Wafer defect detection method and detection device

Abstract

The application discloses a wafer defect detection method and a detection device. The method comprises the following steps: placing a sample on a hollow carrier and adjusting the angle of the hollow carrier to keep the sample horizontal; adjusting the height and angle of a telecentric lens according to the thickness of the sample, the telecentric lens being a 3 times infrared lens, and performing focusing treatment on the telecentric lens to make the imaging area centered; arranging three cameras above the sample, calibrating the positions of adjacent cameras to make the fields of view coincide; adjusting the position of the sample to a measurement initial point based on an XY platform, moving the sample in a preset movement mode by the XY movement platform, scanning the sample by the cameras, and synchronously collecting images to obtain an original defect image set; performing X direction correction splicing on each image in the original defect image set after layer cutting based on an image processing algorithm, and performing denoising filtering and binarization processing to obtain a defect distribution diagram of each layer and output a defect information summary table; and the application can realize simultaneous observation of defects on the upper and lower surfaces and in the interior of a wafer, and improves the detection efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wafer defect detection, and more particularly to a wafer defect detection method and a detection device. BACKGROUND

[0002] With the development and improvement of the manufacturing level of electronic products, the requirements and standards of chip yield are gradually improved. In the product manufacturing process, due to environmental disturbances such as temperature fluctuations and stress changes, cracks may occur in the chip and gradually expand in the manufacturing process, resulting in product failure and becoming waste. Therefore, an efficient and accurate defect detection method is needed to accurately screen and remove unqualified chips in the production link to reduce losses and costs.

[0003] Chip processing belongs to subtractive manufacturing, and the fine microcracks generated in the cutting and polishing process are concentrated on the silicon substrate on the back of the wafer. The size of these cracks is usually in microns, which is not easy to identify by manual observation, and is prone to missed detection and misdiagnosis. The wafer with undetected defects may be broken in the subsequent processing process, resulting in great loss of time and cost.

[0004] At present, the mainstream schemes in the field of wafer defect detection are ultrasonic wave scheme and infrared light scheme. The infrared detection scheme usually adopts front light reflection or backlight transmission composed of infrared or near-infrared waveband. Among them, the front light reflection type is illuminated vertically by an infrared waveband light source, and the reflected light is received for imaging; the backlight transmission type places the light source and the detector on the two sides of the chip respectively, and uses the light transmitted through the chip to realize microcrack detection.

[0005] Disadvantages of the prior art:

[0006] Manual inspection has poor precision, low efficiency, large individual difference, and is extremely dependent on the physiological and psychological state of the detection personnel. It completely relies on the subjective judgment of personnel and cannot mark defects; ultrasonic wave detection has good precision but low detection efficiency. On the other hand, the chip needs to be placed in a liquid, and the chip is easy to be contaminated by the liquid; the infrared light detection has a limited field of view due to the limitation of the camera target surface and the lens magnification, and the detection speed is slow, making it difficult to accurately locate the defects. SUMMARY

[0007] In order to solve at least one of the above technical problems, the present application provides a wafer defect detection method and a detection device.

[0008] The first aspect of the present application provides a wafer defect detection method, characterized in that it comprises the following steps:

[0009] Step S1: placing a sample on a hollow carrier and adjusting the angle of the hollow carrier to keep the sample horizontal;

[0010] Step S2: Adjust the height and angle of the telecentric lens according to the sample thickness, focus the telecentric lens, and center the imaging area;

[0011] Step S3: Place three cameras above the sample, calibrate the positions of adjacent cameras, and make the fields of view coincide;

[0012] Step S4: Adjust the sample position to the initial measurement point based on the XY platform, (the field of view position after focusing is determined, as long as the sample edge enters the field of view, the sample end point can be used as the initial measurement point) the XY motion platform moves the sample according to the preset motion mode, the camera scans the sample, and the image is collected synchronously to obtain the original defect image set;

[0013] Step S5: Based on the image processing algorithm, each image in the original defect image set is cut layer by layer, corrected in the X direction, spliced, denoised, filtered, and binarized to obtain the defect distribution map of each layer, and finally the complete defect distribution map of the XY plane is synthesized and the defect information summary table is output.

[0014] In a preferred embodiment of the application, the focusing process in step S2 includes:

[0015] Adjust the height and angle of the telecentric lens to center the imaging area, the upper surface features are clearly imaged on the right side of the imaging area, the internal features are clearly imaged in the middle of the imaging area, and the lower surface features are clearly imaged on the left side of the imaging area.

