Visual inspection driven aluminum plate surface treatment process parameter regulation method and system

Abstract

The application discloses a visual detection driven aluminum plate surface treatment process parameter regulation method and system, relates to the image processing technical field, and comprises the following steps: deploying an image acquisition device in an aluminum plate surface treatment associated process procedure, guiding real-time processing images collected into an image processing channel of a corresponding process procedure, extracting process image recognition results according to processing rules matched with the associated process procedure, performing multi-granularity window feature analysis according to the process image recognition results, obtaining multi-time sequence surface features, performing process treatment time sequence optimization positioning based on the multi-time sequence surface features, constructing a regulation time sequence chain, and performing fusion analysis on the regulation time sequence chain and current treatment process parameters, combing regulation correction parameters, and generating parameter regulation instructions. The application solves the technical problems that aluminum plate surface treatment process parameter regulation lags behind and is difficult to reflect surface state changes in a timely manner in the prior art, and achieves the technical effects of improving the timeliness of aluminum plate surface treatment process parameter regulation and the stability of surface treatment quality.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method and system for controlling process parameters of aluminum plate surface treatment driven by vision inspection. Background Technology

[0002] In the surface treatment of aluminum plates, process parameters are usually adjusted based on experience or periodic test results, and there is a lack of an effective real-time correlation mechanism between parameter control and changes in the surface condition of the aluminum plate. When the surface condition deviates during processing, the relevant process parameters are difficult to receive timely feedback and make targeted corrections, which can easily lead to lag in control response, resulting in surface quality fluctuations, defect accumulation, or decreased consistency, affecting the stability and controllability of the surface treatment quality of the aluminum plate. Summary of the Invention

[0003] This application provides a vision inspection-driven method and system for controlling the process parameters of aluminum plate surface treatment, which is used to address the technical problems in the prior art where the control of aluminum plate surface treatment process parameters is lagging and it is difficult to reflect changes in surface state in a timely manner.

[0004] In view of the above problems, this application provides a method and system for controlling the process parameters of aluminum plate surface treatment driven by vision inspection.

[0005] The first aspect of this application provides a method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection, the method comprising:

[0006] Image acquisition equipment is deployed in the aluminum plate surface treatment process to collect real-time processing images and import them into the image processing path of the corresponding process. The process image recognition results are extracted according to the processing rules matched with the process. Multi-granularity window feature analysis is performed based on the process image recognition results to obtain multi-temporal surface features. Based on the multi-temporal surface features, the process processing timing is optimized and located to construct a control timing chain. The control timing chain is fused and analyzed with the current processing parameters to sort out the control and correction parameters and generate parameter control instructions.

[0007] A second aspect of this application provides a vision inspection-driven aluminum plate surface treatment process parameter control system, the system comprising: The image recognition module is used to deploy image acquisition equipment in the aluminum plate surface treatment process, import the acquired real-time processing images into the image processing path of the corresponding process, and extract the process image recognition results according to the processing rules matched with the associated process; the feature parsing module is used to perform multi-granularity window feature parsing based on the process image recognition results to obtain multi-temporal surface features; the positioning module is used to perform process processing timing optimization positioning based on the multi-temporal surface features and construct a control timing chain; the analysis module is used to perform fusion analysis based on the control timing chain and the current processing parameters, sort out the control correction parameters, and generate parameter control instructions.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application deploys image acquisition equipment in the aluminum plate surface treatment associated processes, imports the acquired real-time processing images into the image processing path of the corresponding process, and extracts the process image recognition results according to the processing rules matched with the associated processes. Based on the process image recognition results, multi-granularity window feature analysis is performed to obtain multi-temporal surface features. Based on the multi-temporal surface features, process processing timing optimization and positioning are performed to construct a control timing chain. The control timing chain is then fused and analyzed with the current processing parameters to sort out and correct the control parameters, generating parameter control instructions. This invention solves the technical problems of lagging control of aluminum plate surface treatment process parameters and difficulty in timely reflecting changes in surface state in the prior art. By constructing a control timing chain based on process image recognition results and generating parameter control instructions, it achieves the technical effect of improving the timeliness of aluminum plate surface treatment process parameter control and the stability of surface treatment quality. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A schematic diagram of the process parameter control method for visual inspection-driven aluminum plate surface treatment provided in an embodiment of this application; Figure 2 A schematic diagram of the structure of a vision inspection-driven aluminum plate surface treatment process parameter control system provided in an embodiment of this application.

[0011] Figure labeling: Image recognition module 11, feature parsing module 12, localization module 13, analysis module 14. Detailed Implementation

[0012] This application provides a vision-driven method and system for controlling the process parameters of aluminum plate surface treatment. It addresses the technical problems of lagging control of process parameters and difficulty in timely reflecting changes in surface condition in the prior art. By constructing a control timing chain based on process image recognition results and generating parameter control instructions, it achieves the technical effect of improving the timeliness of control of process parameters and the stability of surface treatment quality.

[0013] The technical solutions of the embodiments of this application 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 this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, this application provides a method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection, the method comprising: Step S100: Deploy image acquisition equipment in the aluminum plate surface treatment process, import the acquired real-time processing images into the image processing path of the corresponding process, and extract the process image recognition results according to the processing rules matched with the related process.

[0016] In this embodiment, when deploying image acquisition equipment in the related processes of aluminum plate surface treatment, the process flow nodes of aluminum plate surface treatment are first analyzed. Based on historical data or experimental data, a correlation response relationship between each process flow node and surface treatment quality is established. This correlation response relationship includes at least surface image features, correlation response quality parameters, and their correlation influence. Then, acquisition and recognition analysis is performed based on the surface image features, and process correlation constraint analysis is executed in conjunction with the correlation response quality parameters to determine the deployment location of image acquisition and the matching acquisition equipment. Finally, based on the determined acquisition location and matching acquisition equipment, the image acquisition equipment is deployed at the corresponding process flow node's response process position.

[0017] Next, the acquired real-time processing images are imported into the image processing pathway of the corresponding process, and processed according to the processing rules matched with the associated processes. During this process, based on the correlation response relationship of process flow nodes, the image features of the aluminum plate surface are identified and analyzed to determine the identification sensitive features that can characterize the quality parameters of the correlation response. Then, based on the identification sensitive features, image feature processing rules for stably extracting identification sensitive features are trained or configured using labeled sample data. Next, a dedicated image processing pathway is constructed for different process flow nodes. This image processing pathway consists of one or more image processing modules executed sequentially, each implementing its corresponding image feature processing rules. Finally, the real-time acquired process images are input into the corresponding image processing pathway according to the process node mapping relationship, and after step-by-step processing, the process image recognition result containing quantified feature values ​​is output.

