Gray scale verification method and system

Abstract

The invention provides a gray scale verification method and system, and the method comprises the steps: enabling a target industrial camera to respond to an exposure triggering signal, collecting the actual product image data of a target production node, and transmitting the actual product image data to a production machine; the production machine inputs the actual product image data into the production quality inspection module of the original version for product quality inspection reasoning to obtain a first quality inspection result, the actual product image data and the first quality inspection result are sent to the gray scale machine, the gray scale machine inputs the actual product image data into the production quality inspection module of the latest version for product quality inspection reasoning, and the first quality inspection result is obtained; and performing gray scale verification based on a comparison result of the first quality inspection result and the second quality inspection result to obtain a gray scale verification result. According to the application, verification of the new-version production quality inspection module is completed on the gray scale machine by using the data of the production machine, direct influence on the production machine and a production line is avoided, and verification cost is reduced.

Description

Technical Field

[0001] This application relates to the field of upgrading production quality inspection modules, and in particular to a grayscale verification method and system. Background Technology

[0002] AI machine vision production lines are intelligent production lines that use AI algorithms to empower machine vision technology, enabling automated detection, positioning, and recognition throughout the entire industrial production process. Their core value lies in improving efficiency, reducing errors, and minimizing reliance on manual labor. AI machine vision production lines typically include a Programmable Logic Controller (PLC), production machines, onboard equipment located at different production nodes, and visual monitoring equipment (such as cameras). The production machines run a stable version of a production quality inspection module, which can perform quality inspection inference based on images captured by the visual monitoring equipment. The PLC sends trigger signals to the visual monitoring equipment to control it to take pictures and apply lighting. The product images captured by the visual monitoring equipment are then transmitted to the production quality inspection module on the production machines for quality inspection, and the production process is adjusted based on the inspection results.

[0003] During the production process, production machines inevitably face the need to upgrade the production quality inspection module. One of the existing technology-provided upgrade methods for the production quality inspection module is to upgrade it directly on the production machine. However, since this method cannot directly verify the functionality of the new version of the production quality inspection module, if the new version of the production quality inspection module has bugs, it will lead to the risk of production line shutdown and loss of production efficiency.

[0004] To ensure stable production, another upgrade method for the production quality inspection module provided by existing technology requires setting up a separate test production machine with the same hardware environment as the original production machine. The new version of the production quality inspection module is first deployed on the test production machine, and then the new version of the production quality inspection module is run on the test production machine to verify its stability. This method requires setting up a separate test production machine identical to the original production machine, which increases the cost of the entire upgrade process. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide at least one grayscale verification method and system, which uses data from the production machine to complete the verification of the new version of the production quality inspection module on the grayscale machine, thereby avoiding direct impact on the production machine and production line and reducing verification costs.

[0006] This application mainly includes the following aspects: In a first aspect, embodiments of this application provide a grayscale verification method applied to a grayscale verification system. The grayscale verification system includes a programmable logic controller (PLC), a production machine, a grayscale machine, and industrial cameras set at different production nodes on a production line. The method includes: the PLC executing: generating an exposure trigger signal for a target industrial camera and sending the exposure trigger signal to the target industrial camera; the target industrial camera executing: in response to the exposure trigger signal, acquiring actual product image data at its target production node and sending it to the corresponding production machine on the production line; the production machine executing: inputting the actual product image data into the original version of the production quality inspection module for product quality inspection inference to obtain a first quality inspection result, and sending the actual product image data and the first quality inspection result to the grayscale machine; the grayscale machine executing: inputting the actual product image data into the latest version of the production quality inspection module for product quality inspection inference to obtain a second quality inspection result, and performing grayscale verification based on the comparison result of the first and second quality inspection results to obtain a grayscale verification result. The grayscale verification result indicates whether the latest version of the production quality inspection module can operate stably on the production machine.

[0007] In one possible implementation, the grayscale verification system further includes a lighting device corresponding to each industrial camera, wherein the method further includes: a programmable logic controller, which further performs: generating a lighting signal while generating an exposure trigger signal for the target industrial camera, and sending the exposure trigger signal to the target industrial camera while simultaneously sending the lighting signal to the target lighting device corresponding to the target industrial camera; the target lighting device performs: in response to the lighting signal, lighting the target industrial camera according to the lighting parameters carried by the lighting signal.

[0008] In one possible implementation, the method further includes: a programmable logic controller, which also performs: forwarding the exposure trigger signal corresponding to the target industrial camera and the lighting signal corresponding to the target lighting device to the grayscale machine through the production machine; the grayscale machine performs: in response to the exposure trigger signal and the lighting signal, simulating the product image acquisition process to obtain test product image data, and replacing the test product image data with actual product image data input to the test product image data for quality inspection inference.

