Method and system for the automatic detection of characteristic points of a herringbone fabric with regard to the automatic cutting of parts

Description

Technical Field

[0001] The invention relates to the general field of automatic cutting of pieces in a herringbone fabric. It relates more particularly to a method and a system for automatically detecting certain characteristic points of the herringbone patterns of a fabric.

[0002] Fields of application of the invention include the clothing and furniture industries. Prior art

[0003] When making clothing or furnishing items involves assembling pieces cut from a fabric, there are special constraints if the fabric is patterned. "Patterned fabric" here means any flexible textile material in sheet form printed with a pattern that repeats with regular and predetermined steps.

[0004] In this case, it is then desirable, or even necessary, to respect the continuity of the pattern between two assembled pieces, for example two parts of a garment sewn together, or two pieces intended to be adjacent, for example two parts of a garment located side by side when the garment is worn, or two cushions of a sofa placed side by side.

[0005] In order to respect these constraints, it is known to associate absolute or relative position markers with the parts and to establish a hierarchy between main parts and secondary parts.

[0006] An absolute position reference is normally associated with a main piece. It characterizes the absolute positioning of the main piece relative to the fabric pattern. The position of a piece relative to the pattern is characterized by the fact that a given point on the surface of the piece occupies a determined relative position with respect to the patterns surrounding it. Thus, pieces whose locations on the fabric surface are deduced from each other by translations of an integer number of pattern steps occupy the same position relative to the pattern.

[0007] Relative position markers are associated with two parts to be assembled, taking into account requirements related to the existence of the pattern. They identify the locations of two connection points that must be brought into correspondence when assembling the parts.

[0008] For example, in the case of a jacket, a back piece can be a main piece. An absolute position marker is possibly associated with the back piece, for example, when we want a complete pattern to be visible in a particular location on this piece. A sleeve, the neckline, a front then constitute secondary pieces. For each of these, the location of a connection point is determined to correspond to the location of the associated connection point on the main piece.

[0009] Additionally, a piece with an associated relative positional reference can also be the main piece of one or more other pieces. In this case, we speak of a chain of links. Similarly, a piece can have an absolute, relative, or no positional reference in the weft axis, and another type of positional reference (relative, absolute, or no reference) for the warp axis of the fabric.

[0010] Furthermore, it is known to carry out fabric cutting automatically. Automatic cutting installations have been marketed by the applicant for many years.

[0011] Typically, an automatic cutting process includes a placement operation that consists of optimally determining the positions of the pieces to be cut in a strip of fabric. The placement is chosen so as to minimize fabric waste while respecting certain constraints: respect for the straight grain, sufficient minimum margin between pieces, etc. In the case of a patterned fabric, there are additional constraints related to respecting the locations of the absolute position and relative position markers. Systems allowing an operator to define placements using computer workstations and specialized software are known, including in the case of patterned fabrics.

[0012] To perform the cutting, the fabric is spread on a cutting table in one or more overlapping layers that can be held by suction through the table. The cutting is carried out by means of a tool carried by a head whose movements, relative to the cutting table, are controlled according to the predetermined placement. The cutting can be carried out by vibrating blade, rotary blade, laser, water jet, etc.

[0013] Difficulties arise when the fabric used is a patterned fabric. In particular, the problem arises in practice of the non-coincidence between the fabric model used for placement and the fabric actually spread out on the cutting table. This non-coincidence is expressed in particular as follows. If one places oneself on the cutting table at the coordinates of a reference point of a part of the placement, one notes that the corresponding point on the spread out fabric does not always occupy the desired relative position with respect to the pattern of the actual fabric. These deviations are more or less large and in practice unavoidable. They are due to printing defects, defects in laying the fabric on the cutting table, variations in the density of the loom threads and / or deformations of the fabric which can result in irregularities in the repetition pitch and in the geometry of the pattern.As a result, the pre-established placement, or theoretical placement, must be modified to correspond to the reality of the spread fabric.

[0014] A method for carrying out this placement modification automatically is described in document EP 0,759,708 filed in the name of the applicant. After spreading the patterned fabric on the cutting table, this method provides for detecting a possible discrepancy between the actual pitch of the pattern on the fabric and the theoretical pitch of the latter by capturing images of parts of the spread fabric and then checking on the captured images that locations corresponding to stored information occupy desired positions relative to the actual pattern of the spread fabric. If necessary, the theoretical placement of the parts is modified according to the result of the check in order to adapt it to the actual pitch of the pattern on the spread fabric taking into account the actual characteristics of the fabric.

