Inspection equipment, inspection method, and program

Abstract

To provide a technology with which it is possible to identify a position of a needle tip of a probe needle with good accuracy.SOLUTION: There is provided an inspection device comprising: an imaging unit that captures an image of a needle tip of a probe needle that is provided to a probe card that is used in inspection of an inspection object; a region division unit that inputs a verification image having been captured by the imaging unit to a segmentation model having been trained using training data in which a range of the probe needle and a range of the needle tip are added to a training image derive by capturing the image of the needle tip of the probe needle, so as to recognize a needle region and a needle tip region of the verification image; a position identification unit that identifies a position of the needle tip on the basis of a relationship between the needle region and the needle tip region; and a position adjustment unit that adjusts relative positions of the inspection object and the needle tip of the probe needle on the basis of the needle tip position identified by the position identification unit.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] This disclosure relates to inspection devices, inspection methods, and programs. [Background technology]

[0002] Patent Document 1 discloses an inspection device that inspects an object to be inspected by bringing the tip of a probe needle into contact with each of several test pads provided on the object and supplying an electrical signal to the object to be inspected via the tips of the multiple probe needles. The inspection device disclosed in Patent Document 1 identifies the position of the probe needle tips based on an image taken so that the tips of multiple probe needles are visible in a single image, and adjusts the position of the probe needle tips based on the identified position of the probe needle tips. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2019-102640 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] This disclosure provides a technology that can accurately determine the position of the tip of a probe needle. [Means for solving the problem]

[0005] According to one aspect of the present disclosure, an inspection device is provided comprising: an imaging unit that photographs the tip of a probe needle provided on a probe card used for inspecting an object to be inspected; a region division unit that recognizes a needle region and a needle tip region of a verification image by inputting a verification image taken by the imaging unit into a segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to a training image of the tip of the probe needle; a position identification unit that identifies the position of the needle tip based on the positional relationship between the needle region and the needle tip region; and a position adjustment unit that adjusts the relative position between the object to be inspected and the tip of the probe needle based on the position of the needle tip identified by the position identification unit. [Effects of the Invention]

[0006] One aspect of this approach is that the position of the probe needle tip can be precisely determined. [Brief explanation of the drawing]

[0007] [Figure 1] This is a schematic cross-sectional view showing an example of an inspection apparatus according to one embodiment. [Figure 2] A block diagram showing an example of the hardware configuration of a control device according to one embodiment. [Figure 3] This is a block diagram showing an example of the functional configuration of a control device according to one embodiment. [Figure 4] A flowchart showing an example of a learning method according to one embodiment. [Figure 5] This figure shows an example of a needle tip image according to one embodiment. [Figure 6] This figure shows an example of training data according to one embodiment. [Figure 7] This flowchart shows an example of an inspection method according to one embodiment. [Figure 8] This figure shows an example of a region segmentation image according to one embodiment. [Figure 9] This flowchart shows an example of a location identification process according to one embodiment. [Figure 10]It is a diagram showing an example of a region-divided image according to an embodiment. [Figure 11] It is a graph showing an example of verification results according to an embodiment.

Mode for Carrying Out the Invention

[0008] Hereinafter, embodiments for implementing the present disclosure will be described with reference to the drawings. In each drawing, the same reference numerals are given to the same components, and redundant explanations may be omitted.

[0009] [Embodiment] [Overview] An embodiment of the present disclosure is an inspection device for electrically inspecting an object to be inspected such as a semiconductor wafer. The inspection device according to this embodiment has a probe card provided with a plurality of probe needles. The inspection device brings each probe needle provided on the probe card into contact with a test pad provided on the object to be inspected. Then, the inspection device outputs a predetermined electrical signal from an external tester to the probe card, and evaluates the electrical characteristics of the object to be inspected based on the electrical signal output from the object to be inspected via the probe needle.

[0010] The inspection device adjusts the relative position between the tip of the probe needle and the test pad in order to bring the probe needle into contact with the test pad. At this time, the tip of the probe needle is photographed by a photographing unit such as a camera attached to the mounting table on which the object to be inspected is placed, and the position of the tip of the probe needle is specified by recognizing the tip of the probe needle from the photographed image.

[0011] When recognizing the tip of the probe needle, if an image recognition technique based on rules such as blob analysis or edge analysis is used, the position of the tip of the probe needle may not be detected. The causes include, for example, when a part of the tip of the probe needle is missing due to wear or the like, or when a part of the probe needle is not included in the image due to the positional relationship between the position of the probe needle and the angle of view of the camera.

