Lens extinction control method based on machine vision and visual extinction device

Abstract

This application relates to the field of automation control technology and discloses a lens extinction control method and visual extinction device based on machine vision. This application utilizes images acquired by a vision camera combined with a deep learning model to accurately identify various defect types and key parameters on the lens surface, breaking through the limitations of traditional extinction in judging defects and significantly improving the accuracy and comprehensiveness of defect identification. At the same time, based on defect information, the printing parameters are dynamically adjusted and combined with real-time visual monitoring during the printing process to achieve closed-loop control of "accurate defect identification - dynamic parameter matching - real-time process control". This ensures that the extinction effect is adapted to the actual defect condition of the lens and that the thickness of the extinction layer is uniform and stable, effectively avoiding problems such as uneven extinction and incomplete defect coverage caused by traditional fixed parameter printing, ensuring lens extinction quality, reducing rework rate, and improving production efficiency.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, specifically to a lens extinction control method and visual extinction device based on machine vision. Background Technology

[0002] In the field of optical instruments, the lens, as a core optical component, directly determines the image quality and performance of the entire instrument. Matte finishing, a key process in lens manufacturing, involves forming a uniform matte coating on a specific area of ​​the lens surface (the target matte area). This effectively suppresses stray light reflection, reduces glare interference, and thus improves the lens's contrast and image clarity. It is widely used in the production of various optical products such as cameras, microscopes, and security monitoring equipment.

[0003] Currently, lens matte finishing primarily employs traditional manual operation or semi-automated inkjet printing methods. Manual operation relies on the operator's experience and judgment, using handheld inkjet printing tools to coat the lens surface. This not only makes it difficult to ensure uniform ink thickness and precise positioning of the matte area, but also results in significant fluctuations in product yield due to the instability of human operation. Furthermore, manual processing is inefficient and cannot meet the demands of large-scale industrial production. While semi-automated inkjet printing methods introduce mechanical printing mechanisms and complete the matte finishing process using preset, fixed printing parameters, significant technical bottlenecks still exist. Summary of the Invention

[0004] The main objective of this invention is to provide a lens extinction control method and a visual extinction device based on machine vision, which aims to solve the technical problems of low efficiency and poor effect of lens extinction processing in the prior art.

[0005] To achieve the above objectives, in a first aspect, this application provides a lens extinction control method based on machine vision, applied to a visual extinction device. The visual extinction device includes a printing carriage module and a worktable module. The printing carriage module includes a first running guide rail and a printhead, a vision camera, and a curing UV lamp disposed on the first running guide rail. The worktable module includes a second running guide rail and a vacuum adsorption stage disposed on the second running guide rail. The method includes:

[0006] Obtain the inkjet matte printing command;

[0007] The inkjet printing head is positioned according to the inkjet printing matte command, so that the inkjet printing head is aligned with the target matte area of ​​the lens to be matted.

[0008] The surface state image of the target matte area is acquired by the vision camera, and the defect type and defect parameters in the image are identified by a deep learning model; wherein, the defect type includes scratches, dents and discolored spots, and the defect parameters include defect depth, defect area and defect distribution density;

[0009] Based on the defect type and defect parameters, the preset initial printing parameters are dynamically adjusted in multiple dimensions to obtain the target printing parameters.

[0010] Ink is sprayed onto the lens to be de-luminous according to the target printing parameters, and a vision camera is activated in real time during the ink spraying process to dynamically adjust the layer thickness.

[0011] In one possible implementation, controlling the printhead positioning according to the printhead matting command, so that the printhead is aligned with the target matting area of ​​the lens to be matted, includes:

[0012] According to the inkjet matting command, the vacuum adsorption stage is controlled to vacuum adsorb and fix the lens to be matted, and the vacuum adsorption stage is driven to run to the preset matting area.

[0013] The printing carriage module is controlled to perform a printhead positioning operation within the preset matte area, so that the printhead is aligned with the target matte area of ​​the lens to be matted.

[0014] In one possible implementation, controlling the printing carriage module to perform a printhead positioning operation within the preset matte finish area, aligning the printhead with the target matte finish area of ​​the lens to be matted, includes:

[0015] The visual camera is used to acquire the marking features of the lens to be extinct, and the first coordinates of the marking features in the visual camera coordinate system are extracted.

[0016] Based on the pre-calibrated transformation matrix between the visual camera coordinate system and the printhead coordinate system, the first coordinate is converted into the second coordinate in the printhead coordinate system;

[0017] Based on the second coordinate, the printhead is controlled to move to the target position corresponding to the target matte area;

[0018] The relative positional deviation between the printhead and the target matte area is monitored in real time by the vision camera. The position of the matte lens is then finely adjusted based on the relative positional deviation until the printhead is aligned with the target matte area.

[0019] In one possible implementation, the fine-tuning of the position of the lens to be extinct based on the relative positional deviation includes:

[0020] Based on the relative position deviation, determine the fine-tuning direction and fine-tuning displacement required for the lens to be ablated;

[0021] The distribution of adsorption force of the vacuum adsorption stage is adjusted according to the fine-tuning direction and fine-tuning displacement, so that the vacuum adsorption stage generates different adsorption forces on different areas of the lens, and the deviation force formed by the adsorption force difference drives the lens to produce a small displacement on the vacuum adsorption stage.

[0022] The adjustment effect of the lens position is monitored in real time by a vision camera until the relative positional deviation between the lens to be matted and the print head is less than a preset threshold, thus completing the fine adjustment.

[0023] In one possible implementation, the vacuum adsorption stage includes a placement area for the lens to be exfoliated, the placement area having at least a first adsorption row of holes and a second adsorption row of holes spaced apart from each other. The step of adjusting the adsorption force distribution of the vacuum adsorption stage according to the fine-tuning direction and fine-tuning displacement, so that the vacuum adsorption stage generates differentiated adsorption forces on different areas of the lens, includes:

[0024] The relative size relationship between the first adsorption pore and the second adsorption pore, as well as the difference in adsorption force, are determined based on the fine-tuning direction and the fine-tuning displacement.

