Search

This paper is concerned with the detection of internal defects in hardwood logs. Because the commercial value of hardwood lumber is directly related to the quantity, type, and location of defects in the wood, sawing strategies are typically chosen in an attempt to minimize the defects in the resulting boards. Traditionally, the sawyer makes sawing decisions by visually examining the exterior of log and then revising the sawing strategy as more and more of the log’s interior is exposed. Significantly better results would be expected if internal defects were known, so that a globally optimum solution could be selected in advance. This paper addresses this problem through the analysis of computed tomography (CT) images of logs, for the purpose of detecting important hardwood defects. In particular, the paper considers defect-dependent postprocessing methods, based on mathematical morphology, that show promising results. © 2001 American Institute of Physics. [ABSTRACT FROM AUTHOR]

This paper describes a prototype scanning system that can automatically identify several important defects on rough hardwood lumber. The scanning system utilizes 3 laser sources and an embedded-processor camera to capture and analyze profile and gray-scale images. The modular approach combines the detection of wane (the curved sides of a board, possibly containing residual bark) with classification of defects. For identifying clear (unblemished) wood, a multilayer perceptron network is used; and for other defects, statistically trained radial-basis-function networks are implemented, followed by a competitive decision scheme. The system is among the first to scan and evaluate lumber in its rough (unplaned) state. © 2001 American Institute of Physics. [ABSTRACT FROM AUTHOR]

The initial breakdown of hardwood logs into lumber produces boards with rough (unplanned) surfaces. Because hardwood lumber value is determined by board size and by the percentage of defect-free area (lumber grade), sawmills remove some wood from the edges and/or ends of a board prior to sale. The overall goal of this research project is to develop a prototype scanning system that can automatically identify important defects (knots, wane, and voids) on rough hardwood lumber and can recommend optimal cuts to achieve maximum lumber value for each board. The first step for inspection system development is to integrate off-the-shelf scanning hardware components with in-house developed software to capture high-quality board images. Board illumination is realized by three, 650 mm lasers that cast lines across the width of a board as it moves longitudinally on a conveyor. A matrix array picture processor camera was selected for image capture. It contains an array of 256×256 photodiodes, where each row is addressable and can be manipulated by an on-chip processor. The camera performs initial processing locally, and transfers results to a host PC. A detailed description of instrumentation, system operation, and scanning capabilities are provided, which generate both profile and gray-scale images. © 2000 American Institute of Physics. [ABSTRACT FROM AUTHOR]