Quality detection method and parameter adjustment method based on microlens array exposure process

Abstract

The present application relates to the technical field of exposure, in particular to a quality detection method and a parameter adjustment method based on a microlens array (MLA) exposure process. The quality detection method comprises the following steps: establishing an exposure pattern detection model capable of identifying the category of each exposure point pattern; the pattern category includes qualified patterns and unqualified patterns; setting initial exposure parameters and obtaining an exposure image obtained through the MLA exposure process; inputting the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category of each exposure point; determining the pattern category of the exposure point according to the data information set and counting the number of exposure points of the qualified patterns and the unqualified patterns in the exposure image to obtain the proportion of the number of exposure points of the qualified patterns to the total number of exposure points. Through the above setting, the qualified patterns and the unqualified patterns can be automatically identified, so that the qualified rate of the MLA exposure quality can be automatically obtained, and the quality detection precision and speed are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of exposure technology, and in particular to a quality inspection method and parameter adjustment method based on microlens array exposure process. Background Technology

[0002] Photolithography is a key process technology in micro and nano fabrication. Exposure is one of the most critical steps, directly determining the quality of the lithography. In recent years, maskless lithography has been widely used due to its low cost. Within the crucial exposure stage of this technology, microlens array (MLA) exposure technology, as a novel array-based maskless exposure technique, can simultaneously expose large-area periodic patterns, and has broad application prospects in the future.

[0003] In the microlens array exposure process, the exposure quality pass rate (i.e., the proportion of qualified exposure points out of all exposure points) is a crucial metric for evaluating the effectiveness of the exposure process, and it directly impacts the yield of the photolithography process. Currently, the exposure quality of microlens array exposure processes is primarily assessed by the human eye using a microscope to preliminarily determine the pass rate, which suffers from low automation, poor accuracy, and slow speed. Summary of the Invention

[0004] To address the shortcomings of existing technologies in efficiently detecting the exposure quality of microlens array exposure processes, this invention provides a quality detection method and parameter adjustment method based on microlens array exposure processes, as detailed below:

[0005] In a first aspect, embodiments of the present invention provide a quality inspection method based on a microlens array exposure process, comprising the following steps:

[0006] An exposure pattern detection model is established based on a convolutional neural network to identify the pattern category of each exposure point; the pattern category includes qualified patterns and unqualified patterns.

[0007] Set initial exposure parameters and obtain the exposure image of the exposure point to be detected by the MLA exposure process; input the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point;

[0008] The pattern category of the exposure point is determined based on the data information set, and the number of exposure points of qualified and unqualified patterns in the exposure image is counted to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points.

[0009] In another embodiment of the first aspect, establishing an exposure pattern model includes the following steps:

[0010] A set of images of exposure points is obtained, and the exposure patterns in the image set are labeled with the pattern category to obtain an exposure pattern dataset; the exposure pattern dataset is divided into a training sample dataset and a test sample dataset;

[0011] A target neural network model is established, and the training sample dataset is input into the target neural network model for training to generate an exposure pattern detection model. Then, the test sample dataset is input into the exposure pattern detection model for verification.

[0012] In another embodiment of the first aspect, the target neural network model is one of the target detection models YOLO, SSD, and Faster-RCNN.

[0013] In another embodiment of the first aspect, the data information set includes at least the pattern category, coordinate position, and exposure point size information corresponding to each target exposure point; the data information set is represented by vector representation data (S, X, Y, W, H), where S represents the pattern category, X and Y represent the position of the center of the target exposure point relative to the image, and W and H represent the actual size of the target exposure point.

[0014] In another embodiment of the first aspect, determining the pattern category of the exposure point based on the data information set includes the following steps:

[0015] Set a preset qualified pattern and a corresponding preset size threshold range, and compare the preset size threshold range with the actual size of the target exposure point to determine the pattern category of the target exposure point;

[0016] If the actual size of the target exposure point is within the preset size threshold range, then the target exposure point is a qualified pattern; otherwise, the target exposure point is an unqualified pattern.

