JP2012047463A - Position deviation value detection method and visual inspection method using the position deviation value detection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform a detection of a position deviation value for without causing a problem of memory capacity, and to perform the visual inspection of a substrate by correcting the position of the substrate based on the position deviation value.SOLUTION: On the basis of a reference table 424, an estimated reference point position is set by an estimated reference point position setting part 413, and a temporary reference point position is set by a temporary reference point position setting part 415. A standard deviation of a distance between the estimated reference point position and the temporary reference point position is determined by a distance index value calculation part 414, and the angle of rotation when the standard deviation becomes the minimum and a position difference value between the temporary reference point position and a reference point position is acquired as a position deviation value 425. An alignment of a substrate 6 to be inspected is performed based on the position deviation value 425 so that the alignment and visual inspection of the substrate can be surely performed without increasing the use capacity of a memory.

Description

  The present invention accurately detects a reference position for detecting a displacement amount of a substrate to be inspected, and obtains a displacement amount based on the reference position, and a displacement amount detected by the method. The present invention relates to an appearance inspection method for aligning a substrate to be inspected based on the above and performing an appearance inspection of the substrate to be inspected.

  In recent years, in the manufacture of semiconductor devices, improvement in yield has been demanded in order to improve productivity. However, device defects tend to increase due to the complexity of the manufacturing process and the like, and it is important to accurately detect defects generated in the manufacturing process in order to improve the yield.

  Defects in the manufacturing process include deformation, cutting, short-circuiting, etc. of the wiring pattern, and film thickness fluctuations due to particles and resist residues. In order to detect these defects, a reference image obtained by imaging a reference substrate having no defect is compared with a target image obtained by imaging the target substrate, and a difference between pixel values of the target image and the reference image is calculated. If the target image contains a defect, the pixel information of the defect position is different from the pixel information of the reference image. It was detected as.

  Therefore, in the defect inspection, the accuracy of alignment between the reference substrate and the target substrate is an important factor that affects the inspection accuracy. As a technique for aligning the reference substrate and the target substrate, a reference position for alignment from the target substrate, for example, an alignment mark or a relative position between the rotation center of the stage and the reference point detected from the target substrate is used. A technique is known in which a target substrate is aligned with a reference substrate based on the reference position by detecting with high accuracy.

  For example, as a technique for detecting a reference position, a mark detection technique for detecting the position of an alignment mark formed on a wafer is known, and Patent Document 1 uses an alignment mark to be detected as a template. A technique for performing template matching and detecting the position of an alignment mark on a wafer is disclosed. By using this technique, it is possible to detect the amount of displacement for positioning as a result. However, the template matching has a feature that the detection accuracy of the positional deviation amount with respect to the XY directions is high. However, the template matching has a problem that the processing is time-consuming because it is weak against the rotation and scaling of the target substrate and has a large calculation amount. Furthermore, template matching also has a problem that it is weak against changes in contrast and contrast of a target image due to changes in illumination or the like. Therefore, the position of the alignment mark detected by template matching is not necessarily detected with high accuracy.

  Therefore, as described above, Patent Document 2 discloses a technique that can detect a position serving as a reference and detect a positional deviation amount based on the position serving as a reference even when there is a variation in shading in the target image. A known pattern matching method is known. The pattern matching method described in Patent Document 2 uses the feature pattern extracted from the reference image to detect the position of the feature pattern from the target image by a normalized correlation method which is a kind of template matching. However, as described above, the accuracy is not always detected depending on the state of the target image. Therefore, the candidate position corresponding to the reference point in the reference image is calculated from the target image, and the candidate position most suitable as the reference point in the target image is detected by generalized Hough transform for each candidate position. Pattern matching is performed by detecting the amount of positional deviation with respect to the image in the XY direction, rotation angle, and magnification.

  Specifically, a characteristic region from the reference image is set as a feature pattern, and a reference point that serves as a reference for pattern matching of the target image with respect to the reference image is set, and the relative relationship between each feature pattern and the reference point is set. Position information is generated as a reference table. Then, a position having image information that approximates the feature pattern is acquired from the target image by a normalized correlation method or the like.

  Subsequently, the candidate position of the reference point viewed from the position approximated to the feature pattern in the target image acquired by the normalized correlation method is detected by the generalized Hough transform. In the generalized Hough transform, a parameter space for obtaining a plurality of candidate positions of reference points based on a plurality of combinations of the rotation angle θ and the scaling factor κ of the target image is set as a voting space, and the candidate position that becomes the maximum vote position by voting And the rotation angle θ and the expansion / contraction magnification κ at that time are obtained.

Here, the generalized Hough transform and the normalized correlation method will be briefly described. FIG. 12 is a diagram for explaining the detection of the positional deviation amount in the X and Y directions, the positional deviation amount in the rotational direction, and the positional deviation amount due to the change in the expansion / contraction magnification by the generalized Hough transform. In the generalized Hough transform, first, the inclination A of the tangent at each edge point 911 on the contour of the reference graphic 91 is examined, and the relative position of the reference point 910 of the reference graphic 91 with the edge point 911 as the origin is the relative polar coordinates (r, α). Then, a reference table exemplified in Table 1 is prepared in which the relative position can be referred to using the inclination A as an index.

Here, XY coordinates (x _{n, y n)} of the edge points on the contour of the subject to the subject graphic of recognition are, if the gradient of the tangent at the edge point is A _n, the reference pattern 91 theta Assuming that the object graphic is rotated by κ times and expanded and contracted by κ, the XY coordinates (x, y) of the reference point in the object graphic can be obtained by Equation 1.

Therefore, all combinations of θ and κ in Formula 1 are calculated with a predetermined step size, and voting is performed on a four-dimensional space of (x, y, θ, κ). At this time, r and α in Equation 1 are obtained by obtaining the corresponding relative polar coordinates from the reference table (Table 1) using (A _n −θ) as an index. When a plurality of r and α corresponding to (A _n −θ) are registered in the reference table, voting is performed for all combinations. By voting for all edge points on the contour of the target graphic, a large number of votes are collected near the XY coordinates of the reference point in the target graphic, and the edge point that has obtained the maximum number of votes is the target graphic in the target graphic. Reference point XY coordinates, and θ and κ can be obtained simultaneously.