[0016] In a preferred embodiment of the application, the calibration method in step S3 is as follows:

[0017] The three cameras are respectively referred to as the first camera, the second camera and the third camera, the chessboard test plate is placed on the hollow carrier, the first camera is adjusted to focus on the test plate, and the center of the chessboard pattern is clearly imaged in the center of the field of view.

[0018] Adjust the second camera to focus on the test plate, and fine-tune the second camera to display the same position image as the center of the field of view of the first camera;

[0019] Repeat the focusing of the second camera and the third camera according to the focusing method of the first camera and the second camera, and fine-tune the first camera to display the same position image as the center of the field of view of the third camera.

[0020] In a preferred embodiment of the application, step 4 specifically includes:

[0021] Place the sample to be measured on the hollow carrier, and move the sample through the XY motion platform;

[0022] The interval of the single step is the quotient of the depth of field L of the telecentric lens and the angle γ between the depth of field L and the relative horizontal plane, and then multiplied by the frame rate of the camera and the sine of the angle α at which the first camera is placed, so as to obtain the horizontal line scanning speed;

[0023] The XY motion platform moves horizontally to the left, and 3D information images of each region of the sample are collected;

[0024] After the X direction scanning, the Y direction is changed, and the X direction is scanned again in the reverse direction at the horizontal line scanning speed, and so on, until the entire sample is scanned, and the original defect map set is obtained.

[0025] In a preferred embodiment of the present application, the method for obtaining the complete sample defect distribution map in step 5 is as follows:

[0026] The XY motion platform moves in a preset manner while the camera collects images, and the Z direction is layered according to the layering accuracy and the sample thickness;

[0027] According to the layering condition, the image is segmented, each part of the segmented image corresponds to a depth layer, and the corresponding part of the image scanned at all different XY positions is spliced to obtain the defect distribution map of each depth layer;

[0028] The clarity or sharpness of the defect image in each depth layer image is analyzed, the image is processed, and the clear defect is recognized;

[0029] All the depth layer images are aligned and superimposed to obtain the complete defect distribution map of the sample to be tested.

[0030] In a preferred embodiment of the present application, the defect information summary table includes defect serial number, defect X coordinate, Y coordinate, Z coordinate, defect diameter, defect length, defect width and defect type.

[0031] The second aspect of the present application provides a wafer defect detection device applied to a wafer defect detection method, which comprises an optical device, a motion control module and an image fusion module.

[0032] The optical device comprises a large target surface camera, a telecentric lens, an altimeter, a bright field infrared light source and a dark field infrared light source.

[0033] The large target surface camera and the telecentric lens are arranged above the sample, and the altimeter is used to measure the height between the product and the large target surface camera.

[0034] The bright field infrared light source and the dark field infrared light source are arranged below the sample.

[0035] The motion control module comprises an XY motion platform and a hollow carrier arranged on the XY motion platform, and the XY motion platform is used to control the movement of the hollow carrier along the X direction or the Y direction.

[0036] The image fusion module comprises an image synthesis algorithm and an image stitching algorithm.

[0037] In a preferred embodiment of the present application, the image synthesis algorithm detects the definition of each image collected, and the image synthesis algorithm restores the three-dimensional information of the object by comparing the images of the sample taken at different X positions of the lens using the focusing method.

[0038] The above technical solution of the present application has the following advantages compared with the prior art:

[0039] 1. The present application adopts a large target camera, a high-precision telecentric lens, and a bright-dark field backlight source tilt imaging, which can realize simultaneous observation of the upper and lower surfaces and the internal defects of the wafer, and only needs to determine the pixel position of the defect to locate its depth without layer-by-layer scanning comparison, greatly improving the detection efficiency.

[0040] 2. The present application uses two symmetrically placed cameras to simultaneously scan the same area from two angles to avoid the situation that the upper defects block the imaging of the lower defects, thereby improving the defect detection rate.