[0018] Furthermore, in the method provided in the application embodiments, deploying an image acquisition device in the aluminum plate surface treatment associated process further includes: The process flow nodes of aluminum plate surface treatment are analyzed, and based on historical data or experimental data, the correlation response relationship of each process flow node to surface treatment quality is determined, including surface image features, correlation response quality parameters, and correlation influence relationships. Based on the correlation response relationship, image acquisition and identification analysis is performed according to surface image features, and process correlation constraint analysis is performed according to correlation response quality parameters to locate acquisition positions and acquisition matching devices. According to the acquisition positions and acquisition matching devices, image acquisition devices are deployed at the corresponding process flow node's response process position.

[0019] In this embodiment, the process flow nodes of aluminum plate surface treatment are first analyzed. The continuous processing is clearly divided into process flow nodes such as pretreatment, anodizing, electrolytic coloring, and curing according to the actual production sequence. The corresponding processing time and aluminum plate passage position are recorded for each process flow node. Subsequently, historical data or experimental data are collected based on each process flow node. Historical data includes processing images collected under different process parameter conditions and the corresponding quality inspection results. Experimental data is obtained by changing the parameters of the current process while keeping other process conditions unchanged and recording the processing images and quality inspection results. During data processing, the processing images collected at each process flow node of the same batch of aluminum plates are matched one-to-one with the final quality inspection results of the batch. By summarizing data from multiple batches, the changes in color distribution values, texture statistics values, and surface undulation values ​​in the processing images under different process parameter conditions are statistically analyzed. Image values ​​that can change regularly with process parameter changes are identified as surface image features. At the same time, film thickness values, color difference values, gloss values, and roughness values ​​that can change regularly with process parameter changes in the quality inspection results are identified as associated response quality parameters. By comparing and calculating the changes in surface image feature values ​​and the changes in associated response quality parameters at the same process flow node, the correspondence between the two in terms of the direction and magnitude of the changes is confirmed. This determines the correlation between surface image features and associated response quality parameters at each process flow node, and completes the determination of the correlation response relationship of each process flow node to surface treatment quality.

[0020] Next, based on the correlation response relationship, surface image features are collected, identified, and analyzed, and process correlation constraint analysis is performed according to the correlation response quality parameters. In the pretreatment, anodizing, electrolytic coloring, and curing processes, candidate acquisition locations are set at the process inlet, process outlet, and key processing positions in the middle of the process. Multiple image acquisitions are performed at each candidate acquisition location using the same acquisition parameters. For each set of images acquired at a candidate acquisition location, the numerical results of the determined surface image features are calculated, and the feature values ​​obtained from multiple acquisitions are statistically analyzed to calculate the average value and fluctuation range, used to confirm whether the acquisition location can output consistent surface image feature results under continuous acquisition conditions. Simultaneously, combined with the detection time point corresponding to the correlation response quality parameters, the acquisition location and quality detection results are aligned in time sequence to confirm that the surface image features obtained at the acquisition location can form a correspondence with the subsequently or synchronously obtained correlation response quality parameters. Based on this, combined with the process operating conditions, it is verified whether the acquisition location is located in a stable conveying area, avoids strong reflective areas, and does not affect the normal processing, thus completing the determination of the acquisition location. Subsequently, based on the determined surface image feature types and associated response quality parameter types, the imaging dimensions required for the acquired data are matched. When the surface image features mainly manifest as color and texture changes, the acquisition matching device is determined to be a color area array camera; when the surface image features mainly manifest as surface undulation changes, the acquisition matching device is determined to be a laser scanner; when the associated response quality parameters involve coating composition or film thickness distribution, the acquisition matching device is determined to be a hyperspectral imager, thus completing the determination of the acquisition matching device.

[0021] After determining the acquisition locations and matching equipment, image acquisition devices are deployed at the corresponding process flow nodes according to the acquisition locations. Acquisition locations include process inlet, process outlet, and critical processing positions in the middle of the process. During deployment, a color area array camera, laser scanner, or hyperspectral imager is installed at the corresponding response process location. The installation height and angle of the equipment are adjusted according to the width of the acquisition location and the aluminum plate's running trajectory to ensure the acquisition field of view completely covers the aluminum plate surface. Subsequently, acquisition tests are conducted using a standard template to confirm that the images or data output by the acquisition device at this location can stably obtain the determined surface image feature values. The acquisition results are then compared and verified again with the corresponding associated response quality parameters. After successful verification, the deployment of the image acquisition device at the corresponding process flow node is completed, enabling it to continuously output processing image data that can be used for image recognition and process parameter control in subsequent processes.

[0022] Furthermore, the method provided in the application embodiments also includes: The process flow nodes include the pretreatment process, the anodizing process, the electrolytic coloring process, and the curing process; the collection locations include: the process entry point, the process exit point, and the key processing locations in the middle of the process.

[0023] In this embodiment, the process flow nodes for aluminum plate surface treatment are divided into a pretreatment process, an anodizing process, an electrolytic coloring process, and a curing process. These process flow nodes are used to structurally characterize the main processing stages in the aluminum plate surface treatment process. The pretreatment process is used to complete surface cleaning and ensure consistent state preparation; the anodizing process is used to form an oxide film layer; the electrolytic coloring process is used to achieve the coloring effect; and the curing process is used to stabilize the film layer and color state. Acquisition locations are set for each of the above process flow nodes, including process entry points, process exit points, and key intermediate processing locations. The process entry point corresponds to the position before the aluminum plate enters each process flow node, where image acquisition equipment is deployed to obtain the initial surface state information of the aluminum plate before entering the process. The process exit point corresponds to the position where the aluminum plate completes the corresponding process and leaves the process flow node, where image acquisition equipment is deployed to obtain the surface result state information of the aluminum plate after the process. The key intermediate processing locations correspond to the areas within the process flow node where the aluminum plate receives the main process effects and its operating state is relatively stable, where image acquisition equipment is deployed to obtain surface change information of the aluminum plate during the process.

[0024] Furthermore, the method provided in the application embodiments also includes: The image acquisition device includes one or more of the following: a color area array camera, a laser scanner, and a hyperspectral imager. The high-resolution color area array camera is used to acquire surface color and texture; the 3D line laser scanner is used to acquire the three-dimensional morphology of the surface; and the hyperspectral imager is used to acquire coating composition and film thickness distribution information.

