GB2285861A - Optical detection of grain defects in lumber - Google Patents

Abstract

The detection of grain defects in lumber 2 is accomplished by a pair of plane polarised light detectors 5, 6 viewing a detection point on the surface of the lumber illuminated by a collimated, plane polarised, beam of light 1. The detectors are disposed to view the inspection point coaxially with the incident beam, thus receiving light retroreflected from the surface. The plane of polarisation of the incident beam is set to make an angle alpha with the surface component of the nominal grain direction vector of the lumber, the detector planes of polarisation are symmetrically disposed with respect to the illuminating beam polarisation, the one being set at - alpha with respect to the beam polarisation, the other being set at + alpha . The relative levels of light received by the detectors differs substantially only when normal straight grain wood is present at the point of incidence. Grain defects are discriminated by analysing the calculated ratio of the detector signals. The beam may be scanned across the lumber as it is translated transversely to the scan direction. <IMAGE>

Description

1. Title of invention.

Optical Method and apparatus for the detection of grain defects in lumber.

2. Purpose of invention.

To facilitate the automation of lumber processing operations via the automatic recognition, delineation, and location of certain natural occurring defects in lumber.

3. Background to purpose.

Definitions: LUMBER comprises planks processed from timber logs (tree stems ) by sawing. Prior to eventual use planks may undergo further surface processing for example planing or sanding.

The final value and application of lumber is limited by the presence of defects, both naturally occurring and man made.

Between the rough sawn board and final lumber product, the material passes through a number of processing steps aimed toward maximising the the market value of the final product. At each of these steps the, the defect structure is assessed to determine sawing widths and lengths of various qualities that will yield the highest value in the prevailing market conditions.

The final product is generally quality graded according to the nature and extent of defects remaining in the board, higher qualities generally carrying a higher market value. This quality grading may be against national, regional, or end user standard rules and guidelines. Historically, these assessments and grading tasks have been the exclusive domain of human operators. In more recent times, dimensional defects (departures from the ideal of straight board of rectangular section and specific dimensions) have been detected and evaluated by automatic means. However, relatively little progress has been made toward economically automating the assessment of naturally occurring biological defects (eg. knots, grain distortion, decay etc).

Much effort has been directed toward this goal, and some inroads made, but by and large this assessment remains in the domain of the human operative.

The present invention is intended for application in the automation of this area of defect assessment.

4. Background to the invention.

4.1. Introduction.

At present, for the purpose of assessment, biological defects are described according to their visual manifestations. This description is well suited to application by human operatives. The human eye/brain complex has evolved as a superb machine for analysing visual images and extracting specific information. Many of the approaches applied toward automating the assessment aim toward emulating the human operative processes via complex cameralcomputer machine structures.

This emulation task is a complex one, and even if fully achieved would result in a system performance similar to the human one. While humans can recognise defects with some facility, they show considerable ambiguity in accurately delineating the true extent of defects, which leads to an inefficiency in their assessment. Visual emulation systems are thus ultimately limited in scope.

We believe this visual description of biological defects to be inappropriate to automation. Knots, for example, show a wide variation in visual appearance even within a single wood subspecies, there is no simple visual archetype for the defect. When the scope of the assessment is increased to embrace many species, and many sawing patterns, the diversity in the visual appearance of common defects increases exponentially. However if we look to the underlying physical nature of the defects the situation is much improved. The physical properties exhibit a much smaller variation in parameters than the incidental visual manifestation. In the physical domain, simple archetypes can be found that may be better suited to automation systems.

This thesis can be illustrated by the example of knots. Knots in the surface of lumber arise from the surface intersecting limb growths contained within the tree. Limbs grow radially from the tree stem. When the stem is reduced to boards by sawing processes, limb growth sections become visible at the board surface, as knots of various types. A surface cut orthogonal to the limb growth axis gives rise to face knots of a roughly circular or elliptical shape. A cut parallel to the limb growth axis gives rise to spike knots of a roughly triangular shape. Depending upon the cut and limb growth dispositions a wide range of intermediate shapes can occur within a single board of a single species. In addition, the visual tonal qualities show a very wide range of variation, depending upon species, growing ccnditions, climactic variation etc.

