Forest botanical identification becomes more complex when only the wood is
available to define the species, as this subjective and difficult process requires a
more in-depth analysis of its characteristics by specialists, who are increasingly
scarce and not always available depending on the situation, where the result can
take weeks to be issued. The aim of this work is therefore to use infrared
spectroscopy in an automatic identification process as a guarantee of great accuracy
and speed in recognizing wood species. A portable spectrometer will be used to
capture infrared spectra (NIR - Near Infrared Reflectance), with reflectance values
within 740-1070 nm from a sample of 10 Amazonian species with 100 specimens of
dry wood from various sources at 12% humidity each. The spectrum will be collected
in two cutting planes with 2000 repetitions per species, resulting in 20,000 spectra in
the database. The project began with the collection of spectra, setting up an
extensive database, followed by the verification phase, testing and resulting in
instruments for the creation of an automatic recognition system. The species were
selected according to the criteria of local commercial importance and defined by
commercial groups. It was found that of the 10 species collected there were some
correct identifications in the species called Jatobá, which validates the importance of
developing this project. These collections will serve as a local collection and, with the
creation of a model for identifying each species according to its characteristics
informed by the spectra, it will be possible to develop a technology for automatic
forest identification.
|