[0012] The inspection apparatus according to the present embodiment uses a learned segmentation model to recognize a needle region and a needle tip region from an image of the tip of a probe needle, and specifies the position of the tip of the probe needle based on the positional relationship between the needle region and the needle tip region. The segmentation model according to the present embodiment is trained using training data in which the range of the probe needle and the range of the tip of the probe needle are given to the image without a causal relationship with the image recognition result. By configuring as described above, the inspection apparatus according to the present embodiment can accurately specify the position of the tip of the probe needle.

[0013] Note that not having a causal relationship with the image recognition result means that it is information not based on the needle region and the needle tip region obtained as a result of performing image recognition on the image. For example, the range of the probe needle and the range of the tip of the probe needle are given to the image according to the operation of the user without recognizing the needle region and the needle tip region from the image. Also, for example, the range of the probe needle and the range of the tip of the probe needle may be given according to the operation of the user to an image in which the needle region and the needle tip region recognized from the image are different from the true range of the probe needle and the true range of the tip of the probe needle.

[0014] <Configuration of Inspection Apparatus> FIG. 1 is a schematic cross-sectional view showing an example of the inspection apparatus according to the present embodiment. As shown in FIG. 1, the inspection apparatus 10 according to the present embodiment includes an inspection apparatus main body 20 and a control device 50.

[0015] The inspection apparatus main body 20 has a hollow housing 21. A moving mechanism 23 for moving a mounting table 25 in the vertical direction (z-axis direction shown in FIG. 1) and the horizontal direction (direction in the xy plane parallel to the x-axis and y-axis shown in FIG. 1) is provided substantially at the center inside the housing 21. On the upper surface of the mounting table 25, a semiconductor wafer W, which is an example of a test object, is placed. The mounting table 25 sucks and holds the semiconductor wafer W placed on the upper surface to the upper surface of the mounting table 25 by means of a vacuum chuck or the like.

[0016] A camera 27, which is an example of a shooting unit, is mounted on the side of the mounting base 25. The camera 27 is mounted on the side of the mounting base 25 so that the shooting direction is facing upward. As the mounting base 25 moves by the movement mechanism 23, the camera 27 mounted on the side of the mounting base 25 also moves.

[0017] The moving mechanism 23 is controlled by the control device 50. The amount of movement of the moving mechanism 23 is managed by the control device 50. Therefore, the x, y, and z coordinates of the positions of the mounting platform 25 and the camera 27 within the housing 21 are managed by the control device 50.

[0018] The housing 21 has a substantially circular opening at its top. A test head 30 is provided in this opening. The test head 30 is fixed to a frame 22 provided along the periphery of the opening. Within the test head 30, a plurality of tilt adjustment parts 32 are provided at the position of the frame 22. The plurality of tilt adjustment parts 32 hold a substantially cylindrical holder 34 from above via a shaft 33 below the frame 22.

[0019] The holder 34 detachably holds a probe card 36, which is equipped with a plurality of probe needles 38, at its lower part. The plurality of probe needles 38 provided on the probe card 36 are arranged so that their tips face downward.

[0020] Although the probe card 36 illustrated in Figure 1 shows a cantilever-type probe needle 38, the probe card 36 may also be provided with a vertical-type probe needle 38. Furthermore, the probe card 36 may be provided with both a cantilever-type probe needle 38 and a vertical-type probe needle 38.

[0021] Multiple probe needles 38 are arranged on the probe card 36 such that when the semiconductor wafer W, which is placed on the mounting table 25, moves to the position for inspection, the tips of the probe needles 38 will each contact the test pads provided on the semiconductor wafer W.

[0022] Multiple probe needles 38 are each connected to wiring provided on the probe card 36. The wiring on the probe card 36 is connected to the test head 30 via wiring provided on the holder 34. An external tester 31 is connected to the test head 30.

[0023] Here, due to mounting errors when the probe card 36 is attached to the holder 34, the positions of the tips of the probe needles 38 may be shifted from the positions corresponding to the test pads provided on the semiconductor wafer W. For example, if the probe card 36 is mounted with a lateral shift, the positions of the tips of all the probe needles 38 will be shifted by a certain amount in the lateral direction. If the positions of the tips of the probe needles 38 are shifted by a large amount in the lateral direction, the tips of each probe needle 38 will no longer make contact with the corresponding test pads.

[0024] Therefore, in this embodiment, first, the control device 50 uses the camera 27 to detect the positions of each of the probe needles 38 before the inspection begins. Next, the control device 50 calculates the error between the position of the tip of each probe needle 38 and the position of the test pad. Then, based on the calculated error, the control device 50 adjusts the relative position between the probe needle 38 and the semiconductor wafer W.

[0025] In the inspection apparatus 20 configured in this way, the semiconductor wafer W placed on the mounting table 25 is inspected. First, the control device 50 controls the movement mechanism 23 so that the camera 27 is positioned below the probe needles 38. Next, the control device 50 causes the camera 27 to photograph the probe needles 38. Subsequently, the control device 50 measures the lateral position of the tip of each probe needle 38 based on the image captured by the camera 27. Then, the control device 50 adjusts the lateral position of the mounting table 25 by controlling the movement mechanism 23 to correct for any lateral misalignment of the tip of each probe needle 38.

