Inspection system, inspection device, inspection method, and program

Abstract

This inspection system is for performing inspections of component-mounting substrates on each of which an electronic component having a solder bump formed thereon is mounted, and includes: a three-dimensional data generation means for generating, by using a plurality of X-ray images obtained by capturing the component-mounting substrates, three-dimensional data of a region including at least a solder joint portion formed by the solder bump; a longitudinal cross-sectional image extraction means for extracting, from the three-dimensional data, longitudinal cross-sectional images, each of which shows a longitudinal cross-section passing through the center portion of the solder joint portion; an inspection model generation means for generating an inspection model for the inspections by using the plurality of longitudinal cross-sectional images; and an inspection means for performing the inspections on the basis of the inspection model by using the plurality of longitudinal cross-sectional images that have different observation directions and that have been extracted from the three-dimensional data of the respective component-mounting substrates to be inspected.

Description

Inspection system, inspection device, inspection method and program

[0001] The present invention relates to an inspection system, an inspection device, an inspection method, and a program, and more particularly to a technique for inspecting an object based on three-dimensional data acquired using an X-ray CT.

[0002] In recent years, various products have become smaller and more precise, and for example, the density of components mounted on component-mounted boards has increased, resulting in an increase in the number of parts that are shaded in the imaging field of a visual inspection device, resulting in an increase in the number of parts that cannot be accurately inspected by visual inspection. In response to this, techniques have become known that use X-ray CT inspection to inspect parts that cannot be inspected by visual inspection (for example, Patent Documents 1 and 2).

[0003] JP 2007-17304 A JP 2000-275191 A

[0004] Incidentally, some electronic components used in component mounting boards have solder bumps formed on the bottom surface of the package, such as BGA (Ball Grid Array). For types of defects such as non-wetting of solder bumps or poor soldering, proper pass / fail judgment cannot be performed by inspection using a single feature, such as the solder area in a specific horizontal plane, based on a threshold value. Therefore, complex inspection combining judgments based on multiple feature values ​​is required.

[0005] For such complicated inspections, inspections using machine learning models (so-called AI) trained using techniques such as deep learning are suitable. However, there is a problem in that the computational costs of learning and inspecting (inferring) three-dimensional data (3D voxels) are high, making high-speed automated inspections difficult.

[0006] The present invention has been made in view of the above circumstances, and has as its object to enable high-speed automatic inspection of a substrate on which electronic components including solder bumps are mounted.

[0007] In order to achieve the above object, the present invention employs the following configuration: An inspection system for inspecting component-mounted boards on which electronic components having solder bumps formed thereon are mounted, comprising: three-dimensional data generation means for generating three-dimensional data of an area including at least a solder joint formed by the solder bumps using a plurality of X-ray images of the component-mounted board, longitudinal section image extraction means for extracting, from the three-dimensional data, a longitudinal section image showing a longitudinal section passing through the center of the solder joint, inspection model generation means for generating an inspection model for the inspection using the plurality of longitudinal section images, and inspection means for performing inspection using the inspection model, using the plurality of longitudinal section images observed from different directions extracted from the three-dimensional data of each of the component-mounted boards to be inspected.

[0008] In the above, the term "center" does not refer only to the exact center point, but also includes the area near the center point. For example, in the horizontal direction, the center can be the area from the center point to half the length of the shortest side of the solder joint pad. Furthermore, an "inspection model" is a model (so-called AI) that has been machine-learned using learning data, and generating an inspection model includes causing the model to perform this learning. Note that machine learning may be supervised learning or unsupervised learning.

[0009] In this way, by using longitudinal cross-sectional images (i.e., two-dimensional data) extracted from three-dimensional data of the solder joint to learn an inspection model and to perform inspection using the learned inspection model, it becomes possible to perform high-speed, high-precision inspection even when inspecting component-mounted boards on which components such as BGAs, which have solder bumps formed on the bottom surface of the package, are mounted.

[0010] The longitudinal cross-sectional images used for training by the inspection model generating means may include a plurality of longitudinal cross-sectional images with different observation directions for each of the three-dimensional data. In this way, by extracting a plurality of longitudinal cross-sectional images with different observation directions for each of the three-dimensional data, i.e., for each of the inspection regions, it is possible to improve the efficiency and accuracy of training.

[0011] The longitudinal section image extraction means may also plot brightness variance values ​​of horizontal tomographic images acquired from the three-dimensional data in the vertical direction, and determine the position of the tomographic layer at which the brightness variance value is maximum as the vertical position identifying the central portion. The longitudinal section image extraction means may also plot brightness variance values ​​of horizontal tomographic images acquired from the three-dimensional data in the vertical direction, and determine the position of the tomographic layer at the center of a plurality of tomographic layers at which the brightness variance value exceeds a predetermined threshold as the vertical position identifying the central portion.

[0012] The longitudinal section image extraction means may determine, as the horizontal position specifying the central portion, the center of a label of a binarized image of the horizontal tomographic image having the maximum brightness variance value, or the center of gravity calculated by weighting the horizontal tomographic image having the maximum brightness variance value by brightness.

[0013] Furthermore, the longitudinal section image output means may determine the horizontal position that identifies the central portion as the center of a label of a binarized image of a horizontal tomographic image that is the center of a plurality of horizontal tomographic sections whose brightness variance value exceeds a predetermined threshold, or the position of a center of gravity that is calculated by weighting, by brightness, a horizontal tomographic image that is the center of a plurality of horizontal tomographic sections whose brightness variance value exceeds a predetermined threshold.