[0017] In another embodiment of the first aspect, the preset qualified pattern is set as an axisymmetric annular pattern, and the difference between the inner and outer ring distances of the axisymmetric annular pattern is defined as the preset size threshold range;

[0018] The target exposure point of the defective pattern is determined; if the actual size of the target exposure point of the defective pattern is larger than the preset size threshold range, the target exposure point is determined to be an overexposed pattern; otherwise, the target exposure point is determined to be an underexposed pattern.

[0019] Secondly, embodiments of the present invention provide a parameter adjustment method based on a microlens array exposure process, employing the quality inspection method based on a microlens array exposure process as described in any of the embodiments of the first aspect above, and further including the following steps:

[0020] If the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is greater than or equal to the preset value, then the initial exposure parameters will not be changed.

[0021] If the proportion of the number of exposure points of a qualified pattern to the total number of exposure points is less than the preset value, the initial exposure parameters are modified.

[0022] In another embodiment of the second aspect, in the defective pattern, if the number of exposure points of the overexposed pattern is greater than the number of exposure points of the underexposed pattern, the initial exposure parameters are modified by reducing the laser power or increasing the platform moving speed until the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is greater than or equal to the preset value.

[0023] If the number of exposure points in an overexposed pattern is less than the number of exposure points in an underexposed pattern, the initial exposure parameters are modified by increasing the laser power or decreasing the platform movement speed until the proportion of the number of exposure points in a qualified pattern to the total number of exposure points is greater than or equal to the preset value.

[0024] Thirdly, embodiments of the present invention also provide a quality inspection device based on a microlens array exposure process, comprising:

[0025] Model building module: Based on a convolutional neural network, an exposure pattern detection model is built that can identify the pattern category of each exposure point; the pattern categories include qualified patterns and unqualified patterns;

[0026] The quality inspection module sets initial exposure parameters and acquires the exposure image of the exposure point to be inspected obtained through MLA exposure process; it inputs the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point.

[0027] Pattern statistics module: Based on the data information set, determine the pattern category of the exposure point, and count the number of exposure points of qualified and unqualified patterns in the exposed image to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points.

[0028] Fourthly, embodiments of the present invention also provide a parameter adjustment device based on a microlens array exposure process, comprising:

[0029] Model building module: Based on a convolutional neural network, an exposure pattern detection model is built that can identify the pattern category of each exposure point; the pattern categories include qualified patterns and unqualified patterns;

[0030] The quality inspection module sets initial exposure parameters and acquires the exposure image of the exposure point to be inspected obtained through MLA exposure process; it inputs the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point.

[0031] Pattern statistics module: Based on the data information set, determine the pattern category of the exposure point, and count the number of exposure points of qualified and unqualified patterns in the exposed image to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points;

[0032] The parameter modification module: In the case of unqualified patterns, if the number of exposure points in the overexposed pattern is greater than the number of exposure points in the underexposed pattern, the initial exposure parameters are modified by reducing the laser power or increasing the platform movement speed until the proportion of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to the preset value; if the number of exposure points in the overexposed pattern is less than the number of exposure points in the underexposed pattern, the initial exposure parameters are modified by increasing the laser power or decreasing the platform movement speed until the proportion of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to the preset value.

[0033] Based on the above, compared with the prior art, the quality inspection method based on microlens array exposure process provided by the present invention uses a neural network target detection model to input the exposure image into the exposure pattern model to automatically identify qualified and unqualified patterns, thereby automatically obtaining the pass rate of MLA exposure quality and effectively improving the accuracy and speed of quality inspection.

[0034] Other features and beneficial effects of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other beneficial effects of the invention can be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Unless otherwise specified, the positional relationships shown in the drawings in the following description are based on the direction in which the components are drawn in the figure.

