Method and apparatus for detecting defects in transparent substrates

Abstract

A method and apparatus for detecting defects in transparent substrates are disclosed. The method includes the following steps: Step 1: Automatically focusing the surface of the transparent substrate using a laser confocal microscope to determine the reference zero point in the Z direction; Step 2: Performing a planar scan at the focal plane of the surface to acquire surface defect information of the transparent substrate; Step 3: Driving the transparent substrate upward along the Z direction by a PZT driver to move it a preset distance, and after it reaches the desired position, performing a planar scan to acquire the reflection image of the optical slice layer, obtaining the three-dimensional coordinates and light intensity information of all detection points; Step 4: Determining whether the moving distance of the transparent substrate has reached a preset depth threshold. If not, repeating Step 3; if it has, performing image synthesis processing on all detected data to generate a three-dimensional mapping of the defects and acquire the defect information inside the transparent substrate. The above-described method for detecting defects in transparent substrates has the beneficial effect of being able to detect surface and internal defects of transparent substrates non-destructively and accurately.

Description

Technical Field

[0001] This invention relates to the field of integrated circuit technology, and in particular to a method and apparatus for detecting defects in transparent substrates. Background Technology

[0002] Quartz (Qz) substrates are widely used in semiconductor photomasks, precision optical windows, and laser devices due to their excellent optical transmittance, thermal stability, and chemical inertness. Defects in quartz substrates (including surface scratches, particles, and internal inclusions) can directly affect the quality of subsequent patterning processes and induce laser-induced damage under high-energy laser irradiation, reducing device reliability and lifespan.

[0003] Currently, defect detection of quartz substrates mainly relies on laser confocal optical systems. These systems identify defects by detecting reflected signals from the substrate surface, achieving high-contrast, high-resolution imaging and accurately identifying surface defects. However, while laser confocal optical systems offer high axial resolution, their extremely shallow depth of focus makes it difficult to directly cover the full depth of transparent substrates several millimeters thick. This limits their ability to detect surface defects, preventing the detection of inclusions embedded within the substrate (such as bubbles, impurity particles, and crystallization points). Existing technologies for assessing internal defects in quartz substrates often require destructive sample preparation (such as cutting and grinding), making them unsuitable for inspecting finished or semi-finished products.

[0004] Therefore, there is an urgent need for a method for detecting defects in transparent quartz substrates that can non-destructively detect internal defects. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for non-destructively detecting surface and internal defects of transparent substrates.

[0006] To achieve the above objectives, the present invention provides a method for detecting defects in a transparent substrate, comprising the following steps: Step 1: Use a laser confocal microscope to automatically focus on the surface of the transparent substrate and determine the reference zero point in the Z direction; Step 2: Perform a planar scan at the focal plane of the surface to obtain information on surface defects of the transparent substrate; Step 3: Drive the transparent substrate to move upward along the Z direction by a preset distance using a PZT driver. After it moves into place, perform a planar scan to acquire the reflection image of the optical slice layer and obtain the three-dimensional coordinates and light intensity information of all detection points. Step 4: Determine whether the moving distance of the transparent substrate has reached the preset depth threshold. If the moving distance of the transparent substrate has not reached the preset depth threshold, repeat step 3. Step 5: If the moving distance of the transparent substrate reaches the preset depth threshold, then perform image synthesis processing on all detected data to generate a three-dimensional mapping of the defect and obtain the defect information inside the transparent substrate.

[0007] The aforementioned transparent substrate defect detection method first uses a confocal microscope to detect surface defects on the transparent substrate. Then, a PZT driver drives the substrate to move upwards along the Z-direction to detect internal defects. As the transparent substrate moves upwards, it gradually approaches the objective lens of the confocal microscope, causing the focal plane of the laser confocal microscope to gradually move downwards to optical slice layers at different depths within the transparent substrate. After each upward movement, the confocal microscope performs a planar scan of the transparent substrate to acquire the reflection image of the current optical slice layer, obtaining the three-dimensional coordinates and light intensity information of all detection points. Further, after each movement, it is determined whether the moving distance of the transparent substrate reaches a preset depth threshold to determine if the target detection depth has been reached. If the moving distance does not reach the preset depth threshold, the target detection depth has not been reached, and the movement of the transparent substrate and the planar scan operation continue. If the moving distance reaches the preset depth threshold, the target detection depth has been reached, the movement of the transparent substrate stops, and all detected data undergoes image synthesis processing to generate a three-dimensional mapping of the defects, obtaining defect information within the transparent substrate within the target detection depth, thus completing the detection of internal defects within the transparent substrate within the target detection depth.

