Nozzle atomization identification method and device

Abstract

The invention discloses a nozzle atomization identification method and device, and the method is applied to the technical field of big data, and the method comprises the steps: obtaining a nozzle atomization image, and the nozzle atomization image is used for indicating an image of a nozzle during operation; the nozzle atomization image is input into a recognition model for spraying angle recognition, and the recognition angle of the nozzle atomization image is obtained; comparing the nozzle atomization image with a preset atomization image corresponding to the identification angle to obtain a comparison result; and determining a nozzle atomization identification result based on a comparison result. According to the method, an automatic nozzle atomization recognition scheme is provided, and the recognition efficiency of nozzle atomization recognition can be improved.

Description

Technical Field

[0001] This application relates to the field of big data technology, and in particular, to a nozzle atomization recognition method and device. Background Art

[0002] A coating machine is a device used for organic film coating of materials such as tablets, pills, and candies. Coating can protect drugs and improve their stability. In the coating process of coating machines in the pharmaceutical field, the atomization effects produced by different slurry types of nozzles under different atomization pressures and fog width pressures are different.

[0003] Currently, the atomization effect produced by the nozzle of the coating machine is almost detected manually, and based on the manual detection results, it is judged whether the nozzle system that produces the current nozzle atomization effect needs to be adjusted. There are many influencing factors in the coating process of the coating machine, and abnormal nozzle atomization effects will lead to the failure of the coating process. For example, abnormal nozzle atomization effects can include dripping of coated tablets or unsatisfactory atomization effects.

[0004] However, the recognition efficiency of the above method for detecting the nozzle atomization effect manually is slow, and it cannot detect the coating process in real time, which is very likely to lead to the failure of the final coating process. Summary of the Invention

[0005] In view of this, the embodiments of this application provide a nozzle atomization recognition method and device, aiming to provide an automated nozzle atomization recognition solution that can improve the recognition efficiency of nozzle atomization recognition.

[0006] In a first aspect, the embodiments of this application provide a nozzle atomization recognition method, and the method includes:

[0007] Obtain a nozzle atomization image, where the nozzle atomization image is used to indicate the image of the nozzle during operation;

[0008] Input the nozzle atomization image into a recognition model for jet angle recognition to obtain the recognition angle of the nozzle atomization image;

[0009] Compare the nozzle atomization image with a preset atomization image corresponding to the recognition angle to obtain a comparison result;

[0010] Based on the comparison result, determine the nozzle atomization recognition result.

[0011] In a second aspect, the embodiments of this application provide a nozzle atomization recognition device, and the device includes:

[0012] An acquisition module, configured to acquire a nozzle atomization image, where the nozzle atomization image is used to indicate the image of the nozzle during operation;

[0013] An identification module, configured to input the nozzle atomization image into an identification model for spray angle identification, and obtain the identification angle of the nozzle atomization image;

[0014] A comparison module, configured to compare the nozzle atomization image with a preset atomization image corresponding to the identification angle, and obtain a comparison result;

[0015] A determination module, configured to determine a nozzle atomization identification result based on the comparison result.

[0016] In a third aspect, an embodiment of the present application provides a nozzle atomization identification device, and the device includes:

[0017] A memory, configured to store a computer program;

[0018] A processor, configured to execute the computer program so that the device executes the nozzle atomization identification method described in the foregoing first aspect.

[0019] In a fourth aspect, an embodiment of the present application provides a computer storage medium, and a computer program is stored on the computer-readable storage medium. When the computer program is run, the device running the computer program implements the nozzle atomization identification method described in the foregoing first aspect.

