JP4382649B2 - Alignment mark recognition method, alignment mark recognition device and joining device - Google Patents

Description

  The present invention relates to an alignment mark recognition method for recognizing the position of an alignment mark by comparing an alignment mark image captured by a camera with template data registered in advance.

  This technique can be applied to a bonding apparatus or the like.

  Usually, an alignment mark (a size is about 50 μm to 100 μm, there are various shapes such as a cross shape, a square, etc.) placed on a semiconductor chip or a printed circuit board is very small and has a certain thickness.

  FIG. 12 is a diagram illustrating an example of alignment marks. The alignment mark shown in FIG. 12 is an example of a cross-shaped alignment mark.

  There are various techniques for recognizing the position of such an alignment mark. For example, in Patent Document 1 and Patent Document 2, slit-shaped spot light is projected onto a wafer having an alignment mark or the like, and diffracted light or scattered light from a mark pattern irradiated to the spot light is detected, and alignment is performed. A technique for recognizing a mark or the like is described. The techniques described in Patent Documents 1 and 2 are intended to detect the best focus position by recognizing a focus measurement mark.

  In addition, the alignment mark shape is obtained by imaging the alignment mark with a camera, analyzing the obtained image, and obtaining the alignment mark shape at a certain slice level, and comparing the obtained alignment mark shape with the template data. There is a technology to recognize the position. In this technique, the obtained alignment mark shape and template data are two-dimensional (hereinafter also referred to as 2D) data.

FIG. 13 is a diagram showing a slice image of the alignment mark. FIG. 13 shows a shape when the alignment mark of FIG. 12 is sliced at a certain slice level. The dark part in FIG. 13 is the shape of the alignment mark at this slice level.
JP 09-082620 A Japanese Patent Laid-Open No. 06-216004

  Thick alignment marks such as plating are difficult to sharpen the edges in the material and manufacturing process, and the edges are rounded, so it is not clear where the edges are. May end up. Also in the example of FIG. 12, the edge portion of the alignment mark is rounded. Such roundness of the edge is not uniform due to deformation or scratches in the manufacturing of the alignment mark, and usually there is variation among the alignment marks.

  The technology for recognizing the position of the alignment mark from the shape of the 2D alignment mark described above can recognize the alignment mark at the true position when the alignment mark has a dent or chip at the acquired slice level. There is a possibility of disappearing.

  An object of the present invention is to solve the above problems and to provide an alignment mark recognition method capable of recognizing the position of an alignment mark with high accuracy even with an alignment mark whose shape is not completely uniform. To do.

  In order to solve the above-described problems, the present invention acquires alignment mark three-dimensional (hereinafter also referred to as 3D) data from a plurality of image data, and collates it with three-dimensional template data. It is characterized by recognizing the position of.

  In the present invention, first, images at a plurality of slice levels are acquired by imaging the alignment mark by changing the height of the camera. The shape of the alignment mark is specified for the image captured at each slice level. Next, by generating a 3D model of the alignment mark based on the shape of the alignment mark obtained from the image at each slice level, and collating it with the 3D template data of the alignment mark registered in advance, Recognize the position of the alignment mark.

  In addition, the feature point indicating the feature of the position and shape of the edge of the alignment mark for each slice level, instead of matching the three-dimensional template data using the three-dimensional model generated from the shape of each slice level. The positions of the alignment marks may be recognized by comparing the positions of the edges and feature points with the three-dimensional template data registered in advance.

  Also, rather than simply creating a three-dimensional model using the luminance information of the captured image, the image captured in color is decomposed into images of a plurality of colors (for example, three primary colors of R, G, and B) The position of the alignment mark may be recognized by separately creating a three-dimensional model for each color and collating it with the three-dimensional template data.

  Also, instead of acquiring images at multiple slice levels by changing the camera height, images at multiple slice levels are acquired by changing the focus position of the lens. The position of the alignment mark may be recognized using these images.

  Also, instead of acquiring the three-dimensional data of the alignment mark by imaging with one camera, the three-dimensional data of the alignment mark is acquired by imaging using a plurality of cameras, and the position of the alignment mark is recognized. However, the position can be recognized with high accuracy.