[0014] The longitudinal cross-sectional image extraction means may determine the horizontal position that identifies the central portion as the center of a label of a binarized image of the horizontal cross-sectional image in which the edge strength is maximum, or the center of gravity calculated by weighting the horizontal cross-sectional image in which the edge strength is maximum by brightness.

[0015] The longitudinal section image output means may determine the horizontal position that identifies the central portion as the center of a label of a binarized image of a horizontal tomographic image that is the center of a plurality of horizontal tomographic sections whose edge strength exceeds a predetermined threshold, or the center of gravity that is calculated by weighting the horizontal tomographic image that is the center of a plurality of horizontal tomographic sections whose edge strength exceeds a predetermined threshold by brightness.

[0016] The longitudinal cross-sectional image extracting means may determine, as the horizontal position specifying the central portion, the center of a label of a binarized image obtained by performing maximum intensity projection in the vertical direction on each horizontal tomographic image of a region including a predetermined horizontal range of the solder joint in the three-dimensional data, or the center of gravity calculated by weighting the image obtained by maximum intensity projection by brightness. Here, the predetermined horizontal range is a range set to include a region near the central portion in the three-dimensional data.

[0017] The longitudinal cross-sectional image extracting means may determine, as the horizontal position specifying the central portion, the center of a label of a binarized image obtained by vertically averaging horizontal tomographic images of a region including a predetermined horizontal range of the solder joint in the three-dimensional data, or the center of gravity calculated by weighting the image obtained by the average projection by brightness. Here, the predetermined horizontal range is a range set to include a region near the center of the three-dimensional data.

[0018] With the above configuration, the vertical (hereinafter also referred to as the Z direction) and horizontal (hereinafter also referred to as the XY direction) positions of the center for extracting a longitudinal cross-sectional image can be efficiently identified using a known image processing method with a low processing load.

[0019] The present invention can also be understood as an inspection device that inspects a component-mounted board on which electronic components having solder bumps formed thereon are mounted, wherein the inspection is performed using: an inspection model for inspection that is generated using a longitudinal cross-sectional image that shows a longitudinal cross section passing through the center of the solder joint, the longitudinal cross-sectional image being extracted from three-dimensional data of an area that includes at least the solder joint formed by the solder bumps, and that is generated using a plurality of X-ray images of the component-mounted board; and a plurality of the longitudinal cross-sectional images observed from different directions that are extracted from the three-dimensional data of the component-mounted board that is the inspection target.

[0020] The present invention can also be understood as an inspection method for inspecting a component-mounted board on which electronic components having solder bumps formed thereon are mounted, comprising: a three-dimensional data generation step of generating three-dimensional data of an area including at least a solder joint formed by the solder bumps using a plurality of X-ray images of the component-mounted board; a longitudinal section image extraction step of extracting, from the three-dimensional data, a longitudinal section image showing a longitudinal section passing through the center of the solder joint; an inspection model generation step of generating an inspection model for the inspection using the longitudinal section image; and an inspection step of performing inspection using the inspection model, using the plurality of longitudinal section images observed from different directions extracted from the three-dimensional data of the component-mounted board to be inspected.

[0021] The present invention can also be understood as a program for causing a computer to execute each step described in the inspection method, or as a computer-readable recording medium on which such a program is non-transitoryly recorded.

[0022] The present invention can be achieved by combining the above-described configurations and processes as long as no technical contradiction occurs.

[0023] According to the present invention, it is possible to provide a technique that enables high-speed automatic inspection of a substrate on which electronic components including solder bumps are mounted.

[0024] FIG. 1 is a block diagram showing a schematic functional configuration of an X-ray inspection system according to an embodiment. FIG. 2 is a schematic diagram showing a solder joint of a component mounting board to be inspected according to the embodiment, viewed from a horizontal direction. FIG. 3 is a flowchart showing an example of a process performed in an inspection terminal according to the embodiment. FIG. 4 is a schematic diagram showing the configuration and functional blocks related to X-ray imaging of an X-ray imaging device according to an embodiment. FIG. 5A is a first explanatory diagram showing the relationship between each horizontal tomographic image that can be extracted from three-dimensional data of a solder joint according to the embodiment and the feature amounts of each horizontal tomographic image. FIG. 5B is a second explanatory diagram showing the relationship between each horizontal tomographic image that can be extracted from three-dimensional data of a solder joint according to the embodiment and the feature amounts of each horizontal tomographic image. FIG. 6A is an explanatory diagram showing an example of a horizontal tomographic image that can be extracted from three-dimensional data according to the embodiment and an image obtained by binarizing the horizontal tomographic image. FIG. 6B is an explanatory diagram showing an example of a horizontal tomographic image that can be extracted from three-dimensional data according to the embodiment and an edge image thereof. FIG. 7A is a first explanatory diagram for explaining multiple longitudinal cross-sectional images for which pass / fail judgment is performed by an inspection unit according to the embodiment. FIG. 7B is a second explanatory diagram for explaining a plurality of longitudinal sectional images on which pass / fail determination is performed by the inspection unit according to the embodiment.

[0025] <Application Example> (Configuration of Application Example) An example of an embodiment of the present invention will be described below. The present invention can be applied as an X-ray inspection system including an X-ray inspection device for, for example, taking an X-ray image of a component mounting board on which electronic components, such as BGAs, are mounted, each having metal bumps for solder bonding formed on the bottom surface of the package, and inspecting the component mounting board based on the captured image.

