The goal of our research is to develop machine vision technology for color based inspection and grading of dried lumber boards. The principle used is to eliminate most of the sound wood regions from further scrutiny at an early stage of processing.
The used color features have a basis in spectrophotometric measurements of the material (left). Also a part of our test material is shown (right). The image acquisition has been performed with a three-chip camera at about 0.5 millimeter resolution. Also a prism based color line-scan camera has been used.
The material has been divided to about 2.5-by-2.5 centimeter regions that have been classified manually to about forty categories.
Examples of the most common defects of wood
A database of knot images from spruce and their classifications is available. The database has been prepared carefully to ensure consistency with the Nordic standards. To gain understanding of the defect classes on wood see the examples and look at the explanations for various knots.
We also offer a part of our extensive set of wood images that have been manually classified region-by-region. Larger amounts of data could be made available for scientific users on request. For instance, we have a huge set of manually classified pictures of up to 6 meter long boards. This exceptionally high quality imagery has been acquired using a prism-based color line-scan camera by VTT Building Technology.
If you use the knot or wood samples for any purpose, please, remember to refer to their origin.
Presentation slides (in Finnish) on visual learning
See also our other research activities.