  According to the present invention, it is possible to reduce a position recognition error caused by an unclear outline. In addition, it is possible to reduce position recognition errors caused by illumination conditions and changes in the gloss of the surface of the alignment mark. Furthermore, the difference in the material of the alignment mark can be determined by comparing the information for each color.

  As described above, since the amount of information in the three-dimensional data is larger than that in the two-dimensional data, the stable alignment mark can be recognized even if some information is missing.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[Embodiment 1]
In Embodiment 1 of the present invention, the alignment mark is imaged while changing the height of the camera, the shape of the alignment mark at a plurality of slice levels having different heights is measured from these images, and 3D is obtained from the measurement data. The position of the alignment mark is recognized by generating a model and comparing it with the 3D template data.

  FIG. 1 is a diagram showing a configuration example of an alignment mark recognition apparatus according to Embodiment 1 of the present invention. The alignment mark recognition apparatus in the example of FIG. 1 includes an image processing unit 10 configured by a computer and a software program, an imaging unit 20 including a camera and a lens that captures an alignment mark on the printed circuit board, and a printed circuit board. The alignment stage 30 is configured.

  The image processing unit 10 includes an image input unit 11, a shape recognition processing unit 12, a camera position control unit 13, a measurement data storage unit 14, a 3D model generation unit 15, a 3D correlation calculation unit 16, and 3D template data. And a storage unit 17.

  The image input unit 11 inputs an image of the alignment mark imaged by the imaging unit 20. The shape recognition processing unit 12 measures the shape of the alignment mark from the image input by the image input unit 11 and stores the measurement data in the measurement data storage unit 14. The camera position control unit 13 controls the camera position (height) of the imaging unit 20.

  The 3D model generation unit 15 generates a 3D model of the alignment mark based on the measurement data in the measurement data storage unit 14. The 3D correlation calculation unit 16 performs a correlation process between the generated 3D model and 3D template data stored in advance in the 3D template data storage unit 17 and outputs the result.

  FIG. 2 is an alignment mark recognition processing flowchart according to the first embodiment. Here, the alignment mark is imaged N times by changing the height of the camera. At this time, the focus position of the camera is fixed. Further, the camera height variable i is used to represent the camera height as h [i] (i = 0, 1,..., N−1).

  First, the camera height variable i is initialized (here, i = 0) (step S10). Next, the camera height is set to h [i] (step S11), and the alignment mark is imaged (step S12).

  In the obtained image, an in-focus position is selected, and the shape of the alignment mark in the image is obtained (step S13). On the image, the brightness changes sharply in the in-focus portion of the edge portion, and the out-of-focus portion becomes gentle. Therefore, the shape of the alignment mark can be obtained by performing image analysis and obtaining the position where the change in brightness is steep. The obtained alignment mark shape is stored in the measurement data storage unit 14 as measurement data at the slice level (step S14).

  The camera height variable i is incremented (step S15), and it is determined whether all measurements are completed (i ≧ N) (step S16). If not completed yet, the processing from step S11 to step S15 is performed. repeat. In this way, the shape of the alignment mark at each slice level is obtained in order by changing the height of the camera, and the measurement data is stored in the measurement data storage unit 14.

  If all the measurements are completed in step S16, a 3D model of the alignment mark is generated from the measurement data at each slice level stored in the measurement data storage unit 14 (step S17). Since the measurement data at each slice level is data of the edge portion of the alignment mark, a model of the 3D shape of the alignment mark is completed by stacking them.

  FIG. 3 is a diagram for explaining an example of the 3D model of the alignment mark according to the first embodiment. For example, when a cross-shaped alignment mark as shown in FIG. 12 is viewed from the side, it is as shown in FIG. Here, the height of the camera is changed from high to low in three steps, the alignment mark is imaged, and the slice level at the in-focus position is changed to the slice level 1, slice level 2, slice in descending order of the camera position. Level 3 is assumed.

  The shape of the alignment mark at each slice level is as shown in FIG. By accumulating the shape data, a 3D model of an alignment mark as shown in FIG. 3C can be generated.

  Correlation processing between the generated 3D model and 3D template data registered in advance in the 3D template data storage unit 17 is performed (step S18), and the position with the highest degree of correlation is obtained (step S19). Recognize position.