[0009] In one possible implementation, the grayscale machine simulates the product image acquisition process in the following way: inputting a lighting signal into a virtual lighting device to complete the virtual lighting process; inputting an exposure trigger signal into a virtual camera to obtain test product image data output by the virtual camera.

[0010] In one possible implementation, the product image acquisition process is also simulated in the following way: a lighting signal is input to an experimental lighting device connected to a grayscale machine, so that the experimental lighting device completes the lighting of the experimental camera based on the lighting parameters carried by the lighting signal; an exposure trigger signal is sent to the experimental camera connected to the grayscale machine to obtain the test product image data fed back by the experimental camera.

[0011] In one possible implementation, the programmable logic controller also performs the following actions: receiving a quality inspection trigger signal sent by the production equipment under the target production node; and generating an exposure trigger signal and a lighting signal for the target industrial camera in response to the quality inspection trigger signal.

[0012] In one possible implementation, the first quality inspection result includes multiple first quality inspection indicators, and the second quality inspection result includes a second quality inspection indicator corresponding to each first quality inspection indicator. The grayscale machine obtains the grayscale verification result in the following manner: for each first quality inspection indicator, it invokes the quality inspection strategy corresponding to that first quality inspection indicator to determine the quality inspection comparison result between the first quality inspection indicator and its corresponding second quality inspection indicator. The quality inspection strategy describes the consistency comparison rules for the quality inspection of the first quality inspection indicator and its corresponding second quality inspection indicator. The quality inspection comparison result corresponding to each first quality inspection indicator is pushed to a preset audit object to obtain the grayscale verification result fed back by the preset audit object based on multiple quality inspection comparison results.

[0013] In one possible implementation, the method further includes: a programmable logic controller (PLC) executing: in response to a quality inspection trigger signal, sequentially generating multiple exposure trigger signals within a preset trigger period and sending them to a target industrial camera to trigger continuous image capture; the target industrial camera executing: sequentially responding to each exposure trigger signal, acquiring multiple actual product image data corresponding to a target production node and sending them to the production machine corresponding to the production line; the production machine executing: according to a preset sampling ratio, extracting at least one target actual product image data from the multiple actual product image data, inputting each target actual product data into the original version of the production quality inspection module for product quality inspection inference, obtaining a first quality inspection result corresponding to each target actual product data, and sending each target actual product data and its corresponding first quality inspection result to a grayscale machine; the grayscale machine further executing: inputting multiple target actual product image data into the latest version of the production quality inspection module for product quality inspection inference, obtaining a second quality inspection result corresponding to each target actual product image data, comparing the first and second quality inspection results corresponding to each target actual product image data, and performing grayscale verification based on the comparison results corresponding to each target actual product image data to obtain a grayscale verification result.

[0014] In one possible implementation, the comparison results include multiple quality inspection comparison results corresponding one-to-one with multiple quality inspection indicators. The grayscale machine further performs the following: inputting the multiple quality inspection comparison results corresponding to each target actual product image data into a pre-created consistency comparison model to obtain an overall consistency index for the quality inspection results corresponding to the target actual product image data; calculating the target proportion of target actual product image data whose overall similarity index exceeds a preset overall consistency threshold among all target actual product image data; pushing the overall similarity index and target proportion corresponding to each target actual product image data to a preset review object through a preset channel; and obtaining the grayscale verification results fed back by the preset review object based on the overall similarity index and target proportion corresponding to each target actual product image data.

[0015] Secondly, this application also provides a grayscale verification system, which includes a programmable logic controller (PLC), a production machine, a grayscale machine, and industrial cameras set at different production nodes on the production line. The PLC generates an exposure trigger signal for a target industrial camera and sends the signal to the target camera. The target industrial camera, in response to the exposure trigger signal, acquires actual product image data at its target production node and sends it to the corresponding production machine. The production machine inputs the actual product image data into the original version of the production quality inspection module for product quality inspection inference, obtains a first quality inspection result, and sends the actual product image data and the first quality inspection result to the grayscale machine. The grayscale machine inputs the actual product image data into the latest version of the production quality inspection module for product quality inspection inference, obtains a second quality inspection result, and performs grayscale verification based on the comparison between the first and second results to obtain a grayscale verification result. The grayscale verification result indicates whether the latest version of the production quality inspection module can operate stably on the production machine.