[0026] The control device 50 controls the moving mechanism 23 to raise the mounting table 25 on which the semiconductor wafer W is placed, bringing each test pad on the semiconductor wafer W into contact with the probe needle 38 by a predetermined overdrive amount. The overdrive amount is the amount by which the mounting table 25 is raised after the mounting table 25 on which the semiconductor wafer W is placed has been raised and the tips of the probe needles 38 have been brought into contact with the test pads on the semiconductor wafer W.

[0027] The control device 50 controls the external tester 31 to output a predetermined electrical signal to the test head 30. The test head 30 outputs the electrical signal output from the external tester 31 to the probe card 36 via the wiring in the holder 34. The electrical signal output to the probe card 36 is supplied to each of the multiple probe needles 38 via the wiring in the probe card 36, and output to the test pads of the semiconductor wafer W via the probe needles 38.

[0028] The electrical signals output from the test pads on the semiconductor wafer W are output to the probe needles 38. The electrical signals output to the probe needles 38 are output to the test head 30 via the wiring in the probe card 36 and the wiring in the holder 34. The electrical signals output to the test head 30 are output to the external tester 31. The external tester 31 evaluates the electrical characteristics of the semiconductor wafer W based on the electrical signals output to the test head 30 and the electrical signals output from the test head 30, and outputs the evaluation results to the control device 50.

[0029] Furthermore, if the probe needle 38 is damaged or deformed, it will be difficult to perform the inspection correctly even if its lateral position is adjusted by the moving mechanism 23. Therefore, if damage or deformation of the tip of each probe needle 38 is detected based on the image captured by the camera 27, the control device 50 will notify the operator of the error via a display or the like, prompting them to perform maintenance or replacement of the probe card 36.

[0030] <Control device hardware configuration> Figure 2 is a block diagram showing an example of the hardware configuration of the control device 50 according to this embodiment. As shown in Figure 2, the control device 50 includes a CPU (Central Processing Unit) 500, RAM (Random Access Memory) 501, ROM (Read Only Memory) 502, auxiliary storage device 503, communication interface (I / F) 504, input / output interface (I / F) 505, and media interface (I / F) 506.

[0031] The CPU 500 operates based on programs stored in the ROM 502 or auxiliary storage device 503, and controls each part. The ROM 502 stores boot programs executed by the CPU 500 when the control device 50 starts up, as well as programs that depend on the hardware of the control device 50.

[0032] The auxiliary storage device 503 is, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The auxiliary storage device 503 stores the program executed by the CPU 500 and the data used by the program. The CPU 500 reads the program from the auxiliary storage device 503, loads it onto the RAM 501, and executes the loaded program.

[0033] The communication interface 504 communicates with the inspection device main unit 20 via a communication line such as a LAN (Local Area Network). The communication interface 504 receives data from the inspection device main unit 20 via the communication line and sends it to the CPU 500, and the CPU 500 sends the generated data back to the inspection device main unit 20 via the communication line.

[0034] The CPU 500 controls input devices such as keyboards and output devices such as displays via the I / F 505. The CPU 500 receives signals from input devices via the I / F 505 and sends them to the CPU 500. The CPU 500 also outputs the generated data to the output devices via the I / F 505.

[0035] The media interface 506 reads the program or data stored in the recording medium 507 and stores it in the auxiliary storage device 503. The recording medium 507 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), tape media, magnetic recording medium, or semiconductor memory.

[0036] The CPU 500 of the control device 50 reads the program loaded onto the RAM 501 from the recording medium 507 and stores it in the auxiliary storage device 503. Alternatively, the program may be obtained from another device via a communication line and stored in the auxiliary storage device 503.

[0037] <Functional Configuration of Control Device> Figure 3 is a block diagram showing an example of the functional configuration of the control device 50 according to this embodiment. As shown in Figure 3, the control device 50 according to this embodiment includes an image capture control unit 51, an image storage unit 52, a correct position input unit 53, a model learning unit 54, a model storage unit 55, a region division unit 56, a position identification unit 57, a position adjustment unit 58, and an inspection execution unit 59.

[0038] The image capture control unit 51, the correct position input unit 53, the model learning unit 54, the region division unit 56, the position identification unit 57, the position adjustment unit 58, and the inspection execution unit 59 are implemented, for example, by the CPU 500 shown in Figure 2 executing a program loaded onto the RAM 501. The image storage unit 52 and the model storage unit 55 are implemented, for example, by the RAM 501 or the auxiliary storage device 503 shown in Figure 2.

