Detecção de copas de árvores em imagens de alta resolução espacial

Detalhes bibliográficos
Ano de defesa: 2016
Autor(a) principal: Marília Ferreira Gomes
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Tese
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Minas Gerais
Brasil
IGC - DEPARTAMENTO DE GEOGRAFIA
Programa de Pós-Graduação em Geografia
UFMG
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://hdl.handle.net/1843/53720
https://orcid.org/0000-0001-6853-236X
Resumo: The requirements for advanced knowledge on forest resources have led researchers to develop efficient methods to provide detailed information about trees. Traditionally, this information is obtained from forest inventories, which are expensive and time-consuming procedures that do not usually cover large areas. Since 1999, orbital remote sensing has been providing very high resolution (VHR) image data, which allows the identification of individual tree crowns. The increase in spatial resolution has had a profound effect in image processing techniques and has motivated the development of new object-based procedures to extract information. Tree crown detection has become a major area of research in image analysis considering the complex nature of trees in an uncontrolled environment. However, most of the research for automatic detection of individual tree crowns has been developed for temperate forests. Considering this lag, this thesis undertook the task of developing an automatic method to detect individual tree crowns that is adaptable to different environments and capable of delineating their crown as single objects. The approach, named PPM-TM, integrates geometricaloptical modeling, marked point processes (MPP), and template matching (TM) to individually detect and provide a simplified delineation of tree crowns in VHR images. The algorithm is based on the use of an adaptive metric, which allows the detection of trees, even if they do not have high similarity with the 3D model. It also incorporates functions of spectral thresholds and first-order texture to aid the decision process in the inclusion of objects as individual tree crowns. Phases of birth and death are alternated, in which objects are created and destroyed if they do not meet the criteria determined to be recognized as a tree. A series of post-processing refinements was also incorporated including the redefinition of the tree crown diameter and the relocating of the tree crown center. The PPM-TM was tested in different tree outside environments (TOF) as urban environments, orchards and in natural vegetation of savanna (cerrado sensu stricto) and deciduous tropical forest (mata seca). The mean accuracy of detection was 91.9% and the mean accuracy of the delineation was 67.61%, with best results for isolated trees.