JP2011107348A - Mark recognition device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a mark recognition device which can calculate an accurate reference point of an alignment mark even when the alignment mark includes a set of a plurality of marks. <P>SOLUTION: A substrate having an alignment mark formed thereon is photographed to obtain a photographed image. Each individual mark is detected from the photographed image by template matching. Then a position of a temporary gravity center is determined by voting using a voting space. A detected mark M contributing to the determination of the temporary gravity center is extracted. A mark of a model corresponding to each mark of the detected mark group is extracted. The gravity center of the detected mark group is aligned to the gravity center of the extracted model mark group. Then optimum values of a rotation angle t and a magnification reduction coefficient k of the model mark group are calculated by the method of least squares. Then the position coordinates of a gravity center G of the model mark group after adjusting the posture are calculated as the position coordinates of the true gravity center. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a mark recognition apparatus.

  Conventionally, various exposure apparatuses using photolithography technology have been proposed as apparatuses for drawing a predetermined wiring pattern on a printed wiring board. As an exposure apparatus as described above, for example, a light beam is scanned in a main scanning direction and a sub scanning direction on a substrate coated with a photoresist, and the light beam is modulated based on image data representing a wiring pattern. Thus, there has been proposed a digital exposure apparatus that draws a wiring pattern.

  The wiring pattern of a printed wiring board drawn by a digital exposure apparatus tends to have higher definition. The wiring pattern is drawn at a preset position with respect to the substrate. However, in the case of forming a multilayer printed wiring board, if the wiring pattern of each layer is drawn at a preset position, the drawing position of the wiring pattern of each layer may be shifted. For this reason, it is necessary to perform alignment (alignment) of the wiring pattern of each layer with high accuracy.

  High-precision alignment is necessary not only for digital exposure apparatuses but also for analog exposure apparatuses. In order to perform alignment with high accuracy, an alignment mark is provided on the substrate of each layer based on the design value, and the position of the alignment mark is detected at the time of drawing exposure. The detected position information (read value) and the design value Various exposure apparatuses have been proposed in which a correction amount is obtained based on the above, and the drawing position of the wiring pattern is corrected according to the obtained correction amount.

  The alignment marks include “through holes” such as through holes and “non-through holes (bottomed holes opened in the drawing surface)” such as laser vias. The through hole is formed by mechanically drilling the substrate, such as drilling. On the other hand, since the laser via is formed by drilling or the like by laser processing, the diameter can be reduced to about 100 μm to 200 μm. However, the smaller the alignment mark, the more susceptible to noise and the lower the detection accuracy. That is, it becomes difficult to detect stably.

  Patent Document 1 proposes an alignment mark that is configured as an aggregate of a plurality of marks. Since the alignment mark is composed of an assembly of a plurality of marks in this manner, even if a part of the alignment mark is missing, the influence on the whole is reduced. .

  Patent Document 2 proposes that an alignment mark formed on a substrate is constituted by a through hole and a plurality of non-through holes smaller than the through holes. By using two types of alignment marks in this manner, the influence of the configuration of the through-holes and non-through-holes such as defects and formation position deviations is not directly reflected in the alignment.

Japanese Patent Laid-Open No. 2001-22098 JP 2003-57853 A

  According to the techniques described in Patent Document 1 and Patent Document 2, the alignment mark is constituted by an assembly of a plurality of marks, so that even if there is a defect in part, the influence on the whole is reduced. However, there is still a problem that the reference point (center point or center of gravity) detected by the alignment mark is deviated from the true reference point due to missing some marks or misrecognizing noise as a mark. It has not been solved.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to calculate an accurate reference point of an alignment mark even when the alignment mark is composed of an assembly of a plurality of marks. Another object is to provide a mark recognition device.

  In order to achieve the above object, the invention according to the mark recognition apparatus of the present invention comprises a detection means for detecting positions of a predetermined number or more of marks from a plurality of marks attached to an object based on design information, and the detection Estimation means for estimating a temporary center of gravity position of an assembly composed of the predetermined number or more of the marks based on the positions of the predetermined number or more of the marks detected by the means; and the predetermined number or more of the detection means detected by the detection means A first extraction unit that extracts a mark that contributes to the estimation of the provisional center of gravity position from a mark as a detection mark group; and a detection mark group that is extracted from the plurality of marks based on the design information by the first extraction unit. Second extraction means for extracting a mark corresponding to each of the marks as an alignment mark group, and each of the marks in the alignment mark group is the most similar to each of the marks in the detection mark group Reference position calculation means for calculating a reference point determined by each mark of the detection mark group from a mark group in which each of the mark positions of the alignment mark group is changed so as to match. It is characterized by.

  In the mark recognition apparatus, the reference position calculation means calculates a rotation angle and an enlargement / reduction ratio for making each of the marks of the alignment mark group best match each of the marks of the detection mark group by a least square method. And a reference point determined by each of the marks of the detection mark group from the mark group in which the position of the mark of the alignment mark group is changed by the rotation angle and the scaling factor calculated by the first calculation unit. And a second calculating means for calculating.

  In the mark recognition apparatus, the estimation means uses a plurality of voting spaces formed by dividing a predetermined region into a plurality of areas, and is based on the positions of the predetermined number of marks or more detected by the detection means. Voting in a voting area where the center of gravity of the aggregate composed of the predetermined number or more marks is assumed to exist, and the temporary centroid of the aggregate composed of the predetermined number or more You may estimate as a position where a position exists.

  In the mark recognition apparatus, when there are a plurality of voting spaces having the largest number of votes, the estimation unit estimates a position of the voting space that does not overlap with the position of the mark as a position where the temporary centroid position exists. May be.

  In the mark recognizing device, the estimation unit may not be able to estimate a position where the provisional gravity center position exists when there are a plurality of voting spaces having the largest number of votes.

  According to the mark recognition device of the present invention, there is an effect that an accurate reference point of an alignment mark can be calculated even when the alignment mark is constituted by an assembly of a plurality of marks.