[0015] The automatic cutting process is therefore based on the definition of absolute or relative position markers for the pieces to be cut, and on the detection of patterns on the fabric spread out on the cutting table in order to adjust the positioning and geometry of the pieces to be cut where necessary.

[0016] The methods known from the prior art for automatically modifying the placement and geometry of the pieces to be cut are perfectly suited to striped fabrics, which may be similar to herringbone fabrics.

[0017] Herringbone fabrics have weaves obtained by replicating, after inversion, twill weaves or twill derivatives that give a zigzag or sawtooth effect. These offset herringbone patterns are achieved with weft and warp yarns of different colors and an offset twill weave.

[0018] On such herringbone fabrics, the algorithms currently on the market do not allow automatic recognition of the different characteristic points of the chevrons to digitally construct a plurality of lines passing through the points of the chevrons. Also, to cut such fabrics while respecting the placement constraints, the operator of the cutting machine must manually indicate on the image the position of different characteristic points of the chevrons. However, this manual location operation is long, tiring for the operator and strongly impacts the productivity of the cutting operation. Statement of the invention

[0019] The present invention therefore aims to propose a method for detecting characteristic points of a herringbone fabric for the automatic cutting of pieces which does not have the aforementioned drawbacks.

[0020] According to the invention, this aim is achieved by means of a method for automatically detecting characteristic points of a herringbone fabric for the automatic cutting of pieces, according to claim 1.

[0021] The method according to the invention is remarkable in that it allows, automatically, to detect the position of the lines passing through the tips of the chevrons by optimizing a symmetry criterion of two mirror sub-images acquired along predefined operation lines. This automatic detection results in a considerable saving of time on the cutting process. The productivity of the cutting operation is greatly increased.

[0022] The detection initialization step comprises the determination by an operator of an initial point of coordinates corresponding to the first theoretical crossing point on the first operation line, the determination by the operator or from predefined parameters of the theoretical spacing between two chevron axes, the definition from the initial point of the operation lines perpendicular to the axes of the chevrons, the operation lines being spaced from each other by a step defined by the operator or from predefined parameters, and the creation of an analysis window, initially centered on the initial point, the dimensions of which are determined by the operator or from predefined parameters.

[0023] The set of chevron axis crossing points forms a deformed grid which is used to reposition and / or deform the pieces to be cut.

[0024] In this case, the step of determining the coordinates of the crossing points of the axes of the chevrons advantageously comprises the iterative displacement of the analysis window inside the image along each of the operation lines.

[0025] Still in this case, the step of determining the coordinates of the chevron axis crossing points preferably comprises the calculation of correlation coefficients between the two mirror sub-images entered in the analysis window. The step of determining the coordinates of the chevron axis crossing points may further comprise the recording of the coordinates of the geometric center of the analysis window when the correlation coefficient is maximized in absolute value.

[0026] The analysis window can have a width corresponding to approximately 40% of the gap between two adjacent chevron axes.

[0027] The step of determining the coordinates of the chevron axis crossing points may include calculating the coordinates of the theoretical crossing points along the lines of operation.

[0028] Preferably, in the case of vertical rafters, the operating lines are horizontal, and in the case of horizontal rafters, the operating lines are vertical.

[0029] The invention also relates to a system for the automatic detection of characteristic points of a herringbone fabric for the automatic cutting of pieces, according to claim 7. Brief description of the drawings

[0030] [ Fig. 1A-1B ] THE figures 1A et 1B show examples of herringbone fabrics to which the invention applies. Fig. 2 ] There figure 2 represents an example of an image of a portion of herringbone fabric after application of the method according to the invention. Fig. 3 ] There figure 3 represents an example of initialization of the detection for the implementation of the method of the invention. Fig. 4A-4B ] THE figures 4A et 4B show examples of determination for the implementation of the method of the invention. Fig. 5A-5D ] THE figures 5A à 5D show different examples of implementation of the optimization step of a symmetry criterion. Description of the embodiments

[0031] The invention relates to the automatic detection of characteristic points of a zigzag or sawtooth herringbone fabric, such as the TA and TB fabrics shown in the figures 1A et 1B .

[0032] As is well known, herringbone fabrics are weaving patterns used in weaving and obtained by reproducing, after inversion, twill weaves or twill derivatives.

