Date of Graduation


Document Type



The Automated Lumber Processing System (ALPS) is a lumber processing system developed over the last 27 years in an attempt to provide rough lumber mills with an alternative to traditional, low yield, labor intensive techniques. As lumber prices continue to climb (from {dollar}196.00 per thousand board feet in 1955 to well over \\{dollar}1000.00 in 1996), improving the yield from lumber while minimizing cost has become crucial for manufacturers who wish to remain in business. ALPS addresses these problems as a complete package from the initial grading to the final processing of the lumber into cuttings used by the manufacturer in building final products allowing the manufacturer to minimize the cost of the lumber while maximizing the yield. The two main components of ALPS are the hardwood lumber processing module and the hardwood lumber grading module. Development of the lumber processing module progresses from a research-based prototype to a complete industrial automated prototype including automated scanning, image analysis and laser cutting. Development of the grading module proceeds from initial research to development of the only existing grading program suitable for use in an automated grading system including a new cutting packing algorithm, EPoCH. Although a grading system is not completely developed, proof is given showing that inexpensive sensors can be used to provide sufficiently accurate grading as opposed to perfect grading using very expensive sensor systems. This proof is available based on a study of the sensitivity of grading rules to imperfect sensing based on the developed grading program. Results from this research have led to a fully automated, industrial version of the lumber processing system tested in an industrial environment leading to yield increases on the order of 15%-20%. The developed grading program can grade lumber at the rate of ten boards/second with few or no boards misgraded based on user selected criteria. The sensitivity study using a representative sample of approximately two thousand boards shows that the data collected using simple gray-scale video cameras is sufficient to provide much more accurate grading than the average human grader (approximately 5% error).