Use of computed tomography to detect internal defects in hardwood logs with applications in sawmilling

This dissertation presents characterization of hardwood log CT images, the detectability of various defects from the CT images, algorithms for detecting knots, cracks and piths, algorithms for virtually cutting a log into lumber and veneer, and a solution for the best sawing pattern given a true 3-d log, internal defects and the required lumber sizes. Although many studies claimed that CT imaging can reveal log internal defects, no one has provided how effective CT imaging is for identifying specific defects. A total of 2557 CT images, 256 wood disk optical images, and 76 board optical images was collected from four typical hardwood species covering the whole spectrum of hardwood structures. The defect matching tests show that knots (except for small pin knots), pith, holes, cracks, and advanced decays are clearly revealed in CT-based images. Bark pocket and mineral streak are difficult to recognize in CT cross-sectional images of all hardwood species involved while they may barely be visible in CT synthesized board images. Species and wood moisture content have significant effects on the overall CT image features. Some of the characteristic features derived from the match tests are used in an image analysis system which is to automatically detect knots, cracks and pith. The segmentation process of the system extracts regions from each CT image using a multiple thresholding technique that exploits both the prior knowledge of wood structure and gray level characteristics of the CT image, then retains or discards each region using 2-d geometric features such as region area, shape, and orientation. The 3-d analyzer of the system classifies a retained region as a defect or non-defect via correlation analysis across corresponding regions in neighboring CT images. The graphics system developed furnishes a mathematical basis for cutting a log into lumber or veneer and algorithms for modeling logs, defects, lumber and veneer using a set of log CT images. The first solution of the optimal cutting strategy for live sawing method using information from CT images includes lumber value modeling, the formulation of the optima cutting strategy problem, and the algorithm to solve the problem.