In the normalized correlation method, if the pixel value of the reference image at coordinates (x, y) is f (x, y) and the pixel value of the target image is represented as g (x, y), Using Equation 2, the similarity C (also referred to as a correlation value) between the target images when they are shifted from the reference image by (m, n) and superimposed is obtained.

In Equation 2, f _av and g _av are average values of pixel values in regions corresponding to each other in the reference image and the target image. The calculation according to Expression 2 is performed for all (m, n), and (m, n) when the similarity C is the highest is obtained as the position where the image information of the target image approximates the reference image.

  By combining the above methods, Patent Document 2 detects the amount of positional deviation in the XY direction, the amount of positional deviation in the rotational direction, and the amount of positional deviation due to a change in magnification to perform pattern matching.

JP 2002-210577 A JP 2004-192506 A

  However, in order to improve the detection accuracy in the method of detecting the reference point used in the pattern matching method of Patent Document 2, the step size of the rotation angle θ and the magnification κ is made fine (that is, the resolution is increased). It is necessary to set up a voting space according to. For this reason, when the resolution of the rotation angle and magnification is increased, there is a problem that the memory capacity required to configure the voting space increases in proportion to the resolution. Even if a voting space with a high resolution can be set, there is a problem that the number of votes is likely to be dispersed due to too many candidate positions for voting and the maximum vote position cannot be determined. Further, when a position that matches the feature pattern is acquired from the target image by the normalized correlation method, if there is a defect in a region that matches the feature pattern in the target image, it may not be acquired correctly. Therefore, since the position of the reference point is calculated based on the position that coincides with the erroneously detected feature pattern, there is a problem that the accuracy of the calculated position of the reference point is lowered.

  The present invention has been made in view of the above problems, and detects a position serving as a reference of a substrate to be inspected by a novel method that does not cause a problem in the memory capacity to be used, and detects a displacement amount based on the position serving as a reference. It is an object of the present invention to perform an appearance inspection and a positional deviation amount detection for performing an appearance inspection on a substrate to be inspected by aligning inspection objects.

  In order to solve the above-described problems, the present invention provides a reference image position indicating positions on a stage corresponding to a plurality of reference images obtained by imaging a reference substrate placed on the stage, and a reference that is the rotation center of the stage. A preparatory step of generating relative position information of each point position as a reference table, and a feature image position representing a position on the stage of a feature image whose image information approximates a reference image obtained by imaging the board to be inspected A feature image position acquisition step, an reference reference table, an estimated reference point position setting step for setting a position on each stage corresponding to the reference point position viewed from each feature image position as an estimated reference point position; A temporary reference position setting step for setting the average position of each estimated reference point position acquired in the estimated reference point position setting step as a temporary reference point position, and a reference point position as the center. A distance index value calculating step for calculating a standard deviation or a variance of the distance between the estimated reference point position and the temporary reference point position for each rotation angle while displacing the rotation angle with a predetermined resolution, and a distance index value calculating step A positional deviation amount acquisition step of acquiring, as a positional deviation amount, a rotation angle at which the standard deviation of the distance calculated in step 1 or the distance variance is minimized and a position difference value between the temporary reference point position and the reference point position are provided. It is characterized by that.

  In the invention configured as described above, the standard deviation of distance or the variance of distance is calculated by the distance index value calculating step for each of a plurality of combinations of the rotation angle and the estimated reference point position, and the standard deviation of distance or the variance of distance is calculated. If the amount of misalignment at the time when is minimized is stored in the memory, it is not necessary to provide a voting space corresponding to all combinations in the memory space. Therefore, even when the resolution of the rotation angle is increased, processing that does not cause a problem with the memory capacity can be performed.

  Further, in the misregistration amount detection method according to the present invention, an illegal feature for determining whether or not the feature image position acquired by the feature image position acquisition step is acquired as the position of the feature image approximated by the image information of the reference image. An improper feature image position determining step, and the fraudulent feature image position determining step is performed when the value of the distance between the estimated reference point position and the temporary reference point position is larger than a threshold set based on the standard deviation of the distance. The feature image position corresponding to the point position is determined to be invalid, and the illegal feature image position is deleted.

  In the invention configured as described above, even if an illegal feature image position is acquired due to erroneous detection or the like in the feature image position acquisition step, the illegal feature image position can be deleted by the illegal feature image position determination step. In addition, since the calculation accuracy of the temporary reference point position is not lowered, it is possible to prevent the detection accuracy of the positional deviation amount from being lowered.

  The positional deviation amount detection method according to the present invention includes a pre-alignment step of performing pre-alignment on the substrate to be inspected before the feature image position acquisition step.

In the invention configured as described above, by performing pre-alignment before the feature image position acquisition step, a range in which the rotation angle is displaced is determined based on the pre-alignment performance with respect to the positional deviation amount in the rotation direction of the inspected substrate. Therefore, the amount of misalignment can be calculated at high speed.
Further, the visual inspection method of the present invention uses the above-described positional deviation amount detection method to perform alignment of the board to be inspected based on the positional deviation amount acquired in the positional deviation amount acquisition step. An inspection process for performing the inspection.

  In the invention configured as described above, the standard deviation of distance or the variance of distance is calculated by the distance index value calculating step for each of a plurality of combinations of the rotation angle and the estimated reference point position, and the standard deviation of distance or the variance of distance is calculated. If the amount of misalignment at the time when is minimized is stored in the memory, it is not necessary to provide a voting space corresponding to all combinations in the memory space. Therefore, even when the resolution of the rotation angle is increased, processing that does not cause a problem with the memory capacity can be performed. As a result, the appearance inspection of the substrate can be reliably performed.