[0016] This application provides a grayscale verification method and system, comprising: a target industrial camera responding to an exposure trigger signal to acquire actual product image data at the target production node and sending it to the production machine; the production machine inputting the actual product image data into the original version of the production quality inspection module for product quality inspection inference to obtain a first quality inspection result; sending the actual product image data and the first quality inspection result to a grayscale machine; the grayscale machine inputting the actual product image data into the latest version of the production quality inspection module for product quality inspection inference to obtain a second quality inspection result; and performing grayscale verification based on the comparison between the first and second quality inspection results to obtain a grayscale verification result. This application utilizes the data from the production machine to complete the verification of the new version of the production quality inspection module on the grayscale machine, avoiding direct impact on the production machine and production line, and improving the reliability of module upgrades.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram illustrating the structure of a grayscale verification system provided in an embodiment of this application; Figure 2 This illustration shows an interactive diagram of a grayscale verification system executing a corresponding grayscale verification method, as provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0021] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] AI machine vision production lines are intelligent production lines that use AI algorithms to empower machine vision technology, enabling automated detection, positioning, and recognition throughout the entire industrial production process. Their core value lies in improving efficiency, reducing errors, and minimizing reliance on manual labor. AI machine vision production lines typically include a Programmable Logic Controller (PLC), production machines, onboard equipment located at different production nodes, and visual monitoring equipment (such as cameras). The production machines run a stable version of a production quality inspection module, which can perform quality inspection inference based on images captured by the visual monitoring equipment. The PLC sends trigger signals to the visual monitoring equipment to control it to take pictures and apply lighting. The product images captured by the visual monitoring equipment are then transmitted to the production quality inspection module on the production machines for quality inspection, and the production process is adjusted based on the inspection results.

[0023] During the production process, production machines inevitably face the need to upgrade the production quality inspection module. One of the existing technology-provided upgrade methods for the production quality inspection module is to upgrade it directly on the production machine. However, since this method cannot directly verify the functionality of the new version of the production quality inspection module, if the new version of the production quality inspection module has bugs, it will lead to the risk of production line shutdown and loss of production efficiency.

[0024] To ensure stable production, another upgrade method for the production quality inspection module provided by existing technology requires setting up a separate test production machine with the same hardware environment as the original production machine. The new version of the production quality inspection module is first deployed on the test production machine, and then production is carried out using the test production machine to verify the stability of the new version of the production quality inspection module. This method has at least the following disadvantages: (1) A separate test production machine identical to the original production machine is required, which increases the cost of the entire upgrade process.

[0025] (2) Each upgrade of the production quality inspection module may introduce new bugs, leading to production line shutdown.

[0026] (3) The production machine must be kept running stably, and the new version cannot be verified directly on site.

[0027] Based on this, this application provides a grayscale verification method and system. By utilizing data from the production machine to complete the verification of the new version of the production quality inspection module on a grayscale machine, it avoids direct impact on the production machine and production line, and reduces verification costs. The details are as follows: Please see Figure 1 , Figure 1 This diagram illustrates the structure of a grayscale verification system provided in an embodiment of this application. Figure 1As shown in the embodiment of this application, the grayscale verification system is applied to a complete production line. The grayscale verification system includes a programmable logic controller (PLC), a production machine 1, a grayscale machine 2, and industrial cameras 31, 32...3N set at different production nodes of the production line. The PLC is connected to the industrial cameras 31, 32...3N and the production machine 1, respectively. The production machine 1 is also connected to the grayscale machine 2 and the industrial cameras 31, 32...3N, respectively. Each industrial camera indicates a camera cluster including at least one industrial camera.

[0028] In one specific embodiment provided in this application, each production line is a complete processing and production chain for producing one product, and each production line is divided into multiple production nodes according to the product processing and production sequence. Each production node has corresponding production equipment to complete the product processing of that production node. For the AI ​​machine vision production line, a camera cluster (i.e., at least one industrial camera) is set at the corresponding position of each production node to collect images of the products processed by the production node. The collected product images are used for subsequent evaluation of the quality of the products processed by the production node.

[0029] The programmable logic controller (PLC) provided in this application is the core device of the programmable logic controller (PLC) and is used to control the operation of the corresponding industrial cameras on the production line according to the signals sent by the production equipment at each production node.

[0030] The production machine provided in this application is pre-deployed with multiple first functional modules, including but not limited to at least one of the following: a production quality inspection module, a production management module, and a production monitoring module. The production quality inspection module is responsible for quality inspection reasoning tasks for each production node on the production line, the production management module is responsible for pushing production tasks and controlling production start and stop for each production node on the production line, and the production monitoring module is responsible for monitoring the production nodes using industrial cameras installed on the production nodes.