[0025] In this embodiment, during the aluminum plate surface treatment process, to acquire different types of surface information, one or more of a color area array camera, a laser scanner, and a hyperspectral imager are selected as image acquisition devices. A high-resolution color area array camera images the aluminum plate surface to obtain surface color distribution and texture structure information, reflecting the appearance of the aluminum plate. A 3D line laser scanner scans the aluminum plate surface to obtain surface height variation data, thereby forming three-dimensional surface morphology information, reflecting the surface flatness and morphological characteristics. A hyperspectral imager performs multi-band imaging of the aluminum plate surface to obtain response information at different wavelengths, and based on this, forms coating composition and film thickness distribution information, reflecting the coating state. According to the requirements for acquiring different surface information during the aluminum plate surface treatment process, the above-mentioned image acquisition devices are selected and combined to achieve comprehensive acquisition of the aluminum plate surface state.

[0026] Furthermore, in the method provided in the application embodiments, the acquired real-time processing images are imported into the image processing path of the corresponding process, and the process image recognition results are extracted according to the processing rules matched with the associated process. The method also includes: According to the correlation response relationship of the process flow nodes, the surface image features are identified and analyzed to obtain the identification sensitive features of the quality parameters of the correlation response; based on the identification sensitive features, image feature processing rules for stably extracting the identification sensitive features from the process images are trained or configured based on labeled sample data; a dedicated image processing path is constructed for each process flow node, the image processing path includes one or more image processing modules executed in sequence, each module is used to implement one of the image feature processing rules; the acquired real-time process images are input into the corresponding image processing path according to the mapping relationship of the process nodes, and processed step by step by each image processing module, finally outputting the process image recognition result containing quantized feature values.

[0027] In this embodiment, when identifying and analyzing surface image features according to the correlation response relationship of process flow nodes, firstly, multiple batches of process images collected under the same process flow node are uniformly organized, and a one-to-one correspondence is established between each batch of process images and its corresponding correlation response quality parameters. For each process image, the effective analysis area of ​​the aluminum plate is first determined. The effective analysis area is obtained by removing non-aluminum plate areas from the image, which is used to ensure that subsequent calculations are only performed on the surface of the aluminum plate. Subsequently, the surface image features are quantified and calculated within the effective analysis area. Among them, the color features are obtained by calculating the pixel average values ​​of the red, green, and blue channels within the effective analysis area, respectively, to obtain the three color average values; the texture features are obtained by converting the image into a grayscale image and calculating the average value and standard deviation of the grayscale values, with the grayscale standard deviation used to characterize the degree of surface texture undulation; the defect features are obtained by thresholding the grayscale image, separating areas with grayscale values ​​lower or higher than a set threshold, and counting the separated connected regions to obtain the number of defects; the morphology features, when three-dimensional data exists, are obtained by calculating the difference between the maximum and minimum values ​​of the height data in the corresponding area to obtain the surface undulation height value. Through the above calculation steps, each process image is converted into a set of specific numerical results, and each numerical result corresponds to a specific surface image feature.

[0028] After obtaining the surface image feature values ​​corresponding to multiple batches of process images, the surface image feature values ​​under the same process node are compared and calculated item by item with the associated response quality parameters to determine the identification sensitive features. Specifically, for a certain surface image feature, the values ​​of this feature in all batches are arranged in batch order, and the corresponding associated response quality parameters are arranged in the same order, forming a feature value sequence and a quality parameter sequence, respectively. Then, the Pearson correlation coefficient method is used to calculate the correlation between the feature value sequence and the quality parameter sequence. The correlation coefficient is obtained by dividing the covariance of the two sequences by the product of their respective standard deviations, and the result is a definite value between negative one and positive one. When the absolute value of the correlation coefficient is greater than a preset threshold, it indicates that the change in the value of the surface image feature and the change in the associated response quality parameter have a clear linear correspondence. At the same time, by checking the sign of the correlation coefficient, it is confirmed that there is a single correspondence between the direction of change of the value of the surface image feature and the direction of change of the associated response quality parameter. Under the condition that the absolute value of the correlation coefficient exceeds the threshold and the direction of the correspondence is consistent, the surface image feature is determined as an identification sensitive feature that can respond to the associated response quality parameter.

[0029] After identifying the sensitive features, image feature processing rules for extracting these features are trained or configured based on labeled sample data. The labeled sample data consists of multiple process images with known quality results, with the effective analysis area of ​​the aluminum plate and the calculation area corresponding to the sensitive feature labeled in each sample image. For each sensitive feature, a fixed calculation method is used; for example, color-related sensitive features correspond to the calculation of color channel averages, texture-related sensitive features correspond to the calculation of grayscale standard deviations, defect-related sensitive features correspond to the counting of connected components, and morphology-related sensitive features correspond to the calculation of height differences. The same calculation process is then repeated on the sample data, calculating feature values ​​multiple times for the same sample image under the same parameters, and confirming that the results are consistent each time. This ensures that the calculation method can stably output the numerical results for the sensitive features. After verification, the above fixed calculation steps, parameter settings, and output methods are determined as the image feature processing rules.

[0030] After the image feature processing rules are determined, a dedicated image processing path is constructed for each process node. This path consists of multiple sequentially executed image processing modules, including an image cropping module for determining the effective analysis area, an image preprocessing module for unifying image brightness and size, a feature calculation module for performing sensitive feature calculations, and a result output module for outputting feature values. Each image processing module corresponds to a determined image feature processing rule. Finally, the acquired real-time process images are input into the corresponding image processing path according to the mapping relationship between process nodes. The real-time process images are processed step-by-step by each image processing module, and feature values ​​are output according to the calculation method corresponding to the sensitive features. These feature values, together with the process node identifiers, constitute the process image recognition result, thereby achieving stable parsing and quantification output of the real-time process images.

[0031] Step S200: Perform multi-granularity window feature analysis based on the process image recognition results to obtain multi-temporal surface features.

[0032] Furthermore, in the method provided in the application embodiments, the process of performing multi-granularity window feature parsing based on the process image recognition results to obtain multi-temporal surface features further includes: A response time-series penalty relationship is established between the surface image features and the associated response quality parameters to quantitatively describe the extent to which the impact of image feature deviations in each process on the final product quality will worsen over time or as the process progresses. Using the target quality constraint of aluminum plate surface treatment, the surface image features are adjusted according to the response time-series penalty relationship to determine the adjustment constraint window for each surface image feature, reflecting the granularity of the time constraint for adjusting the surface image features. Based on the adjustment constraint windows of the surface image features included in the process image recognition results, the adjustment time sequence of each surface image feature is labeled and sorted to obtain the multi-time-series surface features.