[0008] The aforementioned method for detecting defects in transparent substrates utilizes the high axial resolution of a confocal microscope combined with PZT depth scanning to extend the detection capabilities of the confocal microscope from the surface to the interior of the transparent substrate, achieving non-destructive detection of internal defects. This method can detect defects on both the surface and interior of the transparent substrate, making it suitable for online inspection during production or finished product inspection, with a wide range of applications. Furthermore, due to the strong penetrating power of the laser and the use of layer-by-layer slicing, overlapping inclusions at different depths can be independently imaged and distinguished, enabling accurate detection of overlapping defects and improving detection accuracy. Further, the method employs an image integration algorithm to generate a three-dimensional defect map, achieving precise quantification of defects and further enhancing detection accuracy.

[0009] In summary, the above-mentioned method for detecting defects in transparent substrates has the advantages of being able to non-destructively detect both surface and internal defects in transparent substrates with high accuracy.

[0010] In one embodiment, the step of performing image synthesis processing on all detected data to generate a three-dimensional map of the defects and obtaining defect information inside the transparent substrate includes the following steps: Construct a pixel window centered on the pixel to be analyzed, and calculate the background mean and standard deviation of the pixel window region. Set the dynamic light intensity threshold according to Formula 1: T = μ + k·σ; [Formula 1] Where T is the dynamic light intensity threshold, μ is the background mean, σ is the standard deviation, and k is an empirical constant with a value range of 3 to 5. The light intensity value of the pixel to be analyzed is compared with the dynamic light intensity threshold. If the light intensity value of the pixel to be analyzed is greater than or equal to the dynamic light intensity threshold, the pixel to be analyzed is marked as a defect point. Check if there are any previously marked defect points within the preset three-dimensional neighborhood of the current defect point. If so, merge the current defect point with the previously marked defect points within its preset three-dimensional neighborhood to form a defect entity.

[0011] In one embodiment, after merging the current defect point with other marked defect points in its preset three-dimensional neighborhood to form a defect entity, the following steps are also included: Extract the outer boundary of the defective entity, calculate the span of the defective entity in the X, Y and Z directions respectively, and calculate the three-dimensional dimensions of the defective entity.

[0012] In one embodiment, after merging the current defect point with other marked defect points in its preset three-dimensional neighborhood to form a defect entity, the following steps are also included: Obtain the three-dimensional coordinates of adjacent defective entities and calculate the three-dimensional Euclidean distance between adjacent defective entities.

[0013] In one embodiment, before the step of constructing a pixel window centered on the pixel to be analyzed, the following steps are included: Each optical slice layer is filtered to remove shot noise.

[0014] In one embodiment, the step of constructing a pixel window centered on the pixel to be analyzed includes: Centered on the pixel to be analyzed, construct a pixel window with N×N points, where N≥3.

[0015] In one embodiment, the preset distance is no greater than the depth of focus of the confocal microscope.

[0016] In one embodiment, the preset depth threshold is not less than 100 μm.

[0017] In one embodiment, the light source wavelength of the laser confocal microscope is ~530nm.

[0018] On the other hand, the present invention also provides a transparent substrate defect detection device, comprising: a laser confocal microscope, a PZT driver and a processor, wherein the PZT driver is connected to the stage of the confocal microscope and is used to drive the stage to move the transparent substrate supported thereon in a stepping motion along the Z direction; the confocal microscope is used to acquire surface reflection images of the transparent substrate and reflection images of the internal optical slice layer of the transparent substrate respectively; and the processor is used to acquire surface defect information and internal defect information of the transparent substrate based on the surface reflection images and the reflection images of the internal optical slice layer of the transparent substrate respectively.

[0019] The aforementioned transparent substrate defect detection device is used to implement the aforementioned transparent substrate detection method, and has the beneficial effects of being able to non-destructively detect surface defects and internal defects of transparent substrates with high detection accuracy. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the principle of a transparent substrate defect detection method according to one embodiment. Detailed Implementation

[0021] The present invention will be described more clearly and completely by way of embodiments and in conjunction with the accompanying drawings, but the present invention is not limited to the scope of the following embodiments.