  Here, a correlation process between the 3D model and the 3D template data will be described.

  FIG. 4 is a diagram illustrating an example of 3D template data. The alignment mark used in the first embodiment has a cross shape, but here, in order to make the explanation easy to understand, it is composed of planes F1 to F5 as shown in FIG. 4 as 3D template data. An example of a structure will be described. Since the bottom surface is a blind spot when imaged from above, it is excluded from the data.

Let the 3D template data be the data of coordinates and brightness constituting each surface. It is good also as a set of point data which constitutes a plane. Let 3D template data be a set of Fn,
Tf = {F1, F2,..., F5}
Fn = {vertex coordinates, surface brightness value}
It is defined as

  The alignment mark measurement data (3D model) is the alignment mark shape information obtained when the camera height is changed by dz in the z direction, that is, the point PD (x, y, z) having a large luminance change. It is a set.

  Here, the 3D template data is set as a set of points PT (x, y, z), and PT (x, y, z) is extracted from the above Tf. At this time, a predetermined luminance value is also set to the coordinates inside the shape of the 3D template data.

  For example, in the case of monochrome, PD (x, y, z) and PT (x, y, z) are represented by brightness of 256 gradations of 0 to 255 with 8-bit data. In the case of color, 24 bit data is used, and R, G, and B are each represented by 8 bits.

  If D (u, v, q, α, β, γ) is D (u, v, q, α, β, γ), D (u, v, q, α, β, γ) u, v, q, α, β, γ) can be expressed by the following equation (1). When the 3D model and the 3D template data are different in size, the following calculation is performed after the 3D template data is enlarged / reduced.

However,
PD (x, y, z): Measurement data (luminance value)
PT (x, y, z): 3D template data (luminance value)
G (α, β, γ): Rotation affine transformation coefficient, G (α, β, γ) ≠ 0.

  In equation (1), the coordinates (u, v, q) and rotational orientation (α, β, γ) that minimize D (u, v, q, α, β, γ) are most correlated. It can be said. From these results, the exact position of the alignment mark on the alignment stage 30 shown in FIG. 1 can be determined.

[Embodiment 2]
In the second embodiment of the present invention, the alignment mark is imaged while changing the height of the camera, the position of the feature point of the alignment mark is measured from these images, the measurement data (position of the 3D feature point) and 3D The position of the alignment mark is recognized by comparing with the template data.

  FIG. 5 is a diagram illustrating a configuration example of the alignment mark recognition apparatus according to the second embodiment of the present invention. The alignment mark recognition apparatus in the example of FIG. 5 includes an image processing unit 40 configured by a computer and a software program, an imaging unit 20 including a camera, a lens, and the like that captures an alignment mark on the printed circuit board, and a printed circuit board. The alignment stage 30 is configured.

  The image processing unit 40 includes an image input unit 41, a feature point recognition processing unit 42, a camera position control unit 43, a measurement data storage unit 44, a matching processing unit 45, and a 3D template data storage unit 46.

  The image input unit 41 inputs an image of the alignment mark imaged by the imaging unit 20. The feature point recognition processing unit 42 measures the position of the feature point of the alignment mark from the image input by the image input unit 41 and stores the measurement data in the measurement data storage unit 44. The camera position control unit 43 controls the camera position (height) of the imaging unit 20.

  The matching processing unit 45 performs a matching process between the measurement data in the measurement data storage unit 44 and the 3D template data stored in advance in the 3D template data storage unit 46 by applying the calculation method of the least square method. Output the result.

  FIG. 6 is a flowchart of alignment mark recognition processing in the second embodiment. In this example, the alignment mark is imaged N times while changing the height of the camera, and the position of the feature point appearing at the edge of the alignment mark of each image is extracted. Here, the camera height variable i is used to represent the camera height as h [i] (i = 0, 1,..., N−1).

  First, the camera height variable i is initialized (here, i = 0) (step S20). Next, the camera height is set to h [i] (step S21), and the alignment mark is imaged (step S22).

  In the obtained image, the position of a feature point such as an edge is calculated (step S23). For example, feature points can be extracted from an image by taking autocorrelation. Various methods for extracting feature points from an image are already known, and any method may be used. The extracted feature point position data is stored as measurement data in the measurement data storage unit 44 (step S24).