[0026] 1 is a schematic block diagram showing the functional configuration of an X-ray inspection system according to an application example. The X-ray inspection system 1 according to the application example includes an X-ray imaging device 11, an inspection terminal 12, and a data server 13, and is an inspection system used to inspect component mounting boards in which soldering bumps are formed on the bottom surface of packages such as BGAs. The X-ray imaging device 11, the inspection terminal 12, and the data server 13 are connected to each other so as to be able to communicate with each other via communication means (not shown).

[0027] Although not shown, the inspection terminal 12 can be a general-purpose computer equipped with a processor such as a CPU or DSP, a storage means, an input means such as a keyboard or a mouse, and an output means such as a liquid crystal display.

[0028] As shown in FIG. 1, the inspection terminal 12 has the following functional units: an image acquisition unit 121 , a three-dimensional data generation unit 122 , an image extraction unit 123 , a model generation unit 124 , an inspection unit 125 , and a storage unit 126 .

[0029] The image acquisition unit 121 acquires a plurality of X-ray image data obtained by capturing an image of an area including a solder joint, which is an inspection target portion of a component mounting board, using the X-ray imaging device 11. The image data may be acquired directly from the X-ray imaging device 11, or may be acquired by first transmitting the image data from the X-ray imaging device 11 to the data server 13 and then storing the data in the data server 13.

[0030] The three-dimensional data generator 122 generates data on the three-dimensional shape of the solder joint (hereinafter simply referred to as three-dimensional data) based on the acquired X-ray image data. Since publicly known techniques such as CT (Computed Tomography) and tomosynthesis can be applied to the method of generating (reconstructing) the data, detailed description thereof will be omitted.

[0031] The image extraction unit 123 extracts a longitudinal cross-sectional image from the three-dimensional data generated by the three-dimensional data generation unit 122, which is used to determine the quality of the component-mounted board. Specifically, it extracts a longitudinal cross-sectional image showing a longitudinal cross section passing through the center of the solder joint. The term "center" does not refer only to the exact center point, but also includes the area near the center point. The center will be explained using FIG. 2. FIG. 2 is a schematic diagram of the solder joint of the component-mounted board as viewed from the horizontal direction. As shown in FIG. 2, for example, in the horizontal direction, the center can be defined as a range of D / 2 from the center point C, where D / 2 is half the length of the shortest horizontal distance D of the pad P with the smallest area to be soldered in the solder joint as the target. Note that the image extraction unit 123 in this application example corresponds to the longitudinal cross-sectional image extraction means of the present invention.

[0032] The model generation unit 124 generates an inspection model for inspection using multiple longitudinal cross-sectional images extracted by the image extraction unit. Specifically, machine learning is performed using a large number of longitudinal cross-sectional image data extracted from three-dimensional data of solder joints of multiple inspection targets, using a method such as deep learning, to generate a trained model (so-called AI). The machine learning method may be supervised learning or unsupervised learning. Furthermore, the model generation unit 124 may be capable of re-training an already trained model. Note that the model generation unit 124 in this application example corresponds to the inspection model generation means according to the present invention.

[0033] The inspection unit 125 performs pass / fail judgment on each of a plurality of longitudinal cross-sectional images extracted from the three-dimensional data of each component mounting board to be inspected, the longitudinal cross-sectional images being observed in different directions and including a common point on the three-dimensional data, using the inspection model generated by the model generation unit 124. For example, if there is even one longitudinal cross-sectional image for one inspection target location (solder joint) with a score below a predetermined value, the inspection unit 125 may be configured to judge the location to be inspected as defective.

[0034] The storage unit 126 is configured by the above-mentioned storage means, and includes, for example, a main storage unit such as a read-only memory (ROM) or a random access memory (RAM), and an auxiliary storage unit such as an EPROM, a hard disk drive (HDD), a removable medium, etc. The auxiliary storage unit stores an operating system (OS), various programs, etc., and the programs are loaded into the working area of ​​the main storage unit and executed, and the inspection terminal 12 is controlled through the execution of the programs, thereby realizing each functional unit.

[0035] Next, the flow of processing performed by the inspection terminal 12 in the X-ray inspection system 1 according to the application example will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of an example of processing performed by the inspection terminal 12. First, the X-ray imaging device 11 captures an X-ray tomographic image of the circuit board, and the image acquisition unit 121 acquires X-ray image data from the X-ray imaging device 11 or the data server 13 (S101). Then, the three-dimensional data generation unit 122 creates three-dimensional data of an area including a solder joint from multiple X-ray images (S102).

[0036] Next, the image extraction unit 123 extracts longitudinal section images showing longitudinal sections passing through the center of the solder joint from the three-dimensional data (S103). The model generation unit 124 then performs machine learning using the extracted longitudinal section images as training data to generate a machine learning model for inspection (S104). The inspection unit 125 then performs pass / fail judgment using the inspection model generated in step S104 for each of the longitudinal section images extracted from the three-dimensional data of each component-mounted board to be inspected, with different observation directions (S105), thereby completing the series of routines. Note that once the inspection model has been generated, the process of step S104 can be prevented from being executed unless some trigger occurs, such as the passage of a predetermined period of time.

[0037] According to the configuration of this application example as described above, even when inspecting a component mounting board on which components such as BGAs, which have solder bumps formed on the bottom surface of the package, are mounted, there is no need to set complicated inspection standards, and it is possible to perform high-speed, high-precision inspection.

[0038] <Embodiments> Hereinafter, embodiments of the present invention will be described in detail by way of example with reference to the drawings (including the drawings already described in the application examples above). However, unless otherwise specified, the specific configurations described in each embodiment are not intended to limit the scope of the present invention to those specific configurations.

