- This article is filed under:
- » View All
Machine Vision Activities in the Wood Industry
by Nello Zuech, Contributing Editor - AIA Posted 03/07/2008
Although 3D-based machine vision technology is only now being widely accepted in other industries, the wood products industry first installed 3D-based systems as far back as 1977. 3D-based systems to optimize the yield out of log or cant have been in use ever since and have generally replaced 2D-based systems in these applications as their use further enhances yield optimization. One finds these 3D systems throughout sawmills and planermills: merchandizing, primary breakdown, curve sawing, edging, trimming.
Some of the earliest 2D-based machine vision capabilities were adapted to the application of automatically reading grade markings applied on the lumber to optimize the grade. As machine vision systems have become more capable with advances in their infrastructure technologies (microprocessors, cameras, lighting and optics), systems have emerged that perform grading on the lumber automatically. This is a very challenging application given the variables in appearance Mother Nature has given us in different species of wood. Nevertheless, there have been some successes.
The combination of machine vision-based grade and yield optimization has lead to value optimization – obtaining the most value out of a piece of wood based on current market pricing for the respective sizes of lumber.
In addition to primary and secondary lumber mills, one also finds machine vision-based graders addressing the requirements of the veneer and plywood industries, as tough as these applications are. Today, grading systems are being embraced that employ color and multispectral analysis, as well as X-Ray and even MRI-based approaches, to identify and classify internal defects.
To gain insights into the latest in machine vision activities in the wood industry, representatives at all known suppliers of such systems were canvassed for input for this article. The following were kind enough to contribute to this article:
- Chris Wells - Product Line Manager, Optimization - Coe Newnes McGehee
- Kristina Swenningsson - Marketing Director - Innovativ Vision AB
- Emmanuel Jolys, - Marketing Director – JVB Vision Inc.
- Rick Massey, – Marketing Manager - Raute
- Todd Buchanan - Business Development - Sicam Systems Corporation
1. What specific generic forest product/lumber industry applications do your 2D- or 3D-based machine vision products address?
[Chris Wells - Coe Newnes McGehee (CNM)] CNM combines laser and machine vision-based technologies to detect, classify and optimally grade natural and manufacturing defects for the purpose of enhancing throughput, recovery, value and processing efficiency. The CNM Optimization product line encompasses optimization systems for all major machine centers in the sawmill and planermill, including merchandizing, primary breakdown, curve sawing, edging, trimming and fully automated grading systems. Furthermore, CNM supplies the plywood sector with optimization of lathes and grading of dry veneer as well as chop and rip saw grade optimization for the valued added industry.
In the sawmill, the Transverse BioScan™ systems currently grade for:
- Geometric properties, wane, skip, hit and miss, hit or miss, etc.
- Knots: center, edger, combination, clusters
- Shake (2 and 3 face)
- Pitch pockets
- Bark pockets
- Sound / unsound knot classification
In the planermill, the Linear High Grader (LHG™) Automated Planermill Grading System currently grades for:
- MSR/MEL strength grading in non-destructive fashion.
- All geometric properties, wane, skip, hit and miss, hit or miss, etc.
- Stain: heart or sap wood (blue stain)
- Defect in stains (i.e. knots, splits, shake, slope, in stained lumber)
- Shake (2 and 3 face)
- Slope of grain
- Rate of growth
- Shallow face wane, or saddle back wane
- Steep edge wane
- Planed bark
- Knots: blond, edge, center, decayed, sound, unsound, bark encased, spike, clusters, multi-face combinations, barrel, shell and MSR edge knot rule.
- Pitch pockets
- Rot with voids (Spruce, Pine, Fir (SPF) only)
- White spec (SPF only)
The BlockPLUS™ Profile and Lathe Charger Control System scans and calculates a highly accurate surface characterization of a block prior to peeling veneer with a lathe. This accurate profile is used by the lathe to optimize spindle centering and carriage positioning. At a resolution of more than 100 data points per square inch, BlockPLUS can accurately model surface features such as knots, holes, dips, catface, butt flare and pistol grip.