[0036] Figure 1 A flowchart of the quality inspection method based on microlens array exposure process provided by the present invention;

[0037] Figure 2 for Figure 1 Flowchart of the specific steps in step S10;

[0038] Figure 3 A schematic diagram of the network structure of the optimized Yolov5 model;

[0039] Figure 4 Flowchart for image training of the optimized Yolov5 model;

[0040] Figure 5 A schematic diagram of the structure of a pre-defined qualified pattern;

[0041] Figure 6 This is a schematic diagram of the pattern at each exposure point under the recognition state of the exposure pattern detection model.

[0042] Figure 7 This is a schematic diagram illustrating the detection performance of the exposure pattern detection model trained using the optimized Yolov5 model.

[0043] Figure 8 A flowchart of the parameter adjustment method based on microlens array exposure process provided by the present invention;

[0044] Figure 9 A schematic diagram of the quality inspection device based on microlens array exposure process provided by the present invention;

[0045] Figure 10 This is a schematic diagram of the parameter adjustment device based on microlens array exposure process provided by the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The technical features designed in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0047] In the description of this invention, it should be noted that all terms used in this invention (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains, and should not be construed as limiting the invention; it should be further understood that the terms used in this invention should be understood to have the same meaning as those in the context of this specification and in the relevant field, and should not be understood in an idealized or overly formal sense, except as expressly defined in this invention.

[0048] To address the shortcomings of existing technologies in efficiently detecting the exposure quality of microlens array exposure processes, this invention provides a quality detection method and parameter adjustment method based on microlens array exposure processes. This method utilizes the training of a neural network target detection model to automatically identify the quality of exposure points, thereby effectively improving the detection accuracy and speed of exposure point quality.

[0049] Example 1

[0050] Please see Figure 1 This embodiment provides a quality inspection method based on microlens array exposure technology, including the following steps:

[0051] Step S10: Establish an exposure pattern detection model based on a convolutional neural network that can identify the pattern category of each exposure point; the pattern category includes qualified patterns and unqualified patterns.

[0052] In practice, the exposure pattern detection model can be obtained by training the image using a CNN object detection model. Please refer to [link / reference]. Figure 2 Specific training methods may include the following steps:

[0053] Step S11: Obtain an image set of exposure points and label the exposure patterns in the image set according to the pattern category to obtain an exposure pattern dataset. The exposure pattern dataset is divided into a training sample dataset and a test sample dataset. The pattern category labeling can be done manually. Both the training sample dataset and the test sample dataset contain exposure points of acceptable and unacceptable patterns.

[0054] Step S12: Establish a target neural network model and input the training sample dataset into the target neural network model for training to generate an exposure pattern detection model.

[0055] Step S13: Input the test sample dataset into the exposure pattern detection model for verification.

[0056] In specific implementation, the target neural network model can be, but is not limited to, target detection models such as RCNN (Region Convolutional Neural Networks), Fast-RCNN, Faster RCNN, YOLO, and SSD. Preferably, this embodiment prefers a one-step target detection model represented by YOLO and SSD, or a two-step target detection model represented by Faster-RCNN, specifically aiming for an exposure point quality detection precision and recall rate of not less than 80%.

[0057] More preferably, the target neural network model in this example can be a YOLOv5 model. Based on the YOLOv5 model, the CSP2 module is transformed into a Transformer encoder-decoder module, and a CBAM (Convolutional Block Attention Module) module is introduced before each convolutional layer to obtain an optimized YOLOv5 model. This increases the model's attention to the target exposure points, enhances the network's recognition of these points, and is more conducive to extracting effective features from the target exposure points, thereby improving the accuracy of quality detection. The specific structural composition is as follows: Figure 3 As shown, the optimized Yolov5 model includes a normalization module, a Transformer encoder module, and a CBAM module. The normalization module includes convolutional layers, batch normalization (BN), and activation layers; the Transformer encoder module includes LayerNorm, multi-head attention, Dropout, and fully connected layers; and the CBAM module includes channel attention and spatial attention modules. For detailed training procedures, please refer to [reference needed]. Figure 4 As shown, the exposure pattern detection model is obtained after multiple iterations of training using the normalization module, Transformer encoder module, and CBAM module. The specific training process can be set according to actual needs and will not be elaborated here.