  According to the present invention, since it is not necessary to provide a voting space in order to detect the positional deviation amount in the XY axes and the positional deviation amount in the rotation direction, which are positional deviation amounts of the substrate to be inspected with respect to the reference substrate, the number of votes is increased. It is possible to prevent the occurrence of a situation where each misalignment cannot be detected due to the fact that the temporary reference point position having the maximum number of votes is not determined, and when the standard deviation of distance or the dispersion of distance is minimized The misregistration amount may be secured in the memory area, and the misregistration amount can be detected so that the memory capacity does not matter. Then, it is possible to achieve an excellent effect that the substrate is aligned based on the detected displacement amount and the appearance inspection of the substrate can be reliably performed.

It is explanatory drawing which shows the structure of an external appearance inspection apparatus. It is a figure which shows the structure of a computer. It is a block diagram which shows the function structure of a computer. It is a figure which shows the flow of a preparatory work. It is a figure which shows an example of a reference board | substrate. It is an external view which shows the rotation center of a reference board | substrate. It is a figure which shows a series of flows of an external appearance test | inspection. It is a figure which shows an example of a to-be-inspected board | substrate. It is a figure which shows the flow of the process which acquires the positional offset amount of 1st Embodiment. It is a block diagram which shows the function structure of the computer of 2nd Embodiment. It is a figure which shows the flow of the process which acquires the positional offset amount of 2nd Embodiment. It is a figure for demonstrating the detection of the amount of position shifts by generalized Hough conversion.

  FIG. 1 is an explanatory diagram showing a configuration of an appearance inspection apparatus 1 according to the present embodiment. The appearance inspection apparatus 1 includes a substrate holding unit 2 that holds and moves the substrate W, an imaging unit 3 that acquires the whole or a part of the substrate W as image data of a two-dimensional image by imaging the substrate W, and a substrate It comprises a holding unit 2 and a computer 4 connected to the imaging unit 3.

  The substrate holding unit 2 includes a stage 21 that horizontally holds the substrate W and a stage driving unit 22 that moves the stage 21 in the X, Y, and θ directions. The stage 21 has a flat plate-like outer shape. Suction grooves (not shown) are provided on the upper surface of the stage 21, and a plurality of suction holes (not shown) are formed dispersedly on the inner bottoms of these suction grooves. These suction holes are connected to a vacuum pump or the like, and the atmosphere in the adsorption groove can be exhausted by operating the vacuum pump. Thereby, when the substrate W is placed on the upper surface of the stage 21, the substrate W can be fixedly held on the upper surface of the stage 21 while being kept horizontal by the suction pressure of the suction holes.

  The stage drive unit 22 is a mechanism for moving the stage 21 in the X axis direction, the Y axis direction, and the rotation direction around the Z axis in FIG. The stage drive unit 22 includes a rotation mechanism 23 that rotates the stage 21, a support plate 24 that rotatably supports the stage 21, an X-direction movement mechanism 25 that moves the support plate 24 in the X-axis direction, and an X-direction movement mechanism 25. And a Y-direction moving mechanism 26 that moves the base plate 27 that supports the support plate 24 in the Y-axis direction.

  The rotation mechanism 23 includes a linear motor 231 and a rotation shaft (not shown) between the lower surface side of the central portion of the stage 21 and the support plate 24. The linear motor 231 includes a mover (not shown) attached to the end of the stage 21 and a stator (not shown) laid on the upper surface of the support plate 24. Therefore, when the linear motor 231 is operated, the mover moves in the X-axis direction along the stator, and the stage 21 rotates within a predetermined angle range around the rotation axis on the support plate 24.

  The X-direction moving mechanism 25 includes a linear motor 251 and a pair of guide portions 252 extending in the X direction between the support plate 24 and the base plate 27. The linear motor 251 includes a mover (not shown) attached to the lower surface of the support plate 24 and a stator (not shown) laid on the upper surface of the base plate 27. For this reason, when the linear motor 251 is operated, the support plate 24 moves in the X-axis direction along the guide portion 252 on the base plate 27.

  The Y-direction moving mechanism 26 includes a linear motor 261 and a guide portion (not shown) extending in the Y-axis direction that guides a part of the base plate 27 between the base plate 27 and the base 28. The linear motor 261 includes a mover (not shown) attached to the lower surface of the base plate 27 and a stator (not shown) laid on the base 28 of the appearance inspection apparatus 1. For this reason, when the linear motor 261 is operated, the base plate 27 moves in the Y-axis direction along a guide portion (not shown) on the base 28.

  The imaging unit 3 is disposed above the stage 21 and is a mechanism for imaging the substrate W held on the stage 21. The imaging unit 3 generates an electrical signal from the light source 31 that emits the illumination light, the optical system 32 that guides the illumination light to the substrate W and the light from the substrate W enters, and the image of the substrate W formed by the optical system 32. And an image pickup device 33 that outputs to the computer 4. As the imaging unit 3, for example, a CCD camera or the like can be used.

  The computer 4 controls the stage driving unit 22 and the imaging unit 3 and performs predetermined processing based on the image of the substrate W captured by the imaging unit 3. Next, the configuration of the computer 4 will be described in detail with reference to FIG.

  As shown in FIG. 2, the computer 4 connects a calculation unit 41 (for example, a CPU) that performs various calculation processes, and a storage unit 42 (for example, a ROM or RAM) that stores basic programs and various types of information to a bus line. The general computer system has a configuration. The bus line further includes a fixed disk 43 (for example, a hard disk drive) for storing information, a display unit 44 (for example, a display) for displaying various information, and an input unit 45 (for example, a keyboard and a mouse) for receiving input from the operator. ), A reading device 47 that reads from a computer-readable recording medium 46 (for example, an optical disk, a magnetic disk, a magneto-optical disk, etc.), and a communication unit that communicates with the stage driving unit 22 and the imaging unit 3 48 are appropriately connected through an interface (I / F) or the like. For example, a touch panel display in which the functions of the display unit 44 and the input unit 45 are integrated may be used.

