SILO



The weekly SILO Seminar Series is made possible through the generous support of the 3M Company and its Advanced Technology Group

3M

with additional support from the Analytics Group of the Northwestern Mutual Life Insurance Company

Northwestern Mutual

The XyloTron: A machine vision-based wood identification device

John Hermanson,

Date and Time: Jun 02, 2016 ( 4:00 PM)
Location: Orchard room (3280) at the Wisconsin Institute for Discovery Building

Abstract:

The worldwide value of illegal logging is 30-100 billion USD annually and within a number of key producer countries an estimated 50-90% of the forestry is illegal. Approximately 80% of the worldwide illegal logging is controlled by organized crime. Rapid, accurate, and inexpensive wood identification is an important factor in keeping illegal wood and wood products out of the supply chain, and thus combating illegal logging. U.S. enforcement of the Lacey Act and the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) have greatly increased the demand for wood identification expertise. The traditional human-mediated wood identification, based upon pattern recognition of anatomical features, has proven too slow and costly to meet the demands of enforcement. Therefore, a netbook based machine vision system (XyloTron) that can quickly identify wood based upon an image of the transverse section was developed. The XyloTron computes the 2D Haar wavelet coefficients of the image of a cross section of wood. The coefficients are squared and summed at each scale to create input parameters for classification. This approach worked well when the breadth of the image collection consisted of 55 genera and 160 genus-species combinations. The accuracy of the system was equal to or better than the average human with a week of training, ~60% at the genus-level. Currently, the image collection has significantly expanded to include 22000+ images of 62 families 190 genera and 898 genus-species combinations. Although accuracy has decreased as a result, research is ongoing to identify additional parameters for a more accurate classification.