JPH07148907A - Method and device for registering image recognizing position - Google Patents

Abstract

PURPOSE:To provide an image recognizing postion registration technology which can register a position with high accuracy without requiring much time for the preparation of position registration by extracting and using a pattern usable for the registration of a base to be processed. CONSTITUTION:When a screen printer is started, first a dictionary zone for a screen mask is set temporarily. Cameras 6a and 6b are moved to a printing station S2 to obtain an image of screen mask 4. A particular pattern in the image, which is not like other patterns in the periphery and can be utilized as a position registration mask, is extracted, and the given zone including the pattern is stored in a RAM 7b of a main control section 7 as a temprary dictionary zone. Then, a temporary printed board 1 of the same kind as a base to be processed is vacuum adsorbed on a table 2 and disposed in a position registration station S1.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition alignment device for positioning a substrate based on image processing of the substrate surface.

[0002]

2. Description of the Related Art Solder is printed on a desired area of a printed circuit board through a screen mask, electronic devices such as semiconductor devices and resistors or capacitors are mounted or mounted, and a photomask is superposed on a semiconductor wafer (semiconductor substrate). It is necessary to accurately align these substrates when photolithography is performed.

Conventionally, alignment has been performed by the following method. That is, an isolated mark is provided in a specific area of the substrate for alignment. On the other hand, the position information of the specific area and the information about the mark shape are stored and set in the image processing unit as reference information. The substrate is conveyed to a position facing an image reading means such as a camera, the read image of the substrate surface is binarized, image processing is performed to detect the mark, and the center position, the center of gravity position, and the like are obtained. The positional deviation is calculated by comparing the positional information with the reference information, and the position of the substrate is corrected based on the deviation. In this way, the substrate is accurately aligned.

In the above method, in order to arrange the isolated mark, it is necessary to have a region around the mark where there is no pattern that is mistaken for the mark. However, as the definition of printed circuit boards and semiconductor wafers has become higher and higher, there have often been situations where there is no space for providing isolated marks.

Therefore, there has been proposed a method for performing alignment using an arbitrary pattern on the substrate surface without providing a special alignment mark. In this method, a pattern to be used as a reference for performing alignment is extracted in advance, and a certain area including the pattern is determined.

The above area of the member to be aligned with the substrate to be processed is read, and its image is stored in the image processing section as a dictionary area (also referred to as a reference model). The substrate to be processed is conveyed to the image reading area, a binary image is created,
By performing image processing, the ratio of the area of the bright part in the processing area to the area of the processing area on the surface of the substrate to be processed (the area ratio of the bright part) is compared with the area ratio of the bright part of the dictionary area, and the bright part is compared. The area having the same area ratio is searched for and the matching area is aligned.

A method for searching for an area in which the area ratio of the dictionary area and the light area is the same is called grayscale pattern matching, grayscale pattern matching, or autocorrelation function method.
Image processing device for computer and its application "(by Eiji Mizutani) (New Technology Communications, Inc." O plus E ")
August 1991 No. 141 pp 141-15
It is introduced in 2) etc.

In the grayscale pattern matching, a normalized correlation coefficient (match rate) is used as the degree of matching between two images. The match rate takes a value from +1 to -1, and when it is +1 it indicates that the two images are completely in agreement.

[0009]

When performing the alignment by the light and shade pattern matching, the alignment is performed by matching the pattern of the dictionary region with the pattern of the region (corresponding region) corresponding to the dictionary region of the surface of the substrate to be processed. It can be carried out. However, when the lightness and darkness of the pattern in the dictionary area and the imaged pattern in the corresponding area on the surface of the processing target substrate are reversed, the match rates almost do not match.

For example, the area ratio of the bright part of the dictionary area is 0.7.
Then, when the area ratio of the bright part of the inverted image pattern is 0.3, the match rate cannot be calculated as it is. Such a thing occurs in a screen printing machine as an example. That is, when the screen mask is read by the camera, the image is dark because the light is transmitted through the apertures, and the image is bright because the light is reflected at the portions other than the apertures. On the other hand, on the printed circuit board, the light is strongly reflected at the land of the wiring that prints the solder, so the image is bright at this part.
It becomes dark at other insulating plates.

