CONTROL METHODS AND APPARATUS
This invention relates to control methods and apparatus, especially to controlling a radiation beam such as a laser beam to effect a power operation such as cutting: the control may involve tracking moving line features such for example as the edge of a sheet of material such as lace or paper: and also to inspection particularly of lace and cut edges thereof.
In WO 94/03301 is disclosed a machine vision controlled technique for guiding a laser beam for cutting (or scalloping) lace. A problem with lace is its flexibility and deformability which result in movement ofthe lace laterally with respect to the guidance system as it is being transported past the cutting station. This can put extreme demands on the guidance system. An edge tracking system could ease the problems thereby encountered.
Tracking systems for detecting lateral shift in edges, and of course, track following systems for guiding robot vehicles, are well known. However, available tracking systems, relying on photocells, lack the required precision for the laser lace cutting application, and the present invention provides an improved system.
In WO 94/03301 the beam is deflected by a moving mirror under the control of a guidance system. As the beam has a high power density, it is expanded prior to impinging on the mirror which could otherwise be damaged. The lace to be scalloped runs at high speed - in the region of ImS'1 beneath the mirror and the beam is brought to a focus on the lace.
As the beam is deflected from side to side to cut the lace, the optical path length varies. Focus is not maintained on the lace across the width over which the beam sweeps. Two measures are proposed to deal with this, namely, first, to arrange the
optical system so that over the sweep width the beam cross-section at the lace is sufficiently small that the beam power density is adequate to cut the lace. and. second, to bow the lace over the sweep width so that lace over the sweep width is equi-distant from the mirror.
The latter solution works for lace, which is flexible, but even then is only possible when the beam sweeps in one plane only, peφendicular to the direction of lace travel. WO 94/03301 describes a two-mirror system in which the beam is deflected over an area for cutting more complicated shapes, and it would be much more difficult to bow a fast-moving lace web in two dimensions.
The first solution can always be made to work, but it requires the use of a higher power laser than would be the case if the beam could always impinge on the lace at its highest possible energy density.
Lace is made ofa series of knitted patterned breadths, the pattern repeating along the lace. After knitting, the breadths have to be separated, and this is done by cutting along the edges ofthe pattern. Usually the edges are more or less convoluted and the cutting line follows the convoluted edge.
Knitted lace comprises solid pattern sections on a background web.
Recently, machinery has been developed capable of cutting lace at high speed using a guided laser beam.
The present invention is concerned with inspecting the cut (or scalloped) edge for accuracy or quality of cut. Errors in cutting comprise "excess cutting" in which the cut is too far away from the pattern section so that background web extends beyond the pattern section edge, and "cutting inside" in which the cut is too close to or even
inside the pattern section edge. A correct cut leaves backing web hairs protruding from the edge.
It is. of course, possible to inspect the edges manually, but this is laborious and time-consuming and prone to error. It is desirable to automate the inspection process not only to provide a quality inspection facility but also to operate in line with the cutting operation for cutting error correction.
The laser cutting technique is capable to operating at high speeds (ImS 1) and for in line operation, an inspection method must be capable of operating at the same speeds. This invention provides a method which can be implemented to operate at such speeds but which, at any speed, gives a reliable indication of correct cutting, excess cutting and cutting inside.
The invention comprises in one aspect a method for tracking lateral shift of a moving Hne feature comprising :
a) forming successive linescan camera images of a field extending transversely of and including said line feature
b) identifying a pixel, by scanning from one end ofthe image, possibly representative of said line feature but possibly due to noise
c) comparing the position of the identified pixel with the position of the identified pixel on a previous image
d) rejecting the pixel as noise if it is not within a tolerance range based on the previous image
On rejecting the pixel, the pixel representative of said line feature is selected as a border pixel of he tolerance range. The border pixel selected will usually be the one nearest the identified (rejected) pixel.
The position ofthe line feature may be computed as the moving average of a plurality of successive images.
