Predição de diâmetros e cálculo de volume de clones de eucalipto: uma abordagem com redes multi layer perceptron e long-short term memory
Saved in:
| Main Author: | |
|---|---|
| Publication Date: | 2019 |
| Format: | Master thesis |
| Language: | por |
| Source: | Repositório Institucional da UFG |
| Download full: | http://repositorio.bc.ufg.br/tede/handle/tede/10005 |
Summary: | Effective management of forest resources is of great importance to the success of a forest enterprise. Obtaining accurate information on planted forests is essential for effective forest activity planning. In this sense, the forest inventory is the procedure used to obtain qualitative and quantitative information from a given region. Through inventory it, is possible, for example, to quantify trees, identify species of a settlement and obtain the total volume to be explored. The total volume is one of the most important elements for the exploration of a given area. Companies use information obtained from forest management inventory to establish the number of trees to be removed without disrupting the natural cycle of forests. For the forest enterprise, it is desirable to obtain the necessary information from a stand without raising costs. Thus, statistical methods provide a way to exploit this information without raising the cost by delivering a near-real result. Several works in the literature apply artificial neural networks in several areas of the forest sector, the results obtained by them have been quite promising for problems of classification and prediction of forest resources. In this context, the present work presents a study on the development of models built through neural networks of different architectures, especially the \ textit {Multi layer Perceptron} and \ textit {Long-Short Term Memory} networks, besides the statistical analysis of the models. For diameter prediction and volume calculation of eucalyptus clones. The results achieved by the models were compared with the values obtained by rigorous cubing and by the Schumacher and Hall model (log). The models built by Long-Short Term Memory networks showed good generalization capacity and were superior for estimating diameters and calculating eucalyptus volume in other sites not available during the training phase. In addition to presenting results quite close to those obtained through rigorous cubing. In general, the results were quite satisfactory concerning the statistical methods present in the literature. |