[0058] Step S20: Set initial exposure parameters and acquire the exposure image of the point to be tested obtained through the MLA exposure process. The initial exposure parameters include the initial laser power and initial platform movement speed set according to actual working requirements. Specifically, the MLA exposure process uses a laser to emit a light source, which is then focused onto the product on a movable platform through an optical system including a microlens array for exposure processing, thereby forming the desired image. The image is then captured by a photographic device to obtain the exposure image of the point to be tested. The minimum resolution of the photographic device should be less than 1 μm.

[0059] In a preferred embodiment, in order to ensure the training effect of the exposure pattern detection model and the recognition effect of the actual exposure point to be detected, after obtaining the image set of exposure points and the exposure image of the exposure point to be detected, image enhancement processing can be performed on both, which may include, but is not limited to, one or more of the following: image cropping, denoising, enhancement, sharpening, restoration, and filtering.

[0060] Next, the exposed image is input into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point. Specifically, the data information set includes at least the pattern category, coordinate position, and exposure point size information corresponding to each target exposure point. To facilitate the classification of data for each exposure point, the data information set can be represented by vector representation data, that is, the vector representation data corresponding to each exposure point is (S, X, Y, W, H); where S represents the pattern category, X and Y represent the position of the center of the target exposure point relative to the image, and W and H represent the actual size of the target exposure point.

[0061] Step S30: Determine the pattern category of the exposure point based on the data information set, and count the number of exposure points of qualified and unqualified patterns in the exposure image to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points.

[0062] The process of determining the pattern category of the exposure point based on the data information set includes the following steps: setting a preset qualified pattern and a corresponding preset size threshold range; comparing the preset size threshold range with the actual size of the target exposure point to determine the pattern category of the target exposure point; if the actual size of the target exposure point is within the preset size threshold range, then the target exposure point is a qualified pattern; otherwise, the target exposure point is an unqualified pattern.

[0063] It should be noted that the preset qualified pattern and the corresponding preset size threshold range can be set according to the exposure points within the actual qualified requirement range. In this embodiment, the preset qualified pattern is set as an axisymmetric annular shape, such as a circular ring, triangular ring, rectangular ring, etc., or it can be an annular shape with an outer ring being a regular polygon and an inner ring being a circle. The distance difference between the inner and outer rings of the axisymmetric annular shape is defined as the preset size threshold range. Preferably, the preset size threshold range is set to 1–10 μm. Please refer to [link to relevant documentation]. Figure 5 In this embodiment, the preferred default pattern is a circular ring, and the default size threshold is the difference B between the inner and outer radii of the ring. Based on the vector representation data (S, X, Y, W, H) corresponding to each exposure point, it is determined whether the actual dimensions of W and H are within the default size threshold range. If so, S = 1, indicating a default pattern; otherwise, S = 0, indicating a non-default pattern. Finally, the number of exposure points with S = 1 and the number with S = 0 are counted. By calculating the proportion of S = 1 to the total number of exposure points, the pass rate is obtained, thus quickly and automatically obtaining the exposure quality detection effect of the microlens array exposure process.

[0064] Example 2

[0065] To more accurately determine the exposure quality, the quality inspection method based on microlens array exposure technology provided in this embodiment also identifies the target exposure points of unqualified patterns; that is, the difference from Embodiment 1 is that Embodiment 2 further subdivides the pattern categories into qualified patterns, overexposed patterns, and underexposed patterns. In the process of establishing the exposure pattern model, the exposure points in the image set are labeled as qualified patterns, overexposed patterns, and underexposed patterns to obtain an exposure pattern dataset, thereby training an exposure pattern detection model capable of identifying whether each exposure point is a qualified pattern, overexposed pattern, or underexposed pattern. For details, please refer to... Figure 6 As shown, Figure 6 This is a schematic diagram of the pattern at each exposure point under the recognition state of the exposure pattern detection model; a schematic diagram of the detection effect of the exposure pattern detection model obtained by training with the optimized Yolov5 model provided by this invention is shown below. Figure 7 As shown, by Figure 7 It can be seen that the recall and precision of qualified patterns, overexposed patterns, and underexposed patterns are all close to 100%. Therefore, it can be seen that the present invention can achieve good detection results by training with the optimized Yolov5 model.

