Portal de Periódicos da CAPES

Identification of planted forests destined to bioenergy production through data mining using satellite images.

2015; Volume: 30; Issue: 3 Linguagem: Inglês

2359-6562

Cláudia Nonato, Y. V. de Abreu,

Agricultural and Food Sciences

In this work, we evaluated the accuracy of the classification and identification of areas cultivated with planted forests by satellite images of Landsat 5 TM, through data mining statistics techniques statistics. The pixels were converted to surface reflectance values from the cities of Sao Miguel do Tocantins, Sao Bento do Tocantins, Araguatins, Babaculândia, Darcinopolis and Wanderlandia, located in northern state of Tocantins. Texture attributes were used to identify improvements in the classification results. The data mining techniques used were effective in the accurate identification of planted forests in Landsat 5 satellite images, both by the classification performance and by reducing the amount of information needed to perform identification. The tree models, through the J48 algorithm, achieved success rates above 90% in the identification of species planted among other targets. Thus, the techniques used in this study enabled the development of robust classification models to assist the planning and decision-making on forest plantations in Brazil.