The method may be used to track an edge of sheet material and may be used in a lace cutting operation in which a cutting device (such as a laser beam) is controlled with reference to a machine vision guidance system, the method supplying edge position information to the guidance system to facilitate its orientation with respect to the lace.
The invention comprises in another aspect a method for directing a radiation beam to effect a power action over a surface comprising focusing a low power density beam deflected by a moving mirror arrangement to a high power density spot at the surface using a beam focusing device which adapts for the optical path difference between different points on the surface as the beam is deflected.
The low power density beam may comprise an expanded high power density beam. The radiation beam may be a laser beam which is expanded into the low power density beam - of course, the invention is not restricted to laser beams, but in general they are preferred because they can be both powerful and highly concentrated.
The beam focusing device may comprise a lens, which may be moved axially ofthe beam to adapt for said path difference. The lens may be moved by a servo motor.
A primary adaptation may be effected according to an optical path difference between a datum beam position and the actual beam position. A secondary
adaptation may be effected in accordance with the system errors. The secondary adaptation may be effected by superimposing a focus correction factor evaluated for each of a plurality of discrete spot positions by means of a focus look-up table of focus correction factors produced by measurement of the actual beam focus at the surface against the desired beam focus.
The method is of particular advantage in connection with a moving mirror arrangement comprising two mirrors rotatable on orthogonal axes. In any event, a one or two mirror system may have the mirror or both mirrors moved by servo motors. Mirror position may be primarily controlled according to a geometrical algorithm relating the mirror deflection to the spot position, and may be secondarily controlled in accordance with system errors, for example by superimposing a position correction factor evaluated for each of a plurality of discrete spot positions by means ofa position look-up table of position correction factors produced by measurement ofthe actual spot position at the surface against the desired spot position.
The mirror control arrangement may be used independently of the focus control arrangement. The secondary correction in each case might in some circumstances also be used as a primary control arrangement, and is if particular advantage inasmuch as it can effectively and inexpensively compensate for imprecision in the mechanical construction ofthe mirror, lens and other parts ofthe system reducing the cost ofthe mechanical arrangements considerably.
The focus control in particular, however, is of greater significance because the method may be adapted for cutting a thick workpiece, such for example as a thick textile fabric or multiple layers of fabric, wood panelling, plastic and composite sheeting and so forth, in which the surface at which the beam is focused to a high density spot is a notional surface which moves through the workpiece from the surface first encountered by the spot as workpiece material is removed from that first-encountered surface down
through the workpiece by the power action ofthe spot. A first pass ofthe beam skims of a shallow surface layer, then the focus is adjusted for a second pass which removes more and so on until the desired depth is reached or the workpiece is cut through.
The method may be adapted for cutting a workpiece moving relatively to the mirror arrangement.
The invention also comprises apparatus for effecting a power action over a surface comprising a radiation beam generator producing a lower power density beams and a moving mirror arrangement deflecting said beam over the surface and a beam focusing device focusing said beam to a spot at the surface and means adapting the said device for the optical path difference between different points on the surface as the beam is deflected.
The beam generator may comprise a laser and a beam expander, and the beam focusing device may comprise a lens, which may be movable axially ofthe beam, for example on a linear slide, to adapt for the path difference. The lens may be moved by a servo motor.
The moving mirror arrangement may comprise two mirrors on orthogonal axes, which may also be moved by servo motor means, as, of course, may be a single mirror arrangement.
The apparatus may also comprise a control arrangement controlling the moving mirror arrangement and the beam focusing device.
The invention in another aspect comprises a method for inspecting cut edges of lace, which comprises solid pattern sections on a backing web, for "excess cutting" and "cutting inside" errors, comprising imaging the web edge to form an image
comprising pattern section, backing web. backing web hair and background image components, examining the image to identify the edge, examining the image a hairs* breadth inside the edge to ascertain if it is background or not and if not measuring the width of the non-background image section along the edge to determine if it is pattern section or hair.
The imaging may be effected by a line scan camera which builds up and updates a two-dimensional image.