[0066] Furthermore, when determining the pattern category of exposure points based on the data information set, the method for judging the target exposure point of unqualified patterns is as follows: if the actual size of the target exposure point of the unqualified pattern is greater than the preset size threshold range, then the target exposure point is determined to be an overexposed pattern; otherwise, the target exposure point is determined to be an underexposed pattern. That is, based on the vector representation data (S, X, Y, W, H) corresponding to each exposure point, it is determined whether the actual size of W and H is within the preset size threshold range. If so, S = 1, which is a qualified pattern; if the actual size of W and H is greater than the preset size threshold range, then S = 0, which is an overexposed pattern; if the actual size of W and H is neither within nor greater than the preset size threshold range, then S = 2, which is an underexposed pattern. Finally, by counting the number of S = 1, S = 0, and S = 2, the exposure quality status of the MLA exposure process can be obtained, which is beneficial for providing direction for subsequent improvements to the MLA exposure process.

[0067] It should be noted that, please refer to Figure 7The structural patterns of overexposed, qualified, and underexposed patterns shown are determined based on their inherent characteristics. Specifically: Overexposed patterns occur when the laser scanning energy of the photoresist is too high, causing an increase in size. This results in the exposure of parts of the ring that were not intended to be exposed. Therefore, patterns with exposure points whose actual size exceeds a preset size threshold are classified as overexposed. Qualified patterns occur when the laser scanning energy of the photoresist is just right, resulting in a complete exposed pattern with an actual size within the preset size threshold. This is specifically manifested in the fact that the center of the ring-shaped pattern is not fully exposed, thus qualifying as a qualified pattern. Underexposed patterns occur when the laser scanning energy of the photoresist is insufficient, resulting in an incomplete or undeveloped pattern. Therefore, exposure points that do not conform to either overexposed or qualified patterns are classified as underexposed.

[0068] Example 3

[0069] Please see Figure 8 The present invention also provides a parameter adjustment method based on microlens array exposure process, which mainly includes the following steps:

[0070] Step T10: Establish an exposure pattern detection model based on a convolutional neural network that can identify the pattern category of each exposure point; the pattern category includes qualified patterns and unqualified patterns.

[0071] Step T20: Set initial exposure parameters and acquire the exposure image of the exposure point to be detected obtained through MLA exposure process; input the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point;

[0072] Step T30: Determine the pattern category of the exposure point based on the data information set, and count the number of exposure points of qualified and unqualified patterns in the exposed image to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points.

[0073] In step T40, if the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is greater than or equal to a preset value, the initial exposure parameters are not changed; if the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is less than the preset value, the initial exposure parameters are modified.

[0074] The specific steps and implementation methods of steps T10 to T30 can be referred to the specific implementation method of Embodiment 1, and will not be repeated here. In step T40, the initial exposure parameters can be changed by modifying the initial laser power and the initial platform moving speed until the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is greater than or equal to a preset value. This indicates that the current exposure parameters are appropriate and no further parameter adjustment is needed. The preset value can be set according to the actual accuracy requirements and is not limited here. In this embodiment, the preset value range is preferably 70% to 90%, for example, the preset value can be set to 80%.

[0075] Example 4

[0076] To more accurately determine the direction of exposure parameter modification, the parameter adjustment method based on microlens array exposure technology provided in this embodiment further distinguishes between overexposed and underexposed patterns. That is, the difference from Embodiment 3 is that the specific steps and implementation methods of steps T10 to T30 can be implemented with reference to the specific implementation method of Embodiment 2.