[0022] This invention provides a method and apparatus for detecting defects in transparent substrates. This method and apparatus can accurately and non-destructively detect surface and internal defects in transparent substrates, achieving non-destructive detection of internal defects. It is suitable for online inspection during production or finished product inspection, and has a wide range of applications. This method and apparatus can be used to detect various transparent substrates, and the transparent substrates can be any shape, such as square or circular. For ease of explanation, the following embodiments use glass substrates as an example, specifically quartz glass substrates (quartz substrates) used as photomasks in the semiconductor field.

[0023] Please see Figure 1 One embodiment of a transparent substrate defect detection method includes the following steps: Step S11: Use a laser confocal microscope to automatically focus on the surface of the transparent substrate and determine the reference zero point in the Z direction.

[0024] Specifically, the upper surface of the quartz substrate is automatically focused using a laser confocal microscope to obtain a reference Z=0 plane and determine the reference zero point in the Z direction.

[0025] Step S12: Perform a planar scan at the focal plane of the surface to obtain surface defect information of the transparent substrate.

[0026] Specifically, the XY plane is scanned at the focal plane of the upper surface of the quartz substrate to collect the reflection signal of the surface defect on the upper surface of the quartz substrate, obtain the surface defect information of the transparent substrate, and record the position and light intensity information of the surface defect.

[0027] Step S13: Drive the transparent substrate to move upward along the Z direction by a preset distance using a PZT driver. After moving into position, perform a planar scan to acquire the reflection image of the optical slice layer and obtain the three-dimensional coordinates and light intensity information of all detection points.

[0028] Specifically, the stage of the confocal microscope is driven by a PZT driver (piezoelectric ceramic driver) to move the transparent substrate upwards along the Z direction by a preset distance, so that the transparent substrate is close to the objective lens of the laser confocal microscope. Then, the focal plane is switched to the optical slice layer inside the transparent substrate. After the transparent substrate is moved into place, the confocal microscope performs a complete XY plane scan of the transparent substrate at the current focal plane position. Due to the transparency of the quartz substrate, the laser of the laser confocal microscope is transmitted through the quartz substrate. If there are inclusions (such as bubbles, impurities, crystal points, etc.) inside the quartz substrate, the inclusions and the quartz material have different refractive indices, forming a reflection interface that can reflect the incident laser back to the confocal optical path. Thus, the confocal microscope can acquire the reflection image of the current optical slice layer by performing a plane scan of the transparent substrate at the current focal plane position, and obtain the three-dimensional coordinates and light intensity information of all detection points.

[0029] In one embodiment, to ensure that the laser from the laser confocal microscope can effectively penetrate a 6.35 mm thick quartz substrate, the wavelength of the light source of the laser confocal microscope is preferably ~530 nm. Furthermore, to ensure full coverage of the internal structure of the transparent substrate and avoid missed detections, the preset distance is preferably no greater than the depth of focus of the confocal microscope. For example, in this embodiment, the light source wavelength is ~530 nm, and the depth of focus adjustment length of the laser confocal microscope is approximately 0.24 μm to 5 μm; preferably, the preset distance is no greater than 0.24 μm.

[0030] Step S14: Determine whether the moving distance of the transparent substrate has reached the preset depth threshold. If the moving distance of the transparent substrate has not reached the preset depth threshold, repeat step S13.

[0031] Specifically, to ensure the target detection depth is reached, after the transparent substrate moves upward to complete the planar scan, the moving distance of the transparent substrate is compared with a preset depth threshold (i.e., the target detection depth). If the moving distance of the transparent substrate is less than the preset depth threshold, the target detection depth has not been reached, and step S13 is repeated. Each time the transparent substrate moves upward, a complete XY plane scan is performed until the total moving distance of the transparent substrate reaches the target detection depth. Then, the movement of the transparent substrate is stopped, and step S15 is executed. Further, the preset depth threshold can be arbitrarily set according to actual detection needs. It can be a depth range covering the entire thickness of the board or a partial depth detection. This embodiment does not impose a specific limitation. Preferably, to ensure effective detection of subsurface defects of the transparent substrate, in one embodiment, the preset depth threshold is preferably not less than 100 μm.

[0032] Step S15: If the moving distance of the transparent substrate reaches the preset depth threshold, then perform image synthesis processing on all detected data to generate a three-dimensional mapping of the defect and obtain the defect information inside the transparent substrate.

[0033] Specifically, if the moving distance of the transparent substrate is greater than or equal to the preset depth threshold, the target detection depth is reached. At this point, all detected data are processed by image synthesis to generate a three-dimensional mapping of the defect, obtain the defect information inside the transparent substrate, and complete the internal defect detection of the transparent substrate.