  The computer 4 reads the program 461 from the recording medium 46 via the reader 47 in advance and stores it in the fixed disk 43. Then, the program 461 is copied to the storage unit 42 (for example, RAM) and the calculation unit 41 executes calculation processing according to the program 461 in the storage unit 42 (that is, the computer 4 executes the program 461). ) The computer 4 executes the inspection of the substrate W.

  FIG. 3 is a block diagram illustrating a functional configuration realized by the calculation unit 41 operating according to the program 461. Note that these functions may be realized by a dedicated electric circuit, or an electric circuit may be partially used.

  The calculation unit 41 includes a reference table generation unit 402, a positional deviation amount acquisition unit 401, a correction amount calculation unit 416, and an inspection unit 404. The reference table generation unit 402 generates relative position information representing the relative positional relationship as the reference table 424 from the reference image position coordinates 422 and the reference point position coordinates 423 acquired in advance, and stores them in the storage unit 42. It has the function to do. In addition, the positional deviation amount acquisition unit 401 acquires a positional deviation amount 425 based on the XY direction, the rotation direction, and the expansion / contraction ratio between the position of the reference substrate and the position of the inspection target substrate. , An estimated reference point position setting unit 413, a distance index value calculation unit 414, and a temporary reference point position setting unit 415.

  The feature image position acquisition unit 412 has a function of using the reference image data 421 to acquire a feature image in which image information matches an image specified from a substrate to be inspected. The estimated reference point position setting unit 413 acquires an estimated reference point position described later from the feature image acquired by the feature image position acquisition unit 412 and the reference table 424, and the temporary reference point position setting unit 415 determines the estimated reference point position. Use to set the temporary reference point position. Then, the distance index value calculation unit 414 calculates a distance index value between each estimated reference point position and the temporary reference point position. By operating these units, the misregistration amount acquisition unit 401 can acquire the misregistration amount 425. A correction amount calculation unit 416 calculates a correction amount for actually correcting the stage 21 based on the positional deviation amount 425. Then, the inspection unit 404 inspects the substrate that has been aligned.

  The flow of preparation work before the inspection is performed by the computer 4 will be described with reference to FIG. FIG. 5 is a view for explaining the reference substrate 5 on which a plurality of dies 50 as semiconductor chips are formed. Hereinafter, the preparation work will be described with reference to FIGS. 3, 4, and 5.

  In the preparatory work, pre-alignment is performed by a pre-alignment apparatus (not shown) on the reference substrate 5 that is separately inspected and has no defects or expansion / contraction changes (step S11). And it is carried in and mounted on the stage 21 of the external appearance inspection apparatus 1 (step S12). In the pre-alignment apparatus, for example, a large positional deviation in the rotation direction of the substrate with respect to a position considered as a reference is corrected to the rotation of the substrate with reference to the orientation flat to the accuracy of the positional deviation in the rotational direction of about ± 1 degree. And it is conveyed on the stage 21 by a conveyance robot etc. from a pre-alignment apparatus.

  Subsequently, the reference substrate 5 is transferred from the pre-alignment apparatus to the stage 21 by the transfer robot so that the continuous arrangement of the dies 50 and the movement directions in the X direction and the Y direction in which the stage 21 moves are vertical and horizontal. The minute rotation of the reference substrate 5 that occurs at this time is corrected (step S13). The specific method will be described with reference to FIG. FIG. 6 is an external view of an image captured by the imaging unit 3 with respect to the reference substrate 5 on the basis of the rotation center of the stage 21. For convenience of explanation, the reference substrate 5 is illustrated by a dotted line in FIG. 6, but an image 502 is displayed on the display unit 44 described later.

  The imaging unit 3 disposed above the stage 21 images the reference substrate 5 placed on the stage 21 moving in the Y direction by the Y direction moving mechanism 26. In FIG. 6, a scribe line formed between the dies 50 aligned on the reference substrate 5 is captured as an image 502. The captured image 502 is displayed on the display unit 44. The operator designates a reference line 501 along the scribe line from one peripheral end of the reference substrate 5 to the other peripheral end via the input unit 45 for the image 502 displayed on the display unit 44.

  That is, by specifying the reference line 501, the coordinate position of the start point 503 and the end point 504 of the reference line 501 with respect to the stage 21, and the intersection point 506 where the perpendicular line 505 dropped from the start point 503 intersects at the peripheral edge of the reference substrate 5. The coordinate position is expressed as the coordinate position with respect to the stage 21, and the rotation angle from the position considered as the reference of the reference substrate 5 placed on the stage 21 using the three-square theorem (the reference line 501 and the perpendicular line 505 form). Corner) can be calculated. Based on the calculated rotation angle, the position of the reference substrate 5 is corrected by rotating the tee 21 around the rotation center of the stage 21 in the direction in which the end point 504 coincides with the intersection point 506 by the rotation mechanism 23.

  The state in which the reference substrate 5 corrected by the above operation is disposed is defined as being disposed on the stage 21 in a state where there is no positional deviation in the rotational direction. Based on this arrangement, a positional deviation amount of a substrate 6 to be inspected, which will be described later, is detected. Note that the amount of displacement detected in the present invention refers to a displacement in the XY direction on the stage 21 of the substrate 6 to be inspected with respect to the reference substrate 5, a displacement in the rotational direction, and / or a displacement due to a change in expansion / contraction magnification. Is.

  Next, the entire image of the reference substrate 5 is displayed on the display unit 44, and an imaging region to be a reference image is designated as a rectangular region via the input unit 45 by the operator (step S14). Also, the coordinates on the stage 21 of the designated area are stored in the storage unit 42. In this embodiment, first, an imaging region 301a indicated by a dotted line on the reference substrate 5 in FIG. Here, as an example of the reference image 49, an image obtained by imaging the imaging region 301a is shown in FIG. FIG. 5B is an external view showing the reference image 49 in an enlarged manner. That is, as shown in FIG. 5B, the imaging region designated by the operator is enlarged and displayed on the display unit 44 as the reference image 49. Note that the reference image 49 may be directly generated from circuit pattern design data or the like on the reference substrate 5.