[0039] 1 is a schematic block diagram showing the functional configuration of an X-ray inspection system 1 according to this embodiment. The X-ray inspection system 1 according to this embodiment is configured to include an X-ray imaging device 11, an inspection terminal 12, and a data server 13, and is an inspection system used to inspect component mounting boards in which metal bumps for soldering are formed on the bottom surface of packages such as BGA. The X-ray imaging device 11, the inspection terminal 12, and the data server 13 are connected to each other so as to be able to communicate with each other via communication means (not shown).

[0040] (Regarding the X-ray Imaging Apparatus) Fig. 4 is a schematic diagram showing the configuration and functional blocks related to X-ray imaging of the X-ray imaging apparatus 11. As shown in Fig. 4, the X-ray imaging apparatus 11 includes an X-ray source 10, an X-ray camera 20, a workpiece stage 30 that holds a component-mounted board (hereinafter referred to as a workpiece) W to be inspected, and a terminal 100, and performs X-ray imaging of the workpiece W transported from an upstream process inside the apparatus.

[0041] The workpiece W is transported from an upstream process by a carry-in conveyor (not shown) and placed on the workpiece stage 30. The workpiece stage 30 is configured to be movable in the horizontal and vertical directions by a stage having XYZ drive axes.

[0042] The X-ray camera 20 is a two-dimensional X-ray detector that detects X-rays irradiated from the X-ray source 10 and transmitted through the workpiece W. As the X-ray camera 20, an image intensifier (I.I.) tube or a flat panel detector (FPD) can be used.

[0043] Both the X-ray source 10 and the X-ray camera 20 are configured to be movable by an X-Y stage, and the X-ray source 10 irradiates the workpiece W with X-rays, and the X-ray image of the transmitted light is taken by the X-ray camera 20. The X-ray source 10 and the X-ray camera 20 revolve on revolving orbits C1 and C2, respectively, and take X-ray images of the workpiece W at multiple positions on the orbits. The movement of each stage is achieved by a combination of a servo motor and a ball screw, but since this technology is well known, its description will be omitted.

[0044] The X-ray imaging device 11 includes a control / operation terminal 100. The terminal 100 is implemented, for example, by a general-purpose computer. Specifically, the terminal 100 may include a control unit 110 implemented by a processor such as a CPU (not shown), a main memory unit 101 and an auxiliary memory unit 102 implemented by storage means (not shown), an input unit 103 implemented by input devices such as a keyboard and a mouse (not shown), an output unit 104 implemented by a liquid crystal display (not shown), and a communication unit 105 that communicates information with the examination terminal 12, the data server 13, etc. Note that the terminal 100 may be configured separately from an imaging unit including the X-ray camera 20, etc.

[0045] Each part of the X-ray imaging device 11 is controlled based on a control signal from the control unit 110 of the terminal 100. The control unit 110 of the X-ray imaging device 11 further includes functional units such as an X-ray source control unit 111, an X-ray source XY stage control unit 112, a workpiece stage control unit 113, a camera XY stage control unit 114, and an X-ray camera control unit 115.

[0046] The camera XY stage control unit 114 drives the camera XY stage (not shown) and transmits a control signal for horizontally moving the X-ray camera 20. The X-ray camera control unit 115 transmits a control signal for causing the X-ray camera 20 to capture an X-ray image.

[0047] The X-ray source control unit 111 transmits signals to start and end irradiation of X-rays by the X-ray source 10 and to adjust the X-ray intensity. The X-ray source XY stage control unit 112 transmits signals to drive the X-ray source XY stage (not shown) and move the X-ray source 10 in the horizontal direction.

[0048] The workpiece stage control unit 113 sends a control signal to the workpiece stage 30 to control the horizontal and vertical positions of the workpiece W to be optimal positions for photographing.

[0049] (Regarding the Data Server) The data server 13 stores imaging conditions for the X-ray imaging device 11, information related to the inspection target board (for example, the type, shape, and dimensions of the component), information related to the inspection conditions, and has an area for storing data (three-dimensional data, longitudinal cross-sectional images, inspection models, etc.) generated by the inspection terminal 12. Note that a program for controlling the inspection terminal 12 may be stored in the data server 13.

[0050] (Regarding the Inspection Terminal) As explained in the application example, the inspection terminal 12 can be configured by a general-purpose computer. The inspection terminal 12 may be configured by a single computer or by multiple computers that work together. The functional units of the inspection terminal 12 are as described above, but the image extraction unit 123 and the inspection unit 125 will be described in more detail below.

[0051] The image extraction unit 123 extracts a longitudinal cross-sectional image showing a longitudinal cross-section passing through the center of the solder joint, which is the inspection target location, from the three-dimensional data generated by the three-dimensional data generation unit 122. In this case, the image extraction unit 123, for example, plots the brightness variance value of the horizontal tomographic image obtained from the three-dimensional data in the vertical direction, and determines the position of the tomographic plane where the brightness variance value is maximum as the vertical position that identifies the "center".

[0052] 5A is an explanatory diagram showing the relationship between the three-dimensional data of a solder joint and the brightness variance of each horizontal tomographic image (XY image) extracted from the data and plotted in the vertical direction (Z direction). As shown in Fig. 5A, the brightness variance increases toward the center of the solder joint (in terms of image characteristics, the contrast between the solder part and other parts increases). Therefore, by setting the position of the horizontal tomographic image where the brightness variance is maximum as the vertical position that identifies the center of the solder joint, the center can be determined with high accuracy.