[0034] In one embodiment, the step of performing image synthesis processing on all detected data to generate a three-dimensional map of the defects and obtain defect information inside the transparent substrate includes the following steps: Step S141: Construct a pixel window centered on the pixel to be analyzed, and calculate the background mean and standard deviation of the pixel window region. Specifically, construct a pixel window with N×N points centered on the pixel to be analyzed, where N≥3.

[0035] Step S142: Set the dynamic light intensity threshold according to Formula 1: T = μ + k·σ; [Formula 1] Where T is the dynamic light intensity threshold, μ is the background mean, σ is the standard deviation, and k is an empirical constant with a value range of 3 to 5.

[0036] Step S143: Compare the light intensity value of the pixel to be analyzed with the dynamic light intensity threshold. If the light intensity value of the pixel to be analyzed is greater than or equal to the dynamic light intensity threshold, then mark the pixel to be analyzed as a defect point.

[0037] Specifically, after calculating the dynamic threshold, the light intensity value of the pixel to be analyzed is compared with the dynamic light intensity threshold. If the light intensity value of the pixel to be analyzed is greater than or equal to the dynamic light intensity threshold, the pixel to be analyzed is marked as a defect point, and the original signal information of the pixel to be analyzed is retained. If the light intensity value of the pixel to be analyzed is less than the dynamic light intensity threshold, the pixel to be analyzed is a defect-free background, and the signal of the pixel is set to 0. This embodiment adaptively calculates the dynamic light intensity threshold to address the background light intensity non-uniformity existing at different depths and in different regions of the quartz substrate, thereby adaptively separating the real defect signal, while effectively suppressing stray light interference, ensuring that the data source for subsequent three-dimensional defect merging has high fidelity, and avoiding missed detections and false detections under the conditions of depth attenuation and local light intensity non-uniformity.

[0038] Step S144: Check if there are any marked defect points in the preset three-dimensional neighborhood of the current defect point. If so, merge the current defect point with the marked defect points in its preset three-dimensional neighborhood to form a defect entity.

[0039] Specifically, after marking the pixel to be analyzed as a defect point, it is detected whether there are any previously marked defect points within the preset three-dimensional neighborhood of the current defect point. If other defect points exist near a defect point within the three-dimensional neighborhood, they are merged with the current defect point to form a complete defect entity. Preferably, the three-dimensional neighborhood range can be configured as "6-neighborhood, 18-neighborhood, or 26-neighborhood", but this embodiment does not impose a specific limitation.

[0040] In one embodiment, after merging the current defect point with its pre-defined three-dimensional neighborhood of marked defect points to form a defect entity, the method further includes the following steps: extracting the outer boundary of the defect entity, calculating the span of the defect entity in the X, Y, and Z directions, and calculating the three-dimensional dimensions of the defect entity. Specifically, after merging the defect entities, the outer boundary of the merged defect region is extracted, the span of the defect entity in the X, Y, and Z directions is calculated, and the dimensional differences in the three directions are recorded as the three-dimensional dimensional features of the defect, thereby achieving accurate quantification of the defect dimensions.

[0041] In one embodiment, after merging the current defect point with its pre-defined three-dimensional neighboring marked defect points to form a defect entity, the method further includes the following steps: obtaining the three-dimensional coordinates of adjacent defect entities and calculating the three-dimensional Euclidean distance between adjacent defect entities. This embodiment calculates the three-dimensional Euclidean distance between any two defects based on the known three-dimensional coordinates of defect entities, providing accurate defect spacing data, which helps assess the interaction risk between defects and ensures stable and reliable product performance.

[0042] Furthermore, in one embodiment, after completing the above-mentioned defect detection and calculation, the system outputs a three-dimensional distribution map of inclusion defects on and inside the quartz substrate, the X / Y / Z three-dimensional dimensions of each defect, volume estimates, and the three-dimensional Euclidean distance between adjacent defects, providing a quantitative characterization of defect shape, size, volume, distribution, and spacing between overlapping defects.

[0043] Furthermore, in one embodiment, before the step of constructing a pixel window centered on the pixel to be analyzed, the following step is included: filtering each optical slice layer to remove shot noise. Specifically, for each Z-layer optical slice layer image, a median filter is applied to remove shot noise introduced by the detection system. This embodiment pre-processes each optical slice layer with median filtering to remove background noise, ensuring that the data source for image synthesis processing has a high signal-to-noise ratio and improving the accuracy of defect detection.