The Real-time Rotation System utilizes machine vision to track the actual rotation of the log during primary breakdown from fixed cameras, and through a closed loop feedback system adjusts and controls the turn to ensure the target is reached, reducing or eliminating the traditional turning error in all rotation systems. CNM’s approach of a closed loop system relies on machine vision as the foundation, tracking surface features as they rotate across the fixed array of the cameras.
[Kristina Swenningson – Innovativ Vision] We have systems that address any saw mill or processing of lumber application that requires an analysis in order to cut or sort or otherwise replace manual inspection.
[Emmanuel Jolys – JVB Vision ] JVB Vision designs and develops vision solutions for the product/lumber industry. VAB Solutions is the company integrating our vision products in their applications.
Lumber Grading System
Grade Mark Reader
Bar Code Reader
[Rick Massey – Raute] Our scanning technology addresses the following applications for the veneer, plywood and LVL industries:
1. Scanning green veneer (A) - Using a camera to detect open defects that need to be clipped from a ribbon of peeled veneer prior to automatic stacking. Mostly used in softwood applications.
2. Scanning green veneer (B) - Using a color camera to detect defects, repeating knots and grain characteristics to determine the best places to clip the ribbon of veneer to extract maximum value from each piece. Applies mostly to hardwood applications.
3. Using a camera to automatically grade dry veneer for industry-standard or proprietary defects to determine grade (A,B,C,D, proprietary).
4. Using a camera to determine the roughness of the dry veneer to make adjustments at the lathe.
5. Using a camera to see open defects in veneer to instruct an automatic patching device.
6. Using a camera to see defects on the surfaces of pressed plywood panels to instruct an automatic polypatch repair robot.
[Todd Buchanan – Sicam Systems] Highly reliable wood products that require specific real-time dimension monitoring and online predictive maintenance technologies, specifically:
- Machine center specific lumber size control, thickness and width verification, scanning systems;
- Pre-trimmer multiple stage lumber fence verification scanning systems;
- Lumber length verification scanning systems;
- Log opening face verification scanning systems;
- Bandsaw deflection monitoring sensing systems;
- Integrated machine speed control systems;
- Veneer full ribbon thickness verification scanning systems;
- Panel (plywood, MDF, OSB) thickness verification scanning systems;
- Engineered wood products dimension verification scanning systems.
2. What are the benefits the forest product/lumber industry derives from the use of your machine vision-based products?
[Kristina] Speed, flexibility, even quality, improved yield, savings in labor costs. Production control.
[Emmanuel] The benefits are mainly cost and labor savings and quality improvements. The return on investment is about 10 weeks for a Lumber Grading System.
1. Repeatability - Grading to an accuracy or 95%+ (98% recorded on a consistent basis).
2. Flexibility - The same basic camera platform can be adapted to a number of different applications.
3. Labor savings - Systems are able to perform without operator intervention.
4. Reporting - All grading data is retrievable.
5. Optimized grading - User can run ‘‘what if’‘ scenarios to determine the most beneficial set-up for grading parameters.
6. Speed - High-speed computing capabilities enable high-speed mechanical operation. Human operators unable to match.
[Todd] Primary benefits are: improved machine reliability, uptime, quality and recovery efficiency improvement and reduced raw material waste.
[Chris] The benefits derived are in some cases as unique as the application they are designed for, such as reduced jam-ups at the downstream unscambler on bug kill BioScan Edger applications. However, in general, all implementations of machine vision for visual defect grading impacts one or more of the following areas; value, recovery, throughput and up-time. Optimization is one of few, if not the only, capital investments that impacts so many independent aspects of the operations.