Across many boards and many species the variety in visual appearance can be very wide indeed.

Now consider the physical domain.

FIGURE 1. refers.

The material structure of wood largely comprises elongated cells, it is this structural feature that gives rise to the grain characteristic of wood, the grain direction axis corresponding to the long axis of the cells.

A tree grows via a continuous deposition of these cells from the cambium layer (or growth layer) 2 situated just under the bark layer 1. The variation in growth (cell deposition) rate varies with seasons and prevailing climactic conditions giving rise to the familiar growth ring pattern 5 seen in stem 3 and limb 4 sections. As noted above limbs grow radially from the tree stem and via the cell deposition growth mechanism the limb cell longitudenal axis exhibits a large deviation angle 6 with respect to the stem cell axis. Note that this large deviation in limb cell axis is fundamental characteristic of limb growth and is thus present in the surface manifestation 7 (knot) independent of the sawing plane 8. In the physical domain then, knots can be concisely defined as surface regions that exhibit a large deviation in grain angle with respect to the stem.This description, unlike the visual description noted above, remains valid independent of species and sawing patterns. In the vast majority of sawing patterns stem saw planes are substantially parallel to the stem axis, so this knot description can be further simplified. Knots can be described as areas showing a large deviation in grain direction with respect to the board axis. In the physical domain then, a simple archetype for knots exists.

This concise description of knots is well suited to automatic recognition and delineation, always provided that the physical information can be gained by automatic detection means.

4.2. Allied techniques.

Over the past decade or so a number of techniques for the optical retrieval of physical information from wood have been proposed, notably: references 1 thru 3 given below. A common feature of these techniques is to derive a signal, characteristic of straight grain wood, that is sensitive to grain direction deviations in one or more dimensions. Deviations of this signal from the norm thus indicate the presence of deviations in grain direction, significant deviation indicating the presence of knotwood.

4.3. Physical background to the present invention.

The reflection of light from wood is a complex physical phenomenon. For the present purposes however, the critical features can be represented by simple models, the validity of which have been supported by experiment.

4.3.1. Geometric reflection of light from wood.

The simple geometric reflection of light from wood can be represented by the model of FIGURE 2.. An aggregate of straight grain wood cells behaves in a similar fashion to a rough dielectric cylinder 1, the axis of the cylinder 5 corresponding to the grain axis of the wood, and lying in the plane of the wood surface 6. There are two major components to such reflected light: a diffuse component which reflects light symmetrically in all directions independent of the incidence angle of the incoming light 7; and a specular component 2, that obeys the laws of simple specular reflection, reflection angle 4 being equal to the incidence angle 3. The rays of specularly reflected light thus lie within the surface of a cone, the axis of this cone being coaxial with the cylinder axis. The relative proportions of these two components are related to the properties of the dielectric material and the roughness of the surface. Grain distortion can be represented in the model by simply deviating the cylinder axis from the surface axis. It should be noted that when the component of the deviation angle in the plane of incidence takes extreme values, the specular component of reflection lies below the material surface; the surface then behaves as a simple diffuse reflector showing no specular component. The techniques of references 1 and 2 capitalise upon the features of this simple model. The method of reference 1 capitalising upon the goniometry of the specular component to gauge grain angle, the method of reference 2 using the relative proportions of the specular and diffuse components to detect grain defects.We should note however, that in a more complete model for geometric reflection of light from knotwood, the reflection maintains a specular component of reflection. In the physical knot this arises from the plane surface of the wood comprising the knot. In this respect the model for geometric reflection of light from knotwood differs from that for straight grain wood. In the case of the knotwood model there is a plane specular reflection element, parallel to the wood surface in addition to the cylindrical element of the basic model of FIGURE 2.

4.3.2. Reflection of polarised light from wood.

The walls of wood cells comprise a number of layers of anisotropic fibrils, a major constituent of which is cellulose, a dielectric material that exhibits optical bi-refringence. This anisotropic structure of bi-refringent material gives rise to polarisation effects that should be considered to derive a more complete reflection model for wood.