[0041] 3. The image synthesis algorithm included in the present application can fuse multiple images into one total image according to the definition algorithm, and locate the XYZ three-dimensional accurate position of the defect. The longitudinal accuracy is the product of the depth of field of the high-precision telecentric lens and the angle β between the main light ray in the medium and the horizontal plane. By determining the relative position of the center pixel of the clear defect image in the picture, the three-dimensional information of the object can be obtained. Specifically, the algorithm analyzes the definition or sharpness of the defect image in different depth layers, identifies the clear defect image, calculates the depth value of the pixel through the relative position of the center pixel in the image. Finally, the clearest part of all the defects scanned is synthesized into one total image. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, some of the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0043] Figure 1 is a schematic diagram of the wafer defect detection method of the embodiment of the present application;

[0044] Figure 2 is a schematic diagram of the scanning method of the embodiment of the present application;

[0045] Figure 3 is a schematic diagram of the XY scanning area of the embodiment of the present application;

[0046] Figure 4 is a schematic diagram of an image processing algorithm of an embodiment of the present application;

[0047] Figure 5 is a schematic diagram of an optical device of an embodiment of the present application;

[0048] Figure 6 is a schematic diagram of a dual-camera defect analysis of an embodiment of the present application;

[0049] Figure 7 is a schematic diagram of an imaging principle of an embodiment of the present application;

[0050] Figure 8 is a schematic diagram of a line-scan detection principle of an embodiment of the present application;

[0051] Figure 9 is a schematic diagram of an optical system of an embodiment of the present application.

[0052] In the figure, 1 is a dark-field infrared light source, 2 is a bright-field infrared light source, 3 is a measured object surface, 4 is a telecentric lens, and 5 is an infrared camera target surface. DETAILED DESCRIPTION

[0053] In order to more clearly understand the above objectives, features and advantages of the present application, the present application will be further described in detail below with specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0054] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, and therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below.

[0055] Embodiment One

[0056] Referring to Figure 1 and Figure 9 shown, the present application proposes a wafer defect detection method, characterized in that it comprises the following steps:

[0057] Step S1: Place the sample on the hollow carrier and adjust the angle of the hollow carrier to keep the sample horizontal;

[0058] Step S2: Adjust the height and angle of the telecentric lens 4 according to the thickness of the sample, the telecentric lens 4 is a 3x infrared lens, focus the telecentric lens 4 to center the imaging area;

[0059] Step S3: Set three cameras above the sample, position calibration is performed on adjacent cameras to make the fields of view coincide;

[0060] Step S4: adjusting the sample position to the measurement initial point based on the XY platform, the XY motion platform drives the sample to move according to the preset motion mode, the camera scans the sample, and the image is collected synchronously to obtain the original defect image set;

[0061] Step S5: performing X direction correction splicing and denoising filtering and binarization processing on each image in the original defect image set after layer cutting based on the image processing algorithm, obtaining the defect distribution map of each layer, finally synthesizing the complete defect distribution map of the XY plane, and outputting the defect information summary table.

[0062] It should be noted that the sample to be measured should not be too thick, and the detection effect is best when the sample thickness is not more than 3 mm. Dust, foreign matter, white spots, scratches and other defects may exist on the upper and lower surfaces and the interior of the sample.

[0063] The lens depth of field L is generally given by the lens supplier, or can be measured.

[0064] According to the embodiment of the present application, the focusing processing in step S2 comprises:

[0065] The height and angle of the telecentric lens 4 are adjusted so that the imaging area is centered, the upper surface features are clearly imaged on the right side of the imaging area, the internal features are clearly imaged in the middle of the imaging area, and the lower surface features are clearly imaged on the left side of the imaging area.

[0066] According to the embodiment of the present application, the calibration method in step S3 is as follows:

[0067] The three cameras are respectively referred to as the first camera, the second camera and the third camera, the chessboard test plate is placed on the hollow carrier, the first camera is adjusted to focus on the test plate so that the center of the chessboard pattern is clear in the central field of view;

[0068] The second camera is adjusted to focus on the test plate, and the second camera is fine-tuned to display the same position as the central field of view of the first camera;

[0069] The second camera and the third camera are repeatedly focused according to the focusing mode of the first camera and the second camera, and the first camera is fine-tuned to display the same position as the central field of view of the third camera.

