NZ242145A - Rotational decision for saw-log based on aggregated cross-sectional knot images - Google Patents

Description

0.C-<\ <v"/ it Wv^ fOi •'<)' A i1-! t CrCtrOl'Z I'i; dJi tl JUL 1993 '2HO Under the provisions of Regulator! 23 (1) the a*,,,,.* Specification has been ani&-dated ™ Initiate Divided Out of N.Z. No 230371 Dated 21 August 1989 NEW ZEALAND Patents Act 1953 COMPLETE SPECIFICATION LOG SCANNING METHOD We. MacMlLLEN BLOEDEL LIMITED, a Canadian corporation of 1075 West Georgia Street, Vancouver, British Columbia V6E 3R9, Canada do hereby declare the invention, for which we pray that a patent may be granted to us, and the method by which it is to be performed, to be particularly described in and by the following statement: - h ? 1 4*1 Backer cur.d cf the Present Invention The disclosure will deal with logs but some cf the technology described will have another application. The tern defects is intended to include one or mo re of stones or nails or other intrusions in the tree and natural defects such as knots or rot or very low density volumes or voids.

It has lone been a desire of the forest products industry to provide a system for internally examining the log to find its defects anc then, based on defects and their location, automatically in "real time" provide a sawinc solution to permit maximization cf the vooc recovery or lumber recovery from the log. Sv "real time" it is meant at a rate that keeps pace with the normal speed of operation of the sawmill particularly the heacrig cf the mill.

In the Fourth Nondestructive Testing of Wood Symposium in August 1973 in a paper entitled "Scanning of and Computing Methods for Measuring Knots and Other Defects in Lumber and Veneer" by Torbjorm Schmidt there was a brief description of the application of tomography to investigate defects in a log.. In that description the exposure time for tomography was 37 seconds anc the computer took two minutes to provide the resulting picture illustrating a cross;section through one section of the log. It will be apparent that, while it was evident one could determine the internal structure of a log using tomography in 1978, it was simply impossible to do this in a time frame that was useful for control in a sawmill. This is particularly true when one considers that only a single cross sectional image was obtained in a two ana half minute time frame.

In Wood Science (Vol 14, No. 3, p 97-104, Jan 1982) an article entitled "Application of Automatic Image Analysis to Wood 2 Science" by Charles W. McMillan discusses automatic image analysis and describes scanning technology for primary Ice breakdown and for cutting clear furniture parts from defective boards. The use of computerized axial tomography "CATSCAN" tc ncncestructively locate defects in the ice interiors is desc r ibed.

The majority of the McMillan paper is directed toward image analysis of photographic images and is simply an indication of what micht be accomplished. However none of the operations are cone in "real time". These teachings are not useful for c. commercial log scanner to determine a sawing solution in "real time".

In the McMillan article the concept of using a CATSCAN to determine the interior of a log is aiscussed as well the use of a plurality of such scans to define the x-v coordinates for a knot in each such cross sectional scan. McMillan suggests that the information from the cross sectional scans then be used in a computer to determine the log positions needed to maximize grade or value yield but provides no teaching on how this might be done .

In Forest Research Bulletin No. 8 (Feb 19, 1982) there is a paper entitled "Computed Tomographic Scanning' for the Detection of Defects within Logs" by Benson-Cooper et all that also suggests that a sawing solution based upon CATSCAN information might be derived, but does not provide any teaching as to how one might obtain this objective. 3 * 1 / ( ■ '! k ^ f ■ Brief Descricticn cf the Present Invention Brcadlv, the present invention relates to a method of identifying elements or defects of different densities in an image generated by projecting electromagnetic energy from a source through a non-uniform body traversing said source ana developing an image representative of the attenuation of the electromagnetic energy in localized areas through the body by detecting the amount of electromagnetic energy passing through the body in such localized areas thereby to gene rate a signal varying in accordance with the density of the material of the body and of the elements obstructing passage of the electromagnetic energy through the body as said body traverses said scanner comprising developing a body geometry related signal indicative of the geometry of the body being sensed by eliminating major fluct u*a tions in signal amplitude anc subtracting said geometry related signal from said signal tc provide a resultant signal ana analyzing said resultant signal for areas of significant differences in signal strength.