[0025] Based on the relative size relationship and the adsorption force difference, the output power of the vacuum pumps corresponding to the first and second adsorption holes is controlled to adjust their actual adsorption forces.

[0026] In one possible implementation, the step of using a deep learning model to identify the type of defect and defect parameters in an image includes:

[0027] The acquired surface state image is input into a pre-trained defect recognition neural network, wherein the defect recognition neural network is a semantic segmentation network based on an encoder-decoder structure;

[0028] The defect recognition neural network outputs a defect segmentation map containing defect category labels and pixel-level location information.

[0029] Based on the defect segmentation map, the pixel area of ​​various defects is counted to determine the defect area, and the proportion of the defect area to the target extinction area is calculated to determine the defect distribution density.

[0030] The depth information of the surface state image is extracted, and the depth features of the defect region are obtained from the depth information to determine the defect depth.

[0031] In one possible implementation, the step of dynamically adjusting the preset initial printing parameters in multiple dimensions based on the defect type and defect parameters to obtain the target printing parameters includes:

[0032] The corresponding printing strategy is matched according to the defect type. Among them, scratch defects correspond to the first strategy of increasing ink filling amount, dent defects correspond to the second strategy of increasing printing layer thickness, and discoloration spot defects correspond to the third strategy of reducing ink coverage density.

[0033] The printing parameter adjustment amount is calculated based on the defect parameters, wherein the printing path and ink volume are dynamically adjusted according to the defect area and defect distribution density, and the number of printing layers is dynamically adjusted according to the defect depth.

[0034] The printing strategy is integrated with the printing parameter adjustment amount to collaboratively optimize the ink volume, printing speed, printing path and layer thickness in the initial printing parameters, thereby generating the target printing parameters.

[0035] In one possible implementation, during the inkjet printing process, a vision camera is activated in real time to dynamically adjust the layer thickness so that the printing thickness meets preset requirements, including:

[0036] After each layer of ink is printed, the three-dimensional morphology data of the current printed layer is acquired by the vision camera based on confocal chromatography or spectral interferometry.

[0037] From the three-dimensional topography data, the height difference between the printed area and the unprinted reference surface is extracted, and this height difference is used as the actual thickness of the current printed layer.

[0038] The actual thickness is compared with the preset thickness threshold of the corresponding layer to obtain the thickness deviation;

[0039] The printing parameters of the next printing layer are dynamically adjusted based on the thickness deviation.

[0040] Repeat the above steps until all layers of printing are completed, so that the cumulative printing thickness reaches the preset requirement.

[0041] In one possible implementation, before acquiring the three-dimensional morphology data of the current printed layer using the vision camera based on confocal chromatography or spectral interferometry, the method further includes:

[0042] The UV curing lamp is turned on to perform UV curing treatment on the ink layer that has been printed.

[0043] Secondly, embodiments of this application also provide a visual extinction device, comprising:

[0044] A printing carriage module, the printing carriage module including a first running guide rail and a print head, a vision camera and a curing UV lamp disposed on the first running guide rail;

[0045] The worktable module includes a second running guide rail and a vacuum adsorption stage disposed on the second running guide rail.

[0046] Memory, the memory being used to store program code; and

[0047] A processor, the processor being configured to invoke the program code to execute the method as described in the first aspect.

[0048] Unlike existing technologies, this application provides a lens de-flare control method based on machine vision. The method first acquires a de-flare printing command, then controls the printing head positioning according to the command, aligning the printing head with the target de-flare area of ​​the lens to be de-flareed. Next, a vision camera acquires an image of the surface state of the target de-flare area, and a deep learning model identifies the defect types and parameters in the image. The defect types include scratches, dents, and discolored spots, and the defect parameters include defect depth, defect area, and defect distribution density. Based on the defect types and parameters, preset initial printing parameters are dynamically adjusted in multiple dimensions to obtain target printing parameters. Finally, ink is printed onto the lens to be de-flareed according to the target printing parameters, and during the ink printing process, the vision camera is activated in real-time to dynamically adjust the layer thickness. Thus, this application utilizes images acquired by a visual camera combined with a deep learning model to accurately identify various defect types and key parameters on the lens surface, breaking through the limitations of traditional exfoliation methods in defect judgment and significantly improving the accuracy and comprehensiveness of defect identification. Simultaneously, by dynamically adjusting printing parameters based on defect information and coordinating with real-time visual monitoring during the printing process, a closed-loop control of "accurate defect identification - dynamic parameter matching - real-time process control" is achieved. This ensures that the exfoliation effect adapts to the actual defect conditions of the lens and that the thickness of the exfoliation layer is uniform and stable, effectively avoiding problems such as uneven exfoliation and incomplete defect coverage caused by traditional fixed-parameter printing. This guarantees the quality of lens exfoliation, reduces rework rates, and improves production efficiency. Attached Figure Description

[0049] 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of the structure of the visual extinction device in some embodiments of this application;

[0051] Figure 2 This is a flowchart illustrating a lens extinction control method based on machine vision in some embodiments of this application;

[0052] Figure 3 This is a flowchart illustrating step S200 of the lens extinction control method based on machine vision in some embodiments of this application;

[0053] Figure 4 This is a schematic diagram of the hardware structure of the visual extinction device in some embodiments of this application.

[0054] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0055] 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0056] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0057] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the term "and / or" throughout the text includes three solutions; taking A and / or B as an example, it includes technical solution A, technical solution B, and a technical solution that simultaneously satisfies A and B. Furthermore, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of a person skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0058] In the field of optical instruments, the lens, as a core optical component, directly determines the image quality and performance of the entire instrument. Matte finishing, a key process in lens manufacturing, involves forming a uniform matte coating on a specific area of ​​the lens surface (the target matte area). This effectively suppresses stray light reflection, reduces glare interference, and thus improves the lens's contrast and image clarity. It is widely used in the production of various optical products such as cameras, microscopes, and security monitoring equipment.