[0070] According to the embodiment of the present application, step 4 specifically comprises:

[0071] The sample to be measured is placed on the hollow carrier, and the XY motion platform moves the sample;

[0072] The quotient of the sine of the angle γ between the depth of field L of the telecentric lens 4 and the relative horizontal plane is the interval of X-axis single step, and then multiplied by the frame rate of the camera and the sine of the angle α at which the first camera is placed to obtain the horizontal line scanning speed;

[0073] The XY motion platform moves horizontally to the left by steps to collect 3D information images of each area of the sample;

[0074] After scanning along the X direction, the Y direction is changed to scan reversely along the X direction at a horizontal line scanning speed, and the scanning is repeated until the whole sample is scanned to obtain a set of original defect maps.

[0075] According to the embodiment of the present application, the method for obtaining the defect distribution map of the whole sample in step 5 is as follows:

[0076] The XY motion platform moves in a preset manner while the camera collects images, and the images are pre-layered according to the Z-direction layering accuracy and the sample thickness;

[0077] The images are segmented according to the layering, each part of the segmented images corresponds to a depth layer, and the corresponding parts of the images scanned at different XY positions are spliced to obtain the defect distribution map of each depth layer;

[0078] The clarity or sharpness of the defect images in each depth layer image is analyzed, the images are processed, and the clear defects are identified;

[0079] The images of all depth layers are aligned and superimposed to obtain the complete defect distribution map of the sample to be measured.

[0080] It should be noted that the depth value is not arbitrarily set during the depth layering process along the Z direction, but is calculated according to the depth of field and the tilt angle of the lens. During the scanning process, if there is no depth of field, a defect at a certain depth is only clearly imaged in a certain column of the image. If there is actually a depth of field, the defect is clearly imaged within a small range. Correspondingly, a clear defect image is seen in the image, and only its depth range can be determined, which is the Z-direction layering accuracy, i.e. the set depth value.

[0081] For example, a defect at depth n should be clearly imaged at position j, but due to the depth of field, it is still clear at a position k after a small X-direction step, and position k corresponds to depth m, so from the perspective of the image we cannot determine whether the defect is at depth n or depth m or other depths within the depth of field. In order to avoid this situation, the range can be skipped by selecting a suitable X-direction step distance, and the Z-direction distance corresponding to the X-direction distance is the Z-direction layering accuracy.

[0082] According to the embodiment of the present application, the defect information summary table includes defect serial number, defect X coordinate, Y coordinate, Z coordinate, defect diameter, defect length, defect width and defect type.

[0083] The second aspect of the present application provides a wafer defect detection device applied to a wafer defect detection method, which comprises an optical device, a motion control module and an image fusion module.

[0084] The optical setup includes a large-area camera, a telecentric lens 4, an altimeter, a bright-field infrared light source 2, and a dark-field infrared light source 1. The large-area camera comprises a first camera, a second camera, and a third camera. The third camera is used to scan from a different angle to avoid upper-layer defects obscuring the imaging of lower-layer defects, thus improving the defect detection rate. The second camera acts as a bridge connecting the first and third cameras, ensuring that both cameras observe the same area of ​​the sample.

[0085] A large-scale camera and a telecentric lens 4 are positioned above the sample, and an altimeter is used to measure the height between the product and the large-scale camera.

[0086] Bright-field infrared light source 2 and dark-field infrared light source 1 are positioned below the sample;

[0087] The motion control module includes an XY motion platform and a hollowed-out carrier mounted on the XY motion platform. The XY motion platform is used to control the hollowed-out carrier to move along the X or Y direction.

[0088] The image fusion module includes image synthesis algorithms and image stitching algorithms.

[0089] It should be noted that the dark field infrared light source 1 and the bright field infrared light source 2 are used to project parallel infrared light through the sample to illuminate the defects. After the beam enters the sample surface, it is refracted and reaches the test surface 3. The defect point on the surface emits a telecentric beam parallel to the principal ray and enters the infrared telecentric lens 4, and finally forms an image on the infrared camera target surface 5.

[0090] This invention employs an infrared camera and a high-magnification, shallow-depth-of-field infrared telecentric lens 4. It utilizes an angle-adjustable tilting optical telecentric imaging system and a combined bright and dark field illumination system to simultaneously observe defects on the upper and lower surfaces and inside the wafer. Specific configurations are shown in Table 1.

[0091] Table 1

[0092]

[0093] According to an embodiment of the present invention, the image synthesis algorithm performs sharpness detection on each acquired image. The image synthesis algorithm recovers the three-dimensional information of the object by comparing the images of the sample taken at different X positions of the lens and using a focusing method.