A method of generating an image comprising passin9 electromagnetic energy from a source through a body containing elements of different density than the average density of the body, a sensor having a plurality of discrete detectors arranged in side by side relationship and adapted to receivo 4 cccy, eacr. saic ciscrete cetector detecting ;ne amount: c: er.e::v passing from said source through said bccy to said detector thereby to generate an image signal based on the degree cf attenuation of electromagnetic energy received by each cf the detectors, said body being cf non-unifcrm thickness measured m the direction cf electromagnetic energy propagation, eliminating a portion of said signal representative cf said body by smoothing the signal from each detector to produce a body signal substantially free of major changes in amplitude and subtracting such body signal from said image signal for each detector to provide signals representative cf said elements contained within said body.

Preferably there will be three such sensors (more may be used but 3 have been found to be adequate) and sources arranged at spaced locations around the body and each adapted to cenerate an image signal and wherein said body is moved relative to said sources to provide continuous plan image signals extending axially of said body in the direction of movement of said bcav relative to said sources, processing of said plan image signals to identify signal areas representing elements, analysing each of said longitudinal plan to identify signals representative of the same element in each of the plans and reconstructing a cross section of said body with said element positioned in said body by a triangulation method.

One mode of identifying the same element in the various plans comprises finding elements in all plan image signals having their end points located in the same pair of spaced planes perpendicular tc the direction of travel of said body past said sensors, determining the approximate size of each said element for each said plan, selecting as representing the same element those elements having their end points in each plan in substantially the same said pair of planes and determined as being essentially the same size.

A mode of identifying the size of an element comprises identifying a longitudinal axis for the element signal, determining the maximum width of said element signal perpendicular to said longitudinal axis, determining the position cf -he end points cf said longitudinal axis, defining a cair cf ccr.es cn said icncit.udir.al axis wi - h the axes c f the ccr.es coinciding with said longitudinal axis ar.c with their -aicr diameter ends abutting and their pointed er.cs coinciding with the end points of said longitudinal axis, said major diameters reing equal and equal to said maximum width, the combined volume cf said pair of cones representing the volume occupied by said element.

In order to eliminate bocv geometry the basic axial density signal from each detector is filtered to suppress the high frequency information representing defects using successive convolutions applied to each channel (signal line extending perpendicular to the longitudinal axis) to provide a bccy geometry signal and subtracting the body geometry signal from the basic signal to provide a defect signal.

Broadly, the present invention also relates to a system for analyzing bodies (logs) to provide a basis for a sawing solution comprising conveyor means for transdorting a leg substantially longitudinally, density scanner means having means for passing electromagnetic waves substantially perpendicular to the direction of travel of said log, said scanner means including at least two discrete sources of electromagnetic energy, said sources being angularly spaced around said conveyer means to pass electromagnetic energy from at least two different directions, through said log as it is conveyed past said scanner means, sensor means for sensing the amount of said electromagnetic energy passing through said log from each of said sources, .each cf said sensor means being composed of a plurality of discrete detectors circumferentiallv positioned relative to said log in side by side relationship opposite their respective of said sources and each adapted to detect the amount of radiation it receives from its respective source to provide discrete values for the degree of attenuation of electromagnetic energy between each said discrete detector r.nd its respective said source, each said sensor generating a longitudinal plan of the radiation detected by said sensor means over the length of said body, means for defining areas depicting different densities representing defects in each said plan, means for identifying areas in each of said plans representing the same element in the bodv in each cf said longitudinal plans and means fcr reconstructing seated discrete cross sections through said body positioning said elements in said cross section each said cross section being representative of a preselected length cf said body traversing said scanner.

Preferably a longitudinal axis is selected for a selected length of said log and said discrete cross sectional views are collapsed along said axis to provide a projected cross sectional view of a selected length of the log identifying the propensity cf such elements in various areas cf such projected cross section.

The propensity of elements in a given area may be determined by providing a selected value per unit length measured ir. such axial direction for each such element in each of said cross sections. Preferably the rate of change in the area volume) as fewer knots are included is used as a means fcr determining an area or volume for a knot core.

Preferably a rotation decision is to rotate said log to a given angular position for presentation to a headrig will be based cn the location of said elements ana the propensity of such elements in areas of said projected cross section preferably on the major diameter of the projected cross section of the knot core.

The system of the present invention also preferably when desired will determine a bucking solution for the log being processed and the length of log for which a sawing solution is to be found will be selected based on the bucking solution. 7 t " r" ' r jL ' .