[0033] More specifically, the chevrons of these TA, TB fabrics are formed by V-shaped patterns with P points that are aligned along a plurality of parallel axes. In the case of vertical chevrons ( figure 1A ), the axes of the rafters Kj passing through the points P of the rafters are vertical axes. In the case of horizontal rafters ( figure 1B ), the axes of the rafters Ki passing through the points P of the rafters are horizontal axes.

[0034] On the example of figures 1A et 1B , the chevrons thus form a succession of parallel bands of different colors and of the same widths which are symmetrical with respect to the axes of the chevrons Kj, Ki, respectively. Of course, the invention applies to other forms of V-shaped patterns, in particular irregular patterns and / or inverse symmetry as detailed later in connection with the figures 5A à 5D .

[0035] The method according to the invention aims to recognize the presence of chevrons in the image of a portion of a herringbone fabric (as described in connection with the figures 1A et 1B ) spread out on a cutting table.

[0036] In particular, as shown in the figure 2 , the method according to the invention aims to determine the position of the intersections of the axes Kj (here vertical) of alignment of the points of the V-shaped patterns of the fabric with a plurality of predefined operating lines Li.

[0037] Operation lines are lines that are predefined by the operator to be perpendicular to the rafter axes, parallel to each other, and spaced (preferably at a regular pitch) from each other.

[0038] So, in the case of the figure 2 which shows a vertical herringbone fabric, the axes Kj of these herringbones are substantially vertical lines and the operation lines Li are horizontal lines.

[0039] Of course, in the case of horizontal chevron detection, the axes of the latter are horizontal lines and the operation lines are defined to be vertical lines.

[0040] This recognition operation according to the invention makes it possible to model a grid reflecting distortions due mainly to the imperfect unrolling of the fabric width on the cutting table, this deformed grid being used to reposition and / or deform the pieces to be cut.

[0041] To this end, the method according to the invention provides the following successive main steps: 1 / A step of acquiring an image of a portion of the fabric: This step is carried out in a manner known per se by spreading the fabric on the cutting table to take an image of a portion of it using an image sensor. 2 / A step of initializing the detection: This step consists of acquiring, from the image acquired previously or from parameters predefined by the operator, geometric parameters of the chevrons, as well as defining the lines of operation. This step is carried out by the operator using a computer embedded in the cutting machine or separate from it. The parameters can be determined at the last moment on the cutting machine or be defined beforehand. 3 / A step of determining the crossing points of the axes of the chevrons along the lines of operation: This step consists of an automatic algorithm for optimizing a symmetry criterion of two mirror sub-images acquired along the lines of operation.For example, it is carried out at the level of the calculator used in the previous step. It allows you to obtain the coordinates of the points of the chevrons in the fabric which constitute characteristic points of the patterned fabric.

[0042] In connection with the figure 3 , we will now describe an example of implementation of the detection initialization step according to the method of the invention.

[0043] This step involves the operator defining a number of parameters relating to the fabric chevrons, namely: the direction of the chevrons which could be vertical (case of the figure 1A ) or horizontal (case of the figure 1B ) an initial point I with coordinates (xI, yI) corresponding to the first theoretical crossing point on the first operating line that the operator will locate approximately on the image (the position of the exact crossing point will be determined by a search algorithm described later) the horizontal pitch dX corresponding, in the case of vertical chevrons to the theoretical horizontal spacing between two chevron axes Kj, and in the case of horizontal chevrons, to the distance between two vertical operating lines Lj the vertical pitch dY corresponding, in the case of horizontal chevrons to the theoretical vertical spacing between two chevron axes Ki, and in the case of vertical chevrons,the distance between two horizontal operating lines Li the maximum angle a of inclination of the chevrons corresponding to the maximum angle deviation authorized with respect to the vertical for vertical chevrons (respectively to the horizontal for horizontal chevrons) the width r of an analysis window F of the image: this value, which depends on the noise level of the acquired image, could be of the order of 40% of the difference between two adjacent chevron axes but could be reduced to improve the calculation times for low noise images,

[0044] This initialization step may advantageously include the verification of a coherent detection of the vertical (respectively horizontal) chevrons by ensuring that the vertical step dX (respectively the horizontal step dY) verifies the following relation: (dY x tan(a) + r) is strictly less than dX / 2 for vertical chevrons, and (dX x tan(a) + r) is strictly less than dY / 2 for horizontal chevrons.