[0059] Currently, lens matte finishing primarily employs traditional manual operation or semi-automated inkjet printing methods. Manual operation relies on the operator's experience and judgment, using handheld inkjet printing tools to coat the lens surface. This not only makes it difficult to ensure uniform ink thickness and precise positioning of the matte area, but also results in significant fluctuations in product yield due to the instability of human operation. Furthermore, manual processing is inefficient and cannot meet the demands of large-scale industrial production. While semi-automated inkjet printing methods introduce mechanical printing mechanisms and complete the matte finishing process using preset, fixed printing parameters, significant technical bottlenecks still exist.

[0060] To address the aforementioned technical problems, this application provides a lens extinction control method based on machine vision. This method can be applied to visual extinction devices, such as... Figure 1 As shown, the visual matting device of this application includes a printing carriage module 100, a worktable module 200, and a cleaning component 300. The printing carriage module 100 is the core execution component, undertaking key operations such as ink printing, visual inspection, and matting layer curing. The worktable module 200 is used to accurately support and move the lens to be matted, working in conjunction with the printing carriage module to complete the full-area matting process. The cleaning component 300 is used to clean the printing components in a timely manner, ensuring the continuity and stability of the printing operation.

[0061] More specifically, the inkjet printing carriage module 100 in this application serves as the core operating unit, and its first running guide rail 110 provides a precise single-axis or multi-axis linear running track for each functional component, ensuring the stability of component displacement and positioning accuracy, and adapting to the requirements of high-precision matte finishing operations. The printhead 120, mounted on the first running guide rail 110, is the actuator for ink printing. It can precisely control the ink output, printing speed, and printing distance according to the target printing parameters to achieve uniform printing on the target matte area of ​​the lens. The vision camera 130 undertakes the dual tasks of image acquisition and real-time monitoring. On the one hand, it acquires images of the surface state of the area to be matted before printing to provide data support for defect identification. On the other hand, it captures images of the layered printing in real time during the printing process to dynamically adjust the printing thickness, thereby ensuring the accuracy of defect identification and thickness detection. The curing UV lamp is used to quickly cure the matte layer after ink printing. By controlling the power and irradiation time of the UV lamp, it ensures that the matte layer is fully cured and has stable performance, avoiding problems such as matte layer peeling due to insufficient curing or lens performance damage due to over-curing.

[0062] The worktable module 200 serves as the lens support and displacement mechanism. The second running guide rail 210 cooperates with the first running guide rail 110 of the printing carriage module to achieve coordinated operation in both the X and Y axes, driving the lens to be exfoliated to complete multi-directional, full-area exfoliation work, meeting the processing needs of lenses of different sizes. The vacuum adsorption stage 220, located on the second running guide rail 210, is used to securely fix the lens to be exfoliated. By extracting air between the adsorption stage and the lens, a negative pressure is created, ensuring the lens is tightly adhered to the surface of the adsorption stage. This prevents lens displacement during printing, which could lead to printing deviations. The adsorption force of the adsorption stage can be adjusted according to the lens material and size, adapting to the fixing requirements of lenses made of different materials such as glass and resin.

[0063] The cleaning assembly 300, as an auxiliary mechanism to ensure the performance of the printhead, is used to rinse the nozzles of the printhead 120 during printing intervals or after printing, removing residual ink from the nozzles to prevent ink from drying and clogging the nozzles, which could lead to uneven printing volume or ink interruptions. The cleaning assembly can use a combination of high-pressure spraying and ultrasonic cleaning. First, high-pressure water is used to rinse away residual ink on the nozzle surface, and then ultrasonic vibration is used to remove ink residue from the tiny channels inside the nozzles. After cleaning, a drying device can be used to dry the printhead to ensure smooth subsequent printing operations.

[0064] like Figures 1-3 As shown, the following explanation uses a visual exfoliation device to execute this machine vision-based lens exfoliation control method as an example. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order. Please refer to the appendix. Figure 2 The method includes the following steps S100-S500:

[0065] Step S100: Obtain the inkjet matting instruction;

[0066] The printing matte command can be triggered automatically or manually by the operator. In actual operation scenarios, the operator can click the lens printing matte start button on the operation screen to trigger the printing matte command. At this time, the controller of the visual matte equipment receives the printing matte command.

[0067] Step S200: Control the printing head to position according to the printing matte command, so that the printing head is aligned with the target matte area of ​​the lens to be matted.

[0068] After obtaining the printing matting command in step S100, the visual matting device can drive the printing head to complete the positioning according to the command, ensuring that it is aligned with the target matting area of ​​the lens to be matted.

[0069] The target matte area refers to the specific area on the lens that needs to be matted, such as the edges of the lens or a local area with minor defects.

[0070] Step S300: Acquire surface state images of the target matte area using the vision camera, and identify defect types and defect parameters in the images using a deep learning model; wherein, the defect types include scratches, dents and discolored spots, and the defect parameters include defect depth, defect area and defect distribution density;

[0071] After the printhead is positioned, the visual matting equipment first acquires an image of the surface condition of the target matting area using a visual camera. Then, a pre-trained deep learning model is used to analyze the image, accurately identifying the defect types and parameters. This data on defect types and parameters provides an accurate basis for targeted adjustments to subsequent printing parameters, ensuring the matting effect.

[0072] The defect types specifically include surface imperfections such as scratches, dents, and discolored spots; the defect parameters include quantifiable defect characteristic indicators such as defect depth, defect area, and defect distribution density.

[0073] Scratches are linear damage to the lens surface caused by friction or contact with hard objects. They can be minor shallow scratches (affecting only the surface finish) or deeper scratches (penetrating the surface layer), which can lead to disordered light refraction.

[0074] A dent is a localized depression on the lens surface caused by impact or pressure. The dented area can change the path of light propagation and may also accumulate impurities, affecting optical performance.