[0044] The aforementioned transparent substrate defect detection method first uses a confocal microscope to detect surface defects on the transparent substrate. Then, a PZT driver drives the substrate to move upwards along the Z-direction to detect internal defects. As the transparent substrate moves upwards, it gradually approaches the objective lens of the confocal microscope, causing the focal plane of the laser confocal microscope to gradually move downwards to optical slice layers at different depths within the transparent substrate. After each upward movement, the confocal microscope performs a planar scan of the transparent substrate to acquire the reflection image of the current optical slice layer, obtaining the three-dimensional coordinates and light intensity information of all detection points. Further, after each movement, it is determined whether the moving distance of the transparent substrate reaches a preset depth threshold to determine if the target detection depth has been reached. If the moving distance does not reach the preset depth threshold, the target detection depth has not been reached, and the movement of the transparent substrate and the planar scan operation continue. If the moving distance reaches the preset depth threshold, the target detection depth has been reached, the movement of the transparent substrate stops, and all detected data undergoes image synthesis processing to generate a three-dimensional mapping of the defects, obtaining defect information within the transparent substrate within the target detection depth, thus completing the detection of internal defects within the transparent substrate within the target detection depth.

[0045] The aforementioned method for detecting defects in transparent substrates utilizes the high axial resolution of a confocal microscope combined with PZT depth scanning to extend the detection capabilities of the confocal microscope from the surface to the interior of the transparent substrate, achieving non-destructive detection of internal defects. This method can detect defects on both the surface and interior of the transparent substrate, making it suitable for online inspection during production or finished product inspection, with a wide range of applications. Furthermore, due to the strong penetrating power of the laser and the use of layer-by-layer slicing, overlapping inclusions at different depths can be independently imaged and distinguished, enabling accurate detection of overlapping defects and improving detection accuracy. Further, the method employs an image integration algorithm to generate a three-dimensional defect map, achieving precise quantification of defects and further enhancing detection accuracy.

[0046] In summary, the above-mentioned method for detecting defects in transparent substrates has the advantages of being able to non-destructively detect both surface and internal defects in transparent substrates with high accuracy.

[0047] On the other hand, the present invention also provides a transparent substrate defect detection device, comprising: a laser confocal microscope, a PZT driver and a processor, wherein the PZT driver is connected to the stage of the confocal microscope and is used to drive the stage to move the transparent substrate supported thereon in a stepping motion along the Z direction; the confocal microscope is used to acquire surface reflection images of the transparent substrate and reflection images of the internal optical slice layer of the transparent substrate respectively; and the processor is used to acquire surface defect information and internal defect information of the transparent substrate based on the surface reflection images and the reflection images of the internal optical slice layer of the transparent substrate respectively.

[0048] Specifically, the transparent substrate defect detection device of this embodiment first performs surface defect detection on the transparent substrate using a confocal microscope to obtain surface defect information. Then, a PZT driver drives the stage to move the transparent substrate upwards along the Z-direction to acquire reflection images of different optical slice layers inside the transparent substrate. Once the transparent substrate has moved to the target detection depth and all optical slice images within the target depth range have been acquired, the processor performs image synthesis processing on the acquired reflection images of the optical slice layers to obtain internal defect information of the transparent substrate. The detailed execution steps of the transparent substrate defect detection device of this embodiment are the same as those of the transparent substrate defect detection method described above, and will not be repeated here. Further, after defect detection, the PZT driver drives the stage to reset the transparent substrate along the Z-direction so that the transparent substrate can be removed. This embodiment uses a PZT driver to drive the stage to move the transparent substrate stepwise along the Z-direction. The PZT driver has high displacement accuracy and fast response speed, ensuring the displacement accuracy of the transparent substrate, effectively avoiding missed detections, and improving detection accuracy and efficiency.

[0049] In one embodiment, to ensure effective penetration of the laser from the laser confocal microscope into a 6.35 mm thick quartz substrate, the wavelength of the laser confocal microscope's light source is preferably ~530 nm. Further, in one embodiment, to ensure full coverage of the transparent substrate and avoid missed detections, the movement step size of the PZT driver is preferably no greater than the depth of focus of the confocal microscope. For example, in this embodiment, the light source wavelength is ~530 nm, the depth of focus adjustment length of the laser confocal microscope is approximately 0.24 μm to 5 μm, and the movement step size of the PZT driver is preferably no greater than 0.24 μm.

