Computed Tomography and Fingerprint Traceability in the Wood Industry

University dissertation from Luleå tekniska universitet

Abstract: The purpose of the work described in this thesis was to develop techniques based on non-invasive measurements of logs and sawn timber that would increase the profitability of the wood industry in general and sawmills in particular. The work has two main focus areas: computed tomography (CT) scanning of logs and traceability of wood products.  The first focus was on detecting knots in CT images of logs and to find ways to use the knot information efficiently. The result is an automated algorithm that can successfully detect knots in CT images of Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies (L.) Karst.). Knots have a negative impact on the bending strength of sawn timber and, since knots can be detected in CT images of logs, it is possible to adapt the sawing process to take into consideration where the knots are located. This thesis includes an investigation of the profitability gain for a sawmill producing strength-graded sawn timber of Norway spruce when detailed knot information from a CT scanner is used. The strategy was to optimize the log breakdown by rotating logs to an optimum position with respect to the sales value of the sawn timber. The investigation was carried out using computer simulations.  The work also includes an investigation into how accurately the bending strength of sawn timber can be predicted from information in CT images of the logs. Features and defects in the CT images were measured and key parameters were inserted into multivariate PLS regression models. The models were calibrated with data from destructive bending tests and, although the results were unclear, there were indications that the bending strength could be predicted with a higher accuracy using CT scanning than by using log scanning techniques that are currently common in the industry.  The second focus area was fingerprint traceability of individual wood products, which is valuable for sawmills since it enables detailed process control. Diagnostics and process surveillance could be based on statistics for each individual piece of sawn timber instead of on statistics at a batch level. Without an automatic recognition system for sawn timber, such studies would involve labor-intensive and possibly process-disruptive manual tests.  The work includes the development of two wood surface recognition systems based on different techniques. One of them uses information about how knots are positioned in relation to each other to construct scale- and rotationally-invariant descriptors. The performance and robustness of this recognition system were tested on 212 edge-glued panel images of Scots pine with different noise levels. The other recognition system was specialized on sawn timber. This particular method considerably reduces the resolution of the board images and matches them using template matching. Tests were performed by matching three sets of 88 board images to a database of 886 Scots pine boards. The recognition systems have different strengths and weaknesses due to their design, but both of them were fast and had high recognition rates in the tests carried out.  Overall, the work led to several computerized methods that enable increases in profitability of sawmills. The proposed knot detection in CT images of logs enables detailed control of the log breakdown process, and the proposed fingerprint traceability methods permit process control based on individual pieces of sawn timber. Results from this thesis also give sawmill managers valuable information on how an industrial CT scanner would affect the profitability of their sawmills.

  This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.