[0045] In connection with the figures 4A et 4B , we will now describe an example of an algorithm for implementing the step of determining the crossing points in accordance with the method according to the invention for vertical rafters ( figure 4A ) and for horizontal rafters ( figure 4B ).

[0046] For vertical rafters ( figure 4A ), the operation lines Li are horizontal lines with regular spacing dY and numbered for example from top to bottom of the image, the first operation line L0 (i=0) being chosen to pass through the initial point I previously defined and with coordinates (xI, yI).

[0047] The chevron axes Kj are vertically trending lines that are numbered, for example from left to right of the image, the first chevron axis being numbered K0 (j=0). The theoretical crossing points N of the j-th chevron axis with the i-th operation line are the coordinate points xA(i,j), yA(i, j) with i varying from 0 to the number of operation lines.

[0048] Likewise, for horizontal rafters ( figure 4B ), the operation lines Lj are vertical lines with regular spacing dX and numbered for example from left to right of the image, the first operation line L0 (j=0) being chosen to pass through the initial point I previously defined and with coordinates (xI, yI).

[0049] The chevron axes Ki are horizontally trending lines that are numbered, for example from top to bottom of the image, the first chevron axis being numbered K0 (i=0). The theoretical crossing points N' of the i-th chevron axis with the j-th operation line are the coordinate points xA(i,j), yA(i, j) with j varying from 0 to the number of operation lines.

[0050] The step of determining the method according to the invention consists of iteratively moving the analysis window F previously defined along the different operation lines Li, Lj inside the image starting from the predefined initial point I.

[0051] The analysis window is moved at each iteration by a distance corresponding to the vertical step dX (for vertical chevrons) and the horizontal step dY (for horizontal chevrons).

[0052] While moving the analysis window, the algorithm calculates the coordinates of the crossing points xA(i,j), yA(i, j) of the chevron axes with the operation lines.

[0053] For this purpose, at each iteration, the algorithm calculates a correlation coefficient between two mirror sub-images inscribed in the analysis window F. When the correlation coefficient thus calculated is maximized in absolute value, the coordinates of the geometric center of the analysis window F correspond to those of the theoretical crossing point N of the axis of the corresponding chevrons with the corresponding operation line.

[0054] In a scanned image, the probability of a vertical chevron axis passing through a passage point will be all the greater if the portions of the image located on either side of the vertical passing through this point are symmetrical to each other (except for the luminance inversion).

[0055] As shown in the figures 5A à 5D , to calculate this probability, the algorithm considers, from the square sub-image M inscribed in the analysis window and centered on a theoretical passage point xA, the two sub-images on the left ML and on the right MR of the axis of symmetry S passing through the theoretical passage point N.

[0056] The algorithm then calculates the correlation coefficient δ between the ML sub-image and the MR sub-image (mirror of ML). The calculation of this correlation coefficient δ is performed from tables containing the luminance series t1, t2 of the two sub-images ML, MR, respectively, and its result is typically given by the following mathematical formula: δ = Cov (t1, t2) / e1 x e2; in which “Cov” denotes the covariance between the series t1 and t2, and e1 and e2 denote the standard deviation of the series t1 and t2, respectively.

[0057] The interpretation of the result obtained is as follows: 1 / When the result obtained is equal to or close to +1 (case of the figure 5A ): this is a symmetry between the two sub-images ML and MR with respect to the S axis. This result therefore indicates the presence of chevrons and the axis of symmetry S is considered as a chevron axis Ki. The coordinates of the geometric center of the analysis window are memorized. 2 / When the result obtained is equal to or close to 0 (positively or negatively - case of figures 5B et 5D ): this is an asymmetry between the two sub-images ML and MR with respect to the S axis. This result indicates that the axis of symmetry S is not a chevron axis. 3 / When the result obtained is equal to or close to -1 (case of the figure 5C ): this is a symmetry with inversion of luminances (i.e. a dark pixel corresponds to a light pixel and vice versa) between the two sub-images ML and MR with respect to the symmetry axis S. This result therefore indicates the presence of chevrons with an inversion of luminance and the symmetry axis S is considered as an axis of chevrons Ki. The coordinates of the geometric center of the analysis window are memorized.

[0058] Thus, the presence of chevrons will be confirmed for a correlation coefficient close to 1 in absolute value. In this situation, the coordinates of the geometric center of the analysis window will be systematically memorized because they correspond to the position of chevron points on the fabric.