[0033] In this embodiment, when performing multi-granularity window feature analysis based on the process image recognition results, a response time penalty relationship between surface image features and associated response quality parameters is first established. Specifically, the sequential order of process flow nodes is used as the time progression order. The quantized values ​​of surface image features obtained at each process flow node of the same batch of aluminum plates are arranged sequentially, while the associated response quality parameter values ​​obtained in the final or subsequent inspections of the batch are also compiled. For each surface image feature, its target reference value is first determined. The target reference value can be taken from the average value under stable production conditions or the statistical value of qualified samples. Then, the surface image feature deviation is calculated at each process flow node. The surface image feature deviation is obtained by the difference between the feature quantization value of the current process and the target reference value. If a surface image feature deviation occurs at a certain process flow node and is not corrected in subsequent processes, the deviation value is recorded at each subsequent process flow node, and the deviation value is accumulated as the process progresses. That is, the initial deviation is recorded at the first process where a deviation occurs, and the current deviation value is added to the cumulative value of the previous process at each subsequent process, thereby obtaining the cumulative influence value that increases with the progress of the process. By comparing the cumulative impact value with the changes in the associated response quality parameters of the corresponding batch, a numerical relationship is established where the impact of surface image feature deviations on the final product quality gradually increases when they are not corrected at different process stages. This establishes a response time-series penalty relationship, which is used to quantitatively describe the amplifying effect of image feature deviations over time or as the process progresses. For example, if a certain color-related surface image feature begins to deviate from the target value at the anodizing process and is not corrected in subsequent electrolytic coloring and curing processes, the deviation value is accumulated in each of the electrolytic coloring and curing processes, forming a cumulative value that increases step by step. Correspondingly, the final color difference detection value increases accordingly, thus reflecting the time-series penalty effect of the deviation.

[0034] After establishing the response time-series penalty relationship, the target quality constraint of aluminum plate surface treatment is used as the control condition to apply the adjustment time-series penalty constraint to the surface image features and determine the adjustment constraint window. Specifically, the target quality constraint is first defined as the allowable range of the associated response quality parameters, such as the upper and lower limits of film thickness, color difference, or roughness. Then, for each surface image feature, using the established response time-series penalty relationship, starting from the process node where the deviation first occurs, the cumulative impact value at each subsequent process node is calculated, and this cumulative impact value is converted into the degree of influence on the associated response quality parameters. The conversion method uses the corresponding ratio between the surface image feature deviation and the change in the associated response quality parameters in historical data, that is, when the surface image feature deviation increases by a certain value, the quality parameter increases by an average value. By comparing the calculated quality impact step-by-step to determine if it remains within the allowable range of the target quality constraints, the latest process node at which the surface image feature can continue to be uncontrolled without violating the target quality constraints is determined. The interval from the process node where the first deviation occurs to this latest allowable process node is defined as the adjustment constraint window for the surface image feature, used to clarify the time range and process range within which the surface image feature must be controlled. For example, if a texture-related surface image feature deviates in the pre-processing step, and calculations show that it will not cause quality exceedances in the anodizing step, but delaying it to the electrolytic coloring step will cause roughness to exceed the allowable range, then the period from the pre-processing step to the anodizing step is defined as the adjustment constraint window for the surface image feature.

[0035] After determining the adjustment constraint windows for each surface image feature, the adjustment timing labels of each surface image feature are sorted according to the surface image features and their corresponding adjustment constraint windows contained in the process image recognition results. In this process, each surface image feature is associated with the end process node of its adjustment constraint window, and the order of these end process nodes is used as the sorting criterion, placing surface image features with earlier end process nodes first, and those with later end process nodes last. When multiple surface image features have the same adjustment constraint window ending process, the growth rate of their cumulative influence value in the response timing penalty relationship is compared, i.e., the increase in cumulative influence value between adjacent processes, and the surface image feature with faster growth is placed in a higher priority position. Through the above sorting process, each surface image feature is assigned a clear adjustment sequence and time position label, ultimately forming a multi-time sequence surface feature that includes the adjustment priority relationship of different surface image features at different process stages.

[0036] Step S300: Based on the multi-time-series surface features, perform process timing optimization and positioning, and construct a control timing chain.

[0037] Furthermore, in the method provided in the application embodiments, the process timing optimization and positioning based on the multi-temporal surface features, and the construction of a control timing chain, further include: Based on the aforementioned multi-temporal surface features and combined with the physicochemical coupling relationship map between each process node, a timing optimization and localization of the control sequence is performed. For continuous processes with strong coupling relationships, a feedforward-feedback composite control link is constructed, using the identification results of the preceding process as the basis for feedforward compensation in the subsequent process. Simultaneously, the quality feedback from the subsequent process can be used to trace and correct the settings of the preceding process. For independent processes exhibiting weak coupling or decoupling characteristics, a parallel independent control link is constructed, and local parameter autonomous optimization is implemented according to the control sequence annotation of surface image features. The temporal attributes of the control response are determined based on the characteristic type, propagation rate, and process repairability assessment of quality anomalies. Based on the determined control link and the temporal attributes of the control response, a control timing chain is constructed, including control process nodes, time windows, parameter types, and intensity coefficients. This control timing chain supports dynamic updates and multi-chain concurrency.

[0038] In this embodiment, when optimizing and locating the process timing based on multi-temporal surface features and constructing a control timing chain, the optimization and location of the control timing is performed based on the multi-temporal surface features and the physicochemical coupling relationship map between each process node. Specifically, the multi-temporal surface features are first classified and organized according to their corresponding process nodes to form a distribution record of surface image features in the pretreatment process, anodizing process, electrolytic coloring process, and curing process. At the same time, the interaction relationships between each process node in the physicochemical coupling relationship map are read. The physicochemical coupling relationship map is obtained from historical or experimental data. It is established by changing the process parameters of the upstream process while keeping the downstream process parameters constant, and observing whether the process image recognition results and associated response quality parameters of the downstream process change. When the change of the upstream process parameters can cause a measurable change in the surface image features or associated response quality parameters in the downstream process, the process node pair is marked as a continuous process with a strong coupling relationship in the physicochemical coupling relationship map. When the change of the upstream process parameters does not cause a change in the corresponding results of the downstream process, or the change is always within the allowable range of the target quality constraint, the process node pair is marked as an independent process with weak coupling or decoupling characteristics. This completes the determination of the coupling relationship of the process nodes.