[0053] The image extraction unit 123 also binarizes the horizontal tomographic image in which the brightness variance value identified as above is maximized, and sets the center position of the label in the binarized image as the horizontal position that identifies the "center." Fig. 6A is an explanatory diagram showing an example of a horizontal tomographic image and an image obtained by binarizing the same. The image extraction unit 123 calculates the center position of the label (white portion) in the binarized image as shown on the right side of Fig. 6A using a desired publicly known technique, and sets this as the horizontal position that identifies the center of the solder joint, thereby reducing the processing load and quickly (efficiently) determining the center.

[0054] The inspection unit 125 performs pass / fail judgment using an inspection model for each of a plurality of longitudinal cross-sectional images extracted from the three-dimensional data of each component mounting board to be inspected, the longitudinal cross-sectional images being observed from different directions but including a common point on the three-dimensional data. The inspection model is stored in the storage unit 126 or the data server 13.

[0055] Here, multiple longitudinal cross-sectional images observed from different directions but containing a common point on the three-dimensional data can be extracted, for example, as follows. Figures 7A and 7B are explanatory diagrams illustrating multiple longitudinal cross-sectional images for which the inspection unit 125 performs a pass / fail judgment. Figure 7A is a horizontal tomographic image including the solder joint (indicated by T in Figure 7A) to be inspected. The image extraction unit 123 determines a central point in the three-dimensional shape data of the solder joint T using the method described above, and determines four longitudinal cross-sections including that point, each shifted in phase by 45°, as shown in Figure 7A. Then, longitudinal cross-sectional images for the four cross-sections are extracted, as shown in Figure 7B.

[0056] The inspection unit 125 judges each of the four longitudinal cross-sectional images thus obtained using the inspection model (specifically, calculates a score related to the quality and judges whether it is below a predetermined score). If the score of any one of the four longitudinal cross-sectional images does not satisfy the standard, the corresponding solder joint is judged to be defective.

[0057] According to the X-ray inspection system 1 of this embodiment as described above, it is possible to perform pass / fail judgment by AI (inspection model) using not three-dimensional data but longitudinal cross-sectional images (i.e., two-dimensional data) including the central portion identified from the three-dimensional data by a known image processing method. In other words, there is no need to set complicated inspection standards, and accurate and fast pass / fail judgment can be performed by AI that has performed machine learning using images similar to those used during inspection.

[0058] (Variant 1) In the above embodiment, an example of a method in which the image extraction unit 123 identifies the position that is the center of the solder joint that is the inspection target is shown, but the vertical and horizontal positions that are the "center" of the solder joint can be identified in various ways other than the above example.

[0059] Specifically, the image extraction unit 123 may vertically plot the brightness variance values ​​of horizontal tomographic images acquired from the three-dimensional data, and determine the vertical position of the tomographic layer at the center of multiple tomographic layers where the brightness variance value exceeds a predetermined threshold value as the vertical position identifying the center. Figure 5B is a graph in which the vertical position of a horizontal tomographic image extracted from the three-dimensional data of the solder joint is plotted on the horizontal axis, and the feature amount (here, brightness variance) of each tomographic layer is plotted vertically. In Figure 5B, the dashed line represents the predetermined threshold value, and the black dots in the figure indicate the (vertical) positions of the horizontal tomographic layers at the center of multiple tomographic layers where the brightness variance value exceeds the predetermined threshold value. That is, in this modification, the image extraction unit 123 determines the position of the horizontal tomographic image located at the point plotted by the black dot in Figure 5B as the vertical position identifying the center. The black square in Figure 5B indicates the position of the horizontal tomographic image at which the feature amount is maximum.

[0060] (Variation 2) The image extraction unit 123 may also plot the edge strength of horizontal tomographic images acquired from three-dimensional data in the vertical direction, and determine the position of the tomographic image where the edge strength is maximum as the vertical position that identifies the center. Figure 6B is a diagram showing horizontal tomographic images extracted from three-dimensional data and edge images obtained by performing image processing on the horizontal tomographic images. The image extraction unit 123 according to this variation calculates the edge strength of each horizontal tomographic image extracted from the three-dimensional data from an edge image such as that shown on the right side of Figure 6B, and determines the position of the horizontal tomographic image where the edge strength is maximum as the vertical position that identifies the center.

[0061] (Variation 3) The image extraction unit 123 may also plot the edge intensities of horizontal tomographic images acquired from three-dimensional data in the vertical direction, and determine the position of the central fault where the edge intensity exceeds a predetermined threshold as the vertical position that identifies the center. Note that when the feature amount in the graph of Fig. 5B is the edge intensity, the "brightness variance" described in Variation 1 can be understood as "edge intensity."

[0062] (Variation 4) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the brightness variance value of the horizontal tomographic image, the image extraction unit 123 may use the position of the center of gravity calculated by weighting the horizontal tomographic image with the brightness that maximizes the brightness variance as the horizontal position that identifies the center. Note that any known technology may be used to calculate the position of the center of gravity of the horizontal tomographic image.

[0063] (Variant 5) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the brightness variance value of the horizontal tomographic image, the image extraction unit 123 may use the center position of the label of the binarized image of the horizontal tomographic image that is the center of multiple horizontal tomographic sections whose brightness variance value exceeds a predetermined threshold as the horizontal position that identifies the center.

[0064] (Variant 6) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the brightness variance value of the horizontal tomographic image, the image extraction unit 123 may use the position of the center of gravity calculated by weighting the horizontal tomographic image that is the center of multiple horizontal tomographic sections whose brightness variance value exceeds a predetermined threshold value by brightness as the horizontal position that identifies the center.

