RI UFLA (Universidade Federal de Lavras) >
Revistas UFLA >

Por favor, utilize esse identificador para citar este item ou usar como link: http://repositorio.ufla.br/jspui/handle/1/14951

Título: Individual tree identification in airborne laser data by inverse search window
Autor(es): Gorgens, Eric Bastos
Rodriguez, Luiz Carlos Estraviz
Silva, André Gracioso Peres da
Silva, Carlos Alberto
Assunto: LiDAR
Local maximum
Canopy height model
Publicador: Editora da Universidade Federal de Lavras
Publicação: 7-Abr-2016
Referência: GORGENS, E. B. et al. Individual tree identification in airborne laser data by inverse search window. Cerne, Lavras, v. 21, n. 1, 2015.
Resumo: The local maximum filtering performance is highly dependent of the window size definition. This paper proposes that the window size should be determined by an inverse relationship to the canopy height model, and test the hypothesis that a windowsize inversely proportional will have better performance than the window proportional to the canopy height model. The study area is located in the southeastern region of the State of British Columbia, Canada. The natural vegetation is the boreal type and is characterized by the dominance of two species Picea engelmannii Parry ex. Engelmann (Engelmann spruce) and Abies lasiocarpa (Hook.) Nutt. (sub-alpine fir). The relief is mountainous with altitudes ranging from 650-2400 meters. 62 plots with 256 square meters were measured in the field. The airborne LiDAR had discrete returns, 2 points per square meter density and small-footprint. The performance of the search windows was evaluated based on success percentage, absolute average error and also compared to the observed values of the field plots. The local maximum filter underestimated the number of trees per hectare for both window sizing methods. The use of the inverse proportional window size has resulted in superior results, particularly for regions with highest density of trees.
Outras Identificações: http://www.cerne.ufla.br/site/index.php/CERNE/article/view/1043
Idioma: eng
Aparece nas coleções: CERNE

Arquivos neste Item:

Não há arquivos associados para este Item.

Itens protegidos por copyright, com todos os direitos reservados, Salvo indicação em contrário.

Mostrar estatísticas


DSpace Software Copyright © 2002-2007 MIT and Hewlett-Packard - Feedback