The resolution of the image is desirably a fraction of the thickness of the backing web hair, which, typically is 0.3mm, when the resolution is 0.04 mm. This is achieved using a 1024 pixel element line scan camera imaging a field of 40 mm.
A binarised image matrix can be formed first (by thresholding and binarisation) having background and non-background pixels - backlighting is suitable. The matrix is (with the 1024 pixel element camera) 1024 elements across and, say, sixteen elements along, the rows and columns having comparable if not equal resolution.
Methods and apparatus according to the invention will now be described with reference to the accompanying drawings, in which:
Figure 1 is a diagrammatic illustration of a lace transport system with the edge tracker system installed;
Figure 2 is a section of a pixel image formed by the method of the invention;
Figure 3 is a diagrammatic illustration of one embodiment of control apparatus;
Figure 4 is a graphical representation of optical path difference to be compensated for in the apparatus of Figure 3:
Figure 5 is a diagrammatic illustration of a control arrangement for the apparatus of Figure 3;
Figure 6 is a diagrammatic representation of an edge of lace showing correct cut and cutting errors:
Figure 7 is a diagrammatic illustration of one embodiment of apparatus for the inspection;
Figure 8 is an image matrix created by the apparatus; and
Figure 9 is a lace matrix created from the image matrix of Figure 8.
Figures 1 and 2 ofthe drawings illustrate a method for tracking lateral shift ofa moving line feature comprising the edge 1 1 of a piece of bulk lace 12 trained over rollers 13 to move it past a cutting station, not shown, at which a machine vision guidance system directs a laser beam to cut the lace along a desired scalloped path. Such a system is described in WO 93/03301.
A linescan camera 14 images a section ofthe lace 12 over a traverse extent including the edge 1 1 , backlit by a light source 15.
Signals from the camera 14 are fed to a data processing system 16 in which images formed on a single scan are binarised and analysed and which controls the linescan camera 14 to read out a line of image data at predetermined time intervals. The
data from each line are processed before the next line is read, resulting in a low-cost PC- based data processing system being adequate.
Figure 2 illustrates a section 21 of the linescan image, showing also the corresponding section 22 ofthe previous image. After binarisation, the pixels are either white or black, the black in this case representing lace, the white, background. The previous image clearly shows the edge 23 ofthe lace as the transition from white to black pixels. The current image merely shows possible positions for the first-encountered black pixel.
Either side of the edge 23 is defined a tolerance range 24 extending five pixels either side. If the first-encountered black pixel, counting in from the left-hand side, is in the range 24 it will be accepted as a new edge position. If it is outside the range 24. it will be rejected as noise.
If the detected pixel is to the left ofthe range 24, as, for example at position X, the new edge position will be taken to be the left hand boundary BL ofthe range, if to the right, the right, as at Y. hand boundary BR.
This philosophy greatly simplifies the processing and reduces processing time so that successive images can be taken and analysed in short time periods - processing must be complete, clearly, before a new image is taken for analysis.
Noise effects can be further reduced - as can spurious edge effects due to imperfections in the edge itself- by defining the edge as the position ofthe mean edge offset taken over a plurality of images.
With a 1024 pixel element linescan camera with a line rate of lKHz it is convenient, for the lace cutting application at least, to use 1024 line scans as the number
over which the mean offset is calculated, giving an information updating time of 1.024 s. By defining the mean as the integer part, processing time is further reduced as compared to using floating point arithmetic.
Figures 4. 5 and 6 of the drawings illustrate a method and apparatus for directing a radiation beam - a laser beam 311 - to effect a power action, such as a cutting action, over a surface 312.
A low power density beam 31 1a produced by a laser 313 and beam expander 314 is deflected by a moving mirror arrangement 315 over the surface 312. The beam 31 la is focused to a high density spot 31 lb at the surface 312 using a beam focusing device - a lens 316 - which adapts for the optical path difference points on the surface 312 as the beam 311 is deflected.
The optical path difference (OPD) is shown graphically in Figure 4, where the curved surface which is the locus of the OPD at each point (x,y) of the plane of surface 312 is, at least to a first approximation, the surface at which the beam 31 1 is focused in the absence of any adaptation by the lens 316. The OPD is, of course, zero at the (x,y) coordinate point directly beneath the mirror arrangement 315.