Key benefits of the planermill LHG are:
- Increase grading consistency
- Increase value recovery
- Recovery gains
- Rationalize below grade
- Minimize above grade
- Reduce trim-loss
- Reduce labor costs
- Increase production speeds
- The issue is not so much consistency as the ability of the grader to optimize the solution for value
- Take market price into account in making cutting decisions
Key benefits of the sawmill Trimmer BioScan are:
Maximum allowable split configuration
Reduce drying costs of low grade fiber
Limit blue stain and similar defects from high grade sorts
Sort for candidate stock
Grade separation at the sawmill
Key benefits of the sawmill Edger BioScan are:
- Remanufacture for value
- Rip for value
- Increase value by edging around biological defects not currently considered in optimization decisions
- Increase value recovery by considering grade limiting defects
- Edge for knot free or defect free clears or high-grade candidate stock
Key benefits of the BlockPLUS lathe charger scanning are:
- Increase veneer recovery
- Shorten cycle times
- Minimize spin-outs and dropped blocks
Key benefits of the Real-time Rotation System are:
- Reduce turning error
- Increase machine consistency
- Increase recovery
- Increase value
- Increase throughput speeds
3. Can you briefly describe your machine vision-based solutions for the forest products/lumber industry? What specific types of machine vision hardware (frame grabbers, smart cameras, embedded vision processors, etc.), lighting arrangement, camera details are used? And, why did you select the specific components for your designs?
[Emmanuel] Lumber Grading System: Downstream from the planer is the VAB Solutions lumber grading system that scans every quarter of an inch for board dimensions and wane at a maximum speed of 2000 feet a minute. The grader is a result of collaboration with JVB Vision Inc. The lumber grading system is coupled with an NMI moisture meter which verifies that boards have been dried to the correct moisture content.
Grade Mark Reader: In the sawmill, the operators mark for the species (fir, tamarack or pine) or the grade (select, #3 or economy). They also identify the trims at each end that will eliminate the rot, the splits or any other biological defect. This improves the efficiency of the drying operation and also the quality at the entry of the planer mill.
In the planer mill, the graders stand along the transverse conveyor, in front of the lug loader. There is no grading zone and the boards are moving freely on the conveyor. The graders can easily grab the boards and make a sign on the board surface, to identify a grade and a trim to be done at the near or the far end.
Bar Code Reader: Downstream from the lumber grading system is the VAB Solutions Bar Code Reader that reads back the code associated with the board in order to retrieve the solution proposed by the lumber grading system. This solution is merged with the decision of graders looking for biological defects and a trim order is sent to the fence. The bar code reader is a result of collaboration with JVB Vision Inc.
[Todd] Multiple Piece Profile (MPP various models) Scanner (by 3DM Devices) - This technology employs a series of laser line emitters and cameras. The number of laser/camera units is indicated by the size of the field-of-view of the target scan zone and the accuracy specification. A 12 inch x12 inch field-of-view requires two laser/camera units. The product is scalable by adding additional laser/camera units. The scan rate is 100 per second. The resolution is +/- .002 inch at the optimum standoff.
Laser Range Sensor (model LRS50) Pairs (by Hermary Opto Electronics) – Short Range - This technology employs a single point laser dot emitter and a camera with a range of 5 inches to 65 inches, a scan rate of 1000 per second and a resolution of +/- .002 inches.
Laser Range Sensor (model SE) Pairs (by Scanmeg) – Long Range - This technology employs a single point laser dot emitter and a camera or multiple cameras with a range of 14 foot 5 inches, a scan rate of 100 per second and a resolution of +/- .100 inches.
[Chris] Any optimization system vendor integrating machine vision must address three distinct and fundamental areas in their design:
- Detect defects of interest through contrast, lighting, cell structure, and density.
- Classify, bound, size and locate the defects; eliminating false hits and miss classifications
- Optimize the entire piece for user-defined value at operating speeds, reliably, consistently and with no failures.
CNM’s approach to machine vision is that superior image data allows for simpler, faster software algorithms, providing more efficient defect detection. The fundamental machine vision concept ‘garbage-in, garbage-out’ applies here.