[0031] The grayscale machine 2 provided in this application is a small portable computer used for grayscale verification of the new version of the production quality inspection module. Its hardware configuration is lower than that of the production machine, but it can complete the simulation of camera control and support the operation of the new version of the production quality inspection module, as well as the comparison and verification process between the quality inspection results corresponding to the production machine and the quality inspection results corresponding to the grayscale machine.

[0032] In a preferred embodiment, please refer to Figure 2 , Figure 2 This illustration shows an interactive diagram of a grayscale verification system executing a corresponding grayscale verification method, as provided in an embodiment of this application. Figure 2 As shown, the method provided in this application includes: Programmable Logic Controller (PLC) executes: S100: Generates an exposure trigger signal for the target industrial camera.

[0033] S101, Send the exposure trigger signal to the target industrial camera.

[0034] Target industrial camera 3X (X is one of 1 to N), execute: S200: In response to the exposure trigger signal, it acquires actual product image data of the target production node where the target industrial camera is located.

[0035] S201. Send the actual product image data to the corresponding production machine on the production line.

[0036] Production machine 1, executing: S300: Input the actual product image data collected by the target industrial camera into the original version production quality inspection module to perform product quality inspection reasoning and obtain the first quality inspection result.

[0037] The original version of the production quality inspection module supports its stable operation on production machine 1.

[0038] S301. Send the actual product image data and the first quality inspection result to the grayscale machine.

[0039] Grayscale machine 2, execute: S400: Input the actual product image data into the latest version of the production quality inspection module to perform product quality inspection reasoning and obtain the second quality inspection result.

[0040] S401. Compare the first quality inspection result and the second quality inspection result to obtain the comparison result.

[0041] S402. Based on the comparison results, perform grayscale verification on the new version of the production quality inspection module to obtain the grayscale verification results.

[0042] The grayscale verification results indicate whether the latest version of the production quality inspection module can run stably on the production machine.

[0043] In a preferred embodiment, in the method provided by this application, specifically in step S1001, for each production node on the production line, during the execution of the corresponding production task, the production equipment on the production node generates a quality inspection trigger signal after completing the processing of one batch of products. The quality inspection trigger signal is sent to the programmable logic controller (PLC) through the corresponding communication tool mounted on the production equipment. The quality inspection trigger signal carries the identifier of the target production node to which it belongs. After receiving the quality inspection trigger signal, the PLC generates an exposure trigger signal for the target industrial camera and a lighting signal in response to the quality inspection trigger signal.

[0044] The grayscale verification system provided in this application also includes lighting devices corresponding to industrial cameras. When the programmable logic controller (PLC) sends the exposure trigger signal to the target industrial camera 3X, it also sends the lighting signal to the target lighting device corresponding to the target industrial camera 3X. In this application, the target lighting device responds to the lighting signal and illuminates the target industrial camera according to the lighting parameters carried by the lighting signal. The lighting parameters are pre-programmed by the programmable logic controller (PLC) and include, but are not limited to, at least one of the following: the operating power of the lighting device, the brightness adjustment value, and the light color control parameters.

[0045] In the preferred embodiments provided in this application, the method further includes: The programmable logic controller (PLC) also forwards the exposure trigger signal corresponding to the target industrial camera and the lighting signal corresponding to the target lighting equipment to the grayscale machine 2 through the production machine 1. The grayscale machine 2 performs the following steps: responding to the exposure trigger signal, simulating the product image acquisition process, obtaining test product image data, replacing the test product image data with the actual product image data transmitted from the production machine 1, and then further executing steps S400 to S402.

[0046] In a preferred embodiment, the grayscale machine 2 of this application provides two methods for simulating the product image acquisition process, one of which is: The lighting signal is input into the virtual lighting device to complete the virtual lighting process. The exposure trigger signal is input into the virtual camera to obtain the test product image data output by the virtual camera.

[0047] In a specific example, this application 2 also pre-deploys a virtual camera and a virtual lighting device. The virtual camera can simulate the camera functions of a real industrial camera, such as exposure response, exposure time, frame rate, cropping, and shooting (achieved by generating virtual images). The virtual lighting device can simulate the lighting process of a real lighting device. In this application, after the grayscale machine 2 receives the exposure trigger signal and lighting signal forwarded from the industrial machine 1, it transmits the lighting signal to the virtual lighting device and uses the virtual lighting device to complete the virtual lighting process. The exposure trigger signal is then transmitted to the virtual camera. The virtual camera responds to the exposure trigger signal and begins to simulate an image acquisition process similar to that of an industrial camera to obtain test product image data. At this time, the obtained test product image data is a purely electronically generated image automatically generated inside the virtual camera.