  The camera height variable i is incremented (step S25), and it is determined whether all measurements are completed (i ≧ N) (step S26). If not completed yet, the processing from step S21 to step S25 is performed. repeat. In this manner, the position of the feature point of the alignment mark is acquired three-dimensionally by changing the height of the camera, and the measurement data is stored in the measurement data storage unit 44.

  If all the measurements are completed in step S26, the distance between the measurement data (feature point position data) stored in the measurement data storage unit 44 and the 3D template data registered in advance in the 3D template data storage unit 46 Is obtained by applying the least squares calculation method (step S27), and the position where the result is minimum is obtained (step S28), thereby recognizing the position of the alignment mark.

  Here, how to determine the feature point in step S23 will be described. Although there are a plurality of methods for determining feature points, here, a method of obtaining using feature of local correlation is shown as an example.

  FIG. 7 is a diagram illustrating an example of local correlation calculation. Consider the local region R (m × n pixels) in the image shown in FIG. The local correlation calculation here refers to an m × n pixel in a region S (−p to + q, −p to + q) around the arbitrary local region R (m × n pixels) of the image shown in FIG. It is to obtain the sum of luminance differences from the area. If the luminance value of the region R is I (x, y) and the luminance value of the m × n pixel region in the peripheral region S is I (x + u, y + v), then m × n in the region R and the peripheral region S. The sum D (u, v) of the luminance difference from the pixel area is expressed by the following equation (2). However, -p <= u <= q and -p <= v <= q.

By calculating D (u, v) for all combinations of u and v satisfying −p ≦ u ≦ q and −p ≦ v ≦ q by the equation (2), the correlation value in the region S of the region R Get the map. An autocorrelation value distribution is used for feature determination. The autocorrelation value distribution here indicates a correlation value map obtained by the above-described local correlation calculation. The following determination process is performed on this correlation value map.
(1) A minimum correlation value D1 near the origin 8 (excluding the origin) of the correlation value map is obtained.
(2) Find the minimum correlation value D0 on the four sides of the rectangle surrounding the range of ± e of the origin.
(3) Let D = D0-D1 be a feature.

  Here, it means that there is a characteristic, so that the characteristic degree D is large. In the case of D> Dt, the coordinates (x, y) are feature points. Note that Dt is a threshold value for determination, and an arbitrary value is set. As described above, a feature point can be obtained by performing feature determination for each local region obtained by subdividing the entire screen.

  Next, a method of matching processing between measurement data and 3D template data in step S27 and step S28 will be described. This method is an application of the least square method.

  FIG. 8 is a diagram for explaining a matching process between measurement data and 3D template data according to the second embodiment. A slice obtained by slicing 3D template data at a certain height can be expressed by a set of line segments calculated based on vertex coordinates constituting the template figure. For example, a 3D template data having a shape as shown in FIG. 4 is considered as a plane when sliced at a height Z = h1. In this case, the template on the slice plane can be expressed by a set of straight lines fn (n = 1, 2, 3, 4) as shown in FIG.

  Here, it is assumed that measurement data (position data of feature points) in a slice having a height Z = h1 is Di (x, y, z). In addition, for each measurement data Di, it is determined which line fn corresponds to each measurement data Di.

  For each straight line fn, the distance to each corresponding measurement data Di is measured, and integrated values Sn (n = 1, 2, 3, 4; corresponding to the straight line fn) of all the measurement data Di corresponding to the straight line fn ) Further, an integrated value ΣSn (= S1 + S2 + S3 + S4) of Sn is obtained and set as Sh1. That is, the distance value when the height Z is Z = h1 is set to Sh1 = ΣSn. Similarly, the distance value in the case of Z = hm (m = 1, 2,...) With the height changed is assumed to be Shm.

  Further, the integrated value of these distance values Shm is S (u, v, q) = ΣShj (j = 1,..., M). Here, (u, v, q) is an offset amount when the initial position of the template is (0, 0, 0). The matching position to be obtained is (u, v, q) in which S (u, v, q) obtained in this way is minimized.

  In this example, the matching position with the 3D template data is obtained using the feature point on the image obtained by imaging the alignment mark. Similarly, the matching position can be obtained from the feature (line segment) of the edge of the alignment mark.

