The Acre unit of the Brazilian Agricultural Research Corporation (Embrapa) announced the creation of Netflora, a methodology that brings together a set of algorithms trained with artificial intelligence (AI) to recognize forest species.
Carried out based on botanical characteristics, available in a database, this learning allows you to identify trees of commercial interest and indicate their exact location in the forest. Species such as chestnut, cumaru-ferro, açaí and cedar are recognized with accuracy rates of 95%, a result that reduces production costs and makes forest management in the Amazon more sustainable.
According to Embrapa Acre researcher Evandro Orfanó, one of the coordinators of these studies, Netflora brings greater automation to the planning of forestry activities and increases the precision and efficiency in the execution of management plans.
“Once trained and specialized, the algorithm also provides metrics, such as diameter and crown area, which make it possible to estimate, through allometric equations (which relate shapes and sizes), the volume of wood in each tree.
These technological tools contribute to increasing forestry production with environmental conservation”, he states.
Research to enable the use of artificial intelligence in the forestry sector has been developed by Embrapa since 2015 and covers different aspects of the activity. In the current phase, studies take place through the project Geotechnologies applied to forestry automation and spatialization of carbon stocks in native and modified land use in the Western Amazon (Geoflora), carried out in Acre, Rondônia, Roraima, Amapá, Pará and Amazonas, in partnership with the JBS Fund for the Amazon.
The adoption of these technologies implies investments in computers, drones, batteries and adequate office structure. According to Orfanó, this initial expense is offset by the drastic reduction in production costs, especially in the forest inventory stage. To give you an idea, in traditional species surveys, with teams in the field, a hectare of mapped forest has an estimated cost of between R$100 and R$140, while with the Netflora methodology this value drops to R$4 to R$6 .
He emphasizes that this reduction is provided by the agility in obtaining and processing information about the area to be managed. “A forestry company that uses traditional management can map up to 10 thousand hectares of forest per year. With the use of AI, the gain in operational capacity can jump to up to one million hectares in the same period,” he adds.
To build the algorithm training database, more than 40 thousand hectares of forest were mapped, in 37 sites (areas) in Acre, Rondônia and southern Amazonas, using drones. In two years of study, around a thousand flight plans were created and each one generated approximately 300 aerial images, which were processed and transformed into orthophotos (high-resolution, georeferenced images). Based on the range of information contained in the orthophotos, nine algorithms were trained, with different purposes and accuracy performances.
“We have algorithms that recognize a single forest species, others have the capacity to identify different groups or the main timber and non-timber trees in Acre and other locations in the Amazon. Some algorithms have already achieved high performance, but this learning will be continuous”, highlights Orfanó, who estimates the project’s mapping goal at 80 thousand hectares of forest, with the insertion of new areas of commercial interest in the Amazon, to expand the construction of the database. data.
- By Edmilson Ferreira, from AC24horas.
Tags: Created Acre methodology identifies forest species commercial
--