[0039] The imaging control unit 51 controls the camera 27 to capture an image of the tip of the probe needle 38 (hereinafter, the image captured by the camera 27 is also called the "needle tip image"). When the imaging control unit 51 captures a needle tip image to be used to train the segmentation model (hereinafter, also called the "training image"), it stores the training image in the image storage unit 52. When the imaging control unit 51 captures a needle tip image to be used to identify the position of the probe needle tip (hereinafter, also called the "verification image"), it sends the verification image to the region division unit 56.

[0040] The image storage unit 52 stores multiple training images captured by the camera 27. The number of training images should be sufficient to train the segmentation model. The number of images sufficient to train the segmentation model varies depending on the type of model, but for example, it is about 50 images.

[0041] The correct position input unit 53 accepts input from the user of information representing the range of the probe needle 38 captured in the training image (correct value of the needle region) and the range of the needle tip of the probe needle 38 (correct value of the needle tip region) (hereinafter also referred to as "correct position information"). The correct position input unit 53 adds the received correct position information to the training image and stores it in the image storage unit 52. Hereinafter, the training image to which the correct position information has been added will also be referred to as "training data". The training data functions as training data. Verification images are not used as training data.

[0042] The model learning unit 54 learns a segmentation model based on the training data stored in the image storage unit 52. The segmentation model takes a needle tip image as input and outputs an image in which each pixel of the needle tip image is divided into a needle region, a needle tip region, and other regions. The needle region is the region in which the probe needle 38 is visible. The needle tip region is the region in which the tip of the probe needle 38 is visible. Other regions are regions that are neither the needle region nor the needle tip region.

[0043] The model memory unit 55 stores the segmentation model learned by the model learning unit 54.

[0044] The region segmentation unit 56 inputs the verification image captured by the camera 27 into a trained segmentation model, thereby dividing the verification image into a needle region, a needle tip region, and other regions. Hereinafter, the image obtained by segmenting the verification image will be referred to as a "region segmentation image".

[0045] The positioning unit 57 determines the position of the needle tip of the probe needle 38 based on the positional relationship between the needle region and the needle tip region included in the region-divided image. When the region-divided image includes multiple needle regions, the positioning unit 57 determines the position of the needle tip for each needle region.

[0046] The position adjustment unit 58 controls the movement mechanism 23 to bring the test pad provided on the object under test into contact with the tip of the probe needle 38. At this time, the position adjustment unit 58 adjusts the relative position between the tip of the probe needle 38 and the test pad based on the information representing the position of the tip of the probe needle 38 identified by the position identification unit 57. In this embodiment, it is assumed that the positional relationship in the Z-axis direction between the tip of the probe needle 38 and the test pad is correctly adjusted.

[0047] The inspection execution unit 59 instructs the external tester 31 to start the inspection, and the inspection of the semiconductor wafer W begins.

[0048] <Processing steps for learning methods> Figure 4 is a flowchart showing an example of a learning method according to this embodiment. The learning method according to this embodiment is a method for learning a segmentation model that divides a needle tip image into regions.

[0049] In step S1, the imaging control unit 51 controls the camera 27 to photograph the tip of the probe needle 38. The imaging control unit 51 may photograph so that the tips of multiple probe needles 38 are visible in one image, or it may photograph so that the tip of one probe needle 38 is visible in one image.

[0050] The shooting control unit 51 stores the training image captured by the camera 27 in the image storage unit 52. At this time, the shooting control unit 51 stores information representing the shooting position in association with the needle tip image. The information representing the shooting position is, for example, the coordinates on the xy plane where the camera 27 was positioned at the time of shooting.

[0051] ≪Image of needle tip≫ Figure 5 shows an example of a needle tip image according to this embodiment. As shown in Figure 5, the needle tip image 100 according to this embodiment captures the needle tip of the probe needle 38. In the example shown in Figure 5, four probe needles 38-1 to 38-4 are captured in the needle tip image 100, but the number of probe needles 38 captured in a single needle tip image 100 is not limited.

[0052] Let's return to Figure 4 for explanation. In step S2, the correct position input unit 53 receives input of correct position information to be added to the training image in response to user operation. The correct position input unit 53 generates training data by adding the received correct position information to the training image. The correct position input unit 53 stores the generated training data in the image storage unit 52.

[0053] The correct position input unit 53 may accept input of correct position information for all training images stored in the image storage unit 52, or it may accept input of correct position information for some of the training images. For example, the correct position input unit 53 may accept input of correct position information only for training images in which the position of the needle tip cannot be correctly recognized by conventional rule-based image recognition technology.

[0054] If multiple probe needles 38 are captured in the training image, the correct position input unit 53 may accept input of correct position information for all probe needles 38 included in the training image, or it may accept input of correct position information for some of the probe needles 38. For example, if the entire range of a probe needle 38 is not included in the training image, as shown in Figure 5 with probe needle 38-4, the input of correct position information for that probe needle 38 does not need to be accepted.