[0075] Discoloration spots refer to discolored deposits or original spots on the lens surface caused by contamination, uneven materials, or processing residues. These spots can cause abnormal local light absorption / reflection, resulting in light spots and shadows.

[0076] Step S400: Based on the defect type and defect parameters, the preset initial printing parameters are dynamically adjusted in multiple dimensions to obtain the target printing parameters;

[0077] After obtaining the defect type and defect parameters of the target matting area in step S300, the visual matting device can dynamically adjust the preset initial printing parameters in multiple dimensions based on these data, and finally determine the target printing parameters.

[0078] In actual operation scenarios, the severity of defects can be determined first: if any of the defects, such as the depth, area, or density, exceeds a preset threshold, the lens is deemed unsuitable for matte finishing. If the defects are minor (all parameters are within the threshold range), the initial printing parameters can be dynamically adjusted based on the type of defect and the specific parameter values, using multiple dimensions such as printing pressure, ink flow rate, and coating thickness, to form suitable target printing parameters, providing a reliable basis for subsequent precise printing of matte ink.

[0079] Step S500: Print ink onto the lens to be matte according to the target printing parameters, and during the ink printing process, activate the vision camera in real time to dynamically adjust the layer thickness.

[0080] After determining the target printing parameters in step S400, the visual matting device can perform ink printing on the lens to be matted according to the target printing parameters. Considering that the actual ink thickness may deviate from the preset value due to factors such as fluctuations in equipment accuracy and changes in ink characteristics during the printing process, this application simultaneously starts a visual camera for real-time monitoring at this stage. By dynamically adjusting the thickness of the printed layer layer, it ensures that the final coating thickness is consistent with the target requirements.

[0081] It should be noted that the vision camera in this application can be a high-precision industrial RGBD camera, or a combination of multiple RGB industrial cameras capable of detecting depth information. The type of vision camera is not limited here.

[0082] Based on this, the lens matte control method of this application first obtains a matte printing command, then controls the printing head positioning according to the matte printing command, so that the printing head is aligned with the target matte area of ​​the lens to be matted; next, the surface state image of the target matte area is acquired by the vision camera, and the defect type and defect parameters in the image are identified by a deep learning model; wherein, the defect type includes scratches, dents and discolored spots, and the defect parameters include defect depth, defect area and defect distribution density; then, based on the defect type and defect parameters, the preset initial printing parameters are dynamically adjusted in multiple dimensions to obtain the target printing parameters; finally, ink is printed on the lens to be matted according to the target printing parameters, and during the ink printing process, the vision camera is activated in real time to dynamically adjust the layer thickness. Thus, this application utilizes images acquired by a visual camera combined with a deep learning model to accurately identify various defect types and key parameters on the lens surface, breaking through the limitations of traditional exfoliation methods in defect judgment and significantly improving the accuracy and comprehensiveness of defect identification. Simultaneously, by dynamically adjusting printing parameters based on defect information and coordinating with real-time visual monitoring during the printing process, a closed-loop control of "accurate defect identification - dynamic parameter matching - real-time process control" is achieved. This ensures that the exfoliation effect adapts to the actual defect conditions of the lens and that the thickness of the exfoliation layer is uniform and stable, effectively avoiding problems such as uneven exfoliation and incomplete defect coverage caused by traditional fixed-parameter printing. This guarantees the quality of lens exfoliation, reduces rework rates, and improves production efficiency.

[0083] In one embodiment, such as Figure 3 As shown, step S200: controlling the printhead positioning according to the printhead matting command, so that the printhead is aligned with the target matting area of ​​the lens to be matted, including:

[0084] S210. According to the inkjet matting command, control the vacuum adsorption stage to perform vacuum adsorption and fixation on the lens to be matted, and drive the vacuum adsorption stage to run to the preset matting area.

[0085] S220. Control the printing carriage module to perform a printing head positioning operation within the preset matte area, so that the printing head is aligned with the target matte area of ​​the lens to be matted.

[0086] Specifically, the visual matting device first controls the vacuum adsorption stage to vacuum adsorb and fix the lens to be matted according to the printing matting instruction. This operation uses negative pressure adsorption to firmly fix the lens on the adsorption stage, preventing the lens from being affected by positional displacement during subsequent printing. At the same time, it drives the vacuum adsorption stage to the preset matting area. This preset area is the printing operation reference area preset by the system, which provides a spatial reference for the precise positioning of the print head and ensures that the lens enters the printable operating range.

[0087] Subsequently, the printing carriage module is controlled to perform a printhead positioning operation within the preset matte area, aligning the printhead with the target matte area of ​​the lens to be matted. The first running guide rail of the printing carriage module can be a single-axis drive rail or a multi-axis drive rail; there is no restriction here.

[0088] Thus, this embodiment of the application achieves precise positioning of the printhead on the target matte area by means of a step-by-step operation of "lens fixing and area transfer - precise alignment of the printhead" according to the print matte command, providing a foundation for subsequent high-quality matte printing, thereby improving the accuracy and consistency of lens matte processing.

[0089] In one embodiment, step S220: controlling the printing carriage module to perform a printhead positioning operation within the preset matte area, so that the printhead is aligned with the target matte area of ​​the lens to be matted, includes: acquiring the marking features of the lens to be matted through the vision camera, and extracting the first coordinates of the marking features in the vision camera coordinate system; converting the first coordinates into a second coordinate in the printhead coordinate system according to a pre-calibrated transformation matrix between the vision camera coordinate system and the printhead coordinate system; controlling the printhead to move to the target position corresponding to the target matte area based on the second coordinates; monitoring the relative position deviation between the printhead and the target matte area in real time through the vision camera, and fine-tuning the position of the lens to be matted according to the relative position deviation until the printhead is aligned with the target matte area.