1 is a perspective view showing a schematic configuration of an exposure apparatus according to an embodiment of the present invention. It is a perspective view which shows the structure of the scanner of exposure apparatus. (A) is a top view which shows the exposed area | region formed on the exposure surface of a board | substrate, (B) is a top view which shows the arrangement | sequence of the exposure area by each exposure head. It is a block diagram which shows the structure of the electric control system of exposure apparatus. (A) is a top view for demonstrating the alignment mark provided in a board | substrate, (B) is sectional drawing along the dotted line of (A), (C) and (D) are the modifications of an alignment mark. It is a figure which shows an example. It is a figure which shows a mode that the drawing position of a drawing area | region is correct | amended and image data is correct | amended. It is a modification of the alignment mark comprised from the aggregate | assembly of several marks. It is a modification of the alignment mark comprised from the aggregate | assembly of several marks. It is a figure which shows the mode of normalization correlation template matching. (A)-(C) are the schematic diagrams explaining the method of the temporary center-of-gravity determination using voting space. (A) And (B) is a figure which shows the number of votes obtained of the voting space around the voting space where a mark exists. (A) And (B) is a schematic diagram which shows the model of a set mark. It is a figure which shows a display screen when some marks of a set mark are not detected. (A)-(D) are the schematic diagrams explaining the process of calculating | requiring a "true centroid". It is a flowchart which shows the routine of a mark recognition process. (A) And (B) is a figure which shows the grid | lattice-like collective mark from which some marks were missing. (A) And (B) is a figure which shows the picked-up image and voting result when the loss | missing of a big mark arises. (A) And (B) is a figure which shows the voting result with a picked-up image and a condition when the loss | missing of a big mark arises. (A)-(C) is a figure which shows the picked-up image when the loss | missing of a big mark arises, the voting result with conditions, and the voting result with no conditions. It is a figure for demonstrating the parameter used for the least squares method applied to a circumferential set mark. It is a figure which shows the coordinate system which makes the origin the center of gravity of a detection mark group.

  Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings. In this embodiment, an example of an exposure apparatus provided with a mark recognition apparatus according to the present invention will be described.

<Schematic configuration of exposure apparatus>
FIG. 1 is a perspective view showing a schematic configuration of an exposure apparatus according to an embodiment of the present invention. This exposure apparatus is an apparatus that draws and exposes images such as a plurality of wiring patterns on a single substrate, and is an apparatus that draws and exposes wiring patterns of each layer of a multilayer substrate such as a multilayer printed wiring board. Note that in this embodiment, a single-layer substrate can also be a target. The substrate may be various structures such as a filter for a display device and a semiconductor.

  As shown in FIG. 1, the exposure apparatus 10 includes a flat moving stage 14 that holds the substrate 12 by adsorbing the substrate 12 to the surface. Two guides 20 extending along the stage moving direction are installed on the upper surface of the thick plate-like installation table 18 supported by the four legs 16. The moving stage 14 is arranged so that the longitudinal direction thereof faces the stage moving direction, and is supported by the guide 20 so as to be reciprocally movable.

  A U-shaped gate 22 is provided at the center of the installation table 18 so as to straddle the moving path of the moving stage 14. Each end of the U-shaped gate 22 is fixed to both side surfaces of the installation base 18. A scanner 24 is provided on one side of the gate 22 and a plurality of cameras 26 are provided on the other side. The plurality of cameras 26 detect the front and rear ends of the substrate 12 and a plurality of alignment marks (not shown in FIG. 1) provided on the substrate 12. In this example, three cameras 26 are provided. The alignment mark will be described later.

  The scanner 24 and the camera 26 are respectively attached to the gate 22 and fixedly arranged above the moving path of the moving stage 14. The scanner 24 and the camera 26 are connected to a controller (described later) that controls them. FIG. 2 is a perspective view showing the configuration of the scanner of the exposure apparatus. FIG. 3A is a plan view showing an exposed area formed on the exposure surface of the substrate, and FIG. 3B is a plan view showing an arrangement of exposure areas by each exposure head. As shown in FIGS. 2 and 3B, the scanner 24 includes ten exposure heads 30 (30A to 30J) arranged in a substantially matrix of 2 rows and 5 columns.

  Each exposure head 30 is provided with a mirror device (not shown) which is a spatial light modulation element (SLM) that spatially modulates an incident light beam. In this embodiment, a reflective spatial light modulation element typified by a digital micromirror device (DMD: registered trademark) is used as the mirror device. It is also possible to employ other SLMs other than DMD (registered trademark).

  The mirror device is mounted such that the pixel column direction forms a predetermined inclination angle θ with the scanning direction. Accordingly, as shown in FIGS. 3A and 3B, the exposure area 32 by each exposure head 30 is a rectangular area inclined at an inclination angle θ with respect to the scanning direction. Along with the movement of the moving stage 14, a strip-shaped exposed region 34 is formed on the substrate 12 for each exposure head 30.

  The mirror devices provided in each of the exposure heads 30 are on / off controlled in units of micromirrors, and the substrate 12 is exposed to a dot pattern (black / white) corresponding to the micromirrors of the mirror device. The exposure area 32 is formed by two-dimensionally arranged dots. The two-dimensional dot pattern is inclined with respect to the scanning direction, so that dots arranged in the scanning direction pass between dots arranged in the direction intersecting the scanning direction. Can be planned.

<Electrical configuration of exposure apparatus>
Next, the electrical configuration of the exposure apparatus according to the present embodiment will be described.
FIG. 4 is a block diagram showing the configuration of the electric control system of the exposure apparatus.
As shown in FIG. 4, the exposure apparatus 10 receives vector data representing a wiring pattern to be exposed, which is output from a data creation apparatus 40 having a CAM (Computer Aided Manufacturing) station, and receives the vector data as raster data (bits). Raster conversion processing unit 50 for converting to (map data), reference position storage means 52 for storing “setting position information” of the alignment mark provided on the substrate 12, and the position of the alignment mark detected by the camera 26 A drawing position correcting unit 54 that corrects the drawing position based on “detected position information” and the like, and a corrected image obtained by correcting the raster data of the wiring pattern (drawing region) based on the drawing position corrected by the drawing position correcting unit 54 The image data correction means 56 for generating data and the data converted by the image data correction means 56 Was corrected image rendering control unit 58 that drives and controls the exposure head 30 on the basis of the data, and a controller 70 for controlling the entire stage control unit 60, and the exposure apparatus for driving and controlling the moving stage 14.