In such a case, first, the dictionary area of the screen mask is determined and its representative point is set. Here, the representative point may be an arbitrary point in the corresponding area and is used for specifying the position. Next, in determining the dictionary area of the printed board, the operator moves the printed board so that the camera can read the area while looking at the monitor screen on which the image of the printed board to be processed is displayed.
When the pattern of the dictionary area matches the pattern of the corresponding area, the cursor on the monitor screen is visually aligned with the position corresponding to the representative point of the dictionary area of the screen mask set in advance.

The corresponding area of the printed circuit board and its representative point are stored in the RAM of the image processing section as reference information. In printing solder, the reference information is used to read the surface of each board that is sequentially conveyed by a camera and perform alignment using representative points. In other words, instead of directly using the dictionary area in the screen mask for alignment,
The corresponding area of the printed circuit board is replaced with the dictionary area to perform the alignment indirectly.

The operation of replacing the corresponding area of the printed circuit board with the dictionary area is performed manually, which takes time. Further, it is easy to make an error in setting the representative point due to the movement of the cursor, and there is a limit in improving the accuracy of alignment. If there is a big mistake in setting the representative point by moving the cursor,
All the products produced by using the dictionary area have low accuracy, and the yield is extremely low.

It is an object of the present invention to provide an image recognition alignment technique that does not require time to prepare for alignment and is capable of performing alignment with high accuracy.

[0015]

An image recognition alignment method of the present invention obtains image data of a surface of a substrate to be processed and a surface of a reference member which serves as a reference for aligning the substrate to be processed. In the image recognition alignment method of comparing both image data and aligning the substrate to be processed, an image of a dictionary area including a pattern that is a part of the surface of the reference member and can be a reference for alignment is displayed. Reading, creating regular image data and storing it as temporary dictionary data; reading the image of the corresponding area on the surface of the processing target substrate corresponding to the dictionary area to obtain image data of the corresponding area; and the temporary dictionary data And a comparison step of comparing the lightness / darkness relationship between the image data of the corresponding area and the lightness / darkness relationship between the temporary dictionary data and the image data of the corresponding area. A step of storing the temporary dictionary data as dictionary area data, and storing image data obtained by reversing the brightness of the temporary dictionary data as dictionary area data when the light-dark relationship is a reverse relationship; and an image of the corresponding area of the processing target substrate. To form processing target image data and align the processing target image data with the dictionary area data.

In the comparing step, in the comparing step, the luminance on the predetermined reference line of the dictionary area is obtained from the temporary dictionary data to obtain a normal luminance distribution, and the corresponding area is obtained from the corresponding area image data. And obtaining the luminance distribution on the line corresponding to the reference line, to obtain the corresponding region luminance distribution.

Further, in the comparing step, in the comparing step, a step of obtaining an inversion luminance distribution obtained by inverting the light-dark relation of the normal luminance distribution, and the normal luminance distribution and the corresponding area luminance distribution are added. And a step of adding the inverted luminance distribution and the corresponding area luminance distribution.

The image recognition alignment apparatus of the present invention acquires image data of the surface of a substrate to be processed and the surface of a reference member which serves as a reference for aligning the substrate to be processed,
In an image recognition alignment device that performs image processing to align the substrate to be processed, image data of a dictionary area including a pattern that is a part of the surface of the reference member and can be a reference for alignment, or its brightness. A temporary dictionary storage means for storing the image data in which the relationship is inverted as a temporary dictionary area, and a corresponding area image data storage means for storing an image of the corresponding area on the surface of the processing target substrate corresponding to the dictionary area, Comparing means for comparing the brightness relation between the image data stored in the temporary dictionary storage means and the image data stored in the corresponding area image data storage means, and the temporary dictionary storage means in the temporary dictionary storage means based on the result of the comparing means. Setting means for adopting stored image data or image data obtained by reversing the brightness of the stored image data as dictionary area data; Including the image data of those areas, and an arithmetic means for performing alignment by using the image data employed as the dictionary area.