[0039] For consecutive processes identified as having strong coupling relationships, a feedforward-feedback composite control link is constructed. Specifically, after the preceding process node is completed, the process image recognition result of that preceding process node is read, which includes the quantized feature values ​​of the surface image features. The difference between the quantized feature value and the target reference value corresponding to the surface image feature is calculated to obtain the current deviation value. Subsequently, based on a correspondence table established in historical or experimental data, the surface image feature is mapped to the adjustable process parameter type in the subsequent process node. The correspondence table is obtained through statistical analysis, and its content is the average change in the quantized feature value of the corresponding surface image feature when a certain process parameter undergoes a unit change during historical production or testing. Based on this correspondence table, the deviation value is divided by the unit change to calculate the parameter adjustment amount that needs to be applied in the subsequent process node, thus forming the feedforward compensation parameter setting for the subsequent process node. After the subsequent process node completes its processing, the process image recognition result or the detection value of the associated response quality parameter is obtained and compared with the target quality constraint. If the detection value still has a deviation, the deviation value is traced back to the previous process node along the control link, and the correction amount of the previous process node parameter is recalculated according to the same correspondence table. In this way, the feedforward-feedback composite control link closed-loop execution of feedforward compensation of previous process and traceback correction of quality feedback of subsequent process is realized.

[0040] For independent processes identified as exhibiting weak coupling or decoupling characteristics, parallel independent control links are constructed. Specifically, within each independent process node, control is performed solely based on the process image recognition results and multi-temporal surface features of that process, without invoking information from other process nodes. Following the control timing sequence indicated in the multi-temporal surface features, the corresponding surface image features are processed one by one. For each surface image feature, the deviation between its quantized feature value and the target reference value is first calculated. Then, based on the established correspondence between parameter changes and feature changes in historical or experimental data, the parameter adjustment amount that the deviation value needs to be converted into in the current process node is calculated, and the corresponding parameter adjustment is executed within the current process node. After one adjustment, the quantized feature value of the surface image feature is recalculated during the next real-time process image acquisition. If the deviation value decreases, the current adjustment direction is maintained; if the deviation value does not decrease, a parameter adjustment is performed in the opposite direction. This achieves autonomous optimization of local parameters within the independent process node without relying on other process nodes. Different independent process nodes maintain their own independent parallel control links, thus achieving parallel execution.

[0041] After determining the control link type, the time attribute of the control response is determined based on the characteristic type of the quality anomaly, its propagation rate, and the process repairability assessment. The characteristic type of the quality anomaly is determined by the surface image feature category corresponding to the deviation in the process image recognition results. The propagation rate is obtained by comparing the deviation value changes of the same quality anomaly in adjacent process nodes, subtracting the deviation value of the previous process node from the deviation value of the subsequent process node, and combining this with the process progression sequence to determine the speed of deviation expansion. The process repairability assessment is completed using historical data or experimental data. The determination method is to find the repair results of anomalies of the same characteristic type after parameter adjustments at the current or subsequent process nodes. If the deviation value can enter the allowable range of the target reference value after a limited number of adjustments, it is determined to be repairable; otherwise, it is determined to be unrepairable. Based on the above three results, when the anomaly is repairable at the current process node and the propagation rate is relatively fast, the control response is determined to be immediate control; when the anomaly is unrepairable at the current process node but repairable at a subsequent process node, the control response is determined to be deferred control; when the anomaly is unrepairable at both the current and subsequent process nodes and has caused the associated response quality parameters to exceed the target quality constraint, the control response is determined to be terminated control.

[0042] After determining the control link and control response time attribute, a control timing chain is constructed based on the determined control link type and control response time attribute. In this process, the control process nodes involved in the feedforward-feedback composite control link and the parallel independent control link are arranged according to the actual process execution order. Combining the time attributes corresponding to immediate control, delayed control, and terminated control, a clear time window is assigned to each control process node. Simultaneously, based on the established correspondence between surface image features and process parameters, the parameter type to be adjusted in each control process node is determined. Furthermore, based on the average impact of unit parameter changes on the quantified feature values ​​of surface image features in historical or experimental data, the required intensity coefficient for parameter adjustment is determined. Subsequently, all control operations, including control process nodes, time windows, parameter types, and intensity coefficients, are chained together according to time and process order to form a control timing chain that reflects the order and conditions of control execution. This allows the control timing chain to logically and completely describe the execution mode of control in different process nodes and different time windows, and supports updating relevant content when new process image recognition results and multi-time-series surface features are subsequently obtained. Multiple control timing chains can also exist and execute in parallel.

[0043] Furthermore, the method provided in the application embodiments also includes: The time attributes of the control response include: immediate control, delayed control, and termination control. The immediate control is for deviations that can be quickly corrected within the current process, triggering millisecond-level parameter fine-tuning. The delayed control is for deviations that require compensation from subsequent processes or whose current correction cost is too high, marking them as control compensation tasks for downstream processes. The termination control is for irreversible quality defects, triggering production line shutdown and incoming material isolation instructions.

[0044] In this embodiment, the time attribute of the control response is used to distinguish the handling methods of quality anomalies in the time dimension. It includes three forms: immediate control, delayed control, and termination control. It is used to clarify at which process node and in what time sequence the deviation is responded to, thereby ensuring the effectiveness of process control and the safety of the production process.

[0045] In this context, real-time control is used to handle deviations that can be quickly corrected within the current process. When the process image recognition results indicate that the surface image features deviate from the target reference value, and historical data or experimental data confirm that the deviation can be corrected by making minor adjustments to the current process parameters, the time attribute of the control response is determined to be real-time control. This triggers millisecond-level parameter fine-tuning within the current process, allowing the surface image features to quickly return to the target range without changing the process flow.

[0046] Deferred control is used to handle deviations that require compensation from subsequent processes, or where continued correction within the current process would significantly increase process costs. When analysis results indicate that the deviation has limited room for adjustment in the current process, but controllability still exists at downstream process nodes, the time attribute of the control response is defined as deferred control, and the corresponding control requirement is marked as a control compensation task, which is then passed down to the downstream process for execution as the process progresses.

[0047] Termination control is used to handle irreversible quality defects. When the process image recognition results and associated response quality parameters jointly indicate that the quality defect corresponding to the current deviation cannot be repaired by the current process or subsequent processes, and continued processing will cause the product quality to continue to fail to meet the target quality constraints, the time attribute of the control response is determined as termination control, and a production line shutdown and incoming material isolation instruction is triggered to prevent non-conforming products from continuing to enter the subsequent production process.

[0048] Step S400: Based on the fusion analysis of the control timing chain and the current processing parameters, sort out the control correction parameters and generate parameter control instructions.