  Next, an area that the operator recognizes as a characteristic area in the reference image 49 from the reference image 49 displayed on the display unit 44 via the input unit 45 is designated as one reference image 491a ( Step S15). The designated reference image 491a is stored in the storage unit 42 as reference image data 421. Note that the rectangular area may be specified so as to include a characteristic area. However, if the rectangular area is too large, the calculation time will be increased in the subsequent processing. The pixel is preferably about 32 pixel square.

  Subsequently, the lower right position in the image region indicating the reference image 491a is set as the reference image position 492a (step S16). This setting is determined in advance so that the lower right position in the reference image 491a is set as the reference image position 492a without depending on the operator. The position set as the reference image position is not limited to the lower right position. For example, the center of gravity position or other four corner points in the reference image may be set as the reference image position. Also, the reference image position may be set by the operator together with the reference image setting.

  Next, if a predetermined number of reference images are not specified, the process returns to step S14 (step S17), and steps S14 to 16 are repeated until a predetermined number of reference images are specified. Therefore, as shown in FIG. 5A, the reference images 491b, 491c, and 491d are designated and the reference image positions 492b, 492c, and 492d are set for the three imaging regions 302b, 302c, and 302d. . The four set reference image positions 491a, 491b, 491c, and 491d are stored as reference image position coordinates 422 in the storage unit 42 as coordinates representing the position on the stage 21.

  When the standard images 491a, 491b, 491c, and 491d are designated for the four imaging regions 301a, 302b, 302c, and 301d, the reference table generation unit 402 corresponds to the standard images 491a, 491b, 491c, and 491d, respectively. Relative position information representing the relative positional relationship between the four reference image positions 492a, 492b, 492c, and 492d and the reference point position 490 that is the rotation center of the stage 21 (also the origin in the coordinate system on the stage 21). A reference table 424 is generated (step S18). Specifically, the reference point position 490 that is the rotation center of the stage 21 is stored in the storage unit 42 as reference point position coordinates 423 as coordinates representing the position on the stage 21.

Then, the relative position is obtained as polar coordinates from the reference image position coordinates 422 and the reference point position coordinates 423 stored in the storage unit 42 in step S <b> 16 and stored in the storage unit 42 as a reference table 424. Thereby, the relative position information of each standard image position 492a, 492b, 492c, 492d and the standard point position 490 can be generated as the reference table 424. Table 2 illustrates the contents of the reference table 424, where P ₀ , P ₁ , P ₂ and P ₃ correspond to the four reference images 491a, 491b, 491c and 491d, and the polar coordinate distance and angle are r, The subscript is added to α. Note that the structure of the reference table 424 is not limited to that shown in Table 2 as long as the structure is substantially the same. For example, the reference table 424 may be an XY coordinate system, and each reference image position based on the reference point position 490. It may represent a relative position with respect to 492a, 492b, 492c, and 492d.

  In the present embodiment, four imaging regions 301a, 302b, 302c, and 301d are designated as the imaging regions. The reason is that, for example, even if reference image information for one location cannot be obtained from a substrate to be inspected, which will be described later, it is possible to accurately detect the amount of displacement using the remaining reference image. Therefore, four or more imaging areas may be designated. In addition, the present invention can be implemented if at least two imaging areas are designated.

  Next, an operation when the inspection of the substrate W is performed in the appearance inspection apparatus 1 will be described. FIG. 7 is a diagram showing a series of flow of appearance inspection. FIG. 8A is a diagram showing an example of the inspected substrate 6 to be inspected, and shows an example in which the substrate 6 is slightly rotated counterclockwise. For convenience of explanation using the drawings, the die 50 formed on the substrate 6 to be inspected is not shown. FIG. 8B is a diagram showing an example of a target image 51 acquired by imaging the board 6 to be inspected. Hereinafter, a series of flow of the appearance inspection will be described with reference to FIG. 3, FIG. 7, and FIG.

  At the time of inspection, the inspected substrate 6 is aligned by a pre-alignment apparatus (not shown) (step S21), and the inspected substrate 6 is carried into the appearance inspection apparatus 1 by a transfer robot (not shown) from the pre-alignment apparatus. It is placed on the stage 21 constituting the holding unit 2 (step S22). Accordingly, the amount of positional deviation in the rotation direction of the substrate 6 to be inspected placed on the stage 21 depends on the alignment accuracy of the pre-alignment apparatus and the transfer accuracy of the transfer robot. For example, the alignment accuracy in the rotation direction of the pre-alignment apparatus Is within ± 1.0 degrees, it can be assumed from experience that the amount of positional deviation in the rotational direction is within about ± 1.5 degrees.

  The stage drive unit 22 moves the substrate 6 to be inspected below the imaging unit 3, and the imaging unit 3 images the substrate 6 to be inspected based on the reference image position coordinates 422 stored in the storage unit 42. Obtain (step S23). The acquired target image 51 is transferred to the computer 4. In this embodiment, as shown in FIG. 8A, by capturing the four reference image position coordinates 422 by the image capturing unit 3, it is not necessary to capture an unnecessary area, and the processing speed can be increased.

  When the target image 51 is acquired in the four imaging regions 302a, 302b, 302c, and 302d, the positional deviation amount 425 of the board 6 to be inspected with respect to the reference board 5 is acquired by the positional deviation amount acquisition unit 401 of the computer 4. (Step S24). The acquisition of the positional deviation amount 425 by the positional deviation amount acquisition unit 401 will be described in detail later.