[0055] When accepting input of correct position information only for training images in which the position of the needle tip cannot be correctly recognized by conventional rule-based image recognition technology, the correct position input unit 53 first recognizes the position of the needle tip of the probe needle 38 from all training images stored in the image storage unit 52. Next, the user selects a training image in which the position of the needle tip of the probe needle 38 obtained as a recognition result differs from the position of the true needle tip of the probe needle 38 captured in the training image. Subsequently, the user inputs the range of the probe needle 38 and the range of the needle tip of the probe needle 38 captured in the training image. Then, the correct position input unit 53 adds the correct position information input by the user to the training image and stores it in the image storage unit 52.

[0056] ≪Training Data≫ Figure 6 shows an example of training data according to this embodiment. Figure 6 shows the range 200 and true needle tip position 210 of the true probe needle 38, and the range 220 and needle tip position 230 of the probe needle 38 recognized by rule-based image recognition technology. The recognized needle tip position 230 is the centroid position of the recognized range 220 of the probe needle 38.

[0057] In the training image shown in Figure 6(A), the recognized probe needle 38 is in the shape of a vertically elongated ellipse, with a portion on the left side not being recognized. As a result, the recognized needle tip position 230 is shifted to the right from the true needle tip position 210.

[0058] In the training image shown in Figure 6(B), the recognized probe needle 38 is U-shaped, and the central portion, which includes the true needle tip position 210, is not recognized. As a result, the recognized probe needle 38 position 230 is significantly different from the true needle tip position 210.

[0059] Thus, rule-based image recognition technology may not correctly recognize the range and position of the needle tip. In this case, training data is generated by adding the range 200 (correct value of the needle region) and the true needle tip position 210 (correct value of the needle tip region) of the true probe needle 38 to the training image in which the needle tip position is not correctly recognized, as correct position information. By training a segmentation model using the training data generated in this way, it becomes possible to recognize the needle region and the needle tip region with high accuracy.

[0060] Let's return to Figure 4 for explanation. In step S3, the model learning unit 54 reads the training data stored in the image storage unit 52. Next, the model learning unit 54 trains a segmentation model using the read-out training data. The segmentation model according to this embodiment takes a needle tip image as input and outputs an image in which each pixel of the needle tip image is divided into a needle region, a needle tip region, and other regions.

[0061] The segmentation model according to this embodiment is, for example, a deep learning model that performs semantic segmentation based on deep learning. An example of semantic segmentation based on deep learning is Unet++ disclosed in Reference 1.

[0062] [Reference 1] Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, and Jianming Liang, "UNet++: A Nested U-Net Architecture for Medical Image Segmentation", [online], [Retrieved June 15, 2020], Internet<URL: https: / / arxiv.org / abs / 1807.10165>

[0063] When the model learning unit 54 learns the segmentation model, it performs deep learning to optimize the error in the position of the tip of the probe needle 38. To this end, the model learning unit 54 uses a loss function including terms representing the error in the centroid position of the needle region and the error in the centroid position of the tip region (hereinafter also referred to as "centroid loss").

[0064] The error in the centroid position of the needle region is the straight-line distance in the xy plane between the centroid position of the range of the probe needle 38 given in the learning image and the centroid position of the needle region recognized from the learning image. The error in the centroid position of the tip region is the straight-line distance in the xy plane between the centroid position of the range of the tip of the probe needle 38 given in the learning image and the centroid position of the tip region recognized from the learning image.

[0065] Specifically, in the loss function loss in this embodiment, the estimated value of the centroid position of the needle region is set as (x needle , y needle ), the correct value of the centroid position of the needle region is set as (^x needle , ^y needle ), the estimated value of the centroid position of the tip region is set as (x tip , y tip ), the correct value of the centroid position of the tip region is set as (^x tip , ^y tip ), and is represented by the following equations (1) to (4). Note that the symbol "ˆ" should be written directly above the following character in principle, but is written immediately before in the text due to text notation limitations. In the mathematical formula, it is written directly above the original character.

[0066]

Equation

[0067] However, BCEDiceLoss is a conventional loss function used in Unet++ described in Reference 1. centerloss needle is the error in the centroid position of the needle region. centerloss tip is the error in the centroid position of the tip region. Therefore, centerloss needle+ centerloss tip This is the center of gravity loss.

[0068] Note that the segmentation model is not limited to the model described above; any segmentation model capable of machine learning based on needle tip images with ground truth position information can be used. Other examples of segmentation models include instance segmentation models and panoptic segmentation.

[0069] In step S4, the model learning unit 54 stores the segmentation model learned in step S3 in the model storage unit 55.