[Embodiment 3]
In Embodiment 3 of the present invention, in the alignment mark recognition method of Embodiment 1 described above, it is assumed that the image obtained by imaging is a color image, and the obtained color image is represented by the three primary colors R, G, and B. Color separation into three images to generate an alignment mark recognition process in the first embodiment described above for each of the R, G, and B images, and the obtained correlation results of R, G, and B are obtained. Based on the alignment mark position.

  FIG. 9 is a diagram illustrating an example in which a color image is divided into three primary colors. In the third embodiment, the image (in this case, the color image) obtained in step S12 in FIG. 2 in the first embodiment is the plane R (red image), the plane, as shown in FIG. G (green image) and plane B (blue image) are respectively decomposed into three primary color images.

  For each obtained image, the shape of the alignment mark is obtained (step S13 in FIG. 2), and the measurement data is stored in the measurement data storage unit 14 (step S14 in FIG. 2). Also in the third embodiment, as in the first embodiment, the shape of alignment marks at a plurality of slice levels is obtained for each of R, G, and B images, and the measurement data is stored in the measurement data storage unit 14. Store.

  A 3D model for each color (R, G, B) is generated from the measurement data stored in the measurement data storage unit 14 (step S17 in FIG. 2), and correlation processing with 3D template data is performed for each 3D model of each color. (Step S18 in FIG. 2), and the position of the alignment mark is recognized based on the correlation result for each color.

  For example, three correlation results (positions with the highest correlation) are obtained by correlation processing for each color, and the position obtained by taking the average of them may be recognized as the position of the alignment mark, or the above formula ( In 1), the minimum D (u, v, q, α, β, γ) for each color is obtained, and the smallest D (u, v, q, α, β, γ) among the three D (u, v, q, α, β, γ) is obtained. The position at u, v, q, α, β, γ) may be recognized as the position of the alignment mark.

[Embodiment 4]
In Embodiment 4 of the present invention, a plurality of images are captured by changing the lens focus position without changing the height of the camera in Embodiments 1, 2, and 3 described above. The position of the alignment mark in the first and third embodiments described above and the position of the feature point of the alignment mark in the second embodiment described above are obtained and compared with the 3D template data to recognize the position of the alignment mark. To do.

  In this case, the camera position control unit 13 in FIG. 1 and the camera position control unit 43 in FIG. 5 control the focus position of the lens instead of controlling the camera position.

[Embodiment 5]
In the fifth embodiment of the present invention, images of the alignment mark are picked up using a plurality of cameras, and the images picked up by the plurality of cameras obtained are used, and the first and third embodiments described above are used. The position of the alignment mark is recognized by obtaining the 3D model of the alignment mark and the feature points of the alignment mark in the second embodiment and comparing them with the 3D template data.

  FIG. 10 is a diagram illustrating an example of imaging by a plurality of cameras according to the fifth embodiment. In the example of FIG. 10, the alignment mark is imaged by two cameras, the imaging unit 20a and the imaging unit 20b. Note that the number of cameras may be three or more.

  For example, an image obtained by imaging an alignment mark using a total of five cameras, one camera that images from the top surface of the alignment mark and four cameras that image four sides of the alignment mark from an oblique direction. The position of the alignment mark is recognized by specifying the shape of the alignment mark from the above, or obtaining the position of the edge portion or feature point. In this example, it is assumed that the position obtained for each camera is replaced with a common spatial coordinate and integrated as one data.

  The embodiments of the present invention have been described with some examples. Such alignment mark recognition technology can be used in a bonding apparatus or the like.

  Bonding equipment that joins semiconductor chips such as LSI on a printed circuit board joins a metal surface and a metal surface by applying pressure from above, or joins a metal surface and a metal surface using ultrasonic waves. There are things.

  FIG. 11 is a diagram illustrating a configuration example of an ultrasonic bonding apparatus. Ultrasonic bonding equipment that joins metal surfaces to each other using ultrasonic waves applies a certain amount of pressure from the top of the semiconductor chip to be bonded to the printed circuit board, vibrates at about 50 kHz to 100 kHz, and fuses the metal and metal. Let them join.