Accordingly, in its broadest respect the invention comprises a method of analyzing a log to provide a rotational decision for sawing comprising determining a plurality of spaced cross-sections of said log having knots positioned therein, defining a longitudinal axis of said log, applying a grey scale intensity to each knot located in each of said cross-sections, projecting said grey scale intensity for said knots parallel to said longitudinal axis of said log to form a grey scale cross-sectional image varying in grey scale intensity in various areas depending on the number of knot representations projected into said various areas of said image. 3rief Description of the Drawings Further features, objects ana advantages will be apparent from the following detailed description of the preferred embodiments of the present invention taken in conjunction with the accompanying drawings in which.

Figure 1 is a schematic representation of scanner system in connection with which the method of the present invention can be used.

Figure 2 is an end view illustrating a scanner having three angularly spaced radiation sources and correspondina 7a A CX 1V p 1 C a _ c/.l : V iX'tT.Cir.C dens i t Y mgtns c: tr.e _cc as cctamec one :::n eacr. of v-.. tr .--.cure is a ::.zersc mace c: :r.e :.ans or Ficure 2 £ hc w: nc kr.c;s . reshclded image derived from the plans u r e c is a recicr. crew cure 6 illustrates three similar clan views after . ng depicting knots. .cure 7 illustrates a common section cf the plan views cf Ficure 5 shewing the same element in each plan.

Figure 8 illustrates a system for delineating a defect kr.cz) ir. a bouncing polygon the sice cf which is determined by the detected extremities of the defects.

Figure 9 illustrates a bouncing polygon in 3 dimensions as well as a second method cf estimating an element size that particularly suited to knots.

Figure 10 illustrates an axiallv projected cross sectional image derived by axiallv projecting to superimpose bounding polygons of knots in a plurality of discrete axially spaced radial cross sectional images.

Figure 11 illustrates different thresholding values for the projected cross sectiona1' images.

Figure 12 is a plot of grey scale threshold values representing number of knots vs volume as a percentage of the whole body.

Figure 13 is a plot of rate of change of volume vs threshold values representing number of knots.

Figure 14 shows a possible sawing solution based on analysis of the log.

Figure 15 is a schematic representation of a log depicting one manner in which the rotational angle for the log may be represented, illustrating a datum for angularly rotating the log and a minimum opening face.

Figure 16 is a schematic plan view illustrating skewing of the log for presentation to the saw. 8 ^ k- i- ■+j ; : r i p t i or. of the Preferred Embed i aer. t s As shewn m Ficure 1 a loc 1C is carried en a convevcr 11 th r ough an inlet he us mc 14 which preferably is ces igned to prevent the escape cf radiation. The lee is carried en ccr.vevcrs through a scanning station 15 which includes preferably at least three scanners 13, 20 and 22 (two may be used but are net recommenced as proper resolution is difficult) each passing electromagnetic waves substantially in a plane perpendicular to the direction of travel cf the conveyor 12 so that the waves pass through the log on paths in a plane substantially radial (perpendicular to the direction of log travel) to the log as the leg is carried by the conveyor 12 in the direction of the arrow 24 through the station 16. Generally each of the scanners 18, 20 anc 22 will pass electromagnetic energy, e.g. x-rays, through the log to determine the local density cf the loc, as will be described below.

Also included within the scanning station 16 is a laser profile scanner 26 which determines the cuter dimensions of the log as it traverses the station 16 on a conveyor 12.

The laser profile scanner 26 may be used to mark the log 10 as it passes using a marker mechanism 28 which may take the form of a router, a paint spray or the like, tracing a line along the log preferably along the line defining the log periphery's maximum spacing from the face of the conveyor 12. This line may be subsequently used either in a bucking solution or as a datum for rotation of the log, as will be described be low.

At least that portion of the conveyor 12 passing through the scanners 18, 20 and 22 preferably is a belt-type conveyor made of suitable material that does not interfere significantly with the operation of the scanners 18, 20 and 22. Some of the electromagnetic waves will be passed through the conveyor 12 to ensure the full cross section of the log is inspected.

The scanners 18, 20 and 22 are spaced along the length of the conveyor 12 however, for convenience in Figure 2 all have been shown in essentially the same plane.

The scanner 18 includes a radiation source 18A and a 9 sensor cr ce:ecicr array idd positioned cirec t_ y ocpcsite ;ne scurce IS.-.. Sensor 135 is composed cf a plurality of discrete detectors ISC which preferably are approximately cne-quarter inch in length measured in the axial direction cf travel of the conveyor 12 and a similar width in the circumferential direction along the curvature cf the detector 135 which preferably will be essentially on an arc the center c f which cc inc ides with the source ISA.