Further, a switch means for approving the adoption of the authorization means may be included.

[0020]

A substrate surface is obtained by extracting a pattern that can be used for alignment from patterns formed on a reference member that serves as a reference for aligning the substrate to be processed, and using the pattern for alignment. There is no need to provide a special pattern for alignment in.

The regular image data of the reference member is stored as a temporary dictionary. Even when the lightness / darkness relationship between the substrate to be processed and the reference member is inverse to each other, the lightness / darkness relationship between the two image data is reversed by inverting the lightness / darkness of the image data of the area including the pattern that can be used for the alignment of the reference member Can be matched.

The image data of the reference member in which the light-dark relationship is matched is stored as dictionary area data, and thereafter, the light and shade pattern matching is performed between the image data of the dictionary area and the image data of the substrate to be processed, whereby the position is quickly aligned. Will be able to do.

By determining the brightness distribution of the image data and comparing the profiles of the brightness distribution, it is possible to easily determine whether or not the light-dark relationship is the same.

[0024]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below based on an embodiment shown in the drawings. FIG. 1 shows a schematic configuration of a screen printing machine using the image recognition alignment apparatus of the present invention.

The printed circuit board 1 is placed on the table 2. The table 2 can be vacuum-sucked to the upper surface of the printed circuit board 1 and can be finely translated in the X-axis direction, which is the left-right direction in the figure, and the Y-axis direction, which is perpendicular to the X-axis, which is the direction perpendicular to the paper surface. Further, it can be finely rotationally moved in the θ direction, which is the rotational direction about the Z axis (the rotational direction in the XY plane). These minute movements are performed by the servo motor 10b.

The table 2 is movable on the rail 3 in the left-right direction (X axis) in the figure by the servomotor 10a, and is stopped on the left side (positioning station S1) to position the printed circuit board 1. Stop on the right side (printing station S2) and perform solder printing. At the position of the printing station S2, a screen mask 4 is arranged slightly above the printed circuit board 1 placed on the table 2.

The screen mask 4 is provided with a squeegee 5 which slides on the upper surface in the left-right (X-axis) direction of the drawing by a servomotor 10d. By sliding the squeegee 5 in the left-right direction, solder can be printed on the surface of the printed board 1 through the openings of the screen mask 4.
Image reading cameras 6a and 6b which can reciprocate between the stations S1 and S2 by a servomotor 10c are arranged above the stage 2.

The operation of the servomotors 10a to 10d is performed via the controller 11 and the driver 12 based on a command from the main controller 7. Each of the motors 10a to 10d is provided with an encoder (not shown) for feeding back the movement amount of the table 2 or the like to the controller 11 so that accurate control can be performed.

An image monitor 8 for displaying image signals obtained from the cameras 6a and 6b is provided, and a touch panel section 9 for stopping operation of the apparatus and inputting various operation signals is provided in the box of the monitor. ing.

The main control unit 7 is an R storing a program.
A RAM 7b for temporarily storing image signals and the like obtained from the OM 7a and the cameras 6a and 6b, and a CPU 7c and camera 6a and 6b for processing and processing image signals and the like obtained from the cameras 6a and 6b according to a program stored in the ROM 7a.
I / O unit 7e for receiving signals from the controller 11
An input / output unit 7f for transmitting / receiving data to and from the image monitor 8 and an input / output unit 7d for transmitting / receiving data to / from the image monitor 8 are provided.

2A, 2B and 2C show the camera 6
The example of the image which reversed the surface image of the screen mask 4 and the printed circuit board 1 read by a or 6b, and the surface image of a screen mask is shown. These images are displayed on the image monitor 8.