  In the present embodiment, a mark recognition method for calculating an accurate reference point of an alignment mark mainly composed of an assembly of a plurality of marks will be described. The drawing position correction means 54 includes a mark recognition device 55 that detects the position of the alignment mark formed from the aggregate. The mark recognition device 55 detects each mark from the captured image captured by the camera 26 and acquires “detected position information” indicating the position of the reference point of the alignment mark. As will be described later, in this embodiment, two types of alignment marks are used (see FIG. 5). Accordingly, the reference position storage means 52 stores set position information regarding two types of alignment marks.

  In the reference position storage means 52, “mark position information” representing the position of the alignment mark on the standard substrate 12 is stored in advance. The mark position information is a design value and is a predetermined value when the alignment mark is provided on the substrate 12. As the mark position information, the coordinate value of the center of gravity of the alignment mark, the coordinate value of the feature point representing the boundary between the alignment mark and the periphery (for example, the coordinate value of each vertex when the alignment mark is a triangle), and the alignment mark are representative Various values such as the coordinate value of the representative point to be performed (for example, the coordinate value of the intersection when the alignment mark is a cross) can be used. In the present embodiment, as an example, a case will be described in which the mark position information represents the coordinate value of the center of gravity position of the alignment mark.

  The reference position storage means 52 stores “mark shape information” representing the shape of the alignment mark in advance in order to detect the mark by a pattern recognition method such as template matching. When the alignment mark is composed of an assembly of a plurality of marks, “mark shape information” representing the shape of each mark is stored in advance. The mark shape information is, for example, image information of a mark used for template matching.

  In addition, the reference position storage means 52 includes “drawing position information” for setting a drawing area in which a wiring pattern is formed, “allowable error threshold information” as threshold information for determining an upper limit value of the shift amount, and the like. Stored in advance. Here, the “shift amount” is a shift amount (error) from the detected position of the corrected position obtained by correcting the set position of the alignment mark. The probability that the corrected position of each mark completely overlaps the detected position is low, and the corrected position of each mark is shifted from the detected position. The “correction amount” for correcting the set position of the alignment mark is calculated based on the above “set position information” and “detected position information”.

Here, the “correction amount” means, for example, an X-direction shift correction amount “ΔX”, a Y-direction shift correction amount “ΔY”, a rotation correction amount “θ”, an X-direction scale ratio correction amount “Δk _X ”, and Y Various correction parameters such as a direction scale ratio correction amount “Δk _Y ”. These correction parameters can be obtained by a conventionally known method. The correction parameter is appropriately selected according to the correction method.

  Each of the mark position information, mark shape information, drawing position information, and allowable error threshold information can be set by the user. For example, the mark position information, mark shape information, and drawing position information may be acquired and set from a captured image obtained by photographing the standard substrate 12 with the camera 26. The permissible error threshold information may be acquired and set from actually accumulated data such as a yield rate when corrected using a predetermined correction amount.

<Alignment mark>
Here, the alignment mark will be described. FIG. 5A is a plan view for explaining alignment marks provided on the substrate. FIG. 5B is a cross-sectional view taken along the dotted line in FIG. FIG. 5C is a diagram illustrating an example of an alignment mark.

As shown in FIG. 5A, in the present embodiment, a substrate 12 having a rectangular shape in plan view is used. The four vertices of the rectangular substrate 12 are designated as B _{1 to} B ₄ . A wiring pattern is formed in the drawing region 24A that is rectangular in plan view surrounded by the alternate long and short dash line on the surface of the substrate 12. In order to specify the drawing position of the drawing area, a plurality of alignment marks are provided on the substrate 12 based on mark position information and mark shape information set in advance.

In each of the four corners of the rectangular substrate 12, first alignment marks DM _{1 to} DM ₄ that are circular holes in plan view are provided. The substrate 12 is provided with a total of four first alignment marks DM _{1 to} DM ₄ . Note that the first alignment marks DM are collectively referred to as the first alignment marks DM when it is not necessary to distinguish each of the first alignment marks DM _{1 to} DM ₄ . The point illustrated in the center of the first alignment mark DM is the center of gravity of the first alignment mark DM. The coordinate value of the barycentric point is position information representing the position of the first alignment mark DM.

In addition, second alignment marks LM _{1 to} LM ₄ are provided at each of the four corners of the rectangular drawing region 24A, in which four marks that are circular in plan view are arranged at the vertices of the square. The substrate 12 is provided with a total of four second alignment marks LM _{1 to} LM ₄ . Note that when there is no need to distinguish each of the second alignment marks LM _{1 to} LM ₄ , they are collectively referred to as a second alignment mark LM.

As shown in FIG. 5C, the point illustrated in the center of the second alignment mark LM is the barycentric point G of the _four marks M _{1 to} M ₄ constituting the second alignment mark LM. When it is not necessary to distinguish each of the marks M _{1 to} M ₄ , they are collectively referred to as a mark M. The coordinate value of the barycentric point G is position information representing the position of the second alignment mark LM. Each of the marks M has a very small diameter of about 100 μm to 200 μm. For this reason, noise on the captured image is easily erroneously detected as a mark.

  As shown in FIG. 5B, it is assumed that the substrate 12 is a multilayer substrate in which the second layer 12B is stacked on the first layer 12A. The first alignment mark DM is a through-hole penetrating the substrate 12, and is formed by mechanically drilling the substrate 12 by drilling or the like. The first alignment mark DM is a so-called through hole. The second alignment mark LM is a bottomed hole (concave portion) opened in the drawing surface of the substrate 12, and the second layer so that the first layer 12A of the substrate 12 is exposed by drilling by laser processing or the like. 12B is formed by perforation. The second alignment mark LM is a so-called laser via.