  The ultrasonic bonding apparatus shown in the example of FIG. 11 includes a pressure mechanism 101, a pressure control unit 102, an ultrasonic head 103, an ultrasonic oscillator 104, an alignment mechanism 105, an alignment mechanism control unit 106, an imaging unit 107, and an imaging unit movement. A mechanism 108, an imaging unit moving mechanism control unit 109, an image processing unit 110, and a main controller 111 are included.

  The pressurizing mechanism 101 is connected to the ultrasonic head 103 and pressurizes with a designated force. The pressurization control unit 102 controls various operations such as pressure, moving speed, and acceleration in the pressurization mechanism 101. The ultrasonic head 103 has a semiconductor chip suction mechanism at the head tip and oscillates at a high frequency. The ultrasonic oscillator 104 oscillates the ultrasonic head 103 with a specified frequency and amplitude.

  The alignment mechanism 105 includes an X-axis, Y-axis, and θ-axis precision stage and a spherical seat, and a printed circuit board is mounted and fixed on the upper surface for alignment. The alignment stage 30 in FIGS. 1 and 5 corresponds to the alignment mechanism 105. The alignment mechanism control unit 106 controls the position / posture of the alignment mechanism 105.

  The imaging unit 107 includes a lens, an illumination, a camera, and the like, and images a semiconductor chip alignment mark attracted to the ultrasonic head 103 and a printed circuit board alignment mark mounted on the alignment mechanism 105. The imaging unit 20 in FIGS. 1, 5, and 10 corresponds to the imaging unit 107.

  The imaging unit moving mechanism 108 includes X-axis, Y-axis, and Z-axis precision stages for adjusting the position of the imaging unit 107. The position of the alignment mark on the printed circuit board fixed to the alignment mechanism 105, ultrasonic waves, and the like. The image pickup unit 107 is moved to the position of the alignment mark of the semiconductor chip adsorbed by the head 103. The imaging unit moving mechanism control unit 109 controls the position of the imaging unit moving mechanism 108.

  The image processing unit 110 measures the position of the alignment mark from the image captured by the imaging unit 107. The image processing units 10 and 40 in FIGS. 1 and 5 correspond to the image processing unit 110.

  The main controller 111 manages the pressure control unit 102, the ultrasonic oscillator 104, the image processing unit 110, the imaging unit moving mechanism control unit 109, and the alignment mechanism control unit 106, and controls the entire ultrasonic bonding apparatus.

  The characteristics of the embodiment of the present invention described above are summarized as follows.

(Appendix 1) An alignment mark recognition method for recognizing the position of an alignment mark,
The process of inputting images of multiple alignment marks taken at different camera heights,
Acquiring the shape of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
Generating a three-dimensional model of the alignment mark from the acquired shape of the alignment mark at a plurality of slice levels;
The three-dimensional model and the three-dimensional template data are matched by collating the generated three-dimensional model of the alignment mark with the alignment mark three-dimensional template data registered in the storage device in advance. And a process for recognizing the position of the alignment mark by obtaining the position.

(Appendix 2) In the alignment mark recognition method according to Appendix 1,
Separating the input plurality of alignment mark images into a plurality of color images for each image, and generating a plurality of alignment mark images for each color;
In the process of obtaining the shape of the alignment mark at the plurality of slice levels, the shape of the alignment mark at the plurality of slice levels for each color is obtained from the plurality of images of the alignment mark for each color,
In the process of generating a three-dimensional model of the alignment mark, a three-dimensional model of the alignment mark for each color is generated from the shape of the alignment mark at a plurality of slice levels for each of the acquired colors.
In the process of recognizing the position of the alignment mark, the three-dimensional model of the alignment mark generated for each color is collated with the three-dimensional template data of the alignment mark registered in the storage device in advance, and the collation result is obtained. An alignment mark recognition method characterized by recognizing the position of an alignment mark.

(Appendix 3) An alignment mark recognition method for recognizing the position of an alignment mark,
The process of inputting images of multiple alignment marks taken at different camera heights,
Obtaining a feature point or edge of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
By comparing the acquired feature point or edge position of the alignment mark at a plurality of slice levels with the three-dimensional template data of the alignment mark registered in the storage device in advance, And a step of recognizing the position of the alignment mark by obtaining a matching position with the three-dimensional template data.