[0065] (Variation 7) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the edge strength of the horizontal tomographic image, the image extraction unit 123 may use the center position of the label of the image obtained by binarizing the horizontal tomographic image in which the edge strength is greatest as the horizontal position that identifies the center.

[0066] (Variant 8) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the edge strength of the horizontal tomographic image, the image extraction unit 123 may use the position of the center of gravity calculated by weighting the horizontal tomographic image at which the edge strength is greatest by brightness as the horizontal position that identifies the center.

[0067] (Variant 9) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the edge strength of the horizontal tomographic image, the image extraction unit 123 may use the center position of the label of the binarized image of the horizontal tomographic image that is the center of multiple horizontal tomographic sections where the edge strength exceeds a predetermined threshold as the horizontal position that identifies the center.

[0068] (Variant 10) Furthermore, when the image extraction unit 123 identifies the vertical position that identifies the center based on the edge strength of the horizontal tomographic image, the image extraction unit 123 may use the position of the center of gravity calculated by weighting the horizontal tomographic image at the center of multiple horizontal tomographic sections where the edge strength exceeds a predetermined threshold value by brightness as the horizontal position that identifies the center.

[0069] (Variation 11) In addition, the image extraction unit 123 may use the position of the center of the label of a binarized image obtained by performing maximum intensity projection in the vertical direction on each horizontal tomographic image of an area in the three-dimensional data that includes a predetermined horizontal range of the solder joint as the horizontal position that identifies the center.

[0070] (Variation 12) Furthermore, the image extraction unit 123 may use the position of the center of gravity calculated by performing maximum intensity projection in the vertical direction on each horizontal tomographic image of an area in the three-dimensional data that includes a predetermined range in the horizontal direction of the solder joint, and weighting the image by brightness as the horizontal position that identifies the center.

[0071] (Variation 13) Furthermore, the image extraction unit 123 may use the position of the center of the label of a binarized image obtained by vertically projecting the average value of each horizontal tomographic image of an area that includes a predetermined horizontal range of the solder joint as the horizontal position that identifies the center.

[0072] (Variation 14) Furthermore, the image extraction unit 123 may determine the position of the center of gravity, which is calculated by weighting the image obtained by vertically projecting the average value of each horizontal tomographic image of an area that includes a predetermined horizontal range of the solder joint, by brightness, as the horizontal position that identifies the center.

[0073] According to the above-described modified examples, it is possible to extract longitudinal cross-sectional images for generating an inspection model and for an inspection object while suppressing the processing load on the processor. Furthermore, since the inspection is performed using longitudinal cross-sectional images (two-dimensional data) passing through the center of the solder joint, it is possible to perform high-speed, high-precision inspection even when inspecting a component-mounted board on which components such as BGAs, which have solder bumps formed on the bottom surface of the package, are mounted.

[0074] <Others> The above examples merely illustrate the present invention, and the present invention is not limited to the specific embodiments described above. Various modifications and combinations of the present invention are possible within the scope of the technical concept. For example, in the above X-ray inspection system 1, the data server 13 is not an essential component, and the data stored in the data server 13 may be saved in the inspection terminal 12 and / or the X-ray imaging device 11.

[0075] In the above embodiment, the X-ray inspection system 1 is described as including the X-ray imaging device 11 and the inspection terminal 12, but each function of the inspection terminal 12 may be incorporated into the X-ray imaging device 11. That is, the present invention can also be applied to an X-ray inspection device in which the terminal 100 of the X-ray imaging device 11 also functions as the inspection terminal 12.

[0076] The pass / fail judgment of the inspection unit 125 can also be set in various ways. For example, instead of determining that even one image among the multiple longitudinal cross-sectional images has a score that does not satisfy the standard as being defective, images with scores that do not satisfy the standard exceeding a predetermined ratio may be determined to be defective. Regarding the multiple longitudinal cross-sectional images used by the inspection unit 125 for inspection, although the above embodiment has described an example in which the image extraction unit 123 extracts four longitudinal cross-sectional images, this is not limitative. The image extraction unit 123 may extract two, three, or five or more longitudinal cross-sectional images, and the inspection unit 125 may perform inspection based on these images.

[0077] <Supplementary Note 1> An inspection system (1) for inspecting a component-mounted board on which electronic components having solder bumps formed thereon are mounted, the inspection system comprising: a three-dimensional data generation means (122) for generating three-dimensional data of an area including at least a solder joint formed by the solder bumps, using a plurality of X-ray images of the component-mounted board; a longitudinal section image extraction means (123) for extracting, from the three-dimensional data, a longitudinal section image showing a longitudinal section passing through a center of the solder joint; an inspection model generation means (124) for generating an inspection model for the inspection, using the plurality of longitudinal section images; and an inspection means (125) for performing an inspection using the inspection model, using the plurality of longitudinal section images having different observation directions extracted from the three-dimensional data of each of the component-mounted boards to be inspected.

[0078] <Supplementary Note 2> The inspection system according to Supplementary Note 1, wherein the longitudinal cross-sectional images used for training of the inspection model generation means include a plurality of longitudinal cross-sectional images with different observation directions for each piece of the three-dimensional data.

[0079] <Supplementary Note 3> The inspection system according to Supplementary Note 1, characterized in that the longitudinal cross-sectional image extraction means plots brightness variance values ​​of the horizontal cross-sectional image acquired from the three-dimensional data in the vertical direction, and determines the position of the cross-section where the brightness variance value is maximum as the vertical position that identifies the center portion.