The other scanners 20 and 22 include similar components each of which is indicated by the number of the scanner followed bv the letter as described for scanner 18.

Each cf the scanners 18, 20 and 22 is used to generate an axial extending density plan based on the attenuation of the electromagnetic energy by the leg as the energy passes from the sources isa, 20a and 2 2a to their respective sensors 18e, 20b and 22b. Such a set of three axially extending density plans or projections are illustrated in Figure 3 for a selected length cf one particular log (axiallv plans cf the whole length of the loc are generated as the log passes through the sensing station 16).

It will be noted that all of the projected plans are different, each being representative of density variations through the loc at the different angles at which the sources project radiation through the log and as sensed by the detectors 18c, 20c and 22c as the log continually passes the scanners 13, 20 and 22. These longitudinally extending density plans are adjusted by calibration factors in the data acquisition computer section 30 (Figure 1) and have been designated as plans 18d; 20d and 22d in Figure 3 (the numeral corresponds with the sensor detecting the particular image). It will be apparent that each discrete axial length of one of the plans is matched (aligned on the same plane) with a corresponding discrete axial length in the other plans.

The images acquired by the computer section 30 are then analyzed, for example, in a further computer section 32 by filtering (Figure 4) and thresholding the images (Figure 5) based on the gray scale analysis, i.e. the images 18D, 20D and 22D vary in brightness depending on the local densities of the log which in turn indicates the degree of attenuation of the radiation r ~ 1 h 5 In Figure 6 rs"ir.ec v? rsicr.s c f the i m a c e s 15 D 2 0 D ^ ^ <■* 222 :r.cvr. as 13", 20F ar.d 22? respectively which clearlv indicate the cutlir.e cf the knets cr high cer.sitv areas in their respective density clans.

It will be apparent that since the loc is of ncn-uniforrt crcss section, fcr example may be substantially oval cr circular in cross sectional shape, the length cf the paths of travel cf the electromagnetic energy rays through the log will be different in different areas of the log. Attention is again directed to Figure 2. The ray 22H which is detected by the cerectcr 22J passes through a thickness of the log 10 as indicated by the distance p whereas the ray 22K detected by the detector 22L passes through a thickness of log 10 indicated at P. It will be apparent that the attenuation of the ray 22H due to the bccy cf the log perse is significant! v less than the attenuation cf the ray 22K simply because ray 2 2H passes through less wocc than does the ray 22K ar.c thus the signal produced by the detector 22J regardless of whether o; not it traverses a defect in the log, will be significantly different than the signal generated by the ray 22K and will bias the scanning results accordingly. It is important that this portion of the signal as determined by body geometry be eliminated or rendered substantially insignificant so that the defects can be discerned.

To eliminate the body geometry portion of the signal as represented by the different thicknesses p ana P the signal generated by each of the discrete detectors 22C for example the detector 22J and 22L are each processed individually along the length cf the scan, i.e. in the direction parallel to' the direction of movement of the log passed the source 22A. Each of these discrete detectors 22C represents a channel in the image generation system and each of these channels is processed independently in a manner to distinguish discrete elements such as knots or rot from the remainder of the body of the log. This can be accomplished by a variety of different techniques including, for example, edge detecting, image shifting and subtraction or multiplication and edge detection or subtraction.

A preferred system of identifying defects within a 11 H 5 : .:c..:3r.: rear. c: a se_ect.ec axial 1 e r. c t h cf : r. e icc is t c *- - z z ~ z z a s : zr.a 1 : :c.t, ea cr. char.r.e 1 by successive!y ccr.vclu;; nc rignal vicr. a sec cf cne-dimens icr.al lew pass filters and cuccrac:::.c the ccr.vclutec signal from the crigina 1 siar.a 1 :r. erecy leaving only the nich frecuentv defect information. Preferably the width cf the low pass filters will be increased by a significant margin cr. each successive pass fcr example, the filtering sequence could be first on twe pixels, i.e. 1/2 filter, next ir. the sequence on 4 pixels, i.e. 1/4 filter, on eight pixels, i.e. 1/8 filter and assuming five passes, 16 pixels and :: pixels and the final convoluted signal subtracted from the original signal to provide a signal indicative cf the defects detected by each discrete detector 22C such as detector 22J or 111, i.e. by each channel.