[0077] In step T40, in addition to judging whether the proportion of the number of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to a preset value, the number of exposure points in the overexposed pattern and the number of exposure points in the underexposed pattern are also judged in the unqualified pattern. Specifically: in the unqualified pattern, if the number of exposure points in the overexposed pattern is greater than the number of exposure points in the underexposed pattern, it is determined to be overexposed. Then, the initial exposure parameters are modified by reducing the laser power or increasing the platform movement speed until the proportion of the number of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to the preset value. If the number of exposure points in the overexposed pattern is less than the number of exposure points in the underexposed pattern, it is determined to be underexposed. Then, the initial exposure parameters are modified by increasing the laser power or decreasing the platform movement speed until the proportion of the number of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to the preset value. The preset value can be set according to the actual accuracy requirements and is not limited here. In this embodiment, the preset value range is preferably 70% to 90%, for example, the preset value can be set to 80%.

[0078] Compared with Example 3, the setting steps of Example 4 above can effectively and accurately provide a direction for adjusting exposure parameters without the need for human experience to make judgments and modifications, thereby further improving the accuracy and speed of exposure parameter adjustment.

[0079] Example 5

[0080] Please see Figure 9 This embodiment also provides a quality inspection device based on a microlens array exposure process, comprising:

[0081] Model building module: Based on a convolutional neural network, an exposure pattern detection model is built that can identify the pattern category of each exposure point; the pattern categories include qualified patterns and unqualified patterns;

[0082] Quality inspection module; sets initial exposure parameters and acquires the exposure image obtained by MLA exposure process; inputs the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point;

[0083] Pattern statistics module: Based on the data information set, determine the pattern category of the exposure point, and count the number of exposure points of qualified and unqualified patterns in the exposed image to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points.

[0084] The specific implementation methods of each module in Embodiment 5 can refer to the implementation methods of Embodiment 1 or Embodiment 2, and will not be repeated here.

[0085] Example 6

[0086] Please see Figure 10 This embodiment also provides a parameter adjustment device based on a microlens array exposure process, including:

[0087] Model building module: Based on a convolutional neural network, an exposure pattern detection model is built that can identify the pattern category of each exposure point; the pattern categories include qualified patterns and unqualified patterns;

[0088] Quality inspection module; sets initial exposure parameters and acquires the exposure image obtained by MLA exposure process; inputs the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category for each exposure point;

[0089] Pattern statistics module: Based on the data information set, determine the pattern category of the exposure point, and count the number of exposure points of qualified and unqualified patterns in the exposed image to obtain the proportion of the number of exposure points of qualified patterns to the total number of exposure points;

[0090] The parameter modification module: In the case of unqualified patterns, if the number of exposure points in the overexposed pattern is greater than the number of exposure points in the underexposed pattern, the initial exposure parameters are modified by reducing the laser power or increasing the platform movement speed until the proportion of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to the preset value; if the number of exposure points in the overexposed pattern is less than the number of exposure points in the underexposed pattern, the initial exposure parameters are modified by increasing the laser power or decreasing the platform movement speed until the proportion of exposure points in the qualified pattern to the total number of exposure points is greater than or equal to the preset value.

[0091] The specific implementation methods of each module in Embodiment Six can be referred to the implementation methods of Embodiment Three or Embodiment Four, and will not be repeated here.

[0092] In summary, compared with existing technologies, the quality inspection and parameter adjustment method based on microlens array exposure technology provided by this invention utilizes a neural network target detection model to input the exposed image into the exposed pattern model to automatically identify the pass rate of the pattern, and then adjusts the exposure parameters according to the pass rate. This not only automates the acquisition of the pass rate of MLA exposure quality, effectively improving the accuracy and speed of quality inspection, but also facilitates the adjustment of exposure parameters, demonstrating promising application prospects.

[0093] Furthermore, those skilled in the art should understand that although many problems exist in the prior art, each embodiment or technical solution of the present invention can be improved in only one or a few aspects, without necessarily solving all the technical problems listed in the prior art or the background art simultaneously. Those skilled in the art should understand that any content not mentioned in a claim should not be construed as a limitation on that claim.