[0020] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0021] The embodiments of the present application provide a nozzle atomization identification method, and the method includes obtaining a nozzle atomization image, where the nozzle atomization image is used to indicate an image of the nozzle during operation, that is, first, an image of the nozzle during operation can be obtained to complete automatic identification based on the nozzle atomization image. Then, the nozzle atomization image can be input into an identification model for spray angle identification to determine the angle of the atomization effect generated by the nozzle during operation, and obtain the identification angle of the nozzle atomization image. After determining the identification angle of the nozzle atomization image, the nozzle atomization image can be compared with a preset atomization image corresponding to the determined identification angle, and a comparison result can be obtained. After determining the comparison result, the nozzle atomization identification result can be determined, that is, based on the comparison result between the nozzle atomization image and the preset atomization image corresponding to the identification angle, a nozzle atomization identification result for indicating whether the nozzle system that generates the current nozzle atomization image needs to be adjusted can be determined. It can be seen that through the above method, an automated nozzle atomization identification solution is provided. By inputting the nozzle atomization image into the identification model, the identification angle of the nozzle atomization image can be determined, and thus the nozzle atomization identification result can be further determined based on the identification angle, without manual detection of the nozzle atomization effect, thereby not only improving the identification efficiency of nozzle atomization identification, but also enabling real-time detection of the atomization effect generated by the nozzle. Description of the Drawings

[0022] To more clearly illustrate the technical solutions in this embodiment or the prior art, the following will briefly introduce the drawings required for the description of the embodiment or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0023] Figure 1 Schematic diagram of an application scenario of a nozzle atomization recognition method provided by an embodiment of the present application;

[0024] Figure 2 Flowchart of a nozzle atomization recognition method provided by an embodiment of the present application;

[0025] Figure 3 Detection schematic diagram of deploying a camera provided by an embodiment of the present application;

[0026] Figure 4 Schematic diagram of the structure of a nozzle system provided by an embodiment of the present application;

[0027] Figure 5a Schematic diagram of a jet atomization effect provided by an embodiment of the present application;

[0028] Figure 5b Another schematic diagram of a jet atomization effect provided by an embodiment of the present application;

[0029] Figure 5c Another schematic diagram of a jet atomization effect provided by an embodiment of the present application;

[0030] Figure 6 Schematic diagram of the structure of the DBNet model provided by an embodiment of the present application;

[0031] Figure 7 Application schematic diagram of a graphics processor provided by an embodiment of the present application;

[0032] Figure 8 Schematic diagram of a nozzle atomization recognition method provided by an embodiment of the present application;

[0033] Figure 9 Schematic diagram of the structure of a nozzle atomization recognition device provided by an embodiment of the present application. Detailed implementation manners

[0034] To enable those skilled in the art to better understand the solution of this application, the following will clearly and completely describe the technical solution in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts belong to the scope of protection of this application.

[0035] Currently, the atomization effect generated by the nozzle of the coating machine equipment is almost detected manually, and it is judged whether the nozzle system that generates the current nozzle atomization effect needs to be adjusted based on the manual detection results. When the coating process is scaled up, the coating liquid formula is fixed and the density remains unchanged, and the variable is the flow rate of the coating liquid. For the gas part, both the volume and density of the compressed air are related to the pressure, and both are variables. For the same nozzle, the gas flow rate, volume flow rate, and mass flow rate are proportional to the gas pressure. Therefore, when the atomization pressure remains unchanged, as the liquid inlet flow rate increases, the droplet diameter will become larger, and the final atomized liquid diameter depends on the complex interaction between the air mass flow rate and the gas flow rate.

[0036] Therefore, there are many influencing factors in the coating process of the coating machine equipment, and abnormal nozzle atomization effect will lead to the failure of the coating process. For example, abnormal nozzle atomization effect can include dripping of coated tablets or unsatisfactory atomization effect.

[0037] However, the recognition efficiency of the above method of detecting the nozzle atomization effect manually is slow, and it cannot detect the coating process in real time, which is very likely to lead to the failure of the final coating process.