Anether approach for cetermining a defect signal from t.-.e cigr.al contaminated with information relating to the body itcelf, is to determine defect edges using an edge detector applied alcng each channel anc masking out each detected defect :r high density area thus generating a defect mask image for that channel. This defect mask image is then subtracted from the original image cr signal to provide a defect signal.

After the defect signal has been gene rated it is preferred to normalize same preferably to value of where all defects are either above or below a selected value, for example in a system with a range of 256 a value of 1/2 the range or 128 might be selected which will confine random noise to about this value. When processing logs, nails and rocks have been found to threshold above 140 on such a normalized signal and knots with rot are above about 130 while dry rot and voids are located below the 128 value at about 125. Binary images can then be produced for each of the defects by thresholding the normalized images at the appropriate level to produce such binary images and region growing the binary images into objects.

In the event the logs being processed or the bodies being processed are all substantially symmetrical about a longitudinal axis, for example as may well be the case for logs produced from properly pruned trees which confine their knot location in the pruned length to a substantially cylindrical 12 n y i a - c ' :ier. : e e a r. c a x ia_ i v scacec s w i r _ s . : ' ' r* 'n - are suf f i c i e r. t i r f e r rr. a 11 e n m a y be a v a 11 a b i e r r em prcc c s s i r. g i r.c 1 e ax: image ar.c i d e n ti tying the size anc tcsi ti c n of t he r. 11 e t r e a r. c swirls i r. t h e o n e image (the othe : image s w 1 _ _ e e -it S ' T. i _ a r ) ar.c use thi s i n f c rma tion in dete : 'mini ng a s a i r.g o _u tier., i . e . only a s ir.gl .e sea r.ne r such as sc a r. n e r 1 8 mav C 6 It will be apparent that in most normal logs, a single axial plan view will net be sufficient and while two views may be usee tc determine the location of knots or other defects and position them in a reconstructed cross section, the accuracy cf such a system is not as good as that obtained by us ing three separate sources anc three sensors to provide three axial plan images. Thus the remainder of this description will relate primarily to the use of three sources and three sensors and detecting and positioning defects within a body (log) based on three circumferentially spaced sensors as shown in Figure 3 generating three axial plan images.

The various defects or high density areas illustrated in the three axial plans 18D, 20Fana 22D are analyzed to determine corresponding areas for the same defect or knot in each cf the plans 18F, 20F ana 22F.

To determine whether areas in the various plans are areas representing the same element, the plans 18F, 20F and 2 2F are analyzed. To illustrate the process attention is directed tc Figure 7 wherein a selected longitudinal segment at the same axial position along the log for each of the plans 18F, 20F and 22F are illustrated by plan segments 18G, 20G and 22G. These plans 18G, 20G and 22G show an element or defect 500 which' has its end points 502 and 504 in corresponding or the same pair of spaced radial planes 506 and 508 respectively in the various plan segments 18G, 20G and 22G.

These end points 502 and 504 may be used to define a selected longitudinal axis indicated by the dash lines 510 in view 18G. Similar selected ax is can be determined for each of the views 20G and 22G for the element 500 . However, it is preferred to determine an axis for the element 500 in all of the views using the well-known technique of a robust estimation to 13 ue:-2 rr.ir.e axis as indicated at 512 where me defect 5 0 C ir. each cf the views 15C-, ZQC- and 2 2G.

In eacr. cr me views 13G, 20G anc 22G the maxi."nun width perpendicular tc the selected axis 512 is deter-inec tc pre •••ice an indication cf the sice cf the defect 500 . By comparison cf the relative sice, location and major axis of the elements in the various view elements having their axial end points 502 and 504 •axial extremities measured substantially axiallv of the view whi ch in turn is axial cf the log) in substantially the same spaced transverse planes 506 and 508 (perpendicular to the axis, i.e. radial planes relative to the view or the log and substantially corresponding in size will be accepted as being the same element).

After the corresponding knots have been detected the extremities of these defects or knots based on the angular projection f rem each cf the sources 18A, 20A ar.c 22A. for each of the respective images is used to define a bounding polygon for the defect (see Figure 8).