[0048] In another preferred embodiment, the grayscale machine 2 of this application provides another method for simulating the product image acquisition process: The lighting signal is input to the experimental lighting equipment connected to the grayscale machine, so that the experimental lighting equipment can illuminate the experimental camera based on the lighting parameters carried by the lighting signal. The exposure trigger signal is sent to the experimental camera connected to the grayscale machine to obtain the test product image data fed back by the experimental camera.

[0049] In another specific example, the grayscale machine 1 of this application 2 is also connected to an experimental lighting device and an experimental camera. The experimental lighting device is of the same model as the lighting device connected to the industrial camera, and the experimental camera is of the same model as the industrial camera. After receiving the lighting signal and the exposure trigger signal, the grayscale machine 1 inputs the lighting signal into the experimental lighting device and the exposure trigger signal into the experimental camera, respectively. The experimental lighting device responds to the lighting signal and completes the lighting of the experimental camera according to the lighting parameters corresponding to the lighting signal. The experimental camera responds to the exposure trigger signal and completes the acquisition of the test product image data. In this application, the experimental camera is set near the grayscale machine 2 and is not set on the corresponding production node of the production line. Therefore, the image captured by the experimental camera is not the production node image, but the image at the position where the camera of the experimental camera is pointed.

[0050] In this application, regardless of which product image acquisition process simulation method is used, the final test product image data is replaced with the actual product image data transmitted from the production machine 1 before steps S400 to S402 are executed.

[0051] In a preferred embodiment, the first quality inspection result includes multiple first quality inspection indicators, and the second quality inspection result includes a second quality inspection indicator corresponding to each first quality inspection indicator, wherein the grayscale machine obtains the grayscale verification result in the following manner: For each first quality inspection indicator, the quality inspection strategy corresponding to the first quality inspection indicator is invoked to determine the quality inspection comparison result between the first quality inspection indicator and its corresponding second quality inspection indicator. The quality inspection strategy describes the consistency comparison rules of the first quality inspection indicator and its corresponding second quality inspection indicator. The quality inspection comparison result corresponding to each first quality inspection indicator is pushed to the preset audit object to obtain the grayscale verification result fed back by the preset audit object based on multiple quality inspection comparison results.

[0052] In one specific embodiment, the quality inspection indicators include, but are not limited to, at least one of the following: classification results, defect detection boxes, and quality inspection inference time. The classification results of this application include products with defects and products without defects. The defect detection boxes are used to locate defects in the input image. The quality inspection inference time describes the time it takes for the production quality inspection module to complete the entire quality inspection inference process and obtain the quality inspection results.

[0053] In this application, different quality inspection strategies are pre-set for different quality inspection indicators. Taking the classification result as an example, the corresponding quality inspection strategy is as follows: if the first classification result output by the original version production quality inspection model and the second classification result output by the new version production quality inspection model are both either defective or non-defective, it is determined that the classification results of the original version production quality inspection model and the new version production quality inspection model are consistent. If the first classification result and the second classification result are mutually exclusive, it is determined that the classification results of the original version production quality inspection model and the new version production quality inspection model are inconsistent.

[0054] Taking a defect detection box as an example, the corresponding quality inspection strategy is as follows: If the size difference between the first defect detection box output by the original version production quality inspection model and the second defect detection box output by the new version production quality inspection model is within a preset size difference range, it is determined that the defect detection box sizes of the original version production quality inspection model and the new version production quality inspection model are consistent. If the size difference between the first and second defect detection boxes exceeds the preset size difference range, it is determined that the defect detection box sizes of the original version production quality inspection model and the new version production quality inspection model are inconsistent. If the center point positioning error between the first and second defect detection boxes is within a preset positioning error range, it is determined that the defect detection box positioning of the original version production quality inspection model and the new version production quality inspection model is consistent. If the center point positioning error between the first and second defect detection boxes exceeds the preset positioning error range, it is determined that the defect detection box positioning of the original version production quality inspection model and the new version production quality inspection model is inconsistent.

[0055] Taking the quality inspection inference time as an example, the corresponding quality inspection strategy is as follows: if the difference between the first quality inspection inference time output by the original version production quality inspection model and the second quality inspection inference time output by the new version production quality inspection model is within the preset time difference range, it is determined that the quality inspection inference time of the original version production quality inspection model and the new version production quality inspection model are consistent; if the difference between the first quality inspection inference time and the second quality inspection inference time exceeds the preset time difference range, it is determined that the quality inspection inference time of the original version production quality inspection model and the new version production quality inspection model are inconsistent.