Linear Applications: The specific CNM linear offerings incorporating machine vision are the AddVantage™ Chop and Rip Optimizers, and the LHG Automated Planermill Grading system. At CNM the linear application of machine vision represents advanced implementation of available technologies and image processing approaches within the suite. In the linear approach, CNM can utilize both surface and internal sensor technologies.
The linear solution consists of a modular approach and flexible, scalable processing and software architecture. Fundamentally, the solution consists of four CNM-designed and manufactured vision sensors featuring 5 channels of multi-spectral data, with LED lighting, x-ray source and detector, and four CNM-designed and manufactured high profile laser sensors for geometric properties. Sensor data is transferred with frame grabbers on fiber to PC-based subsystems, and the LAN is a Gigabit Ethernet. The linear system utilizes a trained ‘classifier’, built from the unique signatures of the 7 distinct data streams.
Knots are denser as compared to decay which is less dense than surrounding wood. CNM uses x-ray imaging for density characteristics of the internal fiber to help pinpoint the locations of knots, decay, pitch and perhaps most importantly, pith.
Key in the design and the grading performance of the machine is the use of the scatter and trachied effects in conjunction with x-ray to locate, classify, bound and size the defects.
In addition to this ‘internal’ view of the lumber, the linear applications also utilize a multi-spectral approach consisting of 5 distinct and independent channels of vision images based on proven concepts such as laser scatter, tracheid effect for dot ratio and dot vector as well as the more tradition light wavelength analysis on two channels. These distinct channels are unique methods of evaluating the fiber structure, cell orientation, density and contrast of the areas of interest, as opposed to just different shades of color.
The approach of verifying and comparing the various channels of image data to eliminate false hits, and improve defect classification is termed DataFusion™, and enables the linear system with very high levels of accuracy and thus grading performance.
The DataFusion approach, based on multi-sensor, multi-spectral data streams, is yielding proven results on hardwood and softwood applications of the chop and rip saw for window and door manufactures, as well as fully automated grading solutions on most North American softwood species, currently in operation at up to 3000 feet/minute.
Transverse Applications: The specific CNM transverse offerings incorporating machine vision are the Trimmer and Edger BioScan systems in the sawmill, the Transverse High Grader in the planermill and the Dry Veneer Grader system. The transverse systems in the sawmill are designed specifically to be a simple, cost effective module that can be added to any existing CNM trimmer or edger scanner frame in the field.
The first tier of transverse offerings is the two-sided BioScan system for lumber, typically applied to edgers, trimmers and stud planermill grading systems. Building on the fundamental architecture and PC-based design from the veneer grader, this design has two banks of cameras and wide spectrum LEDs, one on the top and one on the bottom, to measure defect characteristics on the wide face of lumber.
The next tier of transverse offering is for applications requiring visual defect scanning on all four faces of the lumber, such as sawmill trimmers and planermill grading systems. Building once again on the basic PC-based architecture, the system utilizes four banks of eight cameras to measure each face of the board. Halogen illumination is substituted for the LEDs, due to the requirement for scanning the narrow face.
The third tier is the design for the dry veneer grader consisting of wide spectrum LED illumination and top mounted high resolution digital line scan color cameras. Data from RGB channels feed proprietary image processing algorithms to develop a color-enhanced contrast image utilized to classify defects effectively and consistently. Underside lighting for back-illumination shows brightly through splits and star-checked knots for enhanced detection. All processing subsystems are PC-based for supportability and upgradeability. Grade sorting of sheets is controlled through value optimization based on visual grade, size and shape sensing. In veneer applications, RGB cameras are used for visual grading of dry veneer for all predominant defect classes such as knots, stains, bark pockets, pitch pockets, etc.
Rotational Applications: The BlockPLUS uses four proprietary sensors to collect approximately ½ million data points to determine an accurate representation of an 8 foot block for veneer peeling.
PC – Server Configuration
GrabLink Series (Euresys)
<SPAN lang=EN-CA style="FONT-SIZE: 10pt; COLOR: black
There are currently no comments for this article.
Leave a Comment:
All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.