In Figure 8 a knot 200 has been depicted by crcss-hatchinc and its extremities are detected for example, the x-ray-source 13a determines two extremities as indicated by the lines 18a and 18Y which define the extremities x and v of the' knot 200 as detected by detectors 18C. Similarly the extremities s ana t are determined by the detectors 20C as depicted by the lines 20s ana 20t and similarly the extremities s and t are detected by the detectors 22C based on the lines 22s and 22t. It will be seen that a combination of these lines 18X, 18Y, 20s, 20t and 22s, 22t define the side wall 202, 204, 206, 208, 210 and 212 respectively of a bounding polygon 214 for the knot 200. In many cases the inner extremity (adjacent the - heart of the log) which, in the illustrated arrangement is say boundary s may be confused by the overlap of adjacent, but different knots. In this case, the defined centre line of the log will be used as the inner extremity equivalent to extremity s.

After the knots have been identified and their bounding polygons determined for each of a plurality of discrete axially spaced radial sections, these sections are converted to a binary system wherein each bounding polygon 214 in each discrete radial 14 Gray scale values will depend in part en the length cf the defect measured axially of the icc.

Generally the gray scale value applied to any given cross sectional image will depend cn the axial length of the log represented by a given cross sectional image and the total length of the log tc be processed as will be described herein belcv;. Thus, for example, the gray scale value for a defect might be determined by G = NP where G = grey scale value N = the Number of discernable levels of gray scale, and P = axial lencth cf the discrete cross section axial log length being processed Generally the axial length of a discrete cross section will represent about 4 inches measured in the direction of travel of the log as this length has been found to provide an adequate assessment. One foot has also been used ana found to be satisfactory but it is preferred to use 4 inches as the short length permits better resolution. Similarly shorter axial lengths, i.e. less than 4 inches may be used for each discrete cross section. This will increase the number of cross sections that have to be accumulated as will be described herein below to determine the projected cross sectional image for the log, and will also improve the resolution if required.

Generally shorter than 2 inch axial length sections or slices are not warranted as the time for processing increases with each additional operation, while a slice length of 2 feet or even 1 foot reduces the resolution to the point where the time savings do not warrant the reduction in quality or resolution.

The log analysis may also be take into consideration the frequency of knot occurrences over the length of the log such that if there are a plurality of axially aligned knots with an intervening length between one pair of axially adjacent knots of .I u i r. c - 0 c i 31 c r..

As t n e sccve operations a r ~ fceir.c: carried cut, the profile scanner 15 which will normally be any cr.e cf a number cf commercially available laser scanners provides a sicr.al tc a profile computer 36. This computer is used to interpret the signal from the scanner 26 to select a longitudinal rotational axis for the log as indicated by the rotational axis x-x in ficure 15 and 15.

Various techniques may be used to find this hypothetical axis. One cf the simpler ways is tc determine the centers cf the leading and trailing ends of that portion of the length cf the log that is to be processed and to use as the hypothetical cr longitudinal axis x-:< the line interconnecting tr.ese centers.

Other more complicated techniques may be used tc define the longitudinal rotational axis x-x, for example a least squares method based on the sensed profile cf the log.

It will be apparent that as the length of log being processed changes so will the x-x rotational axis, i.e. the rotational axis will be in part dependent upon a bucking decision, if any, with respect to the log being scanned.

A bucking decision may be made in any suitable or conventional manner, for example, manually or by sensing the curvature of log using the profile scanner 26.

This longitudinal x-x rotational axis as determined above is used with the radial plane images generated by. the computer 34 to axially project or collapse the radial plane images along lines parallel to the longitudinal x-x rotational axis to provide an accumulated cross sectional picture or map for a selected length of the log indicating the accumulation of defects in a given axial line. Such accumulated images are illustrated in Figure 10.

The resulting accumulated radial cross section image or density map is thus composed by superimposing the reconstructed radial images which have been given a signal ratio based on the length of the log being processed such that the axial accumulation of axially overlapping knots results in a particular 16 the accu.T.ulatsd cross sectional accumulating t.te radiai imaces ant o r c c u c ■ - c tne CCCU.T;U -a t e c racial censitv mac is rscr6s0r"cr" by the computer 3 3 (see Ficure 1).

The resultant accumulated racial image is then subjected tc image analysis in the computer section 40 to determine the boundaries which delimit substantially clear vccc from substantially knotty wood. The computer 40 analyzes the gray scale image for example in the image illustrated in Ficure 1j or 11 to determine the knot location anc propensity of knots in given locations and arrives at a rotation decision, i.e. the ancle about rotational axis x-x that the log should be rotated for presentation tc the saw.