(Appendix 4) In the alignment mark recognition method described in appendix 3,
Separating the input plurality of alignment mark images into a plurality of color images for each image, and generating a plurality of alignment mark images for each color;
In the process of obtaining the feature points or edges of the alignment marks at the plurality of slice levels, the feature points or edges of the alignment marks at the plurality of slice levels for each color are obtained from the plurality of images of the alignment marks for each color,
In the process of recognizing the position of the alignment mark, the position of the feature point or edge of the alignment mark acquired for each color is collated with the three-dimensional template data of the alignment mark registered in the storage device in advance. An alignment mark recognition method characterized by recognizing the position of an alignment mark based on a collation result.

(Appendix 5) In the alignment mark recognition method according to any one of Appendix 1 to Appendix 4,
Instead of inputting images of a plurality of alignment marks imaged by changing the height of the camera, inputting images of a plurality of alignment marks imaged by changing the focus position of the camera lens,
The alignment mark recognition method, wherein the plurality of slice levels are determined based on the focus position instead of the height of each camera.

(Appendix 6) In the alignment mark recognition method according to any one of appendices 1 to 5,
In the process of inputting the images of the plurality of alignment marks,
An alignment mark recognition method comprising inputting images picked up by a plurality of cameras.

(Appendix 7) An alignment mark recognition device for recognizing the position of an alignment mark,
Means for inputting images of a plurality of alignment marks taken at different camera heights;
Means for acquiring the shape of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
Means for generating a three-dimensional model of the alignment mark from the acquired shape of the alignment mark at a plurality of slice levels;
The three-dimensional model and the three-dimensional template data are matched by collating the generated three-dimensional model of the alignment mark with the alignment mark three-dimensional template data registered in the storage device in advance. An alignment mark recognition apparatus comprising: means for obtaining a position and recognizing a position of the alignment mark.

(Appendix 8) An alignment mark recognition device for recognizing the position of an alignment mark,
Means for inputting images of a plurality of alignment marks taken at different camera heights;
Means for acquiring feature points or edges of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
By comparing the acquired feature point or edge position of the alignment mark at a plurality of slice levels with the three-dimensional template data of the alignment mark registered in the storage device in advance, An alignment mark recognition apparatus comprising: means for obtaining a matching position with three-dimensional template data and recognizing the position of the alignment mark.

(Additional remark 9) It is the joining apparatus which performs alignment with the 1st joining object and the 2nd joining object using an alignment mark, and joins them,
Means for taking a plurality of images of alignment marks on the first or second joining object by changing the height of the camera;
Means for inputting images of the plurality of imaged alignment marks;
Means for acquiring the shape of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
Means for generating a three-dimensional model of the alignment mark from the acquired shape of the alignment mark at a plurality of slice levels;
The three-dimensional model and the three-dimensional template data are matched by collating the generated three-dimensional model of the alignment mark with the alignment mark three-dimensional template data registered in the storage device in advance. Means for recognizing the position of the alignment mark by determining the position;
A joining apparatus comprising: means for aligning the first joining object and the second joining object based on the recognized position of the alignment mark.

(Additional remark 10) It is the joining apparatus which performs alignment with the 1st joining object and the 2nd joining object using an alignment mark, and joins them,
Means for taking a plurality of images of alignment marks on the first or second joining object by changing the height of the camera;
Means for inputting images of the plurality of imaged alignment marks;
Means for acquiring feature points or edges of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
By comparing the acquired feature point or edge position of the alignment mark at a plurality of slice levels with the three-dimensional template data of the alignment mark registered in the storage device in advance, Means for recognizing the position of the alignment mark by obtaining a matching position with the three-dimensional template data;
A joining apparatus comprising: means for aligning the first joining object and the second joining object based on the recognized position of the alignment mark.

