping of landforms. Among the infinity of techniques that have been and are developed and applied in this field, those based on object-oriented analysis and that more precisely uses the multiresolutional segmentation algorithm have shown important potentialities. Thus, the aim of this research was to develop and evaluate a semiautomated technique based on the use of multiresolutional segmentation for mapping patterns of relief forms compatible with the 4th taxon of Ross's proposal (1992), in the area of Carta Curitiba (eastern Paraná). For this, using a digital terrain model (DTM) with a spatial resolution of 20 meters, the following topographic position index (TPI), altimetric range, slope average, edge density average (regional). Then, the patterns of landforms in the study area were identified, based on the previous mapping carried out within the scope of Paraná of the Paraná geomorphological mapping project on a 1:100.000 scale. The attributes were standardized and then segmented into six levels, thus generating image objects that were classified according to rules adapted from the previous mapping mentioned. In the first segmentation level, TPI and slope average were used, mapping the river plains. In the other levels, the other topographic attributes were used, with the regional attributes calculated from different neighborhood rays for each level. Different segmentation parameters were also used at each level, being appropriate to the respective mapped landforms. The 2nd and 3rd levels classified the areas according to the degree of dissection (low, medium and high), while the other levels mapped the patters of landforms belonging to each of the dissection degrees. 12 patterns of landforms were mapped: 1) river plains, 2) low hills, 3) undulated low hills, 4) hillocks I, 5) hillocks II, 6) hills I, 7) hills II, 8) sloping hills I, 9) sloping hills II, 10) sloping hills III, 11) high hills I and 12) high hills II. The results were consistent when validated in the field campaign. The method showed potential, since it reduced the edge effect and presented relief shapes with sizes appropriate to scale intended in the mapping. However, limitations associated with errors in the segmentation or classification of some landforms have occurred, but without impairing the applicability of the method. Finally, the viability of using multiresolutional segmentation in the mapping of the 4th taxon is evidenced, with this algorithm having the potential to be explored in mapping at the other scales. However, attention is needed in the definition of segmentation parameters, especially the scale parameter, as they present subjectivity. Keywords: Geomorphological mapping. Relief taxonomy. Geomorphometry. Digital relief analysis. Object-oriented analysis
Collections
- Dissertações [111]