[0070] <Processing procedure for inspection method> Figure 7 is a flowchart showing an example of an inspection method according to this embodiment. The inspection method according to this embodiment is a method of inspecting an object by adjusting the relative position between the probe needle and the object to be inspected using a trained segmentation model.

[0071] In step S11, the imaging control unit 51 controls the camera 27 to photograph the tip of the probe needle 38. The imaging control unit 51 sends the verification image captured by the camera 27 to the region division unit 56.

[0072] In step S12, the region segmentation unit 56 receives the verification image from the imaging control unit 51. Next, the region segmentation unit 56 reads the trained segmentation model from the model storage unit 55. Subsequently, the region segmentation unit 56 inputs the received verification image into the read segmentation model. As a result, the region segmentation unit 56 obtains a region segmented image in which the verification image is divided into a needle region, a needle tip region, and other regions. The region segmentation unit 56 sends the obtained region segmented image to the position identification unit 57.

[0073] <<Region Segmentation Image>> Figure 8 shows an example of a segmented image according to this embodiment. Figure 8 shows the result of segmenting the needle tip image 100 shown in Figure 5 using a trained segmentation model.

[0074] As shown in Figure 8, the region segmentation image 300 according to this embodiment includes needle regions 310 (310-1 to 310-4) corresponding to the range of the probe needle tip 38 and needle tip regions 320 (320-1 to 320-3) corresponding to the position of the probe needle tip 38. As shown in Figure 6, the segmentation model according to this embodiment is trained using training data that includes the positional relationship between the range of the probe needle 38 and the range of the probe needle tip 38, so it is expected that the needle tip region 320 will be recognized as being located within one of the needle regions 310.

[0075] Let's return to Figure 7 for explanation. In step S13, the positioning unit 57 receives a region division image from the region division unit 56. Next, the positioning unit 57 determines the position of the needle tip of the probe needle 38 based on the positional relationship between the needle region and the needle tip region included in the received region division image. When the region division image includes multiple needle regions, the positioning unit 57 determines the position of the needle tip for each needle region.

[0076] ≪Procedure for Location Identification≫ Figure 9 is a flowchart showing an example of the location identification process (step S13 in Figure 7) according to this embodiment.

[0077] In step S13-1, the positioning unit 57 receives a region segmentation image from the region segmentation unit 56. The positioning unit 57 selects the needle region to be processed from among the needle regions included in the region segmentation image.

[0078] In step S13-2, the positioning unit 57 obtains the number of needle tip regions included in the needle region selected in step S13-1. If there is only one needle tip region included in the needle region (YES), the positioning unit 57 proceeds to step S13-3. If there are no needle tip regions included in the needle region, or if there are two or more needle tip regions included in the needle region (NO), the positioning unit 57 proceeds to step S13-4.

[0079] In step S13-3, the positioning unit 57 determines the centroid position of the needle tip region included in the needle region selected in step S13-1. The positioning unit 57 identifies the obtained centroid position of the needle tip region as the position of the needle tip of the probe needle 38.

[0080] In step S13-4, the positioning unit 57 determines the centroid position of the needle region selected in step S13-1. The positioning unit 57 identifies the obtained centroid position of the needle region as the position of the needle tip of the probe needle 38.

[0081] If the segmented image obtained in step S13-1 contains multiple needle regions, steps S13-2 to S13-4 are performed for each needle region.

[0082] In step S13-5, the position identification unit 57 outputs information representing the position of the needle tip of the probe needle 38 identified in step S13-3 or S13-4. The needle tip position identified by the position identification unit 57 is the coordinate of the needle tip position in the verification image. Therefore, the position identification unit 57 uses the coordinates representing the shooting position stored in association with the verification image to convert the coordinates of the needle tip position in the verification image into coordinates on the xy plane and outputs this as information representing the position of the needle tip of the probe needle 38.

[0083] Figure 10 shows an example of a region segmentation image according to this embodiment. Figure 10(A) is an example of a region segmentation image in which one needle tip region 320 is included in the needle region 310. As shown in Figure 10(A), when one needle tip region 320 is included in the needle region 310, the position identification unit 57 determines the centroid position of the needle tip region 320 as the position of the needle tip of the probe needle 38.

[0084] Figure 10(B) is an example of a region segmentation image in which the needle region 310 includes multiple needle tip regions 320-1 to 320-2. As shown in Figure 10(B), when the needle region 310 includes multiple needle tip regions 320, the position identification unit 57 determines the centroid position of the needle region 310 as the position of the needle tip of the probe needle 38.

[0085] Figure 10(C) is an example of a region segmentation image in which the needle region 310 does not include the needle tip region. As shown in Figure 10(C), when the needle region 310 does not include the needle tip region, the position identification unit 57 determines the centroid position of the needle region 310 as the position of the needle tip of the probe needle 38.