  In the above description, an example in which the first alignment mark DM is a through hole formed by drilling or the like and the second alignment mark LM is a laser via formed by laser processing or the like has been described. It is not necessarily limited to the above form. For example, as the first alignment mark DM and the second alignment mark LM, lands, vias, etching marks, and the like can be used as appropriate. In the case of a multilayer substrate, a part of the circuit pattern already formed on the layer on which the pattern image is formed may be used as an alignment mark.

  In the above description, an example in which the first alignment mark DM having a circular plan view is provided and the second alignment mark LM in which four marks having a circular plan view are arranged at each vertex of the square has been described. One alignment mark only needs to be composed of an aggregate of a plurality of marks, and the alignment mark is not limited to the above-described form.

  7 and 8 are modifications of the alignment mark composed of an assembly of a plurality of marks. Hereinafter, an alignment mark composed of an aggregate of a plurality of marks is referred to as an “aggregate mark”. FIG. 7A shows an example of a grid-like set mark, and FIG. 7B shows an example of a cross-like set mark. 8A and 8B are examples of circumferential collective marks.

  The “lattice set mark” shown in FIG. 7A is a set mark arranged in a grid form so that a plurality of marks form a quadrangle. In this example, 25 marks are arranged in a grid of 5 rows and 5 columns. In the case of this “lattice set mark”, the reference position storage means 52 is set with the following information as “mark shape information” as shown in Table 1 below. In the present embodiment, these “mark shape information” are used in “provisional centroid determination” described later.

  The “cross-shaped collective mark” shown in FIG. 7B is a collective mark arranged in a lattice shape so that a plurality of marks form a cross shape. In this example, 20 marks are divided into 5 blocks. One block is an aggregate of 2 rows and 2 columns including 4 marks, and 5 blocks are arranged in the center and in the vicinity of 4 so as to form a cross. In the case of this “cross-shaped set mark”, the following information is set as “mark shape information” in the reference position storage means 52 as shown in Table 2 below.

  The “circumferential collective mark” shown in FIG. 8A is a collective mark arranged in a circumferential shape so that a plurality of marks form a circle. In this example, twelve marks are arranged on the same circumference at equal intervals. In the case of this “circumferential set mark”, the reference position storage unit 52 is set with the following information as “mark shape information” as shown in Table 3 below.

  The “circumferential collective mark” shown in FIG. 8B is a collective mark arranged in a circle so that a plurality of marks form a plurality of concentric circles. In this example, twelve marks are arranged on the circumference of the inner circle at equal intervals, and twelve marks are arranged on the circumference of the outer circle at equal intervals. In the case of this “circumferential set mark”, the reference position storage means 52 is set with the following information as “mark shape information” as shown in Table 4 below.

<Operation of exposure apparatus>
Next, the operation of the exposure apparatus according to the present embodiment will be described with reference to FIG.
First, in the data creation device 40, vector data representing the entire image pattern including a plurality of wiring patterns to be exposed on the substrate 12 is created. The generated vector data is input to the raster conversion processing unit 50. In the raster conversion processing unit 50, the vector data is converted into raster data and input to the image data correction unit 56, and the image data correction unit 56 is input. Raster data is temporarily stored.

  When the vector data is input to the raster conversion processing unit 50 as described above, the controller 70 that controls the operation of the entire exposure apparatus 10 outputs an instruction signal to the stage control unit 60, and according to the instruction signal. The stage control unit 60 outputs a control signal to a stage driving device (not shown). In response to the control signal, the stage driving device (not shown) moves the moving stage 14 from the position shown in FIG. 1 along the guide 20 to a predetermined initial position on the upstream side, and then in the stage moving direction. Move at speed.

  When the substrate 12 on the moving stage 14 moving as described above passes under the plurality of cameras 26, the substrate 12 is photographed by these cameras 26, and image data representing the captured images is sent to the drawing position correcting means 54. Entered.

The drawing position correcting unit 54 detects the first alignment marks DM _{1 to} DM ₄ , the second alignment marks LM _{1 to} LM _4, etc. of the substrate 12 placed on the moving stage 14 and detects the center of gravity position. Get location information. As a method for detecting the alignment mark, a conventionally known detection method such as a pattern recognition method using feature extraction can be used. The present embodiment is characterized by a mark recognition method for calculating an accurate reference point of an alignment mark composed of an aggregate of a plurality of marks. That is, a mark recognition processing method for acquiring accurate “detection position information” regarding the barycentric point G of the _four marks M _{1 to} M ₄ constituting the second alignment mark LM from the captured image captured by the camera 26. There are features. The mark recognition process will be described later in detail.

“Setting position information” such as mark position information, mark shape information, drawing position information, and allowable error threshold information, which are alignment mark setting values, is output from the reference position storage means 52 to the drawing position correction means 54. The drawing position correcting unit 54 detects “detection position information” and “setting position information” of the alignment marks acquired from the captured image of the substrate 12 for the first alignment marks DM _{1 to} DM ₄ and the second alignment marks LM _{1 to} LM _4. Based on the above, the “correction amount” for correcting the drawing position of the drawing area 24A is calculated.

Here, correction of the drawing position of the drawing area will be described. FIG. 6 is a diagram illustrating a state where the drawing position in the drawing area 24A is corrected and the image data is corrected. The “setting position information” of the alignment mark is the coordinate value of the alignment mark when the substrate 12 is placed at the normal position with reference to the imaging area 26A of the camera 26 in order to specify the drawing position of the drawing area 24A. Is set as The imaging region 26A is a region that is rectangular in plan view. Four vertices of the rectangular area are A _{1 to} A _4, and one of the four vertices (here, point A ₁ ) is the origin. Since the origin is determined based on the imaging region 26A of the camera 26, the origin of the set position information (coordinate value) and the origin of the detected position information (coordinate value) acquired from the captured image are always the same.

  As shown in FIG. 6, in this example, the substrate 12 and the drawing area 24A move to the upper right and rotate counterclockwise on the drawing. The shape of the substrate 12 is substantially square in design, but is deformed into a rectangle on the captured image due to a change in scale ratio. Accordingly, the “detected position information” acquired from the captured image is different from the preset “set position information”.