Such an analysis of the accumulated racial image may be cone in several different ways. For example, thresholding the accumulated radial image based on a selected degree of brightness (bearing in mind that each knot area had essentially the same brightness in each of the discrete cross sections that are accumulated) to determine clear and knotty areas and classify knotty areas with various degrees of knot propensity ana assign values to the areas. Based on these analyses the quality of the wood that may be cut from any particular section can be cete rrnined .

The preferred system for finding the common knot core anc examining the core includes boundary tracing thresholded areas after the threshold process and further reducing, the boundary points between the thresholded areas and the adjacent areas by a co-linearitv test utilizing the split merge algorithm that 1. sub-divides the boundary points, 2. processes the two segments, 2.1 draws a line through the dividing end points of the segment, 2.2 if the distance of the line to the furthest point of the given segment is greater that a pre-set allowed distance, the segment is split at the farthest point and the process in 2 above is repeated. 1 I iJ / V / ' • | A i--i ■ - '' ' ' fc. 2.2 i f : :ne distance is less than the cre-ss 51 allowed t h 3 P C ~4 er.ts are ~.erc6c. i f t e r the enc points cf the boundary are r* a r o r ne n • e cc-lir.eari ty test a convex r.un a _ - o r i1r.m suer. as the u a rv i - Q - ■: - u _ " n r the identification cf the convex hu. 11 of a ** T n: t set cf points in the plane" (Information Processing Letters, Vcl. 1 paces 13-21 (1S73) is applied. This algorithm finds the lowest point in a data set and uses it as the first current base point, i.e. the first vertex of the hull. The next base point is selected such that it forms the smallest positive angle in relation tc the then current base point anc this next base point then becomes the current base point for the next vertex cf the hull. This procedure is iterated until the next base point is the first base point at which time the complete boundary hull is f c rmec Figure 11 illustrates a series of bounding polygons applied to an image. In the series illustrated the bounding polygon 6 00 bounds that area that contains at least one defect and is based on a threshold value of a grey scale for one defect. Bounding polygon 602 has its periphery based on grey scale value for at most 2 defects. 5ounaing polygon 603 illustrates three defects, i.e. the more defects within a bounding polygon, the darker the images contained therein.

When the convex hulls are identified for each selected thresholding value, their areas are multiplied by the log length to determine volume of the bounded core and the ratio of this bound volume over an estimated log volume based on analysis of the log, may be plotted against its respective threshold value.

The number of thresholding values will vary but it is preferred to take a plurality of such threshold values, determine the volume of the bounded core, and provide a plot of such volume or percent of such volume relative to the total volume of the log against the threshold values as shown for example in Figure 12. This information may then be used by converting same into rate of change, i.e. differentiating the curve of Figure 12 to provide the rate of change of volume as the thresholding value is increased. In the illustration in Figure 13, it can be seen that the rate of change is relatively high or what is defined as high 18 i r*. p r- 'V / j Fv f 'VJ value weed whereas ce~~er. weed shews a rate ef chance versus ::undary between the kr.ee cr common cere anc the high value weed is indicated by the beur.dary level which, in this case, has a t r. r e s h c 1 d value of 4 .

Gene rally it has been found tha t clear c r high value weed has a high rate cf chance of volume, shop wood has a lesser rate cf change anc common wood has a low rate of change. In the example cf Figures 12 and 13 the common core has a rate cf change of less than 0.01.

The above technique provides one way of defining the common knot ccre for a log. It is also possible to determine the knot ccre empirically based on a particular selection cf threshold value and utilize that empirically selected threshold value fer determining the common or knot core for example by iermir.g a bounding polygon based on the selected threshold value. Such a system is less accurate than the th reshcldinc system as described with respect to Figures 12 ana 13, but does provide a simpler method.

Generally the rotational decision will be based on aligning cne of the cutting planes substantially parallel to the longest diagonal of the knot accumulation or common 'core for example the longest diagonal for the bounding polygon for the knot core as defined above.

The other cutting plane will be substantially perpendicular to the longest diagonal and normally a rectangle will be determined that incorporates the common knot core ana has one side parallel to the longest diagonal to redefine the knot core for the sawing solution.

The rotational decision based on such image analysis made by rotation decision computer 40 is fed to a rotational control schematically indicated at 42 and also to the profile computer 36.