  When the positional deviation amount 425 is acquired, the correction amount calculation unit 416 calculates a correction amount for correcting the position of the stage 21 based on the positional deviation amount 425. Then, by driving the stage drive unit 22 based on the calculated correction amount, the stage 21 on which the substrate 6 to be inspected is driven is driven so that the positional deviation of the substrate 6 to be inspected relative to the reference substrate 5 is corrected. Is performed (step S25).

The correction amount can be calculated by Equation 3 from the positional deviation amount (x, y) in the XY direction calculated in step S24, the positional deviation amount θ in the rotational direction, and the positional deviation amount κ due to the change in the expansion / contraction magnification. . Then, after the stage 21 is rotated by the rotational displacement amount θ by the rotation mechanism 23, the X-direction correction value Xa calculated by Equation 3 and the Y-direction correction value Ya are obtained for the positional displacement in the XY directions. Based on the above, the stage 21 can be moved in the X and Y directions by the X direction moving mechanism 25 and the Y direction moving mechanism 26 to perform alignment. That is, the correction amount indicates the amount of movement of the stage 21 in the XY direction and the rotational direction for correcting the positional deviation by moving the stage 21 with respect to the positional deviation of the substrate 6 to be inspected with respect to the reference substrate 5. It is.

  Then, a predetermined inspection is performed on the inspected substrate 6 (step S26). As the predetermined inspection, for example, a method of detecting a defect by obtaining a difference in pixel value between an inspection image obtained by imaging the inspected substrate 6 and a master image having no defect, or other known methods Inspection methods can be applied.

  The inspected substrate 6 after inspection is unloaded from the appearance inspection apparatus 1 by a transfer robot (not shown) (step S27).

  Next, the flow of processing for acquiring the positional deviation amount 425 by the positional deviation amount acquisition unit 401 will be described in detail with reference to FIGS.

  The positional deviation amount acquisition unit 401 has a function of acquiring how much the inspected substrate to be inspected with respect to the reference substrate serving as a reference is displaced due to the XY direction, the rotation direction, and / or expansion and contraction. The position acquisition unit 412, the estimated reference point position setting unit 413, the distance index value calculation unit 414, and the temporary reference point position setting unit 415 are configured.

  The feature image position acquisition unit 412 has a function of acquiring, as a feature image, an image region that approximates the reference image and the image information from the target image 51 based on the reference image data 421 stored in the storage unit 42 by the normalized correlation method. . As the image information, a gradation value or the like in each pixel constituting the reference image can be used. Then, the feature image position acquisition unit 412 acquires the feature image from the target image, thereby acquiring the feature image position (step S231).

  A specific description will be given using this embodiment. FIG. 8B is an external view showing the target image 51 in an enlarged manner. The target image 51 that overlaps the reference image 491a while sequentially moving the reference image 491a in pixel units sequentially from the upper left based on the reference image data 421 stored in the storage unit 42 with respect to the target image 51 acquired in step S22. The similarity ratio in the image area is calculated as the similarity C (value calculated by Expression 2). The similarity C calculated by Expression 2 approaches 1.0 when the correlation between the reference image 491a and the image area of the target image 51 is high, and approaches −1.0 when the correlation is low. In other words, if the image area of the reference image 491a and the target image 51 are similar (that is, the degree of similarity is high), the area approaches 1.0, and in the area that is not similar (that is, the degree of similarity is low), -1. It approaches 0.

  As a result of scanning the reference image 491a with respect to the target image 51, an image area determined in an area where the similarity C is closest to 1.0 is used as a feature image, and a feature image position can be acquired. It should be noted that the region where the similarity C is closest to 1.0 is used as the feature image because the similarity C is 1.0 when there is a defect, an imaging failure, or the like in the image region that matches the reference image in the target image. It is because it is not. Therefore, it is preferable to set a threshold value of similarity C in advance based on an empirical rule of an operator or the like as to how much similarity is used as a feature image. By doing in this way, the image area | region which has predetermined | prescribed similarity can be acquired as a feature image, and acquisition of a positional offset amount can be performed suitably.

  In this embodiment, it is shown in FIG. 8B that the feature image 511a has been acquired. Similarly, FIG. 8A shows how the characteristic images 511b, 511c, and 511d are acquired using the other three reference position images 491b, 491c, and 491d. Accordingly, four feature images 511a, 511b, 511c, and 511d are finally obtained. For convenience of illustration, a characteristic pattern representing each acquired feature image is omitted.

  Then, for the four feature images 511a, 511b, 511c, and 511d, the lower right position is acquired as the feature image positions 512a, 512b, 512c, and 512d (polar coordinates). As the feature image position, a position in the feature image corresponding to the reference image position in the reference image is acquired as the feature image position. Then, the acquired four feature image positions 512a, 512b, 512c, and 512d are temporarily stored in the storage unit 42 as coordinates representing the position on the stage 21 like the reference image position.

  The feature images 511a, 511b, 511c, and 511d acquired by the above-described method are always found as image regions that completely match the reference image due to the rotation of the substrate 6 to be inspected, defects on the substrate 6 to be inspected, and the like. Not necessarily. Therefore, it is necessary to set the temporary reference point position 510 in the subsequent processing steps in order to obtain the positional deviation amount with high accuracy.

  In order to detect the amount of displacement with respect to the reference substrate 5 due to the expansion and contraction and rotation of the substrate 6 to be inspected, a plurality of combinations of the expansion and contraction magnification κ and the rotation angle θ are set (steps S232 and S233). That is, the expansion / contraction magnification κ and the rotation angle θ are displaced with predetermined resolutions Δκ and Δθ. If it is not necessary to consider the expansion / contraction magnification κ, only the rotation angle θ may be set. Further, the expansion / contraction magnification κ represents the expansion / contraction ratio of the substrate to be inspected with respect to the reference substrate, and the rotation angle θ represents a rotation angle around the reference point position 490 that is the rotation center of the stage 21.