Hereinafter, the coordinate value representing the detected position information is referred to as “read value”, and the coordinate value representing the set position information is referred to as “set value” as appropriate. The correction amount is a correction parameter calculated so that the correction value approaches the read value when the set value is corrected with the correction amount. As the “correction parameters”, for example, the above-described correction parameters (ΔX, ΔY, θ, Δk _X , Δk _Y ) are obtained.

Position coordinates of the vertices A _{1 to} A ₄ of the imaging area 26A, position coordinates of the vertices B _{1 to} B ₄ of the substrate 12, position coordinates of the center of gravity of each of the first alignment marks DM _{1 to} DM ₄ (through holes), second The position coordinates of the center of gravity of each of the alignment marks LM _{1 to} LM ₄ (laser vias) and the position coordinates of the four vertices of the drawing area 24A are set in advance. In order to correct the drawing position, it is necessary to obtain a correction value of the position coordinates of each vertex of the drawing area 24A.

If the error between the detected position coordinates (read values) of the vertices B _{1 to} B ₄ and the set position coordinates (set value) exceeds the allowable range, such as when the substrate 12 is out of the imaging area 26A, drawing is impossible. However, in the present embodiment, the substrate 12 does not protrude from the imaging region 16A and cannot be drawn, and the error between the read values and the set values of the vertices B _{1 to} B ₄ is within an allowable range. Will be described.

The drawing position correcting means 54 calculates correction parameters (ΔX, ΔY, θ, Δk _X , Δk _Y ), corrects the set values of the four vertices of the drawing area 24A based on these correction parameters, and sets each vertex. The coordinate value after correction is obtained. As a result, the drawing position of the drawing area 24A is corrected, and the drawing position (set value) before correction is moved to the corrected drawing position. That is, it approaches the detected drawing area 24A (reading value). At the same time, the raster data in the drawing area is corrected, and the raster data corresponding to the drawing position before correction is converted into raster data 34A (image data of the letter “F”) corresponding to the drawing position after correction. That is, the converted raster data 34A is moved (shifted) to the upper right in the drawing and rotated counterclockwise and the scale ratio in the X direction and the Y direction is changed so as to fit in the detected drawing area 24A. (Variable)

  The corrected drawing position information is output to the image data correction unit 56. The image data correction unit 56 corrects the raster data in the drawing area by performing processing such as rotation, shift, and scaling on the raster data temporarily stored in advance based on the input drawing position information after correction. When the corrected raster data is calculated, the moving stage 14 is moved from the downstream position shown in FIG. 1 to the upstream side at a desired speed.

  When the tip of the substrate 12 is detected by the camera 26, exposure is started. Specifically, the corrected raster data calculated as described above is output to the drawing control unit 58, and the drawing control unit 58 controls each exposure head 30 of the scanner 24 based on the input corrected raster data. Based on the control signal, the exposure head 30 turns on / off the micromirror of the mirror device to expose the wiring pattern on the substrate 12 according to the corrected raster data. As the moving stage 14 moves, a control signal is sequentially output to each exposure head 30 to perform exposure. When the rear end of the substrate 12 is detected by the camera 12, the exposure ends.

<Mark recognition processing>
Next, the mark recognition process executed in the present embodiment will be described in detail. FIG. 15 is a flowchart showing a routine of mark recognition processing executed as a function of the mark recognition device (reference numeral 55 in FIG. 4). In the mark recognizing device 55, the position of the center of gravity G of the alignment mark (second alignment mark LM composed of four marks M) composed of an aggregate of a plurality of marks is determined from the captured image captured by the camera 26. It detects (refer FIG.5 (C) and FIG. 6). That is, the position of the center of gravity of each “collection mark” is detected. This processing routine is stored as a program in a storage unit (not shown) such as a hard disk device of the exposure apparatus, and is read from the storage unit and executed by the controller 70.

  First, in step 100, the substrate 12 on which the alignment mark is formed is imaged by the camera 26, and a captured image is acquired from the camera 26. In subsequent step 102, using a template having an image of each mark constituting the collective mark, normalized correlation template matching is performed on the captured image to detect each mark from the captured image.

  FIG. 9 is a diagram showing a state of normalized correlation template matching. The entire imaging region 26A of the camera 26 is displayed on the screen of a display device (not shown). As shown by the dotted arrow, the template 80 having the image of the mark is moved in the horizontal direction of the screen to perform matching, and the mark of the same image as the template 80 is detected. By repeating this matching in the vertical direction of the screen, the template 80 is moved over the entire screen.

All the marks on the screen are detected, and the position coordinates (x _i , y _i ) of each detected mark on the image are obtained. As shown in FIG. 9, for example, the mark M constituting each of the second alignment marks LM _{1 to} LM ₄ is detected from the captured image corresponding to FIG. As described above, since each mark M is small, noise on the captured image may be erroneously detected as a mark, and some marks of the collective mark may not be detected.

  Next, in step 104, the position of the temporary center of gravity is determined by voting using the voting space. FIGS. 10A to 10C are schematic diagrams for explaining a method of determining a temporary center of gravity using a voting space. FIG. 10A is a diagram illustrating a part of a display screen of a captured image. FIG. 10B is a diagram showing a voting space corresponding to a part of the display screen. FIG. 10C is a diagram illustrating a state in which a temporary center of gravity is determined by voting on a voting space.

As shown in FIG. 10A, a part of the display screen of the captured image is divided in a lattice shape so that the detected marks are arranged on the lattice points. The divided blocks are arranged in a matrix of 9 rows and 11 columns. In this example, four marks M _{1 to} M ₄ constituting the set mark and noises M ₅ and M ₆ are detected.

A storage means (memory) (not shown) stores an area having a plurality of voting spaces formed by dividing an area R having the same size as the screen into a plurality of grids. As shown in FIG. 10B, each of the plurality of voting spaces V ₁₁ , V ₁₂ , V ₁₃ ,... V _mn includes lattice points formed by dividing the display screen into a lattice shape. Is formed. The voting space “V _mn ” represents the voting space in the m-th row and the n-th column.