[0049] In this embodiment, when performing fusion analysis based on the control timeline chain and the current processing parameters, the control requirements contained in the control timeline chain are mapped to the corresponding process parameters by comparing the real-time values ​​of the current processing parameters with the set values, thus forming process parameter control relationships. Then, the control requirements for the same process parameter at different time points are fused, and mutually exclusive control requirements are arbitrated, thereby obtaining the control correction parameters corresponding to each process parameter. These control correction parameters include the correction target value, the rate of change constraint, the execution duration, and the safety boundary. Finally, parameter control instructions are generated based on the control correction parameters and sent to the actuator for execution.

[0050] Furthermore, in the method provided in the application embodiments, the method further includes: performing a fusion analysis based on the control timing chain and the current processing parameters to sort out the control correction parameters and generate parameter control instructions; The system acquires the real-time and setpoint values ​​of the current processing parameters, integrates the control requirements in the control timeline with the current parameters, and performs control conflict detection and resolution. Specifically, when there are multiple control requirements for the same parameter at multiple time points in the control timeline, control merging or decomposition is performed based on time proximity and effect accumulation. When there are mutually exclusive control requirements in the control timeline, priority arbitration is performed based on quality impact weight and execution feasibility. Based on the process parameter control relationship analysis of the integrated analysis, the system analyzes the control correction parameters for each process parameter, including the correction target value, change rate constraint, execution duration, and safety boundary, generates parameter control instructions, and issues these instructions to the execution mechanism after verification.

[0051] In this embodiment, the real-time and setpoint values ​​of the current processing parameters are first read and organized. Real-time values ​​are read item by item from the measurement outputs of field instruments or the registers of the controller to form a parameter measurement table. Setpoint values ​​are read item by item from the process formula or control setting table corresponding to the current process to form a parameter target table. Parameter names, units, sampling times, and values ​​are aligned using the same field format, ensuring that each process parameter has corresponding records for both real-time and setpoint values. Subsequently, the control requirements in the control sequence chain are fused and analyzed with the current parameter baseline, and control conflict detection and resolution are performed. In this process, the control sequence chain is parsed line by line, extracting the parameter type, effective time point, start and end of the time window, control direction, and control amplitude for each control requirement. These are then grouped into a control requirement sequence according to parameter type. The control requirement sequences for the same parameter are sorted from earliest to latest effective time point. Then, the time proximity of two adjacent control requirements is calculated. The time proximity is taken as the time difference between the effective times of the two control requirements and compared with a preset time threshold to determine whether they belong to the same control period. When comparing the consistency of control directions, the control direction of each control demand is first converted into a sign value: a positive sign is recorded when the control direction is increasing, a negative sign is recorded when the control direction is decreasing, and a zero sign is recorded when the control direction is maintaining. Then, the sign values ​​of two adjacent control demands are multiplied. If the product is positive and neither sign is zero, the direction is determined to be consistent. If the product is negative, the direction is determined to be opposite. If either sign is zero, the direction is determined not to participate in the merging and is retained separately. The assessment of the cumulative effect is accomplished by algebraically summing the control amplitudes. That is, under the conditions of consistent direction and temporal proximity meeting the threshold, the control amplitudes of the two control demands are algebraically summed to obtain the merged amplitude. The merged amplitude is then compared with the safety boundary of this parameter. If the merged correction target value is still within the safety boundary, control merging is performed, and the earlier effective time point of the merged value is taken. The merged time window is the union of the two time windows. If the merged correction target value will exceed the safety boundary, control decomposition is performed instead of merging. The two control demands are retained, and the amplitude of the subsequent control demand is reduced so that the correction target value is exactly within the safety boundary.

[0052] When mutually exclusive control requirements exist in the control timeline, priority arbitration is performed to resolve the conflict. Specifically, mutually exclusive pairs are first identified. The identification rule is that there are two or more control requirements targeting the same parameter and in opposite directions within the same time window, or there are control requirements targeting related parameters and parameter pairs marked as mutually exclusive in the process mechanism table. For each mutually exclusive control requirement, a quality impact weight is calculated. This involves finding the change in the associated response quality parameter caused by a unit change in the parameter in historical or experimental data, taking the absolute value of this change, and normalizing it according to the target quality constraint allowable range. The normalization method is that the quality impact weight equals the absolute value of the change divided by the width of the target quality constraint allowable range, thus allowing the impact of different quality parameters to be compared on the same scale. Simultaneously, feasibility is determined by comparing the corrected target value corresponding to the control requirement with the safety boundary and comparing the required rate of change with the equipment's allowable rate of change. If any comparison fails, the requirement is deemed infeasible. The priority arbitration rule is to select the control demand with a greater weight in quality impact as the retained item among the feasible control demands, and the other item as the concession item. The concession item is handled as follows: if there is a subsequent time window, its effective time point is shifted to the beginning of the subsequent time window to form a deferred control compensation task; if there is no subsequent time window, its control magnitude is reduced to zero and the control demand is deleted, thereby completing the resolution of mutually exclusive control demands.

[0053] After completing the fusion analysis and establishing the process parameter control relationship, control correction parameters are calculated for each process parameter. Specifically, first, the deviation value is calculated, where the deviation value equals the real-time value minus the setpoint value; then, the correction target value is calculated based on the fused control requirements, which equals the setpoint value plus the control amplitude retained after merging or arbitration; subsequently, the rate of change constraint is determined, which is the smaller of the upper limit of the equipment's allowable rate of change and the upper limit of the process stability requirement. The upper limit of the process stability requirement is obtained statistically from the critical rate of quality fluctuation caused by rapid parameter changes in historical data; the execution duration is calculated by dividing the difference between the correction target value and the real-time value by the rate of change constraint, and compared with the time window length. If the execution duration is longer than the time window length, the rate of change constraint is adjusted to the difference divided by the time window length, and the execution duration is set as the time window length; the safety boundary is composed of the lower and upper limits of the process allowable parameters for this parameter. If the correction target value exceeds the safety boundary, the correction target value is truncated to the boundary value, and the execution duration and rate of change constraint are recalculated to ensure that the control correction parameter meets the safety boundary.

[0054] Finally, parameter control instructions are generated based on the correction target value, rate of change constraint, execution duration, and safety boundary of each process parameter. These instructions are written into an instruction list in a field-based format, with each field including at least the parameter name, start time, end time, correction target value, rate of change constraint, and safety boundary. A verification process is then executed, sequentially checking whether the correction target value is within the safety boundary, whether the rate of change constraint does not exceed the equipment's allowed rate of change, whether the execution duration covers the time window, and whether priority arbitration has been completed for mutually exclusive control requirements. If any check fails, the process reverts to the corresponding step, recalculates, and updates the instruction fields. After all checks pass, the parameter control instructions are written into the control instruction queue and sent to the actuator. The actuator then gradually adjusts the parameters according to the rate of change constraint starting at the start time, reaching the correction target value at the end time, thus completing the execution from the control sequence chain to the parameter control instructions.