[0090] Specifically, the first step is to acquire the marking features of the lens to be ablated using a vision camera. These marking features can be pre-defined positioning marks on the lens surface (such as indentations of a specific shape, scribing lines, or color difference markings). After capturing an image containing these markings, the vision camera uses an image recognition algorithm to extract the first coordinate of the marking features in the vision camera's coordinate system. This coordinate accurately reflects the spatial position of the markings within the camera's field of view, providing a clear original positional reference for subsequent positioning operations and ensuring that the entire positioning process has a traceable reference target.

[0091] Subsequently, based on the pre-calibrated transformation matrix between the vision camera coordinate system and the printhead coordinate system, the first coordinate is converted into the second coordinate in the printhead coordinate system. Due to the physical distance and angular differences in the installation positions of the vision camera and the printhead on the equipment, their measurement coordinate systems have inherent deviations. The transformation matrix, however, is a pre-determined spatial mapping relationship calibrated using a standard calibration plate (e.g., calculated by collecting multiple sets of corresponding point coordinates). Its core function is to accurately convert the two-dimensional / three-dimensional position information detected by the camera into coordinate parameters recognizable by the printhead motion control system, thereby eliminating spatial errors across equipment coordinate systems and providing precise target coordinate guidance for the movement of the printhead.

[0092] Next, based on the second coordinate, the printing carriage module drives the printing head to move to the target position corresponding to the target matte area, completing the coarse positioning operation.

[0093] Finally, a vision camera monitors the relative positional deviation between the printhead and the target matte area in real time. Specifically, this is achieved by setting an auxiliary reference mark on the printhead and comparing the pixel coordinate difference between this auxiliary reference mark and the target matte area mark in the real-time image to calculate the actual spatial deviation value. Based on this deviation value, the vacuum suction stage or print carriage module is driven to fine-tune the position of the matte lens until the vision camera detects that the relative positional deviation between the printhead and the target matte area is less than a preset threshold, indicating that the printhead is precisely aligned with the target matte area.

[0094] It should be noted that if the marker feature is not set at the center of the target matte area (e.g., 2mm to the left and 1mm above the target matte area, or at a fixed point on the edge of the lens), the pre-stored parameter of "relative offset between the marker position and the center of the target matte area" must be called first. This offset is then added to the extracted first coordinates to obtain the corrected coordinates of the target matte area center in the visual camera coordinate system. The purpose of this step is to convert the marker's position information into the actual center position information of the matte area that needs to be aligned through fixed offset compensation, avoiding positioning deviations caused by the marker not being set at the center. If the marker is directly set at the center of the target matte area, this offset compensation step can be skipped, and the first coordinate can be directly used as the camera coordinate system coordinates of the target matte area center.

[0095] Thus, this embodiment of the application achieves high-precision alignment of the print head with the target matte area through a step-by-step positioning strategy of "feature recognition - coordinate transformation - coarse positioning - real-time fine adjustment" combined with visual inspection and coordinate system transformation technology, thereby ensuring the uniformity and consistency of subsequent matte printing and improving the quality of lens matte processing.

[0096] In one embodiment, the step of fine-tuning the position of the lens to be matted based on the relative position deviation includes: determining the required fine-tuning direction and displacement of the lens to be matted based on the relative position deviation; adjusting the adsorption force distribution of the vacuum adsorption stage according to the fine-tuning direction and displacement, so that the vacuum adsorption stage generates differentiated adsorption forces on different areas of the lens, and using the deviation force formed by the adsorption force difference to drive the lens to produce a small displacement on the vacuum adsorption stage; and monitoring the adjustment effect of the lens position in real time through a vision camera until the relative position deviation between the lens to be matted and the printhead is less than a preset threshold, thus completing the fine-tuning.

[0097] Specifically, based on the relative positional deviation between the printhead and the target matte area of ​​the lens to be matted, as monitored by the vision camera, the required fine-tuning direction and displacement of the lens to be matted are determined, providing precise target parameters for subsequent adsorption force adjustment.

[0098] Subsequently, based on the determined fine-tuning direction and displacement, the adsorption force distribution of the vacuum adsorption stage is adjusted. The vacuum adsorption stage includes a placement area for the lens to be fine-tuned, which has at least two spaced-apart first and second adsorption rows, serving as the core execution units for achieving differentiated suction. Specifically, during adjustment, the relative magnitude of the adsorption forces of the first and second adsorption rows and the difference in adsorption forces are first determined based on the fine-tuning direction and displacement. For example, if the lens needs fine-tuning along the negative X-axis, and the first adsorption row is located in the positive X-axis direction while the second adsorption row is located in the negative X-axis direction, then the relative relationship of "the suction force of the first adsorption row is greater than that of the second adsorption row" must be set. Next, based on this relative magnitude and the difference in adsorption forces, the output power of the vacuum pumps corresponding to the first and second adsorption rows is controlled. By adjusting the power, the actual adsorption forces of the two rows are precisely matched, allowing the lens to form a directional differentiated suction field on the surface of the adsorption stage. This suction difference generates a deviation force along the fine-tuning direction. Due to the tiny air gap between the lens and the surface of the adsorption stage (or the use of a low-friction adsorption pad), the deviation force can drive the lens to smoothly produce a tiny displacement on the adsorption stage, achieving precise directional adjustment. The amount of fine-tuning displacement is precisely controlled by adjusting the magnitude and duration of the suction difference. The greater the suction difference and the longer the duration, the greater the displacement, and vice versa.

[0099] Finally, a vision camera monitors the lens position adjustment in real time, forming a closed-loop feedback adjustment. The vision camera captures the relative positions of the lens marking features and the auxiliary reference mark on the printhead at high frequency (e.g., 30 frames per second). The system calculates the deviation between the two in real time and compares it with a preset threshold (e.g., ±0.01mm). If the deviation is still greater than the threshold, the adsorption force distribution is repeatedly adjusted based on the latest deviation data to fine-tune the lens position; if the deviation is less than or equal to the threshold, each adsorption unit is immediately controlled to restore a uniform adsorption force distribution, stabilizing the lens in the adjusted precise position, completing the fine-tuning operation. This closed-loop process effectively avoids the problem of over- or under-adjustment in a single operation, ensuring that the fine-tuning accuracy meets the requirements of high-precision printing.