[0056] For Grayscale Machine 2: After determining the quality inspection comparison results of each first quality inspection indicator and its corresponding second quality inspection indicator, the quality inspection comparison results under each quality inspection indicator are pushed to the preset review object for review through a preset channel. In this application, in order to ensure the effectiveness of grayscale verification, multi-level review is usually involved. Each level of review corresponds to a different preset review object. Multi-level review needs to complete the review process step by step. Each time a level of review is executed, if the review result is passed, the next level of review is executed until all levels of review are completed. That is, if all levels of review are passed, the new version of the production quality inspection module is directly determined to have passed the grayscale verification and the result is fed back to Grayscale Machine 2. If the review result is not passed, the multi-level review process is directly ended, and the new version of the production quality inspection module is determined to have failed the grayscale verification and the result is fed back to Grayscale Machine.

[0057] In this application, for each level of the review process, the corresponding preset review object performs final grayscale verification and confirmation based on the quality inspection comparison results under each quality inspection indicator, and the preset review object issues the corresponding review results through the corresponding review interface.

[0058] In a preferred embodiment, the method provided in this application further includes: The programmable logic controller performs the following actions: in response to a quality inspection trigger signal, it generates multiple consecutive exposure trigger signals for the target industrial camera and sends the multiple exposure trigger signals to the target industrial camera in sequence. The target industrial camera performs the following actions: sequentially responds to each exposure trigger signal, collects multiple actual product image data corresponding to the target production node, and sends them to the corresponding production machine on the production line. The production machine executes the following steps: According to a preset sampling ratio, it extracts at least one target actual product image data from multiple actual product image data. It then performs product quality inspection reasoning on each target actual product data by inputting the original version production quality inspection module, obtains the first quality inspection result corresponding to each target actual product data, and sends each target actual product data and its corresponding first quality inspection result to the grayscale machine.

[0059] In a preferred embodiment, the method provided in this application further includes: The programmable logic controller (PLC) responds to the quality inspection trigger signal and generates multiple exposure trigger signals in sequence within a preset trigger cycle. These signals are then sent to the target industrial camera 3X to trigger continuous photography. The target industrial camera 3X responds to each exposure trigger signal in sequence, collects multiple actual product image data corresponding to the target production node, and sends them to the corresponding production machine 1 on the production line.

[0060] Production machine 1 continuously receives multiple actual product image data sent by target industrial camera 3X within a preset trigger cycle and extracts at least one target actual product image data from the multiple actual product image data according to a preset sampling ratio (this ratio is adjustable). Each target actual product data is input into the original version production quality inspection module for product quality inspection reasoning to obtain the first quality inspection result corresponding to each target actual product data. Each target actual product data and its corresponding first quality inspection result are sent to grayscale machine 2.

[0061] Grayscale machine 2 inputs multiple target actual product image data into the latest version of the production quality inspection module for product quality inspection inference, and obtains the second quality inspection result corresponding to each target actual product image data. For each target actual product image data, the first quality inspection result and the second quality inspection result corresponding to the target actual product image data are compared to obtain the comparison result. Specifically, the process of determining the comparison result corresponding to each target actual product image data (including the quality inspection comparison result corresponding to each quality inspection indicator) is similar to the above, and will not be elaborated on here. Grayscale verification is performed based on the comparison result corresponding to each target actual product image data to obtain the grayscale verification result.

[0062] Based on the above process provided in this application, the grayscale machine 2 of this application also performs: Multiple quality inspection comparison results corresponding to each target actual product image data are input into a pre-created consistency comparison model to obtain the overall consistency index of the quality inspection results corresponding to the target actual product image data. Specifically, the consistency comparison model can be a weighted model. As mentioned above, the quality inspection comparison results are generally described as the consistency or inconsistency of the quality inspection indicators, such as the consistency and inconsistency of the classification results. Therefore, for each quality inspection indicator, when the quality inspection comparison result corresponding to the quality inspection indicator is consistent, a first preset index value (e.g., 1) is assigned to the quality inspection indicator; when the quality inspection comparison result corresponding to the quality inspection indicator is inconsistent, a second preset index value (e.g., 0) is assigned to the quality inspection indicator. In this application, the quality inspection comparison result corresponding to each quality inspection indicator includes at least one comparison dimension (taking the defect detection box as an example, it includes size consistency and positioning consistency). Therefore, each dimension must be assigned a value separately. Then, based on actual needs, the reference weight of the quality inspection comparison result under each dimension corresponding to each quality inspection indicator is pre-set. Then, each quality inspection comparison result corresponding to the target actual product image is weighted according to the corresponding reference weight to obtain the overall consistency index corresponding to the target actual product image data.