[0038] Based on this, to solve the above problems, in the embodiments of the present application, a nozzle atomization image is acquired. The nozzle atomization image is used to indicate the image of the nozzle during operation. That is, first, the image of the nozzle during operation can be acquired to complete automatic recognition based on this nozzle atomization image. Then, the nozzle atomization image can be input into an identification model for jet angle identification to determine the angle of the atomization effect generated by the nozzle during operation, and obtain the identification angle of the nozzle atomization image. After determining the identification angle of the nozzle atomization image, the nozzle atomization image can be compared with a preset atomization image corresponding to the determined identification angle to obtain a comparison result. After determining the comparison result, the nozzle atomization identification result can be determined. That is, based on the comparison result between the nozzle atomization image and the preset atomization image corresponding to this identification angle, the nozzle atomization identification result for indicating whether the nozzle system that generates the current nozzle atomization image needs to be adjusted can be determined. It can be seen that through the above method, an automated nozzle atomization identification solution is provided. By inputting the nozzle atomization image into the identification model, the identification angle of the nozzle atomization image can be determined, and thus the nozzle atomization identification result can be further determined based on this identification angle, without manual detection of the nozzle atomization effect. This not only improves the identification efficiency of nozzle atomization identification but also enables real-time detection of the atomization effect generated by the nozzle.

[0039] For example, one of the scenarios of the embodiments of the present application can be applied to the scenario as Figure 1 shown. This scenario includes a terminal device 101 and a server 102. Among them, the terminal device 101 includes a nozzle atomization image, and the server 102 adopts the implementation method provided by the embodiments of the present application to obtain the nozzle atomization image from the terminal device 101, and then performs nozzle atomization identification based on the obtained nozzle atomization image.

[0040] First, in the above application scenario, although the action description of the implementation method provided by the embodiments of the present application is described as being executed by the server 102; however, the embodiments of the present application are not limited in terms of the execution subject, as long as the actions disclosed in the implementation method provided by the embodiments of the present application are executed.

[0041] Second, the above scenario is only a scenario example provided by the embodiments of the present application, and the embodiments of the present application are not limited to this scenario.

[0042] Next, in conjunction with the accompanying drawings, the specific implementation manners of the nozzle atomization identification method and device in the embodiments of the present application will be described in detail through embodiments.

[0043] Refer to Figure 2 , which is a flowchart of a nozzle atomization identification method provided by an embodiment of the present application. As shown in conjunction with Figure 2 , it may specifically include:

[0044] S201: Obtain the nozzle atomization image, which is used to indicate the image of the nozzle during operation.

[0045] The nozzle atomization image is used to indicate the image of the nozzle during operation, that is, the image of the atomization effect generated by the nozzle during operation. Reference can be made to Figure 3 , Figure 3 which is the detection schematic diagram of deploying the camera provided by the embodiment of the present application. As shown in Figure 3 , a nozzle and a camera can be installed on the nozzle bracket of the coating machine. The nozzle bracket is located inside the coating pan, and each camera can identify the nozzle atomization image of 1 nozzle or the nozzle atomization images of 2 nozzles.

[0046] As an example, the nozzle atomization image generated by the nozzle can be collected by the camera. For example, the camera can take 20 pictures per second. In practical applications, one picture can be selected per second, that is, one picture can be selected from the 20 pictures taken per second as the nozzle atomization image taken in this second, or one picture can also be selected from the pictures taken every 3 seconds as the nozzle atomization image taken in these 3 seconds. In this regard, the present application does not specifically limit the number of nozzle atomization images collected by the camera.

[0047] In a possible implementation manner, a nozzle system provided by the embodiment of the present application can be referred to Figure 4 , Figure 4 which is the structural schematic diagram of the nozzle system provided by the embodiment of the present application. The nozzle system includes a programmable logic controller (PLC) of the coating machine, a camera, an image acquisition card, an industrial personal computer (PC), a human-machine interface (HMI) touch screen, atomization system control, and a peristaltic pump. Among them, the programmable logic controller is a digital operation controller with a microprocessor for automatic control, and can load control instructions into the memory for storage and execution at any time.

[0048] As an example, after the image is collected by the camera, the collected image can be transmitted to the image acquisition card. When the image acquisition card receives the image, the received image signal can be collected into the industrial PC, so that the industrial PC can receive the nozzle atomization image converted into digital image data, and thus the server can obtain the nozzle atomization image from the industrial PC. In this regard, the present application does not specifically limit the operation steps for obtaining the nozzle atomization image.

[0049] S202: Input the nozzle atomization image into the recognition model for spray angle recognition to obtain the recognition angle of the nozzle atomization image.