FIG. 10B shows a region having a plurality of voting spaces corresponding to a part of the display screen shown in FIG. For example, it may be considered that the region R having the same size as the screen is superimposed on the display screen. As shown in FIG. 10C, the detected mark M exists in any voting space. In this example, the mark M ₁ is the voting space _{V 22,} the mark _{M 2} to voting space _{V 24,} the mark _{M 3} are the voting space _{V 42,} the mark _{M 4} is in the voting space _{V 44,} the mark _{M 5} is voting space the V _69, the mark _{M 6} in the voting space _{V 78,} are present each.

For each detected mark M, the position of the temporary center of gravity is determined by voting a predetermined number of votes (one or more votes) in a voting space where a center of gravity is predicted to exist. More specifically, as shown in FIGS. 11A and 11B, when voting one by one on the voting space around the voting space where the mark M exists (here, 8 neighborhoods), the votes of the voting space _{V 33} located in the center of the mark M ₁ ~M ₄ is the most with 4 votes. It is determined that the temporary center of gravity of the set mark exists in the voting space V ₃₃ having the largest number of votes.

Voting space _{V 33} votes in the third column except the space _{V m3,} voting space _{V 3n} the third line except for the voting space _{V 33,} voting space _{V 68,} and also voting space _{V 79,} to vote by each two votes. In addition, the other voting spaces in the vicinity of 8 of the mark M also get one vote each. In FIG. 10C, this is displayed with brightness. The higher the brightness, the greater the number of votes. It can be seen that the voting space V ₃₃ with the largest number of votes is the brightest area, and it is determined that the temporary center of gravity of the set mark exists in the voting space V ₃₃ .

Next, in step 106, the detection mark M that has contributed to the provisional centroid determination is extracted. As shown in FIG. 10C, when it is determined that the temporary center of the set mark exists in the voting space V ₃₃ , the four marks M _{1 to} M ₄ existing around the temporary center of gravity are used to determine the temporary center of gravity. Although contributing, the two marks M ₅ and M ₆ located away from the temporary center of gravity do not contribute to the determination of the temporary center of gravity. Therefore, a detection mark group including _four marks M _{1 to} M ₄ is extracted as a detection mark M that contributes to the determination of the temporary center of gravity. As a result, detection marks (noise due to erroneous detection) that do not constitute the collective mark are excluded, and the detection accuracy of the center of gravity is improved.

Next, in step 108, a model mark corresponding to each mark of the detection mark group is extracted. Here, the model is a set mark model (design drawing) set in advance. As described above, the set mark model is stored in the reference position storage means 52 as “mark shape information”. FIGS. 12A and 12B are schematic views showing a set mark model. The collective mark model shown in FIG. 12A has the same configuration as the second alignment mark LM shown in FIG. 5C, and is composed of four marks M _{1 to} M ₄ arranged at the vertices of a square. ing. The center of gravity G is the center position that is equidistant from the _four marks M _{1 to} M ₄ .

In the example shown in FIG. 10C, all of the _four marks M _{1 to} M ₄ of the collective mark are detected, but some of the collective marks may not be detected. FIG. 13 is a diagram showing a display screen when some marks of the set mark are not detected. This XY coordinate system is the “recognition coordinate system”. In addition, a coordinate system illustrated by arrows perpendicular to the display screen is a “model coordinate system”.

As shown in FIG. 13, looking at the display screen of the captured image, of the four marks M ₁ ~M _4, the mark M ₃ is not detected, the mark M ₁ on the display screen, M 2 _and, Three marks RM ₁ , RM ₂ , and RM ₄ corresponding to each of M ₄ are displayed. The mark RM ₃ corresponding to the mark M ₃ is not displayed on the display screen. When it is not necessary to distinguish each of the marks RM _{1 to} RM ₄ , they are collectively referred to as a mark RM. Here, in order to distinguish from the mark M of the model, the detection mark is indicated as a mark RM.

When the detected marks are three marks RM ₁ , RM ₂ , and RM ₄ , the position of the center of gravity g _m of these marks RM is a position with a cross (+). The position coordinates of the center of gravity g _m are (mg _x , mg _y ). The position coordinate of the center of gravity g _m of the mark RM is substantially deviated from the position coordinates (g _x , g _y ) of the center of gravity g when the _four marks RM ₁ , RM ₂ , RM ₃ , and RM ₄ are detected. It will be. Steps 108 to 114 described below are steps for correcting the above-described deviation amount using a mark model and obtaining a “true center of gravity”.

In step 108, as shown in FIG. 12B, model marks M ₁ , M ₂ , and M ₄ corresponding to the marks RM ₁ , RM ₂ , and RM ₄ of the detection mark group are extracted. The position of the center of gravity of the model is constant regardless of the number of extracted marks M, and is the same as the position of the center of gravity G of the _four marks M _{1 to} M ₄ .

Next, in step 110, the center of gravity of the detected mark group and the center of gravity of the extracted model mark group are matched. FIGS. 14A to 14D are schematic diagrams for explaining a process of obtaining a “true center of gravity”. That is, in step 110, FIG. 14 (A), the detection mark groups (marks RM _1, RM _2, and RM ₄₎ and the center of gravity _{g m} of the extracted model mark groups (mark M _{1, M} 2, And M ₄ ) are overlapped so that the center of gravity G of M ₄ ) coincides.

  Next, in step 112, the attitude of the model mark group (the values of the rotation angle t and the expansion / contraction coefficient k) is adjusted so as to minimize the sum of squares of errors between the detection mark center of gravity and the model mark center of gravity. . That is, the optimum values of the rotation angle t and the scaling factor k of the model mark group are calculated by the least square method.

For example, FIG. 14 (B), the as the inclination of the model mark group and the extracted tilt detection mark group are substantially coincident, while being matched with the center of gravity g _m and the center of gravity G, Model mark group Is rotated by a rotation angle t. Thereafter, as shown in FIG. 14C, the model mark group is enlarged or reduced by the enlargement / reduction coefficient k so that the size of the detected mark group and the size of the extracted model mark group substantially coincide with each other.