[0055] In summary, the embodiments of this application have at least the following technical effects: This application deploys image acquisition equipment in the aluminum plate surface treatment associated processes, imports the acquired real-time processing images into the image processing path of the corresponding process, and extracts the process image recognition results according to the processing rules matched with the associated processes. Based on the process image recognition results, multi-granularity window feature analysis is performed to obtain multi-temporal surface features. Based on the multi-temporal surface features, process processing timing optimization and positioning are performed to construct a control timing chain. The control timing chain is then fused and analyzed with the current processing parameters to sort out and correct the control parameters, generating parameter control instructions. This invention solves the technical problems of lagging control of aluminum plate surface treatment process parameters and difficulty in timely reflecting changes in surface state in the prior art. By constructing a control timing chain based on process image recognition results and generating parameter control instructions, it achieves the technical effect of improving the timeliness of aluminum plate surface treatment process parameter control and the stability of surface treatment quality.

[0056] Example 2, based on the same inventive concept as the vision detection-driven aluminum plate surface treatment process parameter control method in the foregoing examples, such as... Figure 2 As shown, this application provides a vision-inspection-driven aluminum plate surface treatment process parameter control system. The system and method embodiments in this application are based on the same inventive concept. The system includes: Image recognition module 11 is used to deploy image acquisition equipment in the aluminum plate surface treatment process, import the acquired real-time processing images into the image processing path of the corresponding process, and extract the process image recognition results according to the processing rules matched with the associated process; feature parsing module 12 is used to perform multi-granularity window feature parsing based on the process image recognition results to obtain multi-temporal surface features; positioning module 13 is used to perform process processing timing optimization positioning based on the multi-temporal surface features and construct a control timing chain; analysis module 14 is used to perform fusion analysis based on the control timing chain and the current processing parameters, sort out the control correction parameters, and generate parameter control instructions.

[0057] Furthermore, the system is also used to implement the following functions: The process flow nodes of aluminum plate surface treatment are analyzed, and based on historical data or experimental data, the correlation response relationship of each process flow node to surface treatment quality is determined, including surface image features, correlation response quality parameters, and correlation influence relationships. Based on the correlation response relationship, image acquisition and identification analysis is performed according to surface image features, and process correlation constraint analysis is performed according to correlation response quality parameters to locate acquisition positions and acquisition matching devices. According to the acquisition positions and acquisition matching devices, image acquisition devices are deployed at the corresponding process flow node's response process position.

[0058] Furthermore, the system is also used to implement the following functions: The process flow nodes include the pretreatment process, the anodizing process, the electrolytic coloring process, and the curing process; the collection locations include: the process entry point, the process exit point, and the key processing locations in the middle of the process.

[0059] Furthermore, the system is also used to implement the following functions: The image acquisition device includes one or more of the following: a color area array camera, a laser scanner, and a hyperspectral imager. The high-resolution color area array camera is used to acquire surface color and texture; the 3D line laser scanner is used to acquire the three-dimensional morphology of the surface; and the hyperspectral imager is used to acquire coating composition and film thickness distribution information.

[0060] Furthermore, the system is also used to implement the following functions: According to the correlation response relationship of the process flow nodes, the surface image features are identified and analyzed to obtain the identification sensitive features of the quality parameters of the correlation response; based on the identification sensitive features, image feature processing rules for stably extracting the identification sensitive features from the process images are trained or configured based on labeled sample data; a dedicated image processing path is constructed for each process flow node, the image processing path includes one or more image processing modules executed in sequence, each module is used to implement one of the image feature processing rules; the acquired real-time process images are input into the corresponding image processing path according to the mapping relationship of the process nodes, and processed step by step by each image processing module, finally outputting the process image recognition result containing quantized feature values.

[0061] Furthermore, the system is also used to implement the following functions: A response time-series penalty relationship is established between the surface image features and the associated response quality parameters to quantitatively describe the extent to which the impact of image feature deviations in each process on the final product quality will worsen over time or as the process progresses. Using the target quality constraint of aluminum plate surface treatment, the surface image features are adjusted according to the response time-series penalty relationship to determine the adjustment constraint window for each surface image feature, reflecting the granularity of the time constraint for adjusting the surface image features. Based on the adjustment constraint windows of the surface image features included in the process image recognition results, the adjustment time sequence of each surface image feature is labeled and sorted to obtain the multi-time-series surface features.

[0062] Furthermore, the system is also used to implement the following functions: Based on the aforementioned multi-temporal surface features and combined with the physicochemical coupling relationship map between each process node, a timing optimization and localization of the control sequence is performed. For continuous processes with strong coupling relationships, a feedforward-feedback composite control link is constructed, using the identification results of the preceding process as the basis for feedforward compensation in the subsequent process. Simultaneously, the quality feedback from the subsequent process can be used to trace and correct the settings of the preceding process. For independent processes exhibiting weak coupling or decoupling characteristics, a parallel independent control link is constructed, and local parameter autonomous optimization is implemented according to the control sequence annotation of surface image features. The temporal attributes of the control response are determined based on the characteristic type, propagation rate, and process repairability assessment of quality anomalies. Based on the determined control link and the temporal attributes of the control response, a control timing chain is constructed, including control process nodes, time windows, parameter types, and intensity coefficients. This control timing chain supports dynamic updates and multi-chain concurrency.

[0063] Furthermore, the system is also used to implement the following functions: The time attributes of the control response include: immediate control, delayed control, and termination control. The immediate control is for deviations that can be quickly corrected within the current process, triggering millisecond-level parameter fine-tuning. The delayed control is for deviations that require compensation from subsequent processes or whose current correction cost is too high, marking them as control compensation tasks for downstream processes. The termination control is for irreversible quality defects, triggering production line shutdown and incoming material isolation instructions.

[0064] Furthermore, the system is also used to implement the following functions: The system acquires the real-time and setpoint values ​​of the current processing parameters, integrates the control requirements in the control timeline with the current parameters, and performs control conflict detection and resolution. Specifically, when there are multiple control requirements for the same parameter at multiple time points in the control timeline, control merging or decomposition is performed based on time proximity and effect accumulation. When there are mutually exclusive control requirements in the control timeline, priority arbitration is performed based on quality impact weight and execution feasibility. Based on the process parameter control relationship analysis of the integrated analysis, the system analyzes the control correction parameters for each process parameter, including the correction target value, change rate constraint, execution duration, and safety boundary, generates parameter control instructions, and issues these instructions to the execution mechanism after verification.