[0100] Thus, this application embodiment uses a fine-tuning strategy of "deviation analysis - suction difference drive - closed-loop monitoring" to achieve fine-tuning of lens displacement by utilizing the differentiated suction control of the vacuum adsorption stage. This avoids the risk of lens damage that may be caused by mechanical clamping fine-tuning, while ensuring the alignment accuracy between the lens and the printing head after fine-tuning. This provides a solid guarantee for subsequent high-quality matte printing, thereby improving the yield rate of lens matte processing.

[0101] In one embodiment, the step of identifying defect types and defect parameters in an image using a deep learning model includes:

[0102] The acquired surface state image is input into a pre-trained defect recognition neural network, wherein the defect recognition neural network is a semantic segmentation network based on an encoder-decoder structure;

[0103] The defect recognition neural network outputs a defect segmentation map containing defect category labels and pixel-level location information.

[0104] Based on the defect segmentation map, the pixel area of ​​various defects is counted to determine the defect area, and the proportion of the defect area to the target extinction area is calculated to determine the defect distribution density.

[0105] The depth information of the surface state image is extracted, and the depth features of the defect region are obtained from the depth information to determine the defect depth.

[0106] Specifically, in this embodiment, the surface state image of the lens to be inspected is first input into a pre-trained defect recognition neural network, which uses a semantic segmentation network based on an encoder-decoder structure as the core recognition model. Before being put into use, the defect recognition neural network has been trained with a large number of sample images labeled with defect types and locations, and has a mature ability to recognize common defects (such as scratches, bubbles, missing prints, color differences, etc.), laying the foundation for the accuracy of defect recognition.

[0107] The defect recognition neural network then outputs a defect segmentation map containing defect category labels and pixel-level location information. During the network's decoding process, each pixel in the image is classified and labeled as either "background" or a specific type of defect (such as "scratches" or "bubbles"). In the final defect segmentation map, different types of defects are labeled with different colors or grayscale values, along with corresponding category labels. The pixel-level boundaries of each defect are also outlined, clearly showing the specific location distribution of defects within the target extinction area, providing an intuitive and accurate image basis for subsequent defect parameter calculations.

[0108] Next, based on the output defect segmentation map, the defect area and defect distribution density parameters are calculated. To determine the defect area, first, according to the preset "pixel-actual size" mapping ratio during image acquisition (e.g., pre-calibrating 100 pixels to 1 mm²), the total number of pixels occupied by various defects in the segmentation map is counted, and then the actual area of ​​the defect is obtained by converting it using the mapping ratio. To determine the defect distribution density, first, the total pixel area of ​​all defect regions is counted and converted into an actual area, and then the ratio of this area to the total actual area of ​​the target extinction region is calculated to obtain the defect distribution density. This parameter can intuitively reflect the density of defects in the extinction region.

[0109] Finally, depth information is extracted from the surface condition image, and depth features of the defect region are obtained from the depth information to determine the defect depth. The depth information of the surface condition image can be acquired synchronously by a depth camera or extracted from the image's grayscale gradient and texture variations.

[0110] Thus, this application embodiment adopts a step-by-step recognition strategy of "image input - segmentation map output - two-dimensional parameter calculation - three-dimensional depth extraction", combined with the accurate recognition capability of semantic segmentation network based on encoder-decoder structure. It can not only accurately distinguish the defect type and locate the defect location, but also comprehensively obtain key parameters such as defect area, distribution density, and depth. This provides a reliable basis for subsequent defect level determination and rework repair strategy formulation, thereby improving the quality control level of lens exfoliation processing.

[0111] In one embodiment, the step of: dynamically adjusting the preset initial printing parameters in multiple dimensions according to the defect type and defect parameters to obtain the target printing parameters includes:

[0112] The corresponding printing strategy is matched according to the defect type. Among them, scratch defects correspond to the first strategy of increasing ink filling amount, dent defects correspond to the second strategy of increasing printing layer thickness, and discoloration spot defects correspond to the third strategy of reducing ink coverage density.

[0113] The printing parameter adjustment amount is calculated based on the defect parameters, wherein the printing path and ink volume are dynamically adjusted according to the defect area and defect distribution density, and the number of printing layers is dynamically adjusted according to the defect depth.

[0114] The printing strategy is integrated with the printing parameter adjustment amount to collaboratively optimize the ink volume, printing speed, printing path and layer thickness in the initial printing parameters, thereby generating the target printing parameters.

[0115] Specifically, this application first matches a targeted printing strategy based on the defect type identified by the deep learning model. Different defects have different causes and morphological characteristics, requiring precise repair through differentiated strategies: For scratch defects, the core issue is groove-like depressions caused by missing surface material, so the first strategy is to increase the ink filling amount. For depression defects, their depth is usually greater than ordinary scratches, and single filling is insufficient to guarantee flatness, so the second strategy is to increase the printing layer thickness. This reduces the total number of layers by increasing the thickness of each single layer, while ensuring that each layer of ink fully covers the depression area. For discolored spot defects, their surface flatness is basically normal but the color is abnormal, so the third strategy is to reduce the ink coverage density. This covers the discolored area while avoiding surface protrusion caused by ink accumulation, ensuring a uniform matte finish.

[0116] Subsequently, based on the identified defect parameters, the specific adjustment amount of each printing parameter is accurately calculated. For example, the printing path and ink volume are dynamically adjusted according to the defect area and defect distribution density, and the number of printing layers is dynamically adjusted according to the defect depth.

[0117] Finally, the matched printing strategy and the calculated adjustment amount are integrated to collaboratively optimize the ink volume, printing speed, printing path and layer thickness in the initial printing parameters, and generate the target printing parameters.