  Subsequently, the estimated reference position setting unit 413 refers to the relative position information described in the reference table 424 for a plurality of combinations of the respective rotation angles θ and expansion / contraction magnifications κ, and the respective feature image positions 512a and 512b. Four estimated reference point positions 513a, 513b, 513c, and 513d, which are positions corresponding to the reference point positions on the reference substrate 5 as viewed from 512c and 512d, are set as coordinates representing the position on the stage 21 (step S234).

Specifically, it is relative position information (r _(Pi) , ₎ described in the reference table 424 with respect to the coordinates (x _(Pi) , y _(Pi) ) of a plurality of feature image positions on the substrate 6 to be inspected. α _(Pi) ) is associated, and when the inspected substrate 6 is rotated by θ and expanded / contracted by κ to match the reference substrate 5, the coordinates (the estimated reference point position in the inspected substrate 6 are coordinates ( x _i , y _i ) can be calculated by Equation 4. Here, i is a subscript representing an index number assigned to the feature image position, and in the present embodiment, takes a range of 0 ≦ i ≦ 3.

Further, the temporary reference point position setting unit 415 calculates the temporary reference point position 510 from each estimated reference point position 513a, 513b, 513c, 513d (step S235). The coordinates (O _x , O _y ) of the temporary reference point position 510 are the average positions calculated by calculating the average of the x coordinate and the y coordinate from the coordinates of the estimated reference point positions 513a, 513b, 513c, and 513d. Coordinates. Specifically, it can be calculated using Equation 5. Therefore, the temporary reference point position 510 is calculated as coordinates representing the position on the stage 21.

Next, the distance index value calculation unit 414 calculates a distance index value from each estimated reference point position 513a, 513b, 513c, 513d and the temporary reference point position 510 (step S236). The distance index value is a positional deviation between the coordinates (x _i , y _i ) of the estimated reference point positions 513a, 513b, 513c, and 513d on the stage 21 and the coordinates (O _x , O _y ) of the temporary reference point position 510. For example, the standard deviation of the distance between each estimated reference point position 513a, 513b, 513c, 513d and the temporary reference point position 510 can be used. Formula 6 shows the calculation formula when the standard deviation is used as the distance index value. Here, h represents the distance between the estimated reference point position and the temporary reference point position. Note that the dispersion of the distance between each estimated reference point position and the temporary reference point position may be used as the distance index value.

Here, Expression 7 represents an average calculated from the distance between the respective coordinates of the estimated reference point positions 513a, 513b, 513c, and 513d and the temporary reference point position 510 in Expression 6, and Expression 8 represents the estimation in Expression 6. The average of the distance between each coordinate of the reference point position 513a, 513b, 513c, 513d and the temporary reference point position 510 is shown.

  The positional deviation amount acquisition unit 401 assigns the distance index value to the evaluation value f that is a variable, and also includes the rotation angle θ, the expansion / contraction magnification κ, and the reference point position 490 and the temporary reference point position 510 when the distance index value is calculated. Is stored in the storage unit 42 as a positional deviation amount 425 (step S237). The position difference value is a signed difference value between the XY coordinates of the reference point position 490 and the XY coordinates of the temporary reference point position 510. Then, the rotational angle θ and / or expansion / contraction magnification κ is displaced with predetermined resolutions Δκ and Δθ, and the processing is performed until the repetition processing is completed for all combinations of the rotation angle θ and / or expansion / contraction magnification κ (step S238). Step S239).

Here, each time the process is repeated, the evaluation value f stored in the storage unit 42 is compared with the distance index value calculated most recently. If the distance index value is smaller, the evaluation value f is newly set. In addition to substituting a correct distance index value, the positional deviation amount 425 is updated using the rotation angle θ and / or the scaling factor κ and the coordinates (O _x , O _y ) of the temporary reference point position. That is, since the positional deviation amount 425 stored in the storage unit 42 is sequentially updated, it is not necessary to secure a memory space corresponding to the total number of combinations of the rotation angle θ and the expansion / contraction magnification κ. Therefore, even if the number of combinations of the rotation angle θ and the expansion / contraction magnification κ is increased (that is, even when the resolution is increased), only the memory space that can store the positional deviation amount 425 is required, and the memory capacity to be used is increased. Never do.

  By repeating the processing from step S232 to step S239, the positional deviation amount 425 when the minimum evaluation value f is obtained for all combinations of the expansion / contraction magnification κ and the rotation angle θ can be obtained (step S240). When the positional deviation amount 425 is acquired, the process returns to step S25 to calculate a correction amount for alignment of the stage 21, and alignment is performed.

  As described above, by using the distance index value in the appearance inspection apparatus 1, it is not necessary to secure a voting space corresponding to the combination of the rotation angle θ and the expansion / contraction magnification κ, and only the positional deviation amount 425 is stored in the memory area. By ensuring, it is possible to finally align the substrate 6 to be inspected. Therefore, for example, even when the resolution of the rotation angle θ and the expansion / contraction magnification κ is increased in order to perform positioning with high accuracy, it is not necessary to secure a large memory capacity, and the memory capacity does not matter, so even when the resolution is increased. The amount of misalignment can be acquired, and the substrate can be reliably aligned.

  Next, a second embodiment will be described with reference to FIGS. 10 and 11. The first embodiment is different from the first embodiment in that the misregistration image position determination unit 514 is provided in the misregistration amount acquisition unit 401 in FIG. 10. The remaining points are the same, and the same configuration as that in FIG. 3 is the same. The reference numerals are used and the description is omitted. Also, FIG. 11 is different in that it has step S337, and the remaining points are the same. Therefore, the same steps as those in FIG. In other words, the second embodiment is characterized in that the validity of the feature image position acquired by the feature image position acquisition unit 412 is determined. The second embodiment will be described below.

  In the second embodiment, the misregistration image position determination unit 514 is provided in the misregistration amount acquisition unit 401. The unauthorized feature image position determination unit 514 has a function of determining whether or not the feature image position acquired by the feature image position acquisition unit 412 is appropriate. That is, when it is determined that the feature image position acquired by the feature image position acquisition unit 412 is an illegal feature image position due to erroneous detection or the like, the feature image position is deleted so as not to be used in subsequent processing.