The operation and function of the screen printing machine shown in FIG. 1 will be described below with reference to the flow chart of FIG. When the screen printing machine is activated, first, step s1 for temporarily setting the dictionary area of the screen mask is executed. In step s1, the cameras 6a and 6b are moved to the printing station S2 to obtain an image of the screen mask 4. A peculiar pattern in the image, that is, a pattern that does not have a similar pattern in the periphery and can be used as an alignment mark is extracted, and a predetermined area including the pattern is stored in the RAM 7b of the main control unit 7 as a temporary dictionary area. To do.

Next, a temporary printed circuit board 1 of the same type as the substrate to be processed is vacuum-sucked to the table 2 and placed in the alignment station S1. The cameras 6a and 6b are moved to the alignment station S1, the area of the printed board 1 corresponding to the temporary dictionary area of the screen mask 4 stored in the RAM 7b is read, and the image data thereof is stored in the RAM 7b (step s2).

Image data examples of the temporary dictionary area of the screen mask 4 and the corresponding area of the printed circuit board 1 stored in the RAM 7b are shown in FIGS. 2 (A) and 2 (B), respectively. still,
The cameras 6a and 6b read an image in a wider area than the images shown in FIGS. 2A and 2B. Figure 2
(A) and FIG. 2 (B) show the regions extracted as dictionary regions from this.

Here, an image with many bright areas of the image is called a positive screen and an image with many dark areas is called a negative screen. In addition, when viewing the periphery of the screen, a positive screen may be used if the surroundings are bright, a negative screen may be used if the surroundings are dark, or vice versa.

2A and 2B, the camera 6a,
6b, which was read on the other hand. The reason why the two cameras 6a and 6b are provided in FIG. 1 is to see the two points on the diagonal of the printed circuit board 1 and the screen mask 4, respectively, so as to improve the accuracy of alignment.

In the alignment process, the same processing is performed on the images read by these two cameras, so that FIG. 2 shows only the image obtained by one of the cameras.
The temporary dictionary area C1 of the screen mask 4 having the positive screen shown in FIG. 2A (hereinafter, simply referred to as the temporary dictionary area C1
(Referred to as) is a region where light is reflected by the screen mask 4, and a dark portion with hatching is a portion where light is transmitted through an opening provided in the screen mask 4.

The bright portion of the corresponding area C2 (hereinafter simply referred to as the corresponding area C2) of the printed circuit board having the negative screen shown in FIG. 2B is a metal wiring portion on which solder is to be printed, The dark area is an insulating resin area where solder is not printed. There are no holes, but there is light and dark due to the light reflectance.

FIG. 2C shows a reversal area C3 in which the contrast relation of the temporary dictionary area C1 which is a positive screen is reversed. The image processing for screen inversion is used as a screen inversion function in a word processor or the like, and there is the following method as an example. That is, the average value of the brightness of the bright area and the dark area is determined. Next, when the brightness of each pixel forming the screen, that is, the brightness is subtracted from this average value, it has a minus (negative) value in a bright area and a plus (positive) value in a dark area. Assign positive and negative signs to each pixel,
If the plus sign is bright and the minus sign is dark, the resulting screen is a reverse screen.

Next, in step s3, the temporary dictionary area C1
Then, it is determined whether or not both the corresponding area C2 and the corresponding area C2 are positive screens. As shown in FIG. 2, since both areas have a positive and negative relationship, the process proceeds to step s4. In step s4, it is determined whether both areas C1 and C2 are negative screens.
Since the determination is negative, the process proceeds to step s5.

In step s5, the screen of the temporary dictionary area C1 is reversed, and this reversed screen is stored in the RAM 7b of the main controller 7 as a dictionary area. The reverse screen area C3 is the corresponding area C
It agrees with the light-dark relationship of 2. The touch panel section 9 may be provided with an approval (setting) switch, and the dictionary area may be stored in the RAM 7b by pressing the approval switch.

