LDI machine image segmentation and exposure splicing method

Abstract

The invention relates to the technical field of digital photoetching equipment, in particular to an LDI machine image segmentation and exposure splicing method, which comprises the following steps: carrying out primary segmentation on an original complex graph, and carrying out segmentation again after refinement processing; in the exposure process, periodically collecting a gray histogram of an exposure pattern, aligning the exposure images of adjacent areas to form a continuous exposure image, and dynamically adjusting exposure parameters by a light beam modulator according to the characteristics of each second segmentation area to realize closed-loop control on the exposure quality; all exposed images are spliced through a search algorithm and a multi-frame fusion technology, smooth transition of image edges is realized, discontinuity of local features is effectively avoided, the overall consistency and integrity of the spliced images are ensured, the imaging quality and precision are further improved, finally, the spliced images are verified, and the image quality is improved. And seamless splicing of each segmented region is ensured, so that high-quality graphic output is generated.

Description

Technical Field

[0001] The present invention relates to the technical field of digital lithography equipment, and more particularly to an LDI machine image segmentation and exposure splicing method. Background Art

[0002] Existing LDI exposure technology faces significant challenges when processing complex graphics. Particularly during the graphic preprocessing stage, traditional graphic segmentation methods often struggle to accurately demarcate the exposure area when processing complex graphics. Due to inaccurate segmentation, the exposure area is often set too wide. This not only makes it difficult to effectively capture the fine features of the graphic, but also leads to unnecessary material loss and unnecessarily extended exposure times. An overly large exposure area prevents the detailed representation of the graphic, resulting in lost or blurred details, a serious problem for exposure applications requiring high precision and high resolution.

[0003] Furthermore, when processing complex graphics, traditional image stitching methods often fail to fully consider factors such as image matching accuracy, stitching gaps, and overall coordination. This can lead to problems such as misalignment, overlap, or noticeable gaps in the stitched image. These issues not only affect the overall quality of the exposed image but can also lead to unnecessary repeated exposures and wasted time, further extending the exposure cycle and reducing production efficiency.

[0004] The prior art discloses a GDS pattern segmentation method for LDI exposure equipment, comprising: segmenting an original GDS pattern to form a plurality of first rectangular blocks; polygonizing each polygonal circuit pattern within each first rectangular block to decompose each polygonal circuit pattern within each first rectangular block into at least one polygon with a preset number of sides; and segmenting each first rectangular block to form a plurality of second rectangular blocks. This method effectively segments the original GDS pattern into a plurality of small-area patterns by segmenting, decomposing, and then segmenting the original GDS pattern, thereby improving the accuracy of characterizing small patterns within the original GDS pattern and the exposure precision of the LDI exposure equipment. However, using traditional image stitching methods based on this method still results in problems such as misalignment, overlap, or noticeable gaps in the stitched images, resulting in low exposure image quality. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide an LDI machine image segmentation and exposure splicing method to improve the exposure splicing accuracy and quality, ensure seamless splicing of each segmented area, and thus generate high-quality graphic output.

[0006] To solve the above technical problems, the present invention adopts a technical solution: providing an LDI image segmentation and exposure splicing method, comprising the following steps: S1: loading original complex graphic data from a storage medium, the original complex graphic data including polygonal circuit elements of various shapes and sizes, and preprocessing the original complex graphic; S2: Using a first preset rectangular frame as an initial segmentation unit, initially segmenting the original complex graphic to form a plurality of first segmentation regions, each of which contains at least one polygonal circuit element. The polygonal circuit elements in the same first segmentation region collectively form a first-level sub-graphic, and all the polygonal circuit elements collectively form the original complex graphic. S3: performing thinning processing on the polygonal circuit elements within each first segmented area; S4: Using a second preset rectangular frame as a secondary segmentation unit, further segmenting each first segmented area to generate a plurality of second segmented areas, each of which contains at least one polygonal circuit element that has undergone refinement. The polygonal circuit elements in the same second segmented area collectively form a second-level sub-graph, and all second-level sub-graphs in the same first segmented area collectively form a corresponding first-level sub-graph. S5: Based on the segmentation results, combined with the motion characteristics of the laser, micromirror, and projection objective lens and the substrate layout, an exposure device is positioned on the substrate to expose each of the second segmented areas one by one using a serpentine trajectory; during the exposure process, a grayscale histogram of the exposure pattern is periodically collected to align exposure images of adjacent areas to form a continuous exposure image, and a beam modulator dynamically adjusts exposure parameters based on the characteristics of each of the second segmented areas; S6: After the exposure of all the second segmented areas is completed, all the exposed images are stitched together using a search algorithm and a multi-frame fusion technology; S7: Perform image verification on the stitched image and the original exposure image.