It should be noted from the geometric reflection model of 4.3.1. above that the wood surface comprises three reflective components. A diffuse reflector, a plane specular reflector (in the case of knotwood), and a cylindrical specular reflector. Experiment shows that the cylindrical specular component arises from the anisotropic cell, and cell wall, structures; whereas the plane component arises from isotropic structures. Experiment further shows that the cylindrical component of specular reflection exhibits bi-refringence. When the surface is illuminated with plain polarised light, the cylindrical specular component is decomposed into two orthogonal plane polarised components; one component polarised in a plane parallel to the cylinder axis, the other in a plane orthogonal to the cylinder axis.The relative magnitude of these two components relates to the the angle made between the polarisation plane of the incident light and the cylinder axis, an indicator of complex specular reflection (bi-refringent reflection). The plane reflection component gives no such indication, the specularly reflected component maintains the same single plane of polarisation as the incident beam; an indicator of simple specular reflection from an isotropic reflector. The diffuse reflection component shows no preferred polarisation, with no dependence upon the incident beam polarisation, an indicator of simple diffuse (Lambertian) reflection.

In physical terms then the light reflected from the incidence of plane polarised light on wood shows three distinctive components: a complex specular component, a simple specular component, and a diffuse component, each exhibiting distinctive polarisation properties. The present invention capitalises upon distinguishing the first named component from the other two components, to distinguish grain defects from normal straight grain wood.

5. Method of the invention.

The basic method of the invention is illustrated by the arrangement of FIGURE 3..

A collimated beam of, monochromatic, plane polarised light 1, amplitude la, is directed to make normal incidence with the wood surface 2, via a small aperture plane mirror 3; the plane of polarisation of the incident beam making an angle a with the nominal grain axis (straight grain refers) of the wood surface. Light retroreflected from the surface at the point of incidence is collected by the lens 4, the optic axis of the lens being aligned with the axis of the incident beam. The light collected by the lens is focussed onto the polarised detectors 5 and 6 via the non polarising beam splitting mirror 7, which splits the beam into two substantially equal components. The polarisation of the detectors is determined by placing a plane polarising filter in the respective light paths.The polarisation of one detector 5 being set to make an angle of 0 (zero) degrees with the nominal grain axis of the wood surface; the polarisation of the other detector 6 being set to make an angle 2a degrees with the nominal grain axis of the wood surface. The detectors are therefore disposed symmetrically with respect to the incident beam plane of polarisation. The detector elements yield a signal in direct proportion to the amplitude of light received.

The signal generated by detector 5 is designated D1, the signal generated by detector 6 is designated D2.

With respect to the diffuse reflection component, where the reflected light is depolarised , signals D1 and D2 are equal.

With respect to the plane specular reflection component, where the reflected light retains the incident polarisation angle a, signals D1 and D2 are equal (symmetry of detector detector polarisations with respect to a refers).

With respect to the complex specular reflection component, where the reflected light comprises two orthogonally polarised components, signals Dl and D2 differ. The light reflected is decomposed into two components, Ip polarised parallel to the nominal grain axis, and lo polarised orthogonal to the nominal grain axis. The relative amplitudes of Ip and lo are related to the value of angle a.

Mathematically: Ip is proportional to la. cos2a lo is proportional to la. sin2a further, from polarisation filter theory, Dl is proportional to Ip D2 is proportional to (lp.cos22a + lo.sin22a ) therefore, Dl is proportional to Ia. cos2a D2 is proportional to la(cos2a.cos22a + sin2a.sin22a ) The ratio D1/D2 can provide a useful measure for detecting the presence of complex specular reflection. In the absence of a complex specular component the ratio will be unity, in the presence of a complex specular component the ratio will be greater than unity by an amount relating to the relative magnitude of the component, and the value chosen for angle a, independent of the absolute amplitude of the reflected light.