On the other hand, step s3 or step s4
If both screens are positive or negative, the temporary dictionary area C1 is set as the dictionary area in the RAM 7b of the main control unit 7 as it is in step s6. That is, through the steps s3 to s6, the dictionary area and the corresponding area C2
The light and dark relations can be matched. In addition, I explained the case of creating a reverse image only when both screens do not match positive or negative, but first, create a reverse image and decide whether to use a regular image or a reverse image based on whether they match or not match. May be.

The details of steps s3 and s4 will be described later. Next, in step s7, the temporary printed circuit board 1 on the table 2 is removed, and the printed circuit board 1 on which the solder is to be printed is automatically carried in from a previous device (not shown) by a conveyor or the like and placed on the table 2. To do. Further, the printed circuit board 1 is slightly moved and set at a predetermined position while being monitored by the cameras 6a and 6b.

In step s8, the table 2 and the camera 6
A and 6b are respectively moved to the printing station S2, and the screen mask 4 and the printed circuit board 1 are aligned. The screen mask 4 is removed, the printed circuit board 1 is slightly moved while looking at the screen of the monitor 8, and the reversal region C3 of the dictionary region and the corresponding region C2 on the surface of the printed circuit board 1 are removed.
Pattern matching is performed. The alignment is completed when the match rate reaches the maximum.

Solder printing is performed in step s9. First,
Place the screen mask 4 in the specified position and press the squeegee 5
To the left. This allows the screen mask 4
The solder paste above is the opening of the screen mask 4 (see FIG.
Screen printing is performed on the metal wiring portion of the printed circuit board 1 (the bright portion of the area C2 in FIG. 2) through the dark portion of the area C1 in FIG.

After the printing is completed, the table 2 is returned to the alignment station S1 and the printed board 1 is carried out to the subsequent apparatus by the conveyor (step s10). When the squeegee 5 or the like is returned to the initial position, it is judged whether or not there is a subsequent printed circuit board 1 on which the same kind of solder printing should be performed (step s11), and if there is, the process returns to step s7 to perform solder printing in the same process as above. To do. If there is no subsequent printed circuit board, all processing operations are completed.

As described above, the image data of the dictionary area is first set so that the light-dark relationship between the image data of the dictionary area of the screen mask and the image data of the corresponding area of the printed circuit board is set, and thereafter, The position of the substrate can be quickly adjusted without determining the light-dark relationship.

Hereinafter, a method of determining whether positive or negative is performed in steps s3 and s4 will be described. First
In this method, the worker visually compares the temporary dictionary area C1 displayed on the image monitor 8 with the corresponding area C2 or the inverted area C3 to determine whether the image is a positive screen or a negative screen.
In this case, since it is possible to determine whether each area is a positive screen or a negative screen simply by looking at the relationship between lightness and darkness of the image, it is not necessary to utilize the area ratio of lightness and darkness.

Next, the second method will be described with reference to FIG. 5 (A) is a temporary dictionary area C1b, FIG. 5 (B)
Shows an example of the corresponding area C2b, and FIG. 5C shows an example of an inversion area C3b obtained by inverting the contrast relation of the temporary dictionary area C1b. 5D, 5E, and 5F are temporary dictionary areas C, respectively.
1b, the corresponding area C2b, and the inversion area C3b show a luminance distribution on a line passing through substantially the center.

For example, the temporary dictionary area C1b shown in FIG.
Since the central part is dark and the periphery is bright, the brightness distribution is concave as shown in FIG. On the contrary, the brightness distribution is as shown in FIGS. 5 (E) and 5 (F) because the central portion of the corresponding area C2b and the inversion area C3b in FIGS. 5 (B) and 5 (C) is bright and the peripheral portion is dark. Becomes convex.

As described above, the profiles of the brightness distributions of the positive screens or the negative screens are the same or similar. On the other hand, in the case of the combination of the positive screen and the negative screen, the profiles of the luminance distribution have a shape in which the top and bottom are inverted. Therefore, by visually comparing the brightness distribution profiles, it is possible to easily determine whether positive or negative, or a combination of positive and negative.