To effect the adaptation, the lens 316 is moved axially ofthe beam 311a, which emerges from the beam expander 314 as a collimated beam of sufficiently low power density not to damage the mirrors ofthe arrangement 315. Essentially, if the lens is moved towards the mirror arrangement 315 by an amount equal to the OPD. the beam will be focused at the surface 312, as desired.
Movement of the lens 316, as also of the mirrors 15X, 15Y of the arrangement 315, is effected by a dc motor in a servo loop, as shown in Figure 5. where a CAD/CAM system operating in a computer 331 drives motors 332L, 332X and 332Y
through motion controller circuitry 333L. 333X. 333Y (for example Hewlett-Packard HCTL-1 100 chips). The servo control loops are closed by high-resolution incremental encoders 334.
Primary adaptation is effected in accordance with the calculated OPD as graphically represented in Figure 4.
The deflection angles θ and θ, ofthe mirrors are given by Equations 1 and
2.
θ - - tan 1 (2) y 2
where θx- θy = mirror deflection angles [radians] x,y = Cartesian co-ordinates on the cutting plane [mm] h = stand off height (distance between cutting plane and Y-mirror axis [mm] d = distance between X- and Y- mirrors axis [mm]
The rotary nature of the beam movement across a flat cutting plane introduces beam path length difference OPD which is given by Equation 3,
OPL (x,y) = d (h + d) (3)
The optical path difference versus the Carterian co-ordinates (x,y) within a cutting area of 500 mm square is shown in Figure 4. The stand off height is 600 mm. the distance between the mirror axes 25 mm. The maximal path difference, as not unexpected in the corners ofthe square cutting area, is 95 mm.
The cutting pattern data are extracted from a file generated by a CAD system or in real time by a map following system such as is described in WO 93/03301 for a relatively moving workpiece such as the web of lace referred to in that publication. The pattern is given as a series of points in X,Y coordinates, and the data file also contains information for switching the laser on and off, which is appropriate, for example, to the cutting out of discrete shapes.
The given points ofthe cutting perimeter are traced in succession, the laser beam 31 1 being tracked with constant velocity along the desired cutting path.
The control of the constant cutting velocity is achieved indirectly by position control. The command position of each axis is updated every 2 ms so that the tangential velocity along the cutting path is kept constant. The constant speed s, the increment for each axis Δx and Ay between two updating ofthe command positions and the updated frequency , are related in Equation 4.
The increments Ax and Ay are linked by the slope of the straight line between two given data points as
Ay_ k = y y.
(5) Ax
where Xj. v, and x;. y2 are the co-ordinates of two consecutive data points. The number of interpolations m between two given points is found as x - x m = -i— i (6)
Δx
In the case that (x, - x,) < (y2 - v,) the axes are exchanged and the Y-axis is the major axis. The calculated values Δx. Δ and m are stored to respective arrays in addition to the original cutting path point co-ordinates this providing a last command positions generation during cutting. The Cartesian co-ordinates (x,y) need to be transformed to (θx- θ OPD) which are finally written to the command position register ofthe respective motion controller.
Since small imprecisions in the mechanical arrangements can cause large spot positioning errors, secondary adaptation may be required for accurate cutting, and this may be conveniently and inexpensively implemented in software.
The cutting area is divided up by a square mesh when the arrangement is first set up, and the beam 311 aimed at each mesh point. The actual position at which the beam 311 hits the surface 312 will in general, because ofthe mechanical imprecision, be different from the aimed-at-position.
The position error is recorded and a look-up table is created of correction factors which is used to perform the coordinate transformation to ensure that the laser hits the right spot. In one arrangement, a positioning error of 15 mm was reduced by the secondary adaptation to 1.5 mm.