[0007] The LDI machine image segmentation and exposure stitching method of the present invention performs an initial segmentation on the original complex graphics to ensure the preliminary feasibility and accuracy of the graphics segmentation; through refinement processing, the flexibility and accuracy of the graphics processing can be improved, the capacity of the exposure equipment can be more effectively utilized, and the complexity and error during image stitching can be reduced; through re-segmentation, the graphics segmentation is further refined, providing a more accurate basis for subsequent exposure stitching; during the exposure process, the grayscale histogram of the exposure pattern is periodically collected to align the exposure images of adjacent areas to form a continuous exposure image, and at the same time, the beam modulator dynamically adjusts the exposure parameters according to the characteristics of each second segmented area to achieve closed-loop control of the exposure quality; all the exposure images are stitched together through a search algorithm and multi-frame fusion technology to achieve a smooth transition of the image edges, effectively avoid the discontinuity of local features, ensure the overall consistency and integrity of the image after stitching, thereby improving the imaging quality and accuracy, ensuring the seamless stitching of each segmented area, and thus generating high-quality graphics output.

[0008] Preferably, in step S1, the preprocessing includes performing Gaussian filtering denoising and histogram equalization processing on the original complex graphics.

[0009] Preferably, in step S3, the image is processed using a Canny edge detection algorithm, wherein the ratio of the high threshold to the low threshold is 2.3:1, so as to identify the image contour, simplify the polygon contour, and decompose the simplified polygon into triangle / quadrilateral sub-units.

[0010] Preferably, both step S3 and step S4 include processing of incomplete polygonal circuit elements: if the polygonal circuit element in the segmented area is incomplete, that is, a part of the polygonal circuit element exceeds the current segmented area, the current segmented area and the exceeding part of the incomplete polygonal circuit element are first archived as a whole, and the exceeding part is completed to form a closed polygon before segmentation.

[0011] Preferably, step S3 further includes recording the thinned polygons: calculating and recording the number, number and vertex coordinates of the thinned polygons in each first segmented area to provide data support for subsequent exposure stitching and exposure control.

[0012] Preferably, during the segmentation process, based on the bimodal characteristics of the grayscale histogram, the threshold is automatically determined by an algorithm to divide each second segmented area into a bright light area, a dark light area, and a normal light area; the grayscale value of the bright light area is greater than 219, corresponding to the direct area of ​​the laser center spot, which is prone to overexposure and requires reducing the laser power; the grayscale value of the dark light area is less than 80, located at the edge of the spot, and insufficient energy leads to underexposure, requiring an extended exposure time; the grayscale value of the normal light area is greater than or equal to 80 and less than or equal to 219, which is an ideal exposure area with complete graphic features and maintained benchmark parameters; During the exposure process, the exposure area information density evaluation function is introduced , used to quantitatively analyze the complexity of graphic features within the segmented image area; ; Where, Indicates the total number of pixels in the image with a value equal to 0; Indicates the number of exposed images; Indicates the total number of pixels in the exposure image; Indicates the total number of pixels in the exposed image; The value range is , The larger the value, the denser the circuit pattern features in the area and the higher the information complexity; The smaller the value, the larger the background proportion and the more spacious the area; Comprehensively consider the current exposure sub-image area The brightness distribution, contrast and density of the included graphic features are used to evaluate the exposure quality of the current sub-area in real time, and an exposure uniformity evaluation function is introduced. ; ; Where, For the region The number of pixels within ; Indicates area The mean brightness of the inner pixels; Indicates area Internal brightness standard deviation; Indicates the brightness weight coefficient, the value is 0.4; Indicates the contrast weight coefficient, the value is 0.3; Indicates the saturation weight coefficient, the value is 0.3; Calculate the exposure evaluation function of each second segmented area in real time , for normal light area or when When the exposure is qualified, it is judged that the exposure is qualified; for strong optical drive, dark light area or when When the DMD micromirror compensation algorithm is automatically triggered, a 2%-5% overlapping exposure area is generated for light intensity balance.