Next, the third method will be described. Figure 5
5G is an added luminance distribution obtained by adding the luminance distribution of the temporary dictionary area C1b of FIG. 5D and the luminance distribution of the corresponding area C2b of FIG. 5E, and FIG. The luminance distribution of the inversion area C3b of (F) and the corresponding area C2b of FIG. 5 (E)
3 shows an added luminance distribution obtained by adding the luminance distribution of FIG.

As described above, when the positive screens or the negative screens are added together, the profile of the luminance distribution is exaggerated. On the contrary, the profile of the added luminance distribution in the case of the combination of the positive screen and the negative screen becomes almost flat. Therefore, by visually observing the profile of the added luminance distribution, it is possible to determine whether it is positive or negative, or a combination of positive and negative.

Alternatively, for the luminance of the same portion, the value before addition may be subtracted from the value after addition by the CPU 7c to automatically determine. That is, the brightness of the bright part is +1
The brightness of the dark part is assumed to be -1. In the case of positive screens or negative screens, the value before addition of the bright portion is +1 and the value after addition is +2. Therefore, when the value before addition is subtracted from the value after addition, it becomes +1.

In the case of a combination of a positive screen and a negative screen, if one is a bright part and the other is a dark part, the values before addition are +1 and -1, and after addition are 0. Therefore, if the value of the bright part before addition is subtracted from the value after addition, it becomes -1. Therefore, by determining whether the sign is positive or negative,
The following cases are examples of positives or negatives for which the relationship of screens can be easily determined.

Half-etching is performed on the stainless screen mask along the boundary line between the light portion and the dark portion to prepare a temporary screen mask in which the etch groove is filled with a substance having a low reflectance.

FIG. 5A shows a temporary dictionary area C1a obtained by reading the temporary screen mask thus formed with a camera.
Indicates. In this way, the etching groove portion is dark and the other portions are bright. FIG. 5C shows a cross section of the temporary screen mask taken along the line AA of the temporary dictionary area C1a.
A substance having a low reflectance is filled in the etching groove 13 that is etched to the extent that it does not penetrate the screen mask. By darkening the outside of the dark portion of the temporary dictionary area C1a by image processing, it is possible to match the light-dark relationship with the corresponding area C2a of the printed circuit board shown in FIG. 4B.

When the lightness / darkness relationship of the corresponding area of the printed circuit board is opposite, the lightness / darkness relationship can be matched by darkening the inside of the dark portion by image processing. As described above, by darkening a desired region by image processing, it is possible to produce either a positive or negative screen.

Alternatively, the temporary dictionary area and the corresponding area may match in the light / dark relationship as they are. In these cases, if the determination is made in steps s3 and s4,
Since both screens are positive or negative, the temporary dictionary area can be used as a dictionary area without inverting.

In the first to third methods described above, after visually confirming, the switch of the touch panel section 9 may be pressed to set the dictionary area. However, in the third method, the judgment processing program is stored in the ROM 7a. It is also possible to store the dictionary area and automatically set the dictionary area based on the above determination result.

The present invention has been described above with reference to the embodiments.
The present invention is not limited to these. For example, it will be apparent to those skilled in the art that various modifications, improvements, combinations, and the like can be made.

[0062]

As described above, according to the present invention, it is possible to obtain an image recognition alignment device which does not take much time to prepare for alignment and which can perform alignment with high accuracy.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a screen printing machine using an image recognition alignment device of the present invention.

FIG. 2 is a plan view showing an image of the surface of a screen mask and a printed circuit board read by a camera of the screen printing machine shown in FIG. 1 and an inverted image of the image of the screen mask.

FIG. 3 is a flowchart for explaining image processing and solder printing processing of the screen printing machine shown in FIG.

FIG. 4 is a plan view showing a surface image of a screen mask and a printed circuit board which are half-etched, and a sectional view of the screen mask.