[0094] It should be noted that the size of the samples to be tested is generally larger than the field of view of the camera. Therefore, it is necessary to move the XY motion platform and acquire images to obtain an image set. Then, an image stitching algorithm is used to stitch the images together. Specifically, the images are stitched together using a hard stitching method, and a weighted fusion algorithm is used to process the stitching seam problem.

[0095] Taking 4-wheel line scanning splicing as an example, the following is adopted: Figure 2The S-shaped scanning sequence shown can improve the scanning speed.

[0096] It should be noted that the camera relative to the horizontal plane tilt angle α and the defect Z direction positioning accuracy and scanning speed exist monotonic relationship, from the geometric model α is smaller, the defect Z direction positioning accuracy is higher, but the scanning speed is slower; on the other hand, too small α will significantly reduce the contrast and resolution of defect imaging, and the light is difficult to penetrate the deeper defect, resulting in missed detection. After testing and balancing, the tilt angle α is 30°, the Z direction positioning accuracy can be reduced to 33um, and the scanning time is less than 5min.

[0097] In one embodiment of the present application, the implementation process is as follows:

[0098] Place the chessboard test plate on the hollow loading table, adjust the first camera to focus on the test plate, so that the center of the chessboard pattern is clear in the central field of view; then adjust the second camera to focus on the test plate, and fine-tune it to see the same position as the central field of view of the first camera; then repeat the above operation process for the first camera, second camera and third camera, at this time the first camera and third camera observe the same area in the central field of view. Among them, Figure 5 is a schematic diagram of the optical device of the embodiment.

[0099] Replace the test plate with a sample, keep the relative positions of the two cameras unchanged, focus the first camera to make the upper surface information of the sample clearly imaged on the right side of the image, and the lower surface information clearly imaged on the left side of the image; the third camera is just the opposite. The sample moves uniformly to the left on the carrier, and the relative position is regarded as the camera linearly scanning to the right.

[0100] Among them, Figure 6 is a schematic diagram of defect analysis by double cameras, P2 and Q2 are defects missed by camera one and camera three respectively. When the device scans a certain position of the sample, there may be a situation that the upper defect blocks the imaging of the lower defect for camera one. From the image obtained by camera one at this position, the two defects appear to be superimposed together, and only the blurred image of the upper defect is in the position where the clear image of the lower defect should be. Regardless of how the sample is moved, since the two defects are on the same imaging beam, camera one cannot detect the defect. In the perspective of camera three at another angle, the two defects are distinguished and can be normally scanned and detected. Similarly, the defects that cannot be detected by camera three can be detected by camera one. Camera one and camera three respectively obtain the defect distribution map of the sample by image processing, and then compare the two maps to obtain the defects missed by each other. Finally, the total defect distribution map of the sample without missing is obtained by aligning and superimposing them.

[0101] This method is also applicable to transparent thin glass plates, such as Figure 2As shown, the sample size is 247.6*178.5*2mm, the field of view of the present scheme is 5.76mm*4.83mm, the line scanning is performed in the X direction, and considering that there is 5% field of view overlap in the Y direction during the line changing process, 33 rounds of scanning are required in the Y direction.

[0102] First, the sample is scanned for the first round. Assuming that the sample thickness is 2mm, the refractive index is 1.5, the 4X lens depth of field is 0.06mm, and the camera moves the distance of the depth of field / sinγ in the X direction relative to the sample to take one image, this distance multiplied by the frame rate of the camera is the line scanning speed of the system. In this example, the line scanning speed is 450*0.06 / sin35.2°=46.8mm, and the scanning time is 5.3s.

[0103] Then, the second round of scanning is performed by changing the line. The same time is 5.3s; 33 rounds of scanning are performed, and the total scanning time is 176s. The entire sample is scanned by using this method, and the detection atlas of the sample is obtained. The specific scanning rounds and step intervals change when the lens and camera are replaced, and the calculation method does not change.

[0104] The image synthesis algorithm divides and splices the images obtained by line scanning to obtain the defect distribution map of each layer. The clarity of each layer image is analyzed, and the defects in each depth layer are processed and recognized. Then, the processed defect distribution map of each layer is aligned and combined to obtain the defect distribution map of the entire product. Among them, Figure 4 is the flowchart of the image processing algorithm of the present embodiment.

[0105] The technical features of the above embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above embodiments are not described, but as long as the combinations of the technical features do not contradict, they should be considered as the scope of the present disclosure.