[0086] Let's return to Figure 7 for explanation. In step S14, the position adjustment unit 58 controls the movement mechanism 23 to move the mounting table 25 on which the semiconductor wafer W is placed below the probe needle 38. Next, the position adjustment unit 58 adjusts the relative position between the tip of the probe needle 38 and the test pad based on the information representing the position of the tip of the probe needle 38 identified by the position identification unit 57.

[0087] Specifically, first, the position adjustment unit 58 calculates the difference in the position of the needle tip of each probe needle 38 in the xy plane between the needle tip of each probe needle 38 and the test pad on the semiconductor wafer W that is in contact with the needle tip of each probe needle 38, based on the information representing the position of the needle tip of the probe needle 38 identified by the position identification unit 57. Next, the position adjustment unit 58 adjusts the position of the semiconductor wafer W in the xy plane based on the difference in the needle tip positions.

[0088] The position adjustment unit 58 then controls the movement mechanism 23 to raise the mounting table 25 so that the test pads on the semiconductor wafer W and the tips of the probe needles 38 come into contact. As a result, the tips of the multiple probe needles 38 provided on the probe card come into contact with the corresponding test pads.

[0089] In step S15, the inspection execution unit 59 instructs the external tester 31 to start the inspection, and the inspection of the semiconductor wafer W begins. The external tester 31 outputs a predetermined electrical signal to the test head 30. Based on the electrical signals input and output to the semiconductor wafer W via the probe needle 38, the external tester 31 evaluates the electrical characteristics of the semiconductor wafer W and outputs the evaluation result to the inspection execution unit 59.

[0090] <Verification Results> Figure 11 is a graph showing an example of the verification results according to this embodiment. In Figure 11, the box plots represent (1) the error between the position of the needle tip and the true position of the needle tip when the position of the needle tip is the centroid of the needle tip region, (2) the error between the position of the needle tip and the true position of the needle tip when the position of the needle tip is the centroid of the needle tip region, and (3) the error between the position of the needle tip and the true position of the needle tip when the position of the needle tip is determined by the position determination process described above.

[0091] As shown in Figure 11, (1) the error between the needle tip position and the true needle tip position when the needle tip position is defined as the centroid of the needle tip region was an average of 0.66 nanometers. (2) The error between the needle tip position and the true needle tip position when the needle tip position is defined as the centroid of the needle region was an average of 0.71 nanometers. In contrast, (3) the error between the needle tip position and the true needle tip position when the needle tip position is determined by the position determination process was an average of 0.64 nanometers. Therefore, determining the needle tip position by the position determination process yields the result with the smallest error.

[0092] As described above, the verification results demonstrate that the positioning process in this embodiment can accurately determine the position of the probe needle tip. Furthermore, (3) the variation in the error between the position of the needle tip determined by the above positioning process and the true position of the needle tip was ±0.68 nanometers, which is approximately the midpoint of the error variation in (1) and (2), indicating that there is no problem with the degree of variation.

[0093] <Effects of the Embodiment> The inspection device 10 according to this embodiment inputs an image of the tip of a probe needle into a trained segmentation model, recognizes the needle region and the needle tip region of the image, and identifies the needle tip of the probe needle based on the positional relationship between the needle region and the needle tip region. The segmentation model is trained using training data in which the range of the probe needle and the range of the needle tip are added to the image. Therefore, the inspection device 10 according to this embodiment can accurately identify the position of the needle tip of the probe needle. As a result, the inspection device 10 according to this embodiment can reliably bring the needle tip of the probe needle into contact with the test pad of the object being inspected.

[0094] The inspection device 10 according to this embodiment identifies the position of the needle tip based on either the needle region or the needle tip region, depending on the number of needle tip regions included in the needle region. Specifically, when the needle region includes one needle tip region, the inspection device 10 according to this embodiment identifies the centroid of the needle tip region as the position of the needle tip, and in other cases, identifies the centroid of the needle region as the position of the needle tip. As shown in the verification results, the inspection device 10 according to this embodiment can identify the position of the needle tip of the probe needle with even greater accuracy.

[0095] The segmentation model according to this embodiment performs semantic segmentation based on deep learning. Images of the tips of probe needles may capture multiple probe needles, but the tips of different probe needles will not overlap. Therefore, the inspection device 10 according to this embodiment can use semantic segmentation, which requires relatively little processing power.

[0096] The segmentation model according to this embodiment is deep-learned to optimize the error in the centroid position of the needle region and the error in the centroid position of the needle tip region. Therefore, the inspection device 10 according to this embodiment can determine the position of the needle tip of the probe needle with even greater accuracy.