FIG. 5 is a plan view showing a surface image of a temporary dictionary area, a corresponding area, and an inversion area, and a graph showing a luminance distribution.

[Explanation of Codes] 1 Printed circuit board 2 Table 3 Rail 4 Screen mask 5 Squeegee 6a, 6b Camera 7 Main control unit 7a ROM 7b RAM 7c CPU 7d to 7f Input / output unit 8 Image monitor 9 Touch panel unit 10a to 10d Servo motor 11 Controller 12 Driver S1 Alignment station S2 Printing station C1 Temporary dictionary area C2 Corresponding area C3 Inversion area

Front Page Continuation (72) Inventor Masashi Hirosawa 5-2 Koyodai, Ryugasaki, Ibaraki Hitachi Techno Engineering Co., Ltd. Development Laboratory (72) Inosao Abe 5-2 Koyodai, Ryugasaki, Ibaraki Hitachi Techno Engineering Co., Ltd. Ryugasaki Factory

Claims (5)

[Claims] 1. Image data of a surface of a substrate to be processed and a surface of a reference member serving as a reference for aligning the substrate to be processed are acquired, both image data are compared, and the image data of the substrate to be processed is obtained. In an image recognition alignment method for performing alignment, an image of a dictionary area that includes a pattern that is a part of the surface of the reference member and can be a reference for alignment is read, regular image data is created, and stored as temporary dictionary data. And a step of reading an image of a corresponding area on the surface of the processing target substrate corresponding to the dictionary area to obtain image data of the corresponding area, and comparing a light-dark relationship between the temporary dictionary data and image data of the corresponding area. When the comparison step and the contrast relation between the temporary dictionary data and the image data of the corresponding area match, the temporary dictionary data is set as dictionary area data, and When the relationship is an inversion relationship, a step of storing image data in which the lightness and darkness of the temporary dictionary data is inverted as dictionary area data, and reading an image of the corresponding area of the processing target substrate to form processing target image data, An image recognition alignment method including a step of aligning the image data to be processed with the dictionary area data. 2. A step of obtaining luminance on a predetermined reference line of the dictionary area from the temporary dictionary data to obtain a normal luminance distribution in the comparing step; and a step of obtaining the reference of the relevant area from the corresponding area image data. 2. The image recognition alignment method according to claim 1, further comprising a step of obtaining luminance on a line corresponding to the line to obtain a corresponding region luminance distribution. 3. The comparison step further comprises a step of obtaining an inverted luminance distribution by inverting a contrast relationship of the regular luminance distribution, a step of adding the regular luminance distribution and the relevant area luminance distribution, and the inversion. The image recognition alignment method according to claim 2, further comprising the step of adding a luminance distribution and the corresponding area luminance distribution. 4. The image data of the surface of the substrate to be processed and the surface of a reference member serving as a reference for aligning the substrate to be processed are acquired, and image processing is performed to align the substrate to be processed. In the image recognition alignment device for performing, the image data of the dictionary area including a pattern that is a part of the surface of the reference member and can be the reference of the alignment, or the image data in which the brightness relationship is inverted is stored as the temporary dictionary area. A temporary dictionary storage means for storing the corresponding area image data storage means for storing an image of the corresponding area on the surface of the processing target substrate corresponding to the dictionary area, the image data stored in the temporary dictionary storage means and the corresponding Comparing means for comparing the brightness relation with the image data stored in the area image data storing means, and the temporary dictionary storing means based on the result of the comparing means Setting means for adopting stored image data or image data obtained by inverting the brightness of the image data as dictionary area data, image data of the corresponding area of the substrate to be processed, and image data adopted as the dictionary area are used. An image recognition alignment device including a calculation means for performing alignment. 5. The image recognition alignment apparatus according to claim 4, wherein the recognizing means includes a switch means for approving that the image data stored in the temporary dictionary storage means is adopted as dictionary area data.