FIG. _14D is a diagram showing a distance d ₁ between the center of gravity g _RM1 of the detection mark RM _{1 and} the center of gravity g _M1 of the corresponding model mark M ₁ . A distance d ₁ between the center of gravity g _RM1 and the center of gravity g _M1 is a deviation amount (error) of the center of gravity position of the mark. The square value d ₁ ² of the distance (error) d ₁ is defined as e ₁ . The square value e of the distance (error) d is calculated for all the marks RM in the detection mark group, and the sum Σe _i (i = 1 to n) of the square values e for the n marks RM is minimized. The values of the rotation angle t and the expansion / contraction coefficient k are optimized. In the example of FIG. 14C, the value of the sum Σe _i = e ₁ + e ₂ + e ₄ of the square value e is minimized for the three marks RM ₁ , RM ₂ , and RM ₄ .

Next, in step 114, the position coordinates of the center of gravity G of the model mark group after adjusting the posture (that is, after rotating at the rotation angle t and expanding / contracting with the expansion / contraction coefficient k) are the position coordinates of the “true center of gravity”. And the processing routine of the mark recognition process is terminated. As shown in FIG. 14C, the position coordinates of the centroid G (true centroid) of the model mark group after adjusting the posture are detected by the _four marks RM ₁ , RM ₂ , RM ₃ , and RM _4. In this case, the position coordinates (g _x , g _y ) of the center of gravity g substantially overlap. That is, even when some marks of the collective mark are not detected, the accurate position coordinates of the center of gravity of the alignment mark can be calculated.

As shown in FIG. 21, when the origin the center of gravity g _m of the detection mark group, set position coordinates of the center of gravity G of the extracted model mark group is represented by _(-lg x, -lg _y). The coordinate system having the origin of the center of gravity of the detection mark group is also referred to as “model coordinate system” or “design coordinate system”.
The position coordinates of the centroid G of the model mark group (position coordinates of the “true centroid”) when rotated at the rotation angle t and scaled by the scaling coefficient k are the values when all marks are detected in the “recognition coordinate system”. It is substantially the same as the position coordinates (g _x , g _y ) of the center of gravity g. Therefore, using the rotation angle t, the expansion / contraction coefficient k, and the position coordinates (mg _x , mg _y ) of the center of gravity g _m obtained from the detected part of the marks, the following expressions (A) and (B) are used. The position coordinates (g _x , g _y ) of the “true centroid” in the “recognition coordinate system” are calculated by coordinate conversion.

<Modified example of provisional center of gravity determination method>
In the above, a voting space around the detected mark M is voted using a set mark composed of four marks M, and the temporary center of gravity of the set mark exists in the voting space with the largest number of votes Then, the example to decide was explained. However, as shown in FIGS. 7 and 8, in the case of a collective mark including 10 or more marks M, there are cases where the position of the temporary center of gravity can be determined or not determined due to missing of some marks. . If the position of the temporary center of gravity cannot be determined because there are a plurality of voting spaces having the maximum number of votes, the method for determining the temporary center of gravity may be changed.

  FIGS. 16A and 16B are views showing a lattice-like set mark from which some marks are missing. FIG. 16A is a diagram illustrating a state in which 10 marks at random positions are missing from 25 marks arranged in a grid of 5 rows and 5 columns. FIG. 17A is a diagram showing a captured image when the mark is largely lost, and FIG. 17B is a diagram showing a voting result in that case. As shown in FIG. 17B, it can be determined that the lightness of the central voting space is the highest even if a large mark is missing, and that the temporary centroid of the set mark exists in this voting space.

  FIG. 18A is a diagram showing a captured image when a mark is largely lost, and FIG. 18B is a diagram showing a conditional voting result in that case. In this case, a condition is added that the vote is not placed in the voting space where the detected mark M exists. Under this condition, the number of voting spaces that can be voted decreases, so the probability that there are a plurality of voting spaces with the maximum number of votes is reduced. As shown in FIG. 18B, also in this case, it can be determined that the brightness of the central voting space is the highest, and that the temporary center of gravity of the set mark exists in this voting space.

  FIG. 16B shows a state in which five marks in the fifth column are missing from 24 marks arranged in a grid of 5 rows and 5 columns except for the mark in the third row and the third column (center). FIG. FIG. 19A is a diagram illustrating a captured image when the mark in the fifth column is missing. FIG. 19B is a diagram showing a voting result with a condition in that case, and FIG. 19C is a diagram showing an unconditional voting result in that case.

  In the captured image shown in FIG. 19A, it can be seen that there are two voting spaces where the center of gravity is predicted to exist. If voting is performed unconditionally as shown in FIG. 19C, there will be a plurality of voting spaces with the maximum number of votes, and the position of the temporary center of gravity cannot be determined. On the other hand, if a condition that no voting is added to the voting space in which the detected mark M exists is added, as shown in FIG. 19B, the voting space having the maximum number of votes becomes one. It can be determined that there is a temporary center of gravity of the set mark.

<Attitude adjustment by least square method>
Here, the attitude adjustment method by the least square method, that is, the calculation method of the rotation angle t and the expansion / contraction coefficient k will be specifically described.

In order to calculate the posture t and the enlargement / reduction coefficient k, as shown in FIG. 21, consider the following design coordinate point p1 _n and measurement coordinate point pm _n in the coordinate system with the detected mark center of gravity as the origin. .
Design mark coordinates: pl ₁ (lx ₁ , ly ₁ ), pl ₂ (lx ₂ , ly ₂ ),..., Pl _n (lx _n , ly _n )
Measurement mark coordinates: pm ₁ (mx ₁ , my ₁ ), pm ₂ (mx ₂ , my ₂ ), ..., pm _n (mx _n , my _n )
However, it is assumed that the centroids overlap so that Σlx _i = 0, Σly _i = 0, Σmx _i = 0, and Σmy _i = 0.

  The rotation angle t and the expansion / contraction coefficient k that minimize the total sum of the distances between the design coordinate point group and the measurement coordinate point group are calculated by the least square method.