A normalised plot of the calculated value of D1/D2 (R) against incident polarisation angle a. In the event of purely complex specular reflection, is shown in FIGURE 4. the plot shows a pronounced peak 1 for angle a = 30 degrees. This peaking in response at a = 30 degrees is confirmed by the experimental results, gained with wood, shown in FIGURE 5. This indicates that a = 30 degrees represents an optimal design parameter, though other values are suitable in particular circumstances. In practice values lying between 100 and 400 have found useful. Obviously excluded values, from the above analysis, are 00, 450 and 900 which yield a value of unity for ratio D1/D2 in the presence of complex specular reflection.

The inherent symmetry of polarisation relationships allows the use of any of four arrangements of incident polarisation i and related detector polarisations d shown in FIGURE 6..

By this means then straight grain wood can be discriminated from grain defective wood by applying a simple threshold level to the detector signal ratio.

Thus far the method has been described with respect to a single point on the surface normally illuminated, obviously from the reflection model used as a design base, this normal illumination condition can be relaxed. Provided that the incident beam and retroreflection collecting lens axes lie within a plane orthogonal to the nominal grain axis of the surface, specularly reflected light will be received by the detectors (cylindrical reflector model), and the above relationships will hold. This latter observation allows the method to be used to provide a line scanning equipment.

6. The apparatus of the invention.

6.1. Introduction.

Thus far, the discussion of the method of the invention has been confined to a single point of inspection on the surface of the wood. To achieve the target functions of delineating and locating defects appearing on a board surface the apparatus must gain a two dimensional image of the board surface.

The method for achieving two dimensional scanning is a two stage one. The first stage raises the single point method to a one dimensional line scanning method, the second uses board motion to provide the second dimension.

6.2. Basic apparatus.

The basic apparatus employs optolmechanical deflection to provide one dimensional line scanning, and mechanical board translation to provide the second dimension.

Line scanning is achieved by the simple expedient of placing a deflecting mirror in the common illumination and retroreflection path of the apparatus illustrated in FIGURE 3., FIGURE 7. illustrates the arrangement. The axis of the rotatable deflecting mirror 1 is placed in the optic axis of 2 the apparatus of FIGURE 3. to cause the illuminating beam to traverse the incident point in a line across the surface of the board 4. The plane containing the scanning beam is disposed in a plane orthogonal to the nominal grain axis of the board being scanned. The deflecting mirror is driven through an angle sufficient to traverse the board surface with known velocity and a known repetition rate, by motor or galvanometric means.

Second dimension scanning is achieved by translating the board through the deflecting plane with known velocity.

By this means, the entire board surface can be accessed. By continuous computation the position of the point of incidence on the board surface is known at all times. This allows two dimensional mapping of the board surface in terms of the signal responses gained. This in turn allows delineation and location of localised grain defective areas in the board surface.

This information can be stored and used as input data for various sawmill processing equipments.

This format of apparatus is suited for longitudenal translation of the board (in a direction parallel to the nominal grain axis ) through the scan plane.

In some applications a transverse board translation (in a direction orthogonal to the nominal grain axis) may be more desirable than a longitudenal one.

6.3. Extension to transverse scanning.

The line scanning apparatus of FIGURE 7. can be adapted to allow transverse scanning by placing a telecentric element (lens or mirror) in the scan path, this arrangement is illustrated by FIGURE 8.

A wide aperture positive lens (or mirror) 4 is placed in the scan path, the plane of the lens being parallel to the nominal wood surface 3, the optic axis of the lens lying in the plane of the scanning beam. The rotation axis of the deflecting mirror is placed at the focal point of the lens. This disposition of the lens ensures that the scanning beam makes normal incidence with the wood surface at all points along the scan line. This condition allows the method to be used with wood translated in either longitudenal or transverse directions with respect to the scan plane.

6.4. Applications of the apparatus.

The apparatus is intended for use in a variety of forest product industry applications. The apparatus can be used for defect data retrieval from both hardwoods and softwoods, with sawn or machined surfaces.

Example applications are: * automated board edging * automated board trimming * automated board quality grading and trimming * automated rip sawing * automated cutoff sawing * automated optimising sawing * automated process control in composite manufacture REFERENCES CITED: Reference 1. Matthews. P.C. et al.