It is a figure which shows the structural example of the alignment mark recognition apparatus in Embodiment 1 of this invention. It is an alignment mark recognition process flowchart in this Embodiment 1. It is a figure explaining the example of the 3D model of the alignment mark in this Embodiment 1. FIG. It is a figure explaining the example of 3D template data. It is a figure which shows the structural example of the alignment mark recognition apparatus in Embodiment 2 of this invention. It is an alignment mark recognition process flowchart in this Embodiment 2. It is a figure explaining the example of a local correlation calculation. It is a figure explaining the matching process of the measurement data in this Embodiment 2, and 3D template data. It is a figure which shows the example in the case of dividing | segmenting a color image into three primary colors. 10 is a diagram illustrating an example of imaging by a plurality of cameras in Embodiment 5. FIG. It is a figure which shows the structural example of an ultrasonic bonding apparatus. It is a figure which shows the example of an alignment mark. It is a figure which shows the slice image of an alignment mark.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image processing part 11 Image input part 12 Shape recognition process part 13 Camera position control part 14 Measurement data storage part 15 3D model generation part 16 3D correlation calculation part 17 3D template data storage part 20 Imaging unit 30 Alignment stage 40 Image processing part 41 Image Input Unit 42 Feature Point Recognition Processing Unit 43 Camera Position Control Unit 44 Measurement Data Storage Unit 45 Matching Processing Unit 46 3D Template Data Storage Unit 101 Pressure Mechanism 102 Pressure Control Unit 103 Ultrasonic Head 104 Ultrasonic Oscillator 105 Alignment Mechanism 106 Alignment mechanism control unit 107 Imaging unit 108 Imaging unit moving mechanism 109 Imaging unit moving mechanism control unit 110 Image processing unit 111 Main controller

Claims (5)

An alignment mark recognition method for recognizing the position of an alignment mark,
The process of inputting images of multiple alignment marks taken at different camera heights,
Acquiring the shape of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
Generating a three-dimensional model of the alignment mark from the acquired shape of the alignment mark at a plurality of slice levels;
The three-dimensional model and the three-dimensional template data are matched by collating the generated three-dimensional model of the alignment mark with the alignment mark three-dimensional template data registered in the storage device in advance. And a process for recognizing the position of the alignment mark by obtaining the position. An alignment mark recognition method for recognizing the position of an alignment mark,
The process of inputting images of multiple alignment marks taken at different camera heights,
Obtaining a feature point or edge of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
By comparing the acquired feature point or edge position of the alignment mark at a plurality of slice levels with the three-dimensional template data of the alignment mark registered in the storage device in advance, And a step of recognizing the position of the alignment mark by obtaining a matching position with the three-dimensional template data. An alignment mark recognition device for recognizing the position of an alignment mark,
Means for inputting images of a plurality of alignment marks taken at different camera heights;
Means for acquiring the shape of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
Means for generating a three-dimensional model of the alignment mark from the acquired shape of the alignment mark at a plurality of slice levels;
The three-dimensional model and the three-dimensional template data are matched by collating the generated three-dimensional model of the alignment mark with the alignment mark three-dimensional template data registered in the storage device in advance. An alignment mark recognition apparatus comprising: means for obtaining a position and recognizing a position of the alignment mark. An alignment mark recognition device for recognizing the position of an alignment mark,
Means for inputting images of a plurality of alignment marks taken at different camera heights;
Means for acquiring feature points or edges of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
By comparing the acquired feature point or edge position of the alignment mark at a plurality of slice levels with the three-dimensional template data of the alignment mark registered in the storage device in advance, An alignment mark recognition apparatus comprising: means for obtaining a matching position with three-dimensional template data and recognizing the position of the alignment mark. A joining apparatus that performs alignment between a first joining object and a second joining object using an alignment mark, and joins them.
Means for taking a plurality of images of alignment marks on the first or second joining object by changing the height of the camera;
Means for inputting images of the plurality of imaged alignment marks;
Means for acquiring the shape of the alignment mark at a plurality of slice levels according to the height of each camera from the images of the plurality of alignment marks;
Means for generating a three-dimensional model of the alignment mark from the acquired shape of the alignment mark at a plurality of slice levels;
The three-dimensional model and the three-dimensional template data are matched by collating the generated three-dimensional model of the alignment mark with the alignment mark three-dimensional template data registered in the storage device in advance. Means for recognizing the position of the alignment mark by determining the position;
A joining apparatus comprising: means for aligning the first joining object and the second joining object based on the recognized position of the alignment mark.

Images