Surface shape analysis of rough lumber for wane detection

The initial breakdown of hardwood logs into lumber produces boards with rough surfaces. These boards contain wane (missing wood that emanates from the log exterior, often containing residual bark) that is removed by edge and trim cuts prior to sale. Because hardwood lumber value is determined based on board size and quality, knowledge of wane position and defects is essential for selecting cuts that maximize profit. We have developed a system that uses structured light to obtain profile (thickness) images of unplaned boards, in addition to gray-scale images for defect detection. The focus of this paper is to describe a new approach for detecting wane boundaries through the analysis of these profile images. The problem is difficult because bark and other debris adversely affect the laser-based imaging process, and because variations in surface reflectance also cause inaccuracies in the measured thickness values. The problem is compounded by the need to perform wane detection rapidly in a manufacturing environment. The method that we have developed relies on a combination of column-wise image statistics, selective smoothing, and the analysis of surface shape. Initial wane edge estimates that are obtained using the smoothed image are then refined by analysis of the original image data. The paper provides a quantitative evaluation that indicates a dramatic improvement over traditional thresholding techniques.