The design coordinates pl ′ (lx ′, ly ′) after the posture and scale adjustment are expressed by the following equations.
lx '= k * (lx * cos (t) -ly * sin (t))
ly '= k * (lx * sin (t) + ly * cos (t))

The sum Σe _i of the square values of the distance between the design coordinate point group and the measurement coordinate point group is given by the following equation.
_{Σe i = e 1 + e 2} + ... + e n

Here, e _i is given by the following equation.
e _i = (mx _i -k (cos (t) * lx _i -sin (t) * ly _i )) ² + (my _i -k (sin (t) * lx _i + cos (t) * ly _i )) ) ²

As cot (t) = c and sin (t) = s, c, s, and k that minimize Σe _i are calculated. When Σe _i is partially differentiated by c and s, the following equation is derived.

  C and s such that Σδe / δc = 0 and Σδe / δs = 0 are given by the following equations (1A) and (2A).

Similarly, if Σe _i is partially differentiated by the expansion / contraction coefficient k, it can be expanded into the following equation (3A).

  Substituting the results of the above equations (1A) and (2A) into the above equation (3A) to obtain the expansion / contraction coefficient k such that ΣΔe / Δk = 0, the following equation is obtained.

  The rotation angle t is calculated by the following formula.

<Least square method applied to circumferential set marks>
In the case of the “circumferential set mark” shown in FIGS. 8A and 8B, a least square method specific to the circular set mark can be applied to each circle having the same diameter. Even if the plurality of marks shown in FIG. 8B are circumferentially arranged so that a plurality of marks form a plurality of concentric circles, a mark on the same circumference is selected and a unique least square method is selected. Is applied and finally the average is obtained. FIG. 20 is a diagram for explaining parameters used in the least square method applied to the circumferential set mark.

The objective function of the least square method is as follows. The position coordinates of the center point of the circle are (X, Y) and the position coordinates of the center of gravity of the mark on the circumference are (x _i , y _i ). Given in 2).

  X, Y, and R that minimize the above equation (2) are obtained. However, since it is difficult to perform partial differentiation with X, Y, and R, the following equation (3) is introduced.

  The above equation (3) can be transformed into the following equation (4).

  Therefore, X, Y, and R can be replaced by the following formula.

  A value obtained by substituting the above equation (3) into the above equation (1) is used as an objective function of the following equation (5), and a, b, and c that minimize this are obtained.

  The above equation (5) is partially differentiated with respect to each variable, and equations (6), (7), and (8) that are set to 0 are solved for a, b, and c.

  When the above formulas (6) to (8) are arranged, the following formulas (9), (10), and (11) are obtained.

  Since Σc = n * c where n is the total number, the following equation (12) is obtained from the equation (11).

  Substituting Equation (12) into Equations (9) and (10) yields Equations (13) and (14) below.

  Further, when a = 2X and b = 2Y are substituted to form a matrix expression, the above expressions (13) and (14) become the following expressions (15), (16), and (17), and the matrix A and the vector b Is led.

  Therefore, the position coordinates of the center point of the target circle (X, Y) can be calculated by obtaining an inverse matrix of A as shown in the following equation (18).

  In the above embodiment, the example in which the mark recognition apparatus of the present invention is applied to an exposure apparatus that exposes a substrate such as a printed wiring board has been described. However, the mark recognition apparatus of the present invention uses the collective mark as an alignment mark. It is effective when used, and the present invention can be applied to alignment in various applications.

DESCRIPTION OF SYMBOLS 10 Exposure apparatus 12 Substrate 14 Moving stage 16 Leg 18 Installation stand 20 Guide 22 Gate 24 Scanner 24A Drawing area 26 Camera 26A Imaging area 30 Exposure head 32 Exposure area 34A Raster data 34B Raster data 40 Data creation apparatus 50 Raster conversion processing part 52 Reference position storage means 54 Drawing position correction means 55 Mark recognition device 56 Image data correction means 58 Drawing control section 60 Stage control section 70 Controller 80 Template

Claims (5)

Detecting means for detecting positions of a predetermined number or more of marks from a plurality of marks attached to the object based on the design information;
Estimating means for estimating a provisional centroid position of an assembly composed of the predetermined number of marks or more based on the positions of the predetermined number or more marks detected by the detecting means;
First extraction means for extracting, as a detection mark group, a mark that contributed to the estimation of the temporary centroid position from the predetermined number of marks detected by the detection means;
Second extraction means for extracting, as an alignment mark group, a mark corresponding to each of the marks of the detection mark group extracted by the first extraction means from a plurality of marks based on the design information;
From the mark group in which the position of each of the marks in the alignment mark group is changed so that each of the marks in the alignment mark group most closely matches each of the marks in the detection mark group, by each of the marks in the detection mark group A reference position calculating means for calculating a fixed reference point;
A mark recognition device including: The reference position calculating means;
First calculating means for calculating a rotation angle and an enlargement / reduction ratio for making each of the marks of the alignment mark group most coincident with each of the marks of the detection mark group by a least square method;
A second calculation for calculating a reference point determined by each mark of the detection mark group from a mark group in which the position of the mark of the alignment mark group is changed by the rotation angle and the enlargement / reduction ratio calculated by the first calculation means. Means,
The mark recognition apparatus according to claim 1, comprising:   The estimation means uses a plurality of voting spaces formed by dividing a predetermined region into a plurality of areas, and based on the positions of the predetermined number or more of marks detected by the detection means, the predetermined number or more Vote in the voting area where the center of gravity of the aggregate consisting of marks is assumed to exist, and estimate the position of the voting space with the largest number of votes as the position where the temporary center of gravity of the aggregate consisting of the predetermined number of marks or more exists The mark recognition device according to claim 1 or 2.   4. The mark according to claim 3, wherein when there are a plurality of voting spaces having the largest number of votes, the estimating means estimates a position of the voting space that does not overlap with the position of the mark as a position where the temporary centroid position exists. Recognition device. The mark recognizing apparatus according to claim 3, wherein when there are a plurality of voting spaces with the largest number of votes, the estimating unit cannot estimate a position where the temporary center of gravity position exists.