The arrangement may be further modified for the cutting of thick workpieces by several passes ofthe beam around a cutting perimeter, the focus being adjusted for a notional "surface" lowered after each pass by the depth of material
removed. For not-too-thick work pieces in principle all that is required is an indexing ofthe base position of the lens 316 by a distance equal to the depth of material removed at each pass.
Figures 6 to 9 of the drawings illustrate a method and apparatus for inspecting cut edges of lace, typically as shown in Figure 6. The lace 61 1 comprises solid pattern sections 612 on a backing web 613. The lace has been cut along a cutting line 614 (dots) to separate it from other breadths of lace made in the same piece. lhe cutting may have been effected conventionally by a mechanical cutter (or even, in the case of highly convoluted edge, manually by scissors) or by a machine vision controlled laser cutting device as disclosed in WO 93/03301.
The cutting line 614 departs from the ideal cutting line 615 (broken line) in places. At 616 is shown an area of backing web 613 beyond the edge ofthe solid pattern section 612. At 617 the cutting line 614 has gone inside the ideal line 615 and has shaved off part ofthe solid pattern area.
Where the cutting line 614 coincides with the ideal line 615, backing web hairs 618 protrude from the edge ofthe solid sections 612. The method and apparatus ofthe invention essentially determine the presence of these backing web hairs 618 and the absence of backing web to establish a correct cut.
The apparatus is shown in Figure 7 and comprises a line scan camera 621 (e.g. a Fairchild CAM 1350 linescan camera with 1024 pixel elements) situated above a moving lace web so as to be able to image a field of 40 mm to include the cut edge to be inspected. The camera 621 is connected through an interface board 622 to a computer 623 connected to a PC 624, the interface board 622 and computer 623 condition the camera data for parallel transfer to the PC which stores the image on its hard disc for eventual statistical analysis.
The image matrix of Figure 8 is a binary representation of an image of a short section of lace edge, the field being 40 mm (1024 pixels) wide and 16 pixels long (0.64 mm). The data have been binarised so background is represented by "0" and anything which is not background by "1 ".
As each new line is added, as the lace 61 1 moves past the camera 621, the software first locates the edge. i.e. the first "1" encountered starting from the left - if the opposite edge were being inspected, starting from the right, although the software could simply reverse the right hand edge image to leave everything uniform thereafter.
The scanning continues along the new matrix line. A hair is taken to be of maximum width 0.3 mm or approximately eight pixels. So if, after, say, ten pixels, the next pixels contain "0"s, an assumption is made that the feature is a hair extending in the direction of lace travel.
The width ofthe edge ofthe web is somewhat greater at 0.6 - 1.2 mm. If the "l"s pixel count in from the first-encountered "1 " is 16 or greater, a first assumption is made that this is the actual lace edge.
This first "edge" assumption is checked using the two-dimensionality ofthe matrix, counting the number of lines for which the "l"s feature is present. If it persists for no more than a few lines (e.g. up to eight) the "edge" assumption is changed to a "hair" identification.
The image data are then transformed by assigning "B" (background) to all *"0"s. "H" (hair) to all hairs. "L" to all lace pixels and - since once hair and lace edge have been identified, it is pointless to probe further into the lace - "U" to pixels not evaluated (for "unknown"). This generates the lace matrix of Figure 9.
To save unnecessary processing steps, the position ofthe edge in the last line is stored and used as the start point for the search ofthe next line, and the data per line is reduced from 1024 pixels to just two integers of interest, namely "hair begin" and "lace begin".
The "lace begin" data can be regarded as an "ideal" cutting path, but an offset from this is required since the laser is not intended to cut the lace per se.
A desired cutting path can be evaluated from the information, however, and compared with the actual cutting path to provide a statistic related to the quality of the cut with an indication of the relative prevalence of errors of one type over errors of another. In real time, immediately after the cutting operation, the information can be fed back to the beam controller to correct any error tendency.
Of particular interest is the low cost of the hardware required.
Any one aspect of the invention as described with respect to the drawings may be used in conjunction with any ofthe others. Taken together, they provide a low capital cost, highly efficient, high speed cutting mechanism for such difficult products as lace and checking the result of the cut.