[0063] The target proportion of actual product image data whose overall similarity index exceeds the preset overall consistency threshold is calculated among all actual product image data. The overall similarity index and target proportion corresponding to each actual product image data are pushed to the preset review objects through preset channels, and the grayscale verification results fed back by the preset review objects based on the overall similarity index and target proportion corresponding to each actual product image data are obtained.

[0064] In this application, a step-by-step review is also adopted, and the step-by-step review method is similar to that described above, so it will not be elaborated on here. For the pre-set reviewers, they can measure whether the grayscale verification corresponding to the new version of the production quality inspection module has passed by measuring the overall similarity index and target proportion of the actual product image data of each target. The final grayscale verification result of this application is ultimately determined by the review result given by the pre-set reviewers, similar to the review process described above, so it will not be elaborated on here.

[0065] The advantages of this application are: (1) While the production machine 1 is running a stable version of the production quality inspection module, a small grayscale machine 2 is deployed for the verification of the new version of the production quality inspection module. This can ensure the continuity of production, realize the sustainable operation of the production line, avoid direct impact on the production machine and production line, and prevent the entire production line from stopping production due to the upgrade of the production quality inspection software.

[0066] (2) Grayscale machine 2 can simulate the image acquisition process of the production machine through a virtual camera and virtual lighting equipment, and can also connect to a real camera and lighting equipment to simulate the image acquisition process of the production machine. In the inference process, the image acquired by the production machine is used to replace the image, ensuring that the input of the new version of the production quality inspection module is consistent with the input of the original version of the production quality inspection module installed on the production machine. From the aspects of image acquisition simulation and data input, the consistency between the grayscale machine execution process and the production machine execution process is ensured, which makes it easier for the subsequent verification results to be highly close to the production environment and improve reliability.

[0067] (3) The grayscale machine can be configured with a sampling ratio for verification according to actual needs (e.g., based on production line load and its own performance limitations), reducing the impact on the performance and bandwidth of the grayscale machine, and reducing the pressure on network and hardware resources while ensuring the breadth of verification coverage.

[0068] (4) The quality inspection results corresponding to the production machine and the grayscale machine are pushed to a client by a preset reviewer through a preset channel to support manual review. Since the quality inspection results may overlook some potential risks, this application introduces a manual review process after the results, does not rely entirely on the quality inspection results, and uses manual identification of some potential problems in the new version to reduce the upgrade risk.

[0069] (5) The grayscale machine provided in this application is small and portable and can be used on multiple production lines. The factory does not need to equip each production line with complete hardware, thus reducing deployment costs.

[0070] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0071] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0072] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0073] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A grayscale verification method, characterized in that, This technology is applied to grayscale verification systems, which include programmable logic controllers, production machines, grayscale machines, and industrial cameras installed at different production nodes on the production line. The method includes: The programmable logic controller performs the following actions: generating an exposure trigger signal for a target industrial camera and sending the exposure trigger signal to the target industrial camera; The target industrial camera performs the following action: in response to the exposure trigger signal, it acquires actual product image data of the target production node and sends it to the corresponding production machine on the production line. The production machine executes the following steps: inputting the actual product image data into the original version production quality inspection module to perform product quality inspection reasoning, obtaining a first quality inspection result, and sending the actual product image data and the first quality inspection result to the grayscale machine. The grayscale machine performs the following steps: inputting the actual product image data into the latest version of the production quality inspection module to perform product quality inspection reasoning, obtaining a second quality inspection result, and performing grayscale verification based on the comparison result between the first quality inspection result and the second quality inspection result to obtain a grayscale verification result. The grayscale verification result indicates whether the latest version of the production quality inspection module can operate stably on the production machine.

2. The method according to claim 1, characterized in that, The grayscale verification system also includes lighting equipment corresponding to each industrial camera. The method further includes: The programmable logic controller further performs the following actions: while generating an exposure trigger signal for the target industrial camera, it also generates a lighting signal, sends the exposure trigger signal to the target industrial camera, and sends the lighting signal to the target lighting device corresponding to the target industrial camera. The target lighting device performs the following: in response to the lighting signal, it illuminates the target industrial camera according to the lighting parameters carried by the lighting signal.

3. The method according to claim 2, characterized in that, The method further includes: The programmable logic controller also performs the following: forwarding the exposure trigger signal corresponding to the target industrial camera and the lighting signal corresponding to the target lighting equipment to the grayscale machine through the production machine; The grayscale machine performs the following actions: responding to the exposure trigger signal and the lighting signal, simulating the product image acquisition process to obtain test product image data, and replacing the test product image data with actual product image data for quality inspection inference.