[0013] Preferably, during the exposure process, the reflected light intensity distribution map of the current exposure area is collected through the light intensity sensor array, and the light intensity gradient vector field is calculated. When the light intensity gradient vector field is detected to be greater than 5%, a DMD micromirror compensation matrix is ​​generated, and the uneven distribution of the light spot is compensated by the DMD micromirror angle, the laser power is dynamically adjusted, and the scanning speed is adaptively changed.

[0014] Preferably, in step S5, the overlap rate between adjacent exposure positions in the moving direction is set within a range of 3% to 10% based on the layout of the second segmented areas and the pre-planned serpentine trajectory based on the adjacent relationship.

[0015] Preferably, in step S6, the stitching process includes: for the sub-images obtained by exposing adjacent second segmented areas, extracting feature points and feature descriptors using the Canny algorithm; performing nearest neighbor matching of the feature descriptors, and screening out reliable matching point pairs using a ratio test; minimizing the sum of the gradient amplitudes of the pixels through which the stitching line passes, and finding the path with the least obvious image structure; and completing the stitching of the exposed images using histogram matching in the overlapping area.

[0016] Preferably, in step S6, a comprehensive evaluation function of stitching quality is introduced to quantitatively evaluate the light intensity uniformity and edge sharpness of the overlapping area after image stitching; ; Where, Indicates the comprehensive score of splicing quality, with a value range of , the larger the value, the higher the quality; Indicates the light intensity uniformity weight, with a value of 0.6; Indicates the edge sharpness weight, the value is 0.4; Indicates the total number of pixels in the overlapping area; Indicates the The actual light intensity value of each pixel; Indicates the target light intensity value; Represents the gradient modulus and value of the edge area; when When , the splicing is judged to be qualified; when When , the histogram matching correction is triggered; when , the stitching quality is judged to be unqualified, and the corresponding sub-image needs to be re-exposed and stitched.

[0017] Compared with the prior art, the present invention has the following beneficial effects: improving imaging quality and accuracy, ensuring seamless splicing of each segmented area, and thus generating high-quality graphic output. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Flowchart of the LDI image segmentation and exposure stitching method according to an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the original exposure image in an embodiment of the present invention; Figure 3 is a schematic diagram of the segmentation of the original exposure image in an embodiment of the present invention; Figure 4: is a representation diagram of the exposure area information density evaluation function in an embodiment of the present invention; Figure 5 : is a representation diagram of the exposure uniformity evaluation function in an embodiment of the present invention; Figure 6 This is a comparison diagram of images before and after triggering DMD micromirror compensation in an embodiment of the present invention; Figure 7 is the stitched image before correction using histogram matching; Figure 8 is the stitched image after correction using histogram matching. DETAILED DESCRIPTION

[0019] The present invention is further described below with reference to specific embodiments. The accompanying drawings are for illustrative purposes only and are schematic, not actual, representations. They should not be construed as limiting this patent. To better illustrate the embodiments of the present invention, some components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted from the drawings.