[0050] The transparent substrate defect detection device of this embodiment is used to implement the above-described transparent substrate detection method, and has the beneficial effects of being able to non-destructively detect surface defects and internal defects of transparent substrates with high detection accuracy.

[0051] The above description is merely a description of preferred embodiments of the present invention and is not intended to limit the scope of the present invention in any way. Any changes or modifications made by those skilled in the art based on the above disclosure shall fall within the scope of the present invention.

Claims

1. A method for detecting defects in a transparent substrate, characterized in that, Includes the following steps: Step 1: Use a laser confocal microscope to automatically focus on the surface of the transparent substrate and determine the reference zero point in the Z direction; Step 2: Perform a planar scan at the focal plane of the surface to obtain information on surface defects of the transparent substrate; Step 3: Drive the transparent substrate to move upward along the Z direction by a preset distance using a PZT driver. After it moves into place, perform a planar scan to acquire the reflection image of the optical slice layer and obtain the three-dimensional coordinates and light intensity information of all detection points. Step 4: Determine whether the moving distance of the transparent substrate has reached the preset depth threshold. If the moving distance of the transparent substrate has not reached the preset depth threshold, repeat step 3. Step 5: If the moving distance of the transparent substrate reaches the preset depth threshold, then perform image synthesis processing on all detected data to generate a three-dimensional mapping of the defect and obtain the defect information inside the transparent substrate.

2. The method for detecting defects in a transparent substrate according to claim 1, characterized in that, The step of performing image synthesis processing on all detected data to generate a three-dimensional map of defects and obtain defect information inside the transparent substrate includes the following steps: A pixel window is constructed with the pixel to be analyzed as the center, and the background mean and standard deviation of the pixel window region are calculated. Set the dynamic light intensity threshold according to Formula 1: T = μ + k·σ; [Formula 1] Where T is the dynamic light intensity threshold, μ is the background mean, σ is the standard deviation, and k is an empirical constant with a value range of 3 to 5; The light intensity value of the pixel to be analyzed is compared with the dynamic light intensity threshold. If the light intensity value of the pixel to be analyzed is greater than or equal to the dynamic light intensity threshold, the pixel to be analyzed is marked as a defect point. Check if there are any previously marked defect points within the preset three-dimensional neighborhood of the current defect point. If so, merge the current defect point with the previously marked defect points within its preset three-dimensional neighborhood to form a defect entity.

3. The method for detecting defects in a transparent substrate according to claim 2, characterized in that, After the step of merging the current defect point with the previously marked defect points in its preset three-dimensional neighborhood to form a defect entity, the following steps are also included: Extract the outer boundary of the defective entity, calculate the span of the defective entity in the X, Y and Z directions respectively, and calculate the three-dimensional dimensions of the defective entity.

4. The transparent substrate defect detection device according to claim 2, characterized in that, After the step of merging the current defect point with the previously marked defect points in its preset three-dimensional neighborhood to form a defect entity, the following steps are also included: Obtain the three-dimensional coordinates of adjacent defective entities and calculate the three-dimensional Euclidean distance between adjacent defective entities.

5. The method for detecting defects in a transparent substrate according to claim 2, characterized in that, Before the step of constructing a pixel window centered on the pixel to be analyzed, the following steps are also included: Each optical slice layer is filtered to remove shot noise.

6. The method for detecting defects in a transparent substrate according to claim 2, characterized in that, The step of constructing a pixel window centered on the pixel to be analyzed includes: Centered on the pixel to be analyzed, construct a pixel window with N×N points, where N≥3.

7. The method for detecting defects in a transparent substrate according to claim 1, characterized in that, The preset distance is not greater than the depth of focus of the confocal microscope.

8. The method for detecting defects in a transparent substrate according to claim 1, characterized in that, The preset depth threshold is not less than 100μm.

9. The method for detecting defects in a transparent substrate according to claim 1, characterized in that, The light source wavelength of the laser confocal microscope is ~530nm.

10. A device for detecting defects in a transparent substrate, characterized in that, include: The system includes a laser confocal microscope, a PZT driver, and a processor. The PZT driver is connected to the stage of the confocal microscope and is used to drive the stage to move a transparent substrate mounted on it in a stepping motion along the Z direction. The confocal microscope is used to acquire surface reflection images of the transparent substrate and reflection images of the internal optical slice layer of the transparent substrate, respectively. The processor is used to acquire surface defect information and internal defect information of the transparent substrate based on the surface reflection images and reflection images of the internal optical slice layer of the transparent substrate, respectively.

Images