Method for determining localised fibre angle in a three dimensional fibrous material.

U.S. Patent 4,606,645. 1985 Reference 2. Matthews. P.C. et al.

Reflective grain defect characterisation.

European Patent application 92308864.5 1992 Reference 3. Kullervo Hirvonen. et al.

Procede pour identifier les proprietes des pieces de bois, en en vue de leur cissification.

Demande de brevet d'invention No. 82 00007. 1982 France. No. de publication 2499 717

Claims (8)

1. CLAIMS We Claim: 1. A method for identifying regions of grain distortion in the surface of lumber, the said method comprising: directing a collimated beam of plane polarised, monochromatic, light along a line of incidence to an inspection point on the surface of the lumber, the line of incidence being contained within a plane orthogonal to the nominal grain direction axis of the surface of the lumber; the plane of polarisation of the incident light making an angle a degrees with respect to the surface component of the nominal grain direction vector.

   monitoring the first retroreflected light with a plane polarised detection means, the plane of polarisation of the detection means making an angle of 0 (zero) degrees with respect to the surface component of the nominal grain direction vector.

   monitoring the second retroreflected light with a plane polarised detection means, the plane of polarisation of the detection means making an angle 2a degrees with respect to the surface component of the nominal grain direction vector.

   calculating a ratio of the amplitudes of the said first and second retroreflected light to characterise the surface at the point of incidence.

   applying threshold values to the said calculated ratio to distinguish between normal straight grain wood and various types of distorted grain wood situated at the point of incidence.
2. 2. The method according to claim 1 wherein the said angle a has a value lying between a = 10 degrees and a = 40 degrees.
3. 3. The method according to claim 1 wherein the said angle a has a value lying between a = 80 degrees and a = 50 degrees.
4. 4. The method according to claim 1 wherein the said angle a has a value lying between a = 100 degrees and a = 130 degrees.
5. 5. The method according to claim 1 wherein the said angle a has a value lying between a = 140 degrees and a = 170 degrees.
6. 6. The method according to claim 2, or claim 3, or claim 4, or claim 5 wherein: the light beam is directed toward the surface via a small aperture flat mirror, situated within the said orthogonal plane, allowing the light source to be situated outside the said orthogonal plane, whilst maintaining the incident beam within the said orthogonal plane.

   a lens is disposed behind the said mirror, the optical axis of the lens being aligned with the axis of the incident beam, to view the point of incidence coaxially with the incident beam, the lens aperture being substantially larger than the small aperture mirror.

   the retroreflected light collected by the said lens is split into two substantially equal components by a non-polarising beam splitting mirror, disposed to direct each of the said components toward a photodetection element, the paths from lens to each photodetection element being of equal length. The photodetection elements are so disposed to coincide with the position of a focussed image of the point of incidence.

   the detection elements are preferentially polarised by disposing plane polarisation filter elements between the beam splitting element and the respective detection elements.
7. 7. The method of claim 6 wherein: a rotatable mirror, the axis of rotation being orthogonal to the said orthogonal plane of incidence, is disposed in the common incident and retroreflected light paths; rotation of the said mirror causing the point of incidence to be deflected along a line contained within the said orthogonal plane of incidence: the physical location of the point of incidence along the deflection line being related to the rotation angle of the said mirror.
8. 8. The method of claim 7 wherein: the rotatable mirror is repetitively rotated through a known deflection angle, with a known rate of deflection and with a known repetition rate; the deflection angle being of sufficient amplitude to cause the incident point to traverse the full width, or known partial length, of the piece of lumber under inspection.

   the said piece of lumber is translated through the said orthogonal plane in a direction either parallel to the nominal grain direction of the piece, or orthogonal to the nominal grain direction of the piece, with a known velocity.

   the location of the point of incidence in the plane corresponding to the surface of the lumber is continuously computed from the said known angle, the said deflection rate, the said repetition rate and the said lumber translational velocity.

   the location of points of incidence on the lumber surface that are distinguished from normal straight grain wood are recorded, the said recorded data being used to indicate regions of defective lumber to various sawmilling application equipment.