[0080] <Supplementary Note 4> The inspection system according to Supplementary Note 1, characterized in that the longitudinal cross-sectional image extraction means plots brightness variance values ​​of the horizontal cross-sectional image acquired from the three-dimensional data in the vertical direction, and determines the position of a fault that is the center of multiple faults where the brightness variance value exceeds a predetermined threshold as the vertical position that identifies the center portion.

[0081] <Supplementary Note 5> The inspection system according to Supplementary Note 1, characterized in that the longitudinal cross-sectional image extraction means plots edge intensities of horizontal cross-sectional images acquired from the three-dimensional data in the vertical direction, and determines the position of the cross-section at which the edge intensity is maximum as the vertical position that identifies the center portion.

[0082] <Supplementary Note 6> The inspection system according to Supplementary Note 1, characterized in that the longitudinal cross-sectional image extraction means plots edge intensities of horizontal cross-sectional images acquired from the three-dimensional data in the vertical direction, and determines the position of a central fault of a plurality of faults where the edge intensity exceeds a predetermined threshold as the vertical position that identifies the central portion.

[0083] <Supplementary Note 7> The inspection system according to Supplementary Note 3, wherein the longitudinal cross-sectional image extraction means determines the horizontal position that identifies the central portion as the center of a label of a binarized image of the horizontal cross-sectional image having the maximum brightness variance value, or the center of gravity calculated by weighting the horizontal cross-sectional image having the maximum brightness variance value by brightness.

[0084] <Supplementary Note 8> The inspection system according to Supplementary Note 4, characterized in that the longitudinal cross-sectional image output means determines the horizontal position that identifies the central portion as the center of a label of a binarized image of a horizontal tomographic image that is the center of a plurality of horizontal cross sections whose brightness variance value exceeds a predetermined threshold, or the position of a center of gravity that is calculated by weighting the horizontal tomographic image that is the center of a plurality of horizontal cross sections whose brightness variance value exceeds a predetermined threshold by brightness.

[0085] <Supplementary Note 9> The inspection system according to Supplementary Note 5, wherein the longitudinal cross-sectional image extraction means determines the horizontal position that identifies the central portion as the center of a label of a binarized image of the horizontal cross-sectional image in which the edge strength is maximum, or the center of gravity calculated by weighting the horizontal cross-sectional image in which the edge strength is maximum by brightness.

[0086] <Supplementary Note 10> The inspection system according to Supplementary Note 6, characterized in that the longitudinal cross-sectional image output means determines the horizontal position that identifies the central portion as the center of a label of a binarized image of a horizontal cross-section image at the center of a plurality of horizontal cross-sections where the edge intensity exceeds a predetermined threshold, or the center of gravity calculated by weighting the horizontal cross-section image at the center of a plurality of horizontal cross-sections where the edge intensity exceeds a predetermined threshold by brightness.

[0087] <Supplementary Note 11> The inspection system according to any one of Supplementary Notes 1 to 6, characterized in that the longitudinal cross-sectional image extraction means determines, as the horizontal position specifying the central portion, the position of the center of a label of an image obtained by vertically maximum intensity projection of each horizontal tomographic image of an area including a predetermined horizontal range of the solder joint in the three-dimensional data, or the position of a center of gravity calculated by weighting the image obtained by maximum intensity projection by brightness.

[0088] <Supplementary Note 12> The inspection system according to any one of Supplementary Notes 1 to 6, characterized in that the longitudinal cross-sectional image extraction means determines, as the horizontal position specifying the central portion, the position of the center of a label of an image obtained by vertically averaging each horizontal tomographic image of an area including a predetermined horizontal range of the solder joint in the three-dimensional data, or the position of a center of gravity calculated by weighting the image obtained by the average projection by brightness.

[0089] <Supplementary Note 13> An inspection device that inspects a component-mounted board on which an electronic component having solder bumps formed thereon is mounted, the inspection device performing the inspection using: an inspection model for inspection that is generated using longitudinal cross-sectional images that show longitudinal cross sections passing through centers of the solder joints formed by the solder bumps, the longitudinal cross-sectional images being extracted from three-dimensional data of an area that includes at least the solder joints formed by the solder bumps and that is generated using a plurality of X-ray images of the component-mounted board; and a plurality of the longitudinal cross-sectional images with different observation directions that are extracted from the three-dimensional data of the component-mounted board that is an inspection target.

[0090] <Supplementary Note 14> An inspection method for inspecting a component-mounted board on which electronic components having solder bumps formed thereon are mounted, the inspection method comprising: a three-dimensional data generation step (S102) for generating three-dimensional data of an area including at least a solder joint formed by the solder bumps, using a plurality of X-ray images of the component-mounted board; a longitudinal section image extraction step (S103) for extracting, from the three-dimensional data, a longitudinal section image showing a longitudinal section passing through a center of the solder joint; an inspection model generation step (S104) for generating an inspection model for the inspection, using the plurality of longitudinal section images; and an inspection step (S105) for performing inspection using the inspection model, using the plurality of longitudinal section images having different observation directions extracted from the three-dimensional data of the component-mounted board to be inspected.

[0091] <Supplementary Note 15> A program for causing a computer to execute each step of the inspection method described in Supplementary Note 14.