[0020] Example 1 A method for LDI machine image segmentation and exposure splicing, such as Figures 1 to 3 As shown, the following steps are included: S1: loading original complex graphic data from a storage medium, the original complex graphic data including polygonal circuit elements of various shapes and sizes, and preprocessing the original complex graphic; S2: Using a first preset rectangular frame as an initial segmentation unit, initially segmenting the original complex graphic to form a plurality of first segmentation regions, each of which contains at least one polygonal circuit element. The polygonal circuit elements in the same first segmentation region collectively form a first-level sub-graphic, and all the polygonal circuit elements collectively form the original complex graphic. S3: performing thinning processing on the polygonal circuit elements within each first segmented area; S4: Using a second preset rectangular frame as a secondary segmentation unit, further segmenting each first segmented area to generate a plurality of second segmented areas, each of which contains at least one polygonal circuit element that has undergone refinement. The polygonal circuit elements in the same second segmented area collectively form a second-level sub-graph, and all second-level sub-graphs in the same first segmented area collectively form a corresponding first-level sub-graph. S5: Based on the segmentation results, combined with the motion characteristics of the laser, micromirror, and projection objective lens and the substrate layout, an exposure device is positioned on the substrate to expose each of the second segmented areas one by one using a serpentine trajectory; during the exposure process, a grayscale histogram of the exposure pattern is periodically collected to align exposure images of adjacent areas to form a continuous exposure image, and a beam modulator dynamically adjusts exposure parameters based on the characteristics of each of the second segmented areas; S6: After the exposure of all the second segmented areas is completed, all the exposed images are stitched together using a search algorithm and a multi-frame fusion technology; S7: Perform image verification on the stitched image and the original exposure image.

[0021] The above-mentioned LDI machine image segmentation and exposure stitching method performs an initial segmentation on the original complex graphics to ensure the initial feasibility and accuracy of the graphics segmentation; through refinement processing, the flexibility and accuracy of the graphics processing can be improved, the capabilities of the exposure equipment can be more effectively utilized, and the complexity and errors during image stitching can be reduced; through re-segmentation, the graphics segmentation is further refined, providing a more accurate basis for subsequent exposure stitching; during the exposure process, the grayscale histogram of the exposure pattern is periodically collected to align the exposure images of adjacent areas to form a continuous exposure image, and at the same time, the beam modulator dynamically adjusts the exposure parameters according to the characteristics of each second segmented area to achieve closed-loop control of the exposure quality; all exposure images are stitched together through a search algorithm and multi-frame fusion technology to achieve a smooth transition of the image edges, effectively avoid the discontinuity of local features, ensure the overall consistency and integrity of the stitched images, thereby improving the imaging quality and accuracy, ensuring the seamless stitching of each segmented area, and thus generating high-quality graphics output.

[0022] Due to issues with the image file itself, the image file may contain noise, blur, distortion, and other issues, which may interfere with the accuracy of the image segmentation algorithm, leading to segmentation errors and affecting subsequent image processing and analysis results. There is also the issue of image format. Different image formats store and compress images differently, which may cause pixel values ​​in the image file to change, thus affecting the accuracy of the segmentation results. To this end, in step S1, preprocessing includes Gaussian filtering denoising and histogram equalization processing on the original complex image.

[0023] In step S3, the image is processed using the Canny edge detection algorithm, with a high-threshold to low-threshold ratio of 2.3:1. This algorithm identifies image contours, simplifies polygon outlines, and decomposes these simplified polygons into triangular / quadrilateral subunits to improve image processing flexibility and accuracy. This thinning process more effectively utilizes the capabilities of the exposure equipment while reducing complexity and errors in image stitching.

[0024] Both step S3 and step S4 include processing of incomplete polygonal circuit elements: if the polygonal circuit element in the segmented area is incomplete, that is, a part of the polygonal circuit element exceeds the current segmented area, the current segmented area and the exceeding part of the incomplete polygonal circuit element are first archived as a whole, and the exceeding part is completed to form a closed polygon before segmentation.