[0050] The recognition model is used to recognize the angle of the nozzle atomization image. The training process of the recognition model may include the following steps. First, training images, i.e., nozzle atomization images used as training data, can be collected. For example, referring to Table 1, which is a liquid spraying data table, when the spray gun operates according to the liquid spraying flow rate, atomization pressure, and fog width pressure in Table 1, the corresponding atomization and fog width pressure ranges at different spraying angles of the spray gun can be obtained.

[0051] Table 1

[0052] Liquid spraying flow rate g / min Atomization pressure Bar Mist width pressure Bar 0-20 0.4-1 0.4-1 20-60 1-2 1-2 60-100 2 2

[0053] It should be noted that the quality of the atomization effect is related to the distance between the nozzle and the pan bed of the coating machine. This distance can be determined manually or detected by a sensor. During the training process of the recognition model, the distance between the nozzle and the pan bed of the coating machine can be fixed manually. At the same time, the material and the liquid spraying, etc., also need to be fixed. Specifically, the spraying atomization effect generated by the spray gun can be referred to Figure 5a , Figure 5a is a schematic diagram of a spraying atomization effect provided by an embodiment of the present application. This figure is a spraying atomization image generated under the condition of too small compressed air pressure and can be referred to Figure 5b , Figure 5b is another schematic diagram of a spraying atomization effect provided by an embodiment of the present application. This figure is a spraying atomization image generated under the condition of too large compressed air pressure and can also be referred to Figure 5c , Figure 5c is another schematic diagram of a spraying atomization effect provided by an embodiment of the present application. This figure is a spraying atomization image generated under the ideal spraying state, that is, this figure is a preset atomization image. When the determined nozzle atomization image is the same as the preset atomization image, it is considered that the nozzle system generating this nozzle atomization image does not need to be adjusted, and the coating process can be ended.

[0054] As an example, when spraying according to the data in Table 1 above, when the change ranges are 2 g / min and 0.2 Bar respectively, 50 * 8 * 8 = 3200 combinations can be obtained, that is, 3200 images with different angles and different scenarios can be obtained as training images to ensure that the collected training images can cover different scenarios and different angles.

[0055] Then, a labeling tool can be used to label the spraying angles in the training images. For example, angle values can be used for labeling, and the angle labels corresponding to the training images can be obtained. Among them, the labeling tool can be LabelImg for labeling images, VGG Image Annotator. In this regard, the present application does not specifically limit the labeling tool.

[0056] After that, a segmentation-based detection algorithm DBNet model can be built as the recognition model. The DBNet model usually consists of multiple densely connected blocks, and each densely connected block contains multiple convolutional layers and batch normalization layers. Then, a suitable pre-trained model can be selected. The pre-trained model can learn general image features on a large-scale dataset, which can accelerate the convergence of the model and improve the performance of the recognition model. In addition, according to the recognition features of the nozzle atomization image, the structure of the DBNet model can be appropriately adjusted. For example, the number of densely connected blocks can be increased or decreased, or the kernel size and number of convolutional layers can be adjusted. Then, the adjusted DBNet model can be used as the pre-trained recognition model.

[0057] Among them, reference can be made to Figure 6 , Figure 6 which is the structural schematic diagram of the DBNet model provided by the embodiment of this application. The overall structure of the DBNet model adopts the design idea of Feature Pyramid Networks (FPN), which can include bottom-up convolutional operations and top-down upsampling. Based on the above structure, multi-scale image features are obtained. According to Figure 6 the convolutional layers shown, feature maps with proportions of 1 / 2, 1 / 4, 1 / 8, 1 / 16, and 1 / 32 of the original image size can be obtained respectively. In the above structure, 5 downsampling operations and 3 upsampling operations are performed respectively. After 5 downsamplings of the input image, the output results of each layer can be upsampled, and the size of the finally output feature map is 1 / 4 of the size of the input original image.