[0106] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the above embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application should not be limited to the above-described embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features disclosed herein.

[0107] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A wafer defect detection method, characterized by, Comprising the following steps: Step S1: Place the sample on the hollow carrier and adjust the angle of the hollow carrier to keep the sample horizontal; Step S2: Adjust the height and angle of the telecentric lens according to the thickness of the sample, focus the telecentric lens, and center the imaging area; Step S3: Place three cameras above the sample, calibrate the positions of adjacent cameras to make the fields of view coincide; Step S4: Adjust the sample position to the initial measurement point based on the XY platform, the XY motion platform moves the sample according to the preset motion mode, the camera scans the sample, and the image is collected synchronously to obtain the original defect image set; Step S5: Based on the image processing algorithm, each image in the original defect image set is cut layer by layer, then corrected and spliced in the X direction, and denoising and binarization processing is performed to obtain the defect distribution map of each layer. Finally, the complete defect distribution map of the XY plane is synthesized, and the defect information summary table is output; Step S4 specifically comprises: Place the sample to be measured on the hollow carrier and move the sample by the XY motion platform; The quotient of the sine of the angle γ between the depth of field L of the telecentric lens and the relative horizontal plane is the interval of the X-axis single step, and then multiplied by the frame rate of the camera and the sine of the angle α at which the first camera is placed to obtain the horizontal line scanning speed; The XY motion platform moves horizontally to the left, and 3D information images of each region of the sample are collected; After X-direction scanning, the Y-direction is changed, and the X-direction is scanned again along the horizontal line scanning speed, and so on, until the entire sample is scanned, and the original defect image set is obtained; The method for obtaining the complete sample defect distribution map in step S5 is as follows: The XY motion platform moves in a preset manner while the camera collects images, and the sample is pre-layered according to the Z-direction layering accuracy and the sample thickness; According to the layering situation, the image is segmented, each part of the segmented image corresponds to a depth layer, and the corresponding parts of all images scanned at different XY positions are spliced to synthesize the defect distribution map of each depth layer; The clarity or sharpness of the defects in each depth layer image is analyzed, the image is processed, and the clear defects are identified; Align and superimpose all depth layer images to obtain the complete defect distribution map of the sample to be measured.

2. The method of claim 1, wherein The focusing process in step S2 comprises: Adjust the height and angle of the telecentric lens to center the imaging area, and the upper surface features are clearly imaged on the right side of the imaging area, the internal features are clearly imaged in the middle of the imaging area, and the lower surface features are clearly imaged on the left side of the imaging area.

3. The wafer defect detection method according to claim 2, characterized in that, The calibration method in step S3 is as follows: The three cameras are respectively referred to as the first camera, the second camera and the third camera, the chessboard test board is placed on the hollow stage, the first camera is adjusted to focus on the test board, and the center of the chessboard pattern is clearly displayed in the central field of view; Adjust the second camera to focus on the test board, and fine-tune the second camera to display the same position as the central field of view of the first camera; Repeat the focusing of the second camera and the third camera according to the focusing method of the first camera and the second camera, and fine-tune the first camera to display the same position as the central field of view of the third camera.

4. The method of claim 1, wherein The defect information summary table includes defect serial number, defect X coordinate, Y coordinate, Z coordinate, defect diameter, defect length, defect width and defect type.

5. A wafer defect detection apparatus applied to the wafer defect detection method of any one of claims 1-4, comprising: The optical device, the motion control module and the image fusion module are characterized in that The optical device comprises a large target surface camera, a telecentric lens, an altimeter, a bright field infrared light source and a dark field infrared light source. The large target surface camera and the telecentric lens are arranged above the sample, and the altimeter is used for measuring the height between the product and the large target surface camera. The bright field infrared light source and the dark field infrared light source are arranged below the sample. The motion control module comprises an XY motion platform and a hollow carrier arranged on the XY motion platform, and the XY motion platform is used for controlling the hollow carrier to move in the X direction or the Y direction. The image fusion module comprises an image synthesis algorithm and an image stitching algorithm.

6. The wafer defect detection apparatus according to claim 5, wherein The image synthesis algorithm detects the definition of each collected image, and the image synthesis algorithm restores the three-dimensional information of the object by comparing the images of the sample taken at different X positions of the lens by using the focusing method.

Images