[0025] Step S3 also includes recording the thinned polygons: calculating and recording the number, number and vertex coordinates of the thinned polygons in each first segmented area to provide data support for subsequent exposure stitching and exposure control.

[0026] In step S5, based on the layout of the second segmented area and the pre-planned serpentine trajectory of the adjacent relationship, the overlap rate between adjacent exposure positions in the moving direction is set within 3% to 10% to ensure that there is enough overlapping area for alignment and fusion in the image stitching process.

[0027] The size of the first preset rectangular frame is larger than that of the second preset rectangular frame to adapt to the coarse-to-fine segmentation strategy and ensure that the graphic details are fully preserved.

[0028] The first preset rectangular frame and the second preset rectangular frame are in the shape of long strips and rectangles to optimize the stitching accuracy, reduce the movement distance of the ultraviolet laser, the digital micromirror, and the projection objective lens, and improve the exposure efficiency.

[0029] In this embodiment, through the steps of primary segmentation, refinement processing, secondary segmentation and stitching, the original complex exposure image is effectively converted into a series of highly refined images, which significantly improves the ability to capture fine structures in the segmented images.

[0030] Example 2 This embodiment is similar to the first embodiment, except that, during the segmentation process, based on the bimodal characteristics of the grayscale histogram, the threshold is automatically determined by an algorithm to divide each second segmented area into a strong light area, a dark light area, and a normal light area; the grayscale value of the strong light area is greater than 219, corresponding to the direct area of ​​the laser center spot, which is prone to overexposure and requires reducing the laser power; the grayscale value of the dark light area is less than 80, located at the edge of the spot, and insufficient energy leads to underexposure, requiring an extended exposure time; the grayscale value of the normal light area is greater than or equal to 80 and less than or equal to 219, which is an ideal exposure area with complete graphic features and maintained benchmark parameters.

[0031] During the exposure process, the exposure area information density evaluation function is introduced , used to quantitatively analyze the complexity of graphic features within the segmented image area; ; Where, Indicates the total number of pixels in the image with a value equal to 0; Indicates the number of exposed images; Indicates the total number of pixels in the exposure image; Indicates the total number of pixels in the exposed image; The value range is , The larger the value, the denser the circuit pattern features in the area and the higher the information complexity; The smaller the value, the larger the background proportion and the more spacious the area. Figure 4 As shown, the corresponding exposure areas are calculated value; Comprehensively consider the current exposure sub-image area The brightness distribution, contrast and density of the included graphic features are used to evaluate the exposure quality of the current sub-area in real time, and an exposure uniformity evaluation function is introduced. ; ; Where, For the region The number of pixels within ; Indicates area The mean brightness of the inner pixels; Indicates area Internal brightness standard deviation; Indicates the brightness weight coefficient, the value is 0.4; Indicates the contrast weight coefficient, the value is 0.3; Indicates the saturation weight coefficient, the value is 0.3.

[0032] Calculate the exposure evaluation function of each second segmented area in real time , for normal light area or when When the exposure is qualified, it is judged that the exposure is qualified; for strong optical drive, dark light area or when When the DMD micromirror compensation algorithm is automatically triggered, a 2%-5% overlapping exposure area is generated to balance the light intensity. Figure 5 As shown, the current exposure area , the exposure is judged to be qualified; if Figure 6 As shown, the current exposure area , automatically triggering DMD micromirror compensation to balance light intensity .