[0118] Thus, this embodiment of the application achieves precise adaptation between printing parameters and defect status through a type matching strategy, parameter calculation adjustment amount, and multi-dimensional collaborative optimization adjustment logic. This ensures the defect repair effect while avoiding material waste and efficiency reduction caused by over-printing, further improving the quality and economy of lens matte finishing.

[0119] In one embodiment, during the inkjet printing process, a vision camera is activated in real time to dynamically adjust the layer thickness, including:

[0120] After each layer of ink is printed, the three-dimensional morphology data of the current printed layer is acquired by the vision camera based on confocal chromatography or spectral interferometry.

[0121] From the three-dimensional topography data, the height difference between the printed area and the unprinted reference surface is extracted, and this height difference is used as the actual thickness of the current printed layer.

[0122] The actual thickness is compared with the preset thickness threshold of the corresponding layer to obtain the thickness deviation;

[0123] The printing parameters of the next printing layer are dynamically adjusted based on the thickness deviation.

[0124] Repeat the above steps until all layers of printing are completed, so that the cumulative printing thickness reaches the preset requirement.

[0125] Specifically, after each layer of ink is printed, the three-dimensional morphology data of the current printed layer can be acquired by a vision camera based on confocal chromatography or spectral interferometry. Then, the height difference between the printed area and the unprinted reference surface is extracted from the three-dimensional morphology data, and this height difference is used as the actual thickness of the current printed layer. Next, the actual thickness is compared with the preset thickness threshold of the corresponding layer to obtain the thickness deviation. Then, the printing parameters of the next printed layer are dynamically adjusted based on the thickness deviation. The above steps are repeated until all layers are printed, so that the cumulative printing thickness reaches the preset requirement.

[0126] It should be noted that before acquiring the three-dimensional morphology data of the current printed layer using a vision camera based on confocal chromatography or spectral interferometry, the UV curing lamp can be turned on to perform UV irradiation curing on the ink layer that has been printed, ensuring the authenticity and accuracy of the actual thickness detection results and avoiding the adjustment of the next coating parameters based on erroneous data.

[0127] Thus, this embodiment of the application achieves real-time quality control of the printing process through a closed-loop control strategy of layered three-dimensional detection, thickness deviation calculation, and dynamic parameter correction, avoiding the accumulation of multiple layers of deviation that could lead to the final total thickness not meeting the requirements.

[0128] like Figure 4 As shown, Figure 4 The diagram below shows the hardware structure of a visual light-reducing device in some embodiments of this application. The visual light-reducing device provided in the embodiments of this application includes a memory 1000 and a processor 2000. The memory 1000 is used to store computer-readable instructions, and the processor 2000 is used to call the computer-readable instructions to execute the lens light-reducing control method based on machine vision as described above.

[0129] The processor 2000 provides computing and control capabilities to control the visual matting device to perform corresponding tasks. For example, it controls the visual matting device to perform the lens matting control method based on machine vision in any of the above method embodiments. The method includes: acquiring a printing matting command; controlling the printing head to position itself according to the printing matting command, so that the printing head is aligned with the target matting area of ​​the lens to be matted; acquiring a surface state image of the target matting area through the vision camera, and using a deep learning model to identify the defect type and defect parameters in the image; wherein the defect type includes scratches, dents, and discolored spots, and the defect parameters include defect depth, defect area, and defect distribution density; dynamically adjusting the preset initial printing parameters in multiple dimensions based on the defect type and defect parameters to obtain the target printing parameters; printing ink on the lens to be matted according to the target printing parameters, and during the ink printing process, activating the vision camera in real time to dynamically adjust the layer thickness.

[0130] The processor 2000 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0131] The memory 1000, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the lens extinction control method based on machine vision in the embodiments of this application. The processor 2000 can implement the lens extinction control method based on machine vision in any of the above method embodiments by running the non-transitory software programs, instructions, and modules stored in the memory 1000.

[0132] Specifically, memory 1000 may include volatile memory (VM), such as random access memory (RAM); memory 1000 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), solid-state drive (SSD), or other non-transitory solid-state storage devices; memory 1000 may also include combinations of the above types of memory.

[0133] In summary, the visual light-reducing device of this application adopts the technical solution of any of the above-mentioned machine vision-based lens light-reducing control method embodiments. Therefore, it has at least the beneficial effects brought about by the technical solutions of the above embodiments, which will not be elaborated here.

[0134] This application also provides a computer-readable storage medium, such as a memory including program code, which can be executed by a processor to complete the lens extinction control method based on machine vision in the above embodiments. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CDROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0135] This application also provides a computer program product comprising one or more lines of program code stored in a computer-readable storage medium. The processor of the early warning system reads the program code from the computer-readable storage medium and executes the program code to complete the steps of the machine vision-based lens light-reducing control method provided in the above embodiments.

[0136] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program or program code related to hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0137] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0138] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software and a general-purpose hardware platform, or of course, using hardware. Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0139] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A lens extinction control method based on machine vision, applied to a visual extinction device, characterized in that, The visual matting device includes a printing carriage module and a worktable module. The printing carriage module includes a first running guide rail and a printhead, a visual camera, and a curing UV lamp mounted on the first running guide rail. The worktable module includes a second running guide rail and a vacuum adsorption stage mounted on the second running guide rail. The method includes: Obtain the inkjet matte printing command; The inkjet printing head is positioned according to the inkjet printing matte command, so that the inkjet printing head is aligned with the target matte area of ​​the lens to be matted. The surface state image of the target matte area is acquired by the vision camera, and the defect type and defect parameters in the image are identified by a deep learning model; wherein, the defect type includes scratches, dents and discolored spots, and the defect parameters include defect depth, defect area and defect distribution density; Based on the defect type and defect parameters, the preset initial printing parameters are dynamically adjusted in multiple dimensions to obtain the target printing parameters. Ink is sprayed onto the lens to be de-luminous according to the target printing parameters, and a vision camera is activated in real time during the ink spraying process to dynamically adjust the layer thickness.

