Loading...
Please wait, while we are loading the content...
ESTIMATIVA DA ALTURA DE ÁRVORES DE Tectona grandis L.f. UTILIZANDO REGRESSÃO E REDES NEURAIS ARTIFICIAIS
| Content Provider | Semantic Scholar |
|---|---|
| Author | Vendruscolo, Diogo Guido Streck Chaves, Arthur Guilherme Schirmbeck Medeiros, Reginaldo Antonio Silva, Raiane Scandiane Da Souza, Hudson Santos Drescher, Ronaldo Leite, Helio Garcia |
| Copyright Year | 2017 |
| Abstract | O objetivo deste estudo foi avaliar a modelagem por regressao e por redes neurais artificiais na estimativa da altura total de arvores de teca em diferentes espacamentos em Caceres, MT. A base de dados foi proveniente da medicao do dap por meio de censo florestal. Posteriormente, estes foram agrupados em classes de diâmetro, com amplitude de 5 cm. Foi medida a altura total ( h ) de 20% dos individuos em cada espacamento e classe de diâmetro. Para estimativa da altura total por regressao foram utilizados modelos nao lineares e lineares, enquanto que para a estimativa por redes neurais artificiais foram testadas redes do tipo Multilayer Perceptron . Na modelagem por regressao, os modelos nao lineares foram superiores aos lineares, com destaque para o modelo de Gompertz. As duas tecnicas foram eficientes na estimativa da altura total de Tectona grandis , no entanto, a modelagem por redes neurais artificiais apresentou erro inferior a 10% em todos os espacamentos analisados. Palavras-chave: Teca, identidade de modelos, inteligencia artificial. HEIGHT ESTIMATIVE OF Tectona grandis L. f. TREES USING REGRESSION AND ARTIFICIAL NEURAL NETWORKS ABSTRACT The aim of this study was to evaluate the modeling regression and artificial neural networks to estimate the total of teak trees in different spacings in Caceres, MT, Brazil. The database was derived from the measurement of dbh by forest census. Subsequently, these were grouped into classes of diameter, with an amplitude of 5 cm. Was measured the overall height ( h ) of 20% of subjects in each spacing, and diameter class. To estimate the total height per regression were used nonlinear and linear models, while for the estimated by artificial neural networks of the type were tested Multilayer Perceptron. In regression modeling, the non-linear models were superior to linear, highlighting the Gompertz model. Both techniques were efficient for estimates the total teak height, however, through the modeling for artificial neural network, the error values were lower than 10% in all evaluated spacings. Keywords: Teak, identity models, artificial intelligence. DOI: http://dx.doi.org/10.5935/2318-7670.v05n01a09 |
| Starting Page | 52 |
| Ending Page | 58 |
| Page Count | 7 |
| File Format | PDF HTM / HTML |
| DOI | 10.5935/2318-7670.v05n01a09 |
| Volume Number | 5 |
| Alternate Webpage(s) | http://www.bibliotekevirtual.org/revistas/NATIVA/v05n01/v05n01a09.pdf |
| Alternate Webpage(s) | https://doi.org/10.5935/2318-7670.v05n01a09 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |