You are here: Home / Research Topics / Research Highlights / Individual Highlight

Research Highlights

Individual Highlight

Machine Vision Wood Identification of Endangered Tropical Woods

Photo of (A) A representation of a deep neural network: the input to the neural network is an image of the transverse section of wood and the output is a confidence score over the woods on which the model was trained. (B) Conceptualizes the similarity of woods for a wood anatomist, showing which species are most likely to be confused with each other. (C) The results of the predictions of a model – the correct identification is shown in the rows, and the values in the cells are the proportion of all the images classified as the species listed in the columns, with the diagonal showing correct predictions.(A) A representation of a deep neural network: the input to the neural network is an image of the transverse section of wood and the output is a confidence score over the woods on which the model was trained. (B) Conceptualizes the similarity of woods for a wood anatomist, showing which species are most likely to be confused with each other. (C) The results of the predictions of a model – the correct identification is shown in the rows, and the values in the cells are the proportion of all the images classified as the species listed in the columns, with the diagonal showing correct predictions.Snapshot : Field identification of wood (screening) is the first step in a forensic workflow to combat illegal logging. Similarly, industries interested in proactively managing their supply chains to ensure only legally logged material also need access to such field screening expertise. Screenings are currently performed by people trained in wood anatomy, but the demand for trained exerts far outpaces availability. A reliable, consistent, and cost effective field screening method is necessary for effective global scale enforcement of international treaties and national laws governing timber trade and imports.

Principal Investigators(s) :
Wiedenhoeft, Alex C. 
Research Location : Center for Wood Anatomy Research, Forest Products Laboratory, Madison, WI
Research Station : Forest Products Laboratory (FPL)
Year : 2018
Highlight ID : 1505

Summary

The current state-of-the-art for field wood identification to combat illegal logging relies on experienced practitioners using hand lenses, specialized identification keys, atlases of woods, and field manuals. Accumulation of this expertise is time-consuming and access to training is relatively rare compared to the international demand for field wood identification. Researchers have presented highly effective models to discriminate woods based on digital images of the transverse surface of solid wood blocks, which are surfaces and images that can be prepared and captured in the field.   The woods of 10 neotropical species in the family Meliaceae were targeted, including endangered species Swietenia macrophylla, Swietenia mahagoni, Cedrela fissilis, and Cedrela odorata. Researchers developed models to classify the 10 woods at the species and genus levels, with image-level model accuracy ranging from 87.4 to 97.5%, with the strongest performance by the genus-level model. This work represents a strong proof-of-concept and first scholarly demonstration of using machine vision and convolutional neural networks to develop practical models for field screening timber and wood products to combat illegal logging.

Forest Service Partners

External Partners

  • None.
  • Dr. Prabu Ravindran, Mr. Richard Soares both of the University of Wisconsin-Madison Department of Botany.  Dr. Adriana Costa, Visiting Scientist in the Center for Wood Anatomy Research