4. The method according to claim 3, characterized in that, The grayscale machine simulates the product image acquisition process in the following way: The lighting signal is input into the virtual lighting device, and the virtual lighting process is completed using the virtual lighting device. The exposure trigger signal is input into the virtual camera to obtain the test product image data output by the virtual camera.

5. The method according to claim 3, characterized in that, The product image acquisition process is also simulated in the following ways: The lighting signal is input to the experimental lighting equipment connected to the grayscale machine, so that the experimental lighting equipment can complete the lighting of the experimental camera based on the lighting parameters carried by the lighting signal. The exposure trigger signal is sent to the experimental camera connected to the grayscale machine to obtain the test product image data fed back by the experimental camera.

6. The method according to claim 2, characterized in that, The programmable logic controller also performs: Receive quality inspection trigger signals sent by production equipment under the target production node; In response to the quality inspection trigger signal, an exposure trigger signal and a lighting signal are generated for the target industrial camera.

7. The method according to claim 1, characterized in that, The first quality inspection result includes multiple first quality inspection indicators, and the second quality inspection result includes the second quality inspection indicators corresponding to each first quality inspection indicator. The grayscale machine obtains the grayscale verification result in the following manner: For each first quality inspection indicator, the quality inspection strategy corresponding to the first quality inspection indicator is invoked to determine the quality inspection comparison result between the first quality inspection indicator and its corresponding second quality inspection indicator. The quality inspection strategy describes the consistency comparison rules between the first quality inspection indicator and its corresponding second quality inspection indicator. The quality inspection comparison results corresponding to each first quality inspection indicator are pushed to the preset review object, and the gray-scale verification results fed back by the preset review object based on multiple quality inspection comparison results are obtained.

8. The method according to claim 1, characterized in that, The method further includes: The programmable logic controller performs the following actions: in response to the quality inspection trigger signal, it sequentially generates multiple exposure trigger signals within a preset trigger period and sends them to the target industrial camera to trigger the target industrial camera to take continuous pictures. The target industrial camera performs the following actions: sequentially responds to each exposure trigger signal, collects multiple actual product image data corresponding to the target production node, and sends them to the corresponding production machine on the production line. The production machine executes the following steps: According to a preset sampling ratio, it extracts at least one target actual product image data from multiple actual product image data, inputs each target actual product data into the original version production quality inspection module for product quality inspection reasoning, obtains the first quality inspection result corresponding to each target actual product data, and sends each target actual product data and its corresponding first quality inspection result to the grayscale machine. The grayscale machine also performs the following steps: inputting multiple target actual product image data into the latest version of the production quality inspection module to perform product quality inspection reasoning, obtaining a second quality inspection result corresponding to each target actual product image data, comparing the first quality inspection result and the second quality inspection result corresponding to each target actual product image data for each target actual product image data, and performing grayscale verification based on the comparison result corresponding to each target actual product image data to obtain a grayscale verification result.

9. The method according to claim 8, characterized in that, The comparison results include multiple quality inspection comparison results that correspond one-to-one with multiple quality inspection indicators. The grayscale machine also performs the following: Input the multiple quality inspection comparison results corresponding to each target actual product image data into the pre-created consistency comparison model to obtain the overall consistency index of the quality inspection results corresponding to the target actual product image data. The percentage of actual product image data whose overall similarity index exceeds the preset overall consistency threshold is calculated among all actual product image data. The overall similarity index and the target proportion corresponding to the actual product image data of each target are pushed to the preset review objects through preset channels; Obtain the grayscale verification results of the preset audit objects based on the overall similarity index corresponding to the actual product image data of each target and the target proportion.

10. A grayscale verification system, characterized in that, The grayscale verification system includes a programmable logic controller, a production machine, a grayscale machine, and industrial cameras installed at different production nodes on the production line. The programmable logic controller (PLC) performs the following actions: generating an exposure trigger signal for the target industrial camera and sending the exposure trigger signal to the target industrial camera; The target industrial camera performs the following action: in response to the exposure trigger signal, it acquires actual product image data of the target production node and sends it to the corresponding production machine on the production line. The production machine executes the following steps: inputting the actual product image data into the original version production quality inspection module to perform product quality inspection reasoning, obtaining a first quality inspection result, and sending the actual product image data and the first quality inspection result to the grayscale machine. The grayscale machine performs the following steps: inputting the actual product image data into the latest version of the production quality inspection module to perform product quality inspection reasoning, obtaining a second quality inspection result, and performing grayscale verification based on the comparison result between the first quality inspection result and the second quality inspection result to obtain a grayscale verification result. The grayscale verification result indicates whether the latest version of the production quality inspection module can operate stably on the production machine.

Images