[0033] Exposure evaluation function It can help identify which areas need more detailed processing and which areas can be simplified, thereby effectively improving processing efficiency while ensuring image quality; it can deeply analyze each segmented image block to dynamically monitor the quality of sub-areas during the exposure process, and for high-risk areas, High or low values, bright and dark areas, often accompanied by low or borderline values , requiring more refined exposure control or priority processing during stitching; for low-risk areas, The exposure is relatively high, and the exposure is relatively stable, so more efficient algorithms can be applied during stitching. In addition, the introduction of the exposure evaluation function has also brought beneficial effects to image stitching. By pre-classifying the image blocks, it is possible to more accurately determine which areas need to be processed first and which areas can be processed later during the fusion process, as well as how to adjust the stitching strategy to adapt to the information characteristics of different areas. This not only helps to improve the quality of the fused image but also reduces unnecessary computational overhead and improves overall processing efficiency. The adaptive prioritization and strategy adjustment of evaluation results avoid unnecessary waste of computing resources in low-risk areas, such as using the highest-precision algorithm to process all blocks, while ensuring that high-risk areas receive sufficient processing intensity, improving the overall consistency and visual quality of the stitched image, reducing problems such as misalignment and gaps, and improving the processing efficiency of the entire exposure stitching process.

[0034] During the exposure process, the light intensity sensor array collects the reflected light intensity distribution map of the current exposure area and calculates the light intensity gradient vector field. When the light intensity gradient vector field is detected to be greater than 5%, a DMD micromirror compensation matrix is ​​generated. The DMD micromirror angle is used to compensate for uneven spot distribution, dynamically adjust laser power, and adaptively change scanning speed. Through a dual detection system of light intensity sensor array and image analysis, the actual light intensity distribution data of each exposure area is acquired in real time and dynamically compared with the preset exposure model. This enables dynamic adjustment of the exposure process, ensuring that the LDI machine can efficiently and accurately complete exposure and stitching operations, improving precision and quality.

[0035] Example 3 This embodiment is similar to the second embodiment, except that, in step S6, the stitching process includes: for the sub-images obtained by exposing the adjacent second segmented area, the Canny algorithm is used to extract feature points and feature descriptors; the nearest neighbor matching of the feature descriptors is performed, and the ratio test is used to screen out reliable matching point pairs; the sum of the gradient amplitudes of the pixels through which the stitching line passes is minimized to find the path with the least obvious image structure; and the histogram matching is used in the overlapping area to complete the stitching of the exposed images. In this embodiment, before fusion, the best stitching line search algorithm is used to select the path with the smallest gradient change in the overlapping area of ​​adjacent images as the stitching seam to achieve a natural transition of the image edge and improve the overall consistency, and perform photometric consistency correction. After fusion, based on the statistical analysis of the pixel values ​​in the overlapping area, histogram matching is applied to correct the brightness and color consistency of the stitched overall image, such as Figure 7 、 Figure 8 shown.

[0036] In step S6, a comprehensive evaluation function for stitching quality is introduced to quantitatively evaluate the light intensity uniformity and edge sharpness of the overlapping area after image stitching; ; Where, Indicates the comprehensive score of splicing quality, with a value range of , the larger the value, the higher the quality; Indicates the light intensity uniformity weight, with a value of 0.6; Indicates the edge sharpness weight, the value is 0.4; Indicates the total number of pixels in the overlapping area; Indicates the The actual light intensity value of each pixel; Indicates the target light intensity value; Represents the gradient modulus and value of the edge area; when When , the splicing is judged to be qualified; when When , the histogram matching correction is triggered; when , the stitching quality is judged to be unqualified, and the corresponding sub-image needs to be re-exposed and stitched.

[0037] In this embodiment, each segmented image block is first subjected to a detailed brightness and contrast analysis to identify bright and shadowed areas, as well as the contrast within the image. The image's color saturation, hue distribution, and texture characteristics are further analyzed to capture detailed information and overall visual quality. The complexity of the image content, such as edge density and the number and distribution of objects, is then used to determine the specific processing strategy. A comprehensive stitching quality evaluation function calculates a composite score that reflects the overall quality of the stitched image, providing an important reference for subsequent image processing and fusion.