[0094] Although this paper frequently uses terms such as exposure pattern detection model, pattern category, qualified pattern, unqualified pattern, exposure image, data information set, image set, exposure pattern dataset, training sample dataset, test sample dataset, target neural network model, preset qualified pattern, preset size threshold range, overexposed pattern, underexposed pattern, preset value, etc., the possibility of using other terms is not excluded. The use of these terms is merely for the convenience of describing and explaining the essence of the invention; interpreting them as any additional limitation would contradict the spirit of the invention. The terms "first," "second," etc. (if present) in the specification, claims, and accompanying drawings of the embodiments of the invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A quality detection method based on a microlens array exposure process, characterized in that, The method comprises the following steps: establishing an exposure pattern detection model capable of identifying the category of each exposure point pattern based on a convolutional neural network; the pattern category comprises a qualified pattern and an unqualified pattern; setting an initial exposure parameter and obtaining an exposure image of an exposure point to be detected through an MLA exposure process; inputting the exposure image into the exposure pattern detection model to generate a data information set corresponding to the pattern category of each exposure point; the data information set at least comprises the pattern category, coordinate position and exposure point size information corresponding to each target exposure point; the data information set is represented by a vector representation data (S, X, Y, W, H), wherein S represents the pattern category, X and Y represent the position of the center of the target exposure point relative to the image, and W and H represent the actual size of the target exposure point; judging the pattern category of the exposure point according to the data information set and counting the number of exposure points of the qualified pattern and the unqualified pattern in the exposure image to obtain the proportion of the number of exposure points of the qualified pattern in the total number of exposure points; judging the pattern category of the exposure point according to the data information set comprises the following steps: setting a preset qualified pattern and a corresponding preset size threshold range, comparing the preset size threshold range with the actual size of the target exposure point to determine the pattern category of the target exposure point; the preset qualified pattern is set as an axisymmetric ring pattern, and the difference between the inner and outer ring distances of the axisymmetric ring pattern is defined as the preset size threshold range; if the actual size of the target exposure point is within the preset size threshold range, the target exposure point is a qualified pattern; otherwise, the target exposure point is an unqualified pattern; judging the target exposure point of the unqualified pattern; if the actual size of the target exposure point of the unqualified pattern is greater than the preset size threshold range, the target exposure point is determined to be an overexposure pattern; otherwise, the target exposure point is determined to be an underexposure pattern. 2.The quality detection method based on microlens array exposure process of claim 1, wherein, establishing an exposure pattern model comprises the following steps: obtaining an image set of exposure points and labeling the exposure patterns in the image set according to the pattern category to obtain an exposure pattern data set; dividing the exposure pattern data set into a training sample data set and a test sample data set; establishing a target neural network model, inputting the training sample data set into the target neural network model for training to generate an exposure pattern detection model, and then inputting the test sample data set into the exposure pattern detection model for verification. 3.The quality detection method based on microlens array exposure process of claim 2, wherein: The target neural network model is one of Yolo, SSD and Faster-RCNN target detection models.

4. A method for adjusting parameters of a microlens array exposure process, characterized in that The quality detection method based on the micro-lens array exposure process according to any one of claims 1-3 further comprises the following steps: if the proportion of the number of exposure points of the qualified pattern in the total number of exposure points is greater than or equal to a preset value, the initial exposure parameter is not changed; if the proportion of the number of exposure points of the qualified pattern in the total number of exposure points is less than the preset value, the initial exposure parameter is modified.

5. The method of claim 4, wherein: In the unqualified pattern, if the number of exposure points of the overexposed pattern is greater than the number of exposure points of the underexposed pattern, the initial exposure parameters are modified by reducing the laser power or increasing the platform moving speed until the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is greater than or equal to the preset value; If the number of exposure points of the overexposed pattern is less than the number of exposure points of the underexposed pattern, the initial exposure parameters are modified by increasing the laser power or reducing the platform moving speed until the proportion of the number of exposure points of the qualified pattern to the total number of exposure points is greater than or equal to the preset value.

Images