In the second embodiment, step S337 is added as a processing flow. In step 337, the validity of the acquired feature image position is determined. Specifically, the value of the distance between the estimated reference point position and the temporary reference point position determines that the estimated reference point position, the temporary reference point position, and the feature image position are incorrect, and deletes the incorrect feature image position. Next, the setting of the threshold value corresponding to the estimated reference point position when it is larger than the threshold value set based on the standard deviation of the distance will be described. Assuming that the error for detecting the position of the feature image is a uniform random number of [−0.5, +0.5], the standard deviation is about 0.28555. Therefore, if all the positions of the feature images can be detected normally, the estimated reference point position variation is also within 1.75 (0.50000 ÷ 0.28555) times the standard deviation. Can be expected. As a result, it is experimentally possible to obtain a result that it is appropriate to use a value obtained by multiplying the standard deviation of the distance between the estimated reference point position and the temporary reference point position by a factor of 1.40 to 1.50. I was able to. Further, as an optimum value, it is appropriate to use a threshold value obtained by multiplying the standard deviation of the distance between the estimated reference point position and the temporary reference point position by √2 (route 2).
Thus, by setting the threshold value and deleting the illegal feature image position, the illegal feature image position is not used in the subsequent iterative processing. As a result, it is possible to prevent a decrease in the accuracy of the positional deviation amount when the position acquired as the feature image position is erroneous detection or the like.
In addition, it is possible to adjust the precision which deletes an illegal feature image position by setting a coefficient suitably in a threshold value. For example, by increasing the coefficient, the criterion for determining whether or not the feature image position is incorrect can be relaxed.
<Modification>

  Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and various modifications can be made.

  Further, in the above-described embodiment, the position shift amount is detected for alignment in the semiconductor substrate appearance inspection method, but the object to be inspected is not limited to this. For example, in the inspection of a printed wiring board, a glass substrate for a flat panel display, a photomask, or the like, the above-described positional deviation amount detection method and appearance inspection method may be used.

  In the above embodiment, the operator designates a reference image that is a characteristic image region for the image displayed on the display unit, but the present invention is not limited to this. For example, a known alignment mark can be used as the reference image.

  Further, the positional deviation amount detection technique may be used in fields other than the appearance inspection, and may be used for exposure of, for example, a semiconductor substrate, a printed wiring board, a wiring pattern of a flat panel display board, or the like.

  Moreover, in the said embodiment, although the normalized correlation method is used as a suitable example for acquiring a feature image position, it is not restricted to this, You may use another known method.

  The visual inspection method and the positional deviation detection method according to the present invention require alignment with high resolution in the XY and rotational directions with respect to a reference position in order to perform visual inspection of a substrate to be inspected. It is useful in such cases.

DESCRIPTION OF SYMBOLS 1 Appearance inspection apparatus 2 Board | substrate holding | maintenance part 3 Imaging part 4 Computer 5 Reference board 6 Inspected board 21 Stage 22 Stage drive part 41 Calculation part 42 Storage part 44 Display part 45 Input part 49 Reference image 51 Target image 401 Misregistration amount acquisition part 402 Reference table generation unit 403 Inspection unit 412 Feature image position acquisition unit 413 Estimated reference position setting unit 414 Distance index value calculation unit 415 Temporary reference point position setting unit 421 Reference image data 422 Reference image position coordinates 423 Reference point position coordinates 424 Reference table 425 Position shift amount 490 Reference point position 491a, 491b, 491c, 491d Reference image 492a, 492b, 492c, 492d Reference image position 510 Temporary reference point position 511a, 511b, 511c, 511d Feature image 512a, 512b, 512c, 512d Feature image Position 513 , 513b, 513c, 513d estimated reference point position 514 unauthorized feature image position determination unit W substrate θ rotation angle κ stretching magnification

Claims (4)

Relative position information of a reference image position indicating a position on the stage corresponding to a plurality of reference images acquired by imaging a reference substrate placed on the stage, and a reference point position that is the rotation center of the stage As a reference table,
A feature image position obtaining step for obtaining a feature image position representing a position on the stage of a feature image whose image information approximates to the reference image obtained by imaging the substrate to be inspected;
With reference to the reference table, an estimated reference point position setting step for setting a position on the stage corresponding to the reference point position viewed from each feature image position as an estimated reference point position;
A temporary reference position setting step for setting an average position of each of the estimated reference point positions acquired in the estimated reference point position setting step as a temporary reference point position;
A distance for calculating the standard deviation of the distance between the estimated reference point position and the temporary reference point position or the variance of the distance for each rotation angle while displacing the rotation angle around the reference point position with a predetermined resolution. An index value calculation step;
A positional deviation that obtains the standard deviation of the distance calculated in the distance index value calculating step or the position difference value between the rotation angle and the temporary reference point position and the reference point position when the dispersion of the distance is minimized. A displacement acquisition method comprising: an amount acquisition step. A fraudulent feature image position determination step for determining whether the feature image position acquired by the feature image position acquisition step is acquired as a position of a feature image approximated by image information of the reference image;
The fraudulent feature image position determining step corresponds to the estimated reference point position when a distance value between the estimated reference point position and the temporary reference point position is larger than a threshold set based on a standard deviation of the distance. 5. The positional deviation amount detection method according to claim 4, wherein the feature image position to be determined is determined to be illegal, and the illegal feature image position is deleted.   The positional deviation amount detection method according to claim 4, further comprising a pre-alignment step of performing pre-alignment on the substrate to be inspected before the feature image position acquisition step. Using the positional deviation amount detection method according to any one of claims 1 to 3,
Based on the displacement amount acquired by the displacement amount acquisition step, the inspection substrate is aligned, and an inspection step of performing an appearance inspection of the inspection substrate,
An appearance inspection method comprising:

Images