[0092] DESCRIPTION OF SYMBOLS 1: X-ray inspection system 10: X-ray source 11: X-ray imaging device 12: Inspection terminal 13: Data server 20: X-ray camera 30: Work stage 100: Terminal 110: Control unit C1, C2: Rotation path O: Center point P: Pad W: Work

Claims

1. An inspection system for inspecting a component mounting board on which electronic components having solder bumps are mounted, comprising: a three-dimensional data generation means for generating three-dimensional data of an area including at least a solder joint formed by the solder bump using a plurality of X-ray images of the component mounting board; a longitudinal section image extraction means for extracting a longitudinal section image showing a longitudinal section passing through the center of the solder joint from the three-dimensional data; an inspection model generation means for generating an inspection model for the inspection using the plurality of longitudinal section images; and an inspection means for performing inspection using the inspection model using the plurality of longitudinal section images with different observation directions extracted from the three-dimensional data of each of the component mounting boards to be inspected.

2. The inspection system according to claim 1, characterized in that the longitudinal section images used for training the inspection model generation means include a plurality of the longitudinal section images having different observation directions for each of the three-dimensional data.

3. The inspection system according to claim 1, wherein said longitudinal section image extraction means plots the brightness variance value of the horizontal tomographic image obtained from said three-dimensional data in the vertical direction, and determines the position of the tomographic section where the brightness variance value is maximum as the vertical position that identifies said central portion.

4. The inspection system of claim 1, wherein the longitudinal section image extraction means plots the brightness variance value of the horizontal tomographic image obtained from the three-dimensional data in the vertical direction, and determines the position of a fault that is the center of multiple faults whose brightness variance value exceeds a predetermined threshold value as the vertical position that identifies the center portion.

5. The inspection system according to claim 1, characterized in that said longitudinal section image extraction means plots the edge intensity of the horizontal tomographic image obtained from said three-dimensional data in the vertical direction, and determines the position of the tomographic section where the edge intensity is maximum as the vertical position that identifies said central portion.

6. The inspection system according to claim 1, characterized in that said longitudinal section image extraction means plots the edge intensity of the horizontal section image obtained from said three-dimensional data in the vertical direction, and determines the position of a fault that is the center of multiple faults where the edge intensity exceeds a predetermined threshold as the vertical position that identifies said central portion.

7. The inspection system of claim 3, wherein the longitudinal section image extraction means determines, as the horizontal position specifying the central portion, the center of the label of a binarized image of the horizontal section image in which the brightness variance value is maximum, or the center of gravity calculated by weighting the horizontal section image in which the brightness variance value is maximum by brightness.

8. The inspection system of claim 4, characterized in that the longitudinal section image output means determines, as the horizontal position specifying the central portion, the center of the label of a binarized image of a horizontal tomographic image which is the center of a plurality of horizontal tomographic images whose brightness variance value exceeds a predetermined threshold value, or the center of gravity which is calculated by weighting, by brightness, a horizontal tomographic image which is the center of a plurality of horizontal tomographic images whose brightness variance value exceeds a predetermined threshold value.

9. The inspection system of claim 5, wherein the longitudinal section image extraction means determines, as the horizontal position specifying the central portion, the center of the label of a binarized image of the horizontal section image in which the edge strength is maximum, or the center of gravity calculated by weighting the horizontal section image in which the edge strength is maximum by brightness.

10. The inspection system of claim 6, characterized in that the longitudinal section image output means determines, as the horizontal position identifying the central portion, the center of the label of a binarized image of a horizontal tomographic image that is the center of a plurality of horizontal tomographic images whose edge strength exceeds a predetermined threshold value, or the center of gravity calculated by weighting, by brightness, a horizontal tomographic image that is the center of a plurality of horizontal tomographic images whose edge strength exceeds a predetermined threshold value.

11. The inspection system of any one of claims 1 to 6, characterized in that the longitudinal section image extraction means determines, as the horizontal position identifying the central portion, the center of a label of a binarized image obtained by vertically maximum projection of each horizontal tomographic image of an area including a predetermined horizontal range of the solder joint in the three-dimensional data, or the position of a center of gravity calculated by weighting the image obtained by maximum projection by brightness.

12. The inspection system of any one of claims 1 to 6, characterized in that the longitudinal section image extraction means determines, as the horizontal position identifying the central portion, the center of the label of a binarized image obtained by vertically averaging each horizontal tomographic image of an area including a predetermined horizontal range of the solder joint in the three-dimensional data, or the position of the center of gravity calculated by weighting the image obtained by the average projection by brightness.

13. An inspection device for inspecting a component mounting board on which electronic components having solder bumps are mounted, characterized in that the inspection is performed using: an inspection model for inspection generated using a longitudinal section image showing a longitudinal section passing through the center of the solder joint extracted from three-dimensional data of an area including at least the solder joint formed by the solder bump, which is generated using a plurality of X-ray images taken of the component mounting board; and a plurality of the longitudinal section images with different observation directions extracted from the three-dimensional data of the component mounting board to be inspected.

14. An inspection method for inspecting a component mounting board having electronic components with solder bumps formed thereon, comprising: a three-dimensional data generation step of generating three-dimensional data of an area including at least a solder joint formed by the solder bump using a plurality of X-ray images of the component mounting board; a longitudinal section image extraction step of extracting, from the three-dimensional data, a longitudinal section image showing a longitudinal section passing through a center of the solder joint; an inspection model generation step of generating an inspection model for the inspection using the plurality of longitudinal section images; and an inspection step of performing inspection using the inspection model, using the plurality of longitudinal section images with different observation directions extracted from the three-dimensional data of the component mounting board to be inspected.

15. A program for causing a computer to execute each step of the inspection method according to claim 14.

Images