[0038] The evaluation function system in this embodiment integrates understanding of segmentation, real-time monitoring and adaptive control of the exposure process, and finally, verification of stitching quality. By dynamically analyzing and classifying block image features, it achieves closed-loop control of exposure quality and implements resource allocation and strategy selection during the image stitching process. These dual functions ensure efficient completion of the entire adaptive segmentation and exposure stitching process while improving exposure accuracy and stitching quality. This method not only considers the geometric shape and positional relationship of each segmented area to ensure visual continuity and consistency of the stitched image, but also closely integrates the physical characteristics and kinematic constraints of the LDI exposure equipment to achieve efficient and accurate exposure stitching.

[0039] In the specific contents of the above-mentioned specific implementation methods, the various technical features can be combined in any non-contradictory manner. In order to make the description concise, not all possible combinations of the above-mentioned technical features are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0040] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the claims of the present invention.

Claims

1. A method for LDI image segmentation and exposure splicing, characterized in that: The following steps are involved: S1: loading original complex graphic data from a storage medium, the original complex graphic data including polygonal circuit elements of various shapes and sizes, and preprocessing the original complex graphic; S2: Using a first preset rectangular frame as an initial segmentation unit, initially segmenting the original complex graphic to form a plurality of first segmentation regions, each of which contains at least one polygonal circuit element. The polygonal circuit elements in the same first segmentation region collectively form a first-level sub-graphic, and all the polygonal circuit elements collectively form the original complex graphic. S3: performing thinning processing on the polygonal circuit elements within each first segmented area; S4: Using a second preset rectangular frame as a secondary segmentation unit, further segmenting each first segmented area to generate a plurality of second segmented areas, each of which contains at least one polygonal circuit element that has undergone refinement. The polygonal circuit elements in the same second segmented area collectively form a second-level sub-graph, and all second-level sub-graphs in the same first segmented area collectively form a corresponding first-level sub-graph. S5: Based on the segmentation results, combined with the motion characteristics of the laser, micromirror, and projection objective lens and the substrate layout, an exposure device is positioned on the substrate to expose each of the second segmented areas one by one using a serpentine trajectory; during the exposure process, a grayscale histogram of the exposure pattern is periodically collected to align exposure images of adjacent areas to form a continuous exposure image, and a beam modulator dynamically adjusts exposure parameters based on the characteristics of each of the second segmented areas; S6: After the exposure of all the second segmented areas is completed, all the exposed images are stitched together using a search algorithm and a multi-frame fusion technology; S7: Perform image verification on the stitched image and the original exposure image.

2. The LDI image segmentation and exposure splicing method according to claim 1, characterized in that: In step S1 , the preprocessing includes performing Gaussian filtering denoising and histogram equalization processing on the original complex graphics.

3. The LDI image segmentation and exposure splicing method according to claim 1, characterized in that: In step S3, the image is processed using the Canny edge detection algorithm, where the ratio of the high threshold to the low threshold is 2.3:1, to identify the image contour, simplify the polygon contour, and decompose the simplified polygon into triangle / quadrilateral sub-units.

4. The LDI image segmentation and exposure splicing method according to claim 3, characterized in that: Both step S3 and step S4 include processing of incomplete polygonal circuit elements: if the polygonal circuit element in the segmented area is incomplete, that is, a part of the polygonal circuit element exceeds the current segmented area, the current segmented area and the exceeding part of the incomplete polygonal circuit element are first archived as a whole, and the exceeding part is completed to form a closed polygon before segmentation.

5. The LDI image segmentation and exposure splicing method according to claim 4, characterized in that: Step S3 also includes recording the thinned polygons: calculating and recording the number, number and vertex coordinates of the thinned polygons in each first segmented area to provide data support for subsequent exposure stitching and exposure control.