[0065] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0066] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for controlling process parameters of aluminum plate surface treatment driven by vision inspection, characterized in that, include: Image acquisition equipment is deployed in the aluminum plate surface treatment process to acquire real-time processing images and import them into the image processing path of the corresponding process. The image recognition results of the process are extracted according to the processing rules matched with the associated process. Based on the image recognition results of the aforementioned process, multi-granularity window feature analysis is performed to obtain multi-temporal surface features; Based on the aforementioned multi-temporal surface features, process timing optimization and positioning are performed to construct a control timing chain; Based on the fusion analysis of the control timing chain and the current processing parameters, control and correction parameters are sorted out and parameter control instructions are generated.

2. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 1, characterized in that, Deploying image acquisition equipment in aluminum plate surface treatment-related processes includes: The process flow nodes of aluminum plate surface treatment are analyzed, and based on historical data or experimental data, the correlation response relationship of each process flow node to surface treatment quality is determined, including surface image features, correlation response quality parameters, and correlation influence relationships. Based on the aforementioned correlation response relationship, the acquisition and identification analysis are performed according to the surface image features, and the process correlation constraint analysis is performed according to the correlation response quality parameters to locate the acquisition position and the acquisition matching device. Based on the acquisition location and the acquisition matching device, the image acquisition device is deployed at the corresponding process location of the corresponding process flow node.

3. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 2, characterized in that, The process flow nodes include the pretreatment process, the anodizing process, the electrolytic coloring process, and the curing process; the collection locations include: the process entry point, the process exit point, and the key processing locations in the middle of the process.

4. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 2, characterized in that, The image acquisition device includes one or more of the following: a color area array camera, a laser scanner, and a hyperspectral imager. The high-resolution color area array camera is used to acquire surface color and texture; the 3D line laser scanner is used to acquire the three-dimensional morphology of the surface; and the hyperspectral imager is used to acquire coating composition and film thickness distribution information.

5. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 2, characterized in that, The acquired real-time processing images are imported into the image processing pathway of the corresponding process. Following the processing rules matched with the associated process, the process image recognition results are extracted, including: Based on the correlation response relationship of the process flow nodes, the surface image features are identified and analyzed to obtain the identification sensitive features of the quality parameters of the correlation response. Based on the identified sensitive features, image feature processing rules for stably extracting the identified sensitive features from process images are trained or configured based on labeled sample data. A dedicated image processing path is constructed for each process node. The image processing path includes one or more image processing modules executed sequentially, and each module is used to implement one of the image feature processing rules. The acquired real-time process images are input into the corresponding image processing path according to the mapping relationship of the process nodes. They are processed step by step by each image processing module, and finally the process image recognition result containing quantized feature values ​​is output.

6. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 2, characterized in that, Based on the image recognition results of the aforementioned process, multi-granularity window feature analysis is performed to obtain multi-temporal surface features, including: Establish the response time penalty relationship between the surface image features and the associated response quality parameters to quantitatively describe the extent to which the impact of image feature deviations in each process on the final product quality will worsen over time or as the process progresses if they are not corrected. Using the target quality constraint of aluminum plate surface treatment, the surface image features are adjusted and time-series penalty constraint according to the response time-series penalty relationship, and the adjustment constraint window of each surface image feature is determined to reflect the time constraint granularity of the surface image feature adjustment. Based on the adjustment constraint window of the surface image features contained in the process image recognition result, the adjustment time sequence labeling and sorting of each surface image feature is performed to obtain the multi-time sequence surface features.

7. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 6, characterized in that, Based on the aforementioned multi-temporal surface features, process timing optimization and localization are performed to construct a control timing chain, including: Based on the aforementioned multi-temporal surface features and combined with the physicochemical coupling relationship map between each process node, a timing optimization and localization of the control sequence is performed. Specifically, for continuous processes with strong coupling relationships, a feedforward-feedback composite control link is constructed, using the identification results of the preceding process as the basis for feedforward compensation of the subsequent process. At the same time, the quality feedback of the subsequent process can be used to trace and correct the settings of the preceding process. For independent processes exhibiting weak coupling or decoupling characteristics, a parallel independent control link is constructed, and local parameter autonomous optimization is implemented according to the control timing annotation of the surface image features. Based on the characteristics, propagation rate, and process repairability assessment of the quality anomaly, determine the time attribute of the control response; Based on the defined control link and the time attribute of the control response, a control timing chain is constructed, which includes control process nodes, time windows, parameter types and intensity coefficients. The control timing chain supports dynamic updates and multi-chain concurrency.

8. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 7, characterized in that, The time attributes of the control response include: immediate control, delayed control, and termination control. The immediate control is for deviations that can be quickly corrected within the current process, triggering millisecond-level parameter fine-tuning. The delayed control is for deviations that require compensation from subsequent processes or whose current correction cost is too high, marking them as control compensation tasks for downstream processes. The termination control is for irreversible quality defects, triggering production line shutdown and incoming material isolation instructions.

9. The method for controlling the process parameters of aluminum plate surface treatment driven by vision inspection according to claim 1, characterized in that, Based on the fusion analysis of the aforementioned control timing chain and the current processing parameters, control correction parameters are identified, and parameter control instructions are generated, including: Obtain the real-time and set values ​​of the current processing parameters, integrate the control requirements in the control timing chain with the current parameters, and perform control conflict detection and resolution. Among them, when there are multiple time-point control requirements for the same parameter in the control time series, the control is merged or decomposed based on time proximity and effect accumulation. When there are mutually exclusive regulatory needs in the regulatory time-series chain, priority arbitration is carried out based on the quality impact weight and execution feasibility. Based on fusion analysis, the process parameter control relationship analysis is used to determine the control and correction parameters for each process parameter, including the correction target value, change rate constraint, execution time and safety boundary. Parameter control instructions are then generated and sent to the actuator after verification.

10. A vision-inspection-driven aluminum plate surface treatment process parameter control system, characterized in that, The system is used to execute the vision-driven aluminum plate surface treatment process parameter control method as described in any one of claims 1-9, and the system includes: The image recognition module is used to deploy image acquisition equipment in the related processes of aluminum plate surface treatment, import the acquired real-time processing images into the image processing path of the corresponding process, and extract the process image recognition results according to the processing rules matched with the related process. The feature parsing module is used to perform multi-granularity window feature parsing based on the process image recognition results to obtain multi-temporal surface features; The positioning module is used to optimize and locate the process processing timing based on the multi-temporal surface features and construct a control timing chain. The analysis module is used to perform fusion analysis based on the control timing chain and the current processing parameters, sort out the control correction parameters, and generate parameter control instructions.

Images