[0097] [supplement] The inspection apparatus and inspection method according to the embodiments disclosed herein are illustrative and not restrictive in all respects. The embodiments can be modified and improved in various ways without departing from the scope and spirit of the appended claims. The matters described in the above embodiments can be otherwise configured and combined in a non-consistent manner. [Explanation of Symbols]

[0098] W Semiconductor wafer 10 Inspection equipment 20 Inspection device main unit 21 cabinets 22 frames 23 Moving mechanism 25 Mounting platform 27 Cameras 31 External Tester 32 Tilt adjustment section 33 shafts 34 holder 36 Probe Cards 38 Probe needles 50 Control device 51 Imaging Control Unit 52 Image storage unit 53 Correct position input section 54 Model Learning Department 55 Model Memory Unit 56 Area division part 57 Location identification part 58 Position adjustment section 59 Inspection Execution Department

Claims

1. An imaging unit that photographs the tip of the probe needle attached to the probe card used for inspecting the object being examined, A segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to a training image of the tip of the probe needle is input to a verification image captured by the imaging unit, thereby enabling a region segmentation unit to recognize the needle region and the needle tip region of the verification image. A position determination unit that determines the position of the needle tip based on the positional relationship between the needle region and the needle tip region, A position adjustment unit adjusts the relative position between the object to be inspected and the tip of the probe needle based on the position of the needle tip identified by the position identification unit, Equipped with, The position identification unit identifies the position of the needle tip based on either the needle region or the needle tip region, according to the number of needle tip regions included in the needle region. Inspection device.

2. An inspection apparatus according to claim 1, The position-determining unit identifies the position of the needle tip as the centroid of the needle tip region when the needle region includes one needle tip region, and identifies the position of the needle tip as the centroid of the needle region when the needle region does not include a needle tip region or when the needle region includes multiple needle tip regions. Inspection device.

3. An imaging unit that photographs the tip of the probe needle attached to the probe card used for inspecting the object being examined, A segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to a training image of the tip of the probe needle is input to a verification image captured by the imaging unit, thereby enabling a region segmentation unit to recognize the needle region and the needle tip region of the verification image. A position determination unit that determines the position of the needle tip based on the positional relationship between the needle region and the needle tip region, A position adjustment unit adjusts the relative position between the object to be inspected and the tip of the probe needle based on the position of the needle tip identified by the position identification unit, Equipped with, The segmentation model performs semantic segmentation based on deep learning and is deep-learned to optimize the error in the centroid position of the needle region and the centroid position of the needle tip region recognized from the training images. Inspection device.

4. A process of photographing the tip of a probe needle attached to a probe card used for inspecting an object under test, The process involves inputting a verification image captured in the aforementioned imaging step into a segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to a training image of the tip of the probe needle, thereby recognizing the needle region and the needle tip region of the verification image. A step of determining the position of the needle tip based on the positional relationship between the needle region and the needle tip region, A step of adjusting the relative position between the object to be inspected and the tip of the probe needle based on the position of the needle tip identified by the above-mentioned identification step, Execute, The aforementioned identification step involves identifying the position of the needle tip based on either the needle region or the needle tip region, depending on the number of needle tip regions included in the needle region. Testing method.

5. A process of photographing the tip of a probe needle attached to a probe card used for inspecting an object under test, The process involves inputting a verification image captured in the aforementioned imaging step into a segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to a training image of the tip of the probe needle, thereby recognizing the needle region and the needle tip region of the verification image. A step of determining the position of the needle tip based on the positional relationship between the needle region and the needle tip region, A step of adjusting the relative position between the object to be inspected and the tip of the probe needle based on the position of the needle tip identified by the above-mentioned identification step, Execute, The segmentation model performs semantic segmentation based on deep learning and is deep-learned to optimize the error in the centroid position of the needle region and the centroid position of the needle tip region recognized from the training images. Testing method.

6. In the control device, A procedure for recognizing the needle region and needle tip region of a verification image by inputting a verification image of the needle tip of a probe needle into a segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to training images of the needle tip of a probe needle provided on a probe card used for inspecting an object under test, and A procedure for determining the position of the needle tip based on the positional relationship between the needle region and the needle tip region, Make it run, The identifying procedure involves identifying the position of the needle tip based on either the needle region or the needle tip region, depending on the number of needle tip regions included in the needle region. program.

7. In the control device, A procedure for recognizing the needle region and needle tip region of a verification image by inputting a verification image of the needle tip of a probe needle into a segmentation model trained using training data in which the range of the probe needle and the range of the needle tip are assigned to training images of the needle tip of a probe needle provided on a probe card used for inspecting an object under test, and A procedure for determining the position of the needle tip based on the positional relationship between the needle region and the needle tip region, Make it run, The segmentation model performs semantic segmentation based on deep learning and is deep-learned to optimize the error in the centroid position of the needle region and the centroid position of the needle tip region recognized from the training images. program.

Images