6. The LDI image segmentation and exposure splicing method according to claim 1, characterized in that: During the segmentation process, based on the bimodal characteristics of the grayscale histogram, the algorithm automatically determines the threshold value and divides each second segmentation area into a bright light area, a dark light area, and a normal light area. The grayscale value of the bright light area is greater than 219, corresponding to the direct area of ​​the laser center spot, which is prone to overexposure and requires reducing the laser power. The grayscale value of the dark light area is less than 80, located at the edge of the spot, and insufficient energy leads to underexposure, requiring a longer exposure time. The grayscale value of the normal light area is greater than or equal to 80 and less than or equal to 219, which is the ideal exposure area with complete graphic features and maintained benchmark parameters. During the exposure process, the exposure area information density evaluation function is introduced , used to quantitatively analyze the complexity of graphic features within the segmented image area; ； Where, Indicates the total number of pixels in the image with a value equal to 0; Indicates the number of exposed images; Indicates the total number of pixels in the exposure image; Indicates the total number of pixels in the exposed image; The value range is , The larger the value, the denser the circuit pattern features in the area and the higher the information complexity; The smaller the value, the larger the background proportion and the more spacious the area; Comprehensively consider the current exposure sub-image area The brightness distribution, contrast and density of the included graphic features are used to evaluate the exposure quality of the current sub-area in real time, and an exposure uniformity evaluation function is introduced. ; ； Where, For the region The number of pixels within ; Indicates area The mean brightness of the inner pixels; Indicates area Internal brightness standard deviation; Indicates the brightness weight coefficient, the value is 0.4; Indicates the contrast weight coefficient, the value is 0.3; Indicates the saturation weight coefficient, the value is 0.3; Calculate the exposure evaluation function of each second segmented area in real time , for normal light area or when When the exposure is qualified, it is judged that the exposure is qualified; for strong optical drive, dark light area or when When the DMD micromirror compensation algorithm is automatically triggered, a 2%-5% overlapping exposure area is generated for light intensity balance.

7. The LDI image segmentation and exposure splicing method according to claim 6, characterized in that: During the exposure process, the light intensity sensor array is used to collect the reflected light intensity distribution map of the current exposure area and calculate the light intensity gradient vector field. When the light intensity gradient vector field is detected to be greater than 5%, a DMD micromirror compensation matrix is ​​generated. The uneven distribution of the light spot is compensated by the DMD micromirror angle, the laser power is dynamically adjusted, and the scanning speed is adaptively changed.

8. The LDI image segmentation and exposure splicing method according to claim 1, wherein: In step S5, the overlap rate between adjacent exposure positions in the moving direction is set within a range of 3% to 10% according to the layout of the second segmented areas and the pre-planned serpentine trajectory based on the adjacent relationship.

9. The LDI image segmentation and exposure splicing method according to any one of claims 1 to 8, characterized in that: In step S6, the stitching process includes: for the sub-images obtained by exposing the adjacent second segmented areas, extracting feature points and feature descriptors using the Canny algorithm; performing nearest neighbor matching of the feature descriptors and screening out reliable matching point pairs using a ratio test; minimizing the sum of the gradient amplitudes of the pixels through which the stitching line passes, and finding the path with the least obvious image structure; and completing the stitching of the exposed images using histogram matching in the overlapping areas.

10. The LDI image segmentation and exposure splicing method according to claim 9, characterized in that: In step S6, a comprehensive evaluation function for stitching quality is introduced to quantitatively evaluate the light intensity uniformity and edge sharpness of the overlapping area after image stitching; ； Where, Indicates the comprehensive score of splicing quality, with a value range of , the larger the value, the higher the quality; Indicates the light intensity uniformity weight, with a value of 0.6; Indicates the edge sharpness weight, the value is 0.4; Indicates the total number of pixels in the overlapping area; Indicates the The actual light intensity value of each pixel; Indicates the target light intensity value; Represents the gradient modulus and value of the edge area; when When , the splicing is judged to be qualified; when When , the histogram matching correction is triggered; when , the stitching quality is judged to be unqualified, and the corresponding sub-image needs to be re-exposed and stitched.

Images