[0059] Note that the analysis window can be moved along an operation line from pixel to pixel. In this case, for each pixel, the algorithm calculates the correlation coefficient of the two sub-images on either side of the symmetry axis, which gives a curve of the correlation coefficient as a function of the position along the operation line. This curve has local maxima whenever the analysis window is centered on a chevron axis. The coordinates of the geometric center of the analysis window when these maxima are reached are systematically memorized.

[0060] It will also be noted that the set of crossing points N, N' of the axes of the chevrons forms a grid as represented on the figure 2which is deformed relative to a theoretical grid which would have been developed from a regular pitch between the rafter points and a regular alignment of the rafter points along parallel lines. This deformed grid is thus used to reposition and / or deform the pieces to be cut.

Claims

1. A method of automatically detecting characteristic points of a herringbone pattern fabric (T-A; T-B) for automatically cutting pieces, the herringbone pattern being formed by V-shaped patterns with apexes (P) which are aligned along a plurality of parallel axes, the method comprising: - a step of acquiring an image of a portion of the fabric; - a detection initialization step comprising: ∘ the acquisition, from parameters predefined by the operator, of geometric parameters of the herringbones comprising the vertical or horizontal direction of the herringbones, an initial point (I) of coordinates (xI, yI) corresponding to the first theoretical crossing point on the first line of operation, and a theoretical spacing (dX, dY) between two herringbone axes (Ki, Kj), and ∘ the definition of operating lines (Li, Lj) perpendicular to the herringbone axes (Ki, Kj) and spaced from each other by an operator-defined pitch or from predefined parameters; and ∘ the creation of an analysis window (F), initially centered on the initial point (I), the dimensions of which are determined by the operator or from predefined parameters; - a step of determining in the image the coordinates of the crossing points (N, N') of herringbone axes along the lines of operation by iteratively displacing the analysis window (F) inside the image along each of the lines of operation and optimizing a symmetry criterion of two mirror sub-images (MR, ML) acquired along the lines of operation, the set of crossing points (N, N') of the herringbone axes forming a deformed grid which is used to reposition and / or deform the parts to cut.

2. The method according to claim 1, wherein the step of determining the coordinates of the crossing points of the herringbone axes comprises the calculation of correlation coefficients between the two mirror sub-images inscribed in the analysis window.

3. The method according to claim 2, wherein the step of determining the coordinates of the crossing points of the herringbone axes further comprises the saving of coordinates of the geometric center of the analysis window when the correlation coefficient is maximized in absolute value.

4. The method according to any of claims 1 to 3, wherein the analysis window (F) has a width (r) corresponding to about 40% of the distance between two adjacent herringbone axes.

5. The method according to any of claims 1 to 4, wherein the step of determining the coordinates of the crossing points of the herringbone axes comprises the calculation of the coordinates of theoretical crossing points along the lines of operation.

6. The method according to any of claims 1 to 5, wherein in the case of vertical herringbones the lines of operation are horizontal, and in the case of horizontal herringbones, the lines of operation are vertical.

7. A system for the automatic detection of characteristic points of a herringbone pattern fabric (T-A; T-B) for the automatic cutting of pieces, the herringbones being formed by V-shaped patterns with apexes (P) which are aligned along a plurality of parallel axes, the system comprising: - means of acquiring an image of a portion of the fabric; - means of initializing the detection comprising, from parameters predefined by the operator, the acquisition of geometric parameters of the herringbones comprising the vertical or horizontal direction of the herringbones, an initial point (I) of coordinates (xl,yl), corresponding to the first theoretical crossing point on the first line of operation, and a theoretical spacing (dX,dY) between two herringbone axes (Ki,Kj), and the definition of lines of operation (Li,Lj) perpendicular to the herringbones axes (Ki,Kj) and spaced from each other by a pitch defined by the operator of from predefined parameters; and the creation of an analysis window (F), initially centered on the initial point (I), the dimensions of which are determined by the operator or from predefined parameters; - means of determining in the image, the coordinates of crossing points (N,N') of the herringbone axes along the lines of operation by iteratively moving the analysis window (F) within the image along each of the lines of operation and optimizing a symmetry criterion of two mirror sub-images (MR,ML) acquired along the lines of operation, all crossing points (N,N') of the herringbone axes forming a distorted grid which is used to cut / reposition the parts.