2. The lens extinction control method based on machine vision as described in claim 1, characterized in that, The step of controlling the printhead positioning according to the printhead matting command, so that the printhead is aligned with the target matting area of ​​the lens to be matted, includes: According to the inkjet matting command, the vacuum adsorption stage is controlled to vacuum adsorb and fix the lens to be matted, and the vacuum adsorption stage is driven to run to the preset matting area. The printing carriage module is controlled to perform a printhead positioning operation within the preset matte area, so that the printhead is aligned with the target matte area of ​​the lens to be matted.

3. The lens extinction control method based on machine vision as described in claim 2, characterized in that, The control of the printing carriage module to perform a printhead positioning operation within the preset matte area, aligning the printhead with the target matte area of ​​the lens to be matted, includes: The visual camera is used to acquire the marking features of the lens to be extinct, and the first coordinates of the marking features in the visual camera coordinate system are extracted. Based on the pre-calibrated transformation matrix between the visual camera coordinate system and the printhead coordinate system, the first coordinate is converted into the second coordinate in the printhead coordinate system; Based on the second coordinate, the printhead is controlled to move to the target position corresponding to the target matte area; The relative positional deviation between the printhead and the target matte area is monitored in real time by the vision camera. The position of the matte lens is then finely adjusted based on the relative positional deviation until the printhead is aligned with the target matte area.

4. The lens extinction control method based on machine vision as described in claim 3, characterized in that, The step of fine-tuning the position of the lens to be extinct based on the relative positional deviation includes: Based on the relative position deviation, determine the fine-tuning direction and fine-tuning displacement required for the lens to be ablated; The distribution of adsorption force of the vacuum adsorption stage is adjusted according to the fine-tuning direction and fine-tuning displacement, so that the vacuum adsorption stage generates different adsorption forces on different areas of the lens, and the deviation force formed by the adsorption force difference drives the lens to produce a small displacement on the vacuum adsorption stage. The adjustment effect of the lens position is monitored in real time by a vision camera until the relative positional deviation between the lens to be matted and the print head is less than a preset threshold, thus completing the fine adjustment.

5. The lens extinction control method based on machine vision as described in claim 4, characterized in that, The vacuum adsorption stage includes a placement area for the lens to be exfoliated. The placement area has at least a first adsorption row of holes and a second adsorption row of holes spaced apart from each other. Adjusting the adsorption force distribution of the vacuum adsorption stage according to the fine-tuning direction and fine-tuning displacement, so that the vacuum adsorption stage generates differentiated adsorption forces on different areas of the lens, includes: The relative size relationship between the first adsorption pore and the second adsorption pore, as well as the difference in adsorption force, are determined based on the fine-tuning direction and the fine-tuning displacement. Based on the relative size relationship and the adsorption force difference, the output power of the vacuum pumps corresponding to the first and second adsorption holes is controlled to adjust their actual adsorption forces.

6. The lens extinction control method based on machine vision as described in claim 1, characterized in that, The method of using a deep learning model to identify defect types and defect parameters in images includes: The acquired surface state image is input into a pre-trained defect recognition neural network, wherein the defect recognition neural network is a semantic segmentation network based on an encoder-decoder structure; The defect recognition neural network outputs a defect segmentation map containing defect category labels and pixel-level location information. Based on the defect segmentation map, the pixel area of ​​various defects is counted to determine the defect area, and the proportion of the defect area to the target extinction area is calculated to determine the defect distribution density. The depth information of the surface state image is extracted, and the depth features of the defect region are obtained from the depth information to determine the defect depth.

7. The lens extinction control method based on machine vision as described in claim 1, characterized in that, The step of dynamically adjusting the preset initial printing parameters in multiple dimensions according to the defect type and defect parameters to obtain the target printing parameters includes: The corresponding printing strategy is matched according to the defect type. Among them, scratch defects correspond to the first strategy of increasing ink filling amount, dent defects correspond to the second strategy of increasing printing layer thickness, and discoloration spot defects correspond to the third strategy of reducing ink coverage density. The printing parameter adjustment amount is calculated based on the defect parameters, wherein the printing path and ink volume are dynamically adjusted according to the defect area and defect distribution density, and the number of printing layers is dynamically adjusted according to the defect depth. The printing strategy is integrated with the printing parameter adjustment amount to collaboratively optimize the ink volume, printing speed, printing path and layer thickness in the initial printing parameters, thereby generating the target printing parameters.

8. The lens extinction control method based on machine vision as described in claim 1, characterized in that, During the inkjet printing process, a vision camera is activated in real time to dynamically adjust the layer thickness, including: After each layer of ink is printed, the three-dimensional morphology data of the current printed layer is acquired by the vision camera based on confocal chromatography or spectral interferometry. From the three-dimensional topography data, the height difference between the printed area and the unprinted reference surface is extracted, and this height difference is used as the actual thickness of the current printed layer. The actual thickness is compared with the preset thickness threshold of the corresponding layer to obtain the thickness deviation; The printing parameters of the next printing layer are dynamically adjusted based on the thickness deviation. Repeat the above steps until all layers of printing are completed, so that the cumulative printing thickness reaches the preset requirement.

9. The lens extinction control method based on machine vision as described in claim 8, characterized in that, Before acquiring the three-dimensional morphology data of the current printed layer using the vision camera based on confocal chromatography or spectral interferometry, the process also includes: The UV curing lamp is turned on to perform UV curing treatment on the ink layer that has been printed.

10. A visual extinction device, characterized in that, include: A printing carriage module, the printing carriage module including a first running guide rail and a print head, a vision camera and a curing UV lamp disposed on the first running guide rail; The worktable module includes a second running guide rail and a vacuum adsorption stage disposed on the second running guide rail. The memory is used to store program code; as well as A processor, the processor being configured to invoke the program code to perform the method as described in any one of claims 1 to 9.

Images