To facilitate operation of the rotational control either manually or automatically, the location of the axial center line x-x (rotational axis) relative to one point on the periphery of the log at least one end of the log 10 must be known so that the angular rotation or displacement of the log 10 19 r :unc tr. e axis x-x as :r.d; :a~ec bv the ancle A ir. Ficure 15 car. relative t: a datum. I r. "icure 15 the 11 r. e :.e3::i:ec cy :.a:ke: -8 r.a; been indicated at 54 ar.c the junction c: tr.i: _ir.e ••■•ith the leading face 5c is indicated at 55. This junction is connected by a line 51 tc the axis x-x in the selected cen:e: as indicated at 60 of the front face 56. The ancle A is the ancle between the line 62 anc the rotational decision which is depicted by the line 64 extending from the center 50 tc the cuter periphery of loc.

One of the opening faces or cuts is mace substantially perpendicular to the line 64. If this first cut is to be parallel tc this opening face the line 64 must be oriented bv rotation c f the loc to be substantially pe rpenaicular to that c f the cut. This then defines the angular or rotational position of the log relative to the saw. Alternatively the first cut cr opening face may be parallel to line 64 and the log will be rotated so that the plane cf the saw is parallel to the line 64.

Once the rotation decision is made the information from the profile computer and rotation decision are fed to the skew decision computer section 44 which adjusts the x-x axis of the log to the sawing plane of the saw to ensure that the first cut provides for a preselected minimum width board as shown by the dimension Z of the opening face 46 in Figures 15 and 16. The opening face 46 is parallel to the direction of cut of the saw 48, i.e. is parallel to the direction of feed as indicated by the arrow 50 in Figure 16. This skew decision is then fed to the skew control as depicted at 52 in Figure 1.

The skew control is then exercised 'for example by adjusting the relative positions of pushers or abutments 66' ana 68 as indicated by the arrows- 70 ana 72 to align the face 46, i.e. minimum cut width face with the direction of travel of the log 10 to the saw 48.

This skew decision and rotational decision together with the information of the image analysis is fed to a further computer section 74 which determines the sawing solution. The equipment may automatically control the saw lines by both lateral adjustment of the log relative to the saw (as indicated by arrow 76 in Figure 15) ana rotation of the log (as indicated It has been found that the present invention provides a method for scanning and determining the locations and size of elements within the body.

It has also been found that the present invention provides a log scanning system operable in real lime to determine a sawing solution for a log based on the location of internal defects.

It has also been found that the present invention provides a method for separating a signal representative of elements of a selected density in an irregular shaped body of a different density wherein the generated signal incorporates a component representative of the shape of the body and a further component representative of the elements by determining that portion of the signal generated by the body and subtracting from the overall signal to provide a signal representative of the elements.

It has also been found that the present invention provides a simplified method for identifying objects and their location within a body based on determining the axial ends of said elements in at least two projected plan views of the body determining the approximate size of the elements in each said plan and selecting as the same element those elements having substantially the same size and their end points located in said plan to use in the same pair of axially spaced planes said planes being substantially perpendicular to the longitudinal axis of said plans. 21 f The description has referred :: loca::r.c knots as -he material cf high density, other imperfections cr inclusions such as metal, rocks and rot can be located and taken into consideration in the sawing cecisicr..

The above description has dealt with logs as this is the intent cf the equipment but could be used tc detect ar.c locate areas of different densities in other bodies.

Having described the invention modifications will be evident to those skilled in the art without departing from the scirit cf the invention as defined in the aDoenaec claims. 22

Claims (4)

WHAT WE CLAIM IS:

1 A method of analyzing a log to provide a rotational decision for sawing comprising determining a plurality of spaced cross-sections of said log having knots positioned therein, defining a longitudinal axis of said log, applying a grey scale intensity to each knot located in each of said cross-sections, projecting said grey scale intensity for said knots parallel to said longitudinal axis of said log to form a grey scale cross-sectional image varying in grey scale intensity in various areas depending on the number of knot representations projected into said various areas of said image.

2. A method as defined in claim 1 further comprising determining a knot core in said cross-section by selecting a threshold grey scale intensity for said knot core and determining a bounding polygon based on said thresholding intensity.

3. A method as defined in claim 2 further comprising determining a maximum diagonal length for said knot core and providing a rotation decision based on the angular position of said maximum diagonal.

4. A method of analyzing a log according to claim 1 substantially as hereinbefore described with reference to the accompanying drawings. MaeM+kfctN BLOEDEL LIMITED By their attorneys HENRY HUGHES LTD Per: /' / ( / / H.