exit bar

 

 

report logo
 

Special Report
Interdisciplinary Activities and Programs

 April 1998

 

Scanning Lumber To Make the Grade

lumber photoA Virginia Tech computer engineering/forestry science team is developing a low-cost computer vision system for sawmills to evaluate defects in rough lumber and determine how it should be cut for maximum value and minimum waste. Left: Farzad Valad (G) works on the image processing software in the computer engineering laboratory.

Work being done by a team of forestry, software, and computer engineering experts may help the lumber industry realize value gains of up to 30 percent while conserving existing and future timber supplies.

The team is working on a low-cost computer vision system for sawmills that would automatically identify and locate external defects in rough lumber and determine how it should be cut for maximum value and minimum waste. "At present, sawmills determine the edging and trimming of each log manually, which usually results in excessive wood being removed," said Dr. Daniel L. Schmoldt, a research scientist with the USDA Forest Service and a principal investigator on the project. "In extreme cases, some sawmills convert only 55 percent of a log's original volume into lumber."

Rough lumber is the result of the initial log breakdown in a sawmill. Logs are first cut into boards, which often contain residual bark, called wane, plus defects, such as knots and discolorations. The boards then go through edging, which reduces the width of the board, and trimming, which reduces the length. At this point, varying degrees of wane and some defects are removed, leaving the boards more valuable, even though they have a decreased surface area.

Edger and trimmer operators make judgments about the placement of edge and trim saws based on their knowledge of lumber grades and on current lumber prices. "However, money and wood material are lost on almost every board that is processed, because manually, it is not possible to make such critical lumber processing decisions accurately and reliably in only a few seconds," Schmoldt said. Removing too little wood results in a low grade board that has low value despite a large surface measure, and removing too much wood results in a higher grade board, but may lower its surface measure to the point that its value is lower than a larger, lower-grade board.

"To use each board to its potential, all possible edging and trimming solutions must be taken into account," said Professor Lynn Abbott, who leads the project's computer engineering team. "Each potential edging/trimming option for a board generates a dollar value that must be calculated and compared to all other possible options in order to select the optimal case. Manually, it is not possible to make such critical lumber processing decisions accurately and reliably in only a few seconds.

"This is a prime example where computer technology can help," he said.. "An accurately configured vision system can spot defects and calculate the optimal cuts in seconds. The industry has reached the point where the raw material is getting scarce, and the labor is expensive- but the cost of computer technology has become affordable and can provide a solution."

Abbott and Schmoldt are collaborating with Phil Araman, a forest products scientist on a system that would insert a scanner/marker system before the trimmer and edger. Their system would detect defects using a commercially available "smart camera," that has an onboard processor and uses laser light to illuminate the boards. The camera would preprocess the data (thereby greatly condensing it) and send the data to a host computer, eliminating several transfers that would be necessary in older vision systems, and would be much faster. "We're hoping that by using a faster camera, with specific vision and application techniques, we'll be able to develop a process that will be fast enough and cheap enough to be used in most sawmills," Abbott said. "Also, the camera only requires laser light, rather than expensive or difficult-to-control lighting systems. This is an important advantage in the dusty sawmill environment."

The big challenge for the team is developing the software and equipment to analyze rough lumber. A number of systems have been developed in the past decade working on planed lumber, but rough lumber has particular difficulties. "We are working on creating innovative image processing and interpretation software that can handle the 'messiness' of rough lumber and that can integrate multiple images," Schmoldt said.

"Rough lumber is a tough vision problem, and we do not expect to solve all the rough lumber scanning problems in this project," he said. "However, we do expect to be able to provide the level of discrimination necessary to identify many rough lumber defects at real-time rates and thereby improve sawmill edging/trimming operations."

The Bradley Department
of Electrical and Computer Engineering
Virginia Tech


Last Updated, May 10, 1998
Questions or comments about the content: eqb@rightwordonline.com
Technical questions or comments: webmaster@birch.ee.vt.edu
http://www.ece.vt.edu/ecenews/ar98/lumber.html