[0058] In the Head part of the DBNet model, two branches are respectively led out. One branch is used to predict the probability map based on the output feature map. The size of the predicted probability map is the same as that of the output feature map, which is 1 / 4 of the size of the original input image, that is, the scale of the predicted probability map can be (w / 4)*(h / 4)*1, where w / 4 represents the width is 1 / 4 of the width of the original input image, and h / 4 represents the height is 1 / 4 of the height of the original input image. The probability value corresponding to each pixel point in the predicted probability map represents the probability that this position belongs to the nozzle atomization effect. The other branch is used to predict the threshold map based on the output feature map. The size of the predicted threshold map is the same as that of the output feature map, which is 1 / 4 of the size of the original input image, that is, the scale of the predicted threshold map can be (w / 4)*(h / 4)*1, where w / 4 represents the width is 1 / 4 of the width of the original input image, and h / 4 represents the height is 1 / 4 of the height of the original input image. The predicted threshold map is used to determine the binarization threshold of each pixel point. Finally, the differentiable binarization algorithm can be used. After binarization based on the predicted threshold map and the predicted probability map, the corresponding approximate binary map is obtained. The size of the approximate binary map is the same as that of the output feature map, that is, the size of the approximate binary map can be (w / 4)*(h / 4)*1. Finally, the size of the approximate binary map can be mapped to the size of the original input image, and then the spraying angle in the input image can be recognized.

[0059] Among them, the basic idea of the above feature pyramid is to construct a series of images or feature maps with different scales for model training and testing, aiming to improve the robustness of the detection algorithm for detection targets of different sizes. At the same time, in order to save computing resources, the feature pyramid adopted in the embodiments of the present application adopts a multi-scale feature fusion method, which can significantly improve the scale robustness of feature expression without significantly increasing the amount of calculation.

[0060] After determining the pre-trained recognition model, the training images can be preprocessed. For example, the preprocessing methods can include but are not limited to normalization processing, cropping, and scaling operations to ensure the consistency and effectiveness of the input training images. Data augmentation techniques such as random rotation, flipping, and scaling can also be used to perform data augmentation on the preprocessed training images to enhance the diversity of the input training images. For example, the image size of the training images input to the pre-trained recognition model can be 640*640.

[0061] In addition, an appropriate loss function can be selected. The loss function is used to measure the difference between the recognition angle predicted by the recognition model and the angle label. For example, loss functions such as Mean Squared Error (MSE), Mean Absolute Error (MAE), etc. can be used as the loss function during the training process. An appropriate optimization algorithm can also be selected. For example, the Stochastic Gradient Descent (SGD) algorithm, etc., to adjust parameters such as the model parameters or learning rate of the recognition model based on the loss value calculated by the optimization algorithm based on the loss function.

[0062] Then, the training images after the above-mentioned preprocessing and data augmentation can be input into the pre-trained recognition model for training. The training process may include feature extraction of the input training images, that is, performing convolution operations through the convolutional layers in the pre-trained recognition model to extract the image features corresponding to the training images. After that, the image features output by each convolutional layer in the pre-trained recognition model can be fused to obtain fused features to enhance the feature representation ability of the image features. Moreover, target segmentation recognition can be performed on the fused features to obtain a segmentation result, that is, the atomization effect can be separated from the image represented by the fused features, and the segmented atomization effect, that is, the segmentation result, can be obtained. After determining the segmentation result, further edge detection can be performed based on the segmentation result, that is, edge detection can be performed on the segmentation result. For example, an edge detection algorithm can be used to detect the edge points in the atomization effect of the segmentation result, and the corresponding border of the atomization effect can be determined by connecting these edge points. Further, the spraying angle can be recognized based on the border corresponding to the atomization effect, that is, the predicted recognition angle can be obtained. At the same time, during the training process, the mini-batch training method can be used, that is, training is performed based on a preset number of training images each time to improve the training efficiency. After determining the predicted recognition angle output by the pre-trained recognition model, the difference between the predicted recognition angle and the angle label corresponding to the training image can be calculated based on the loss function to obtain a loss value, and thus the recognition model can be trained based on the calculated loss value.

[0063] After calculating the loss value based on the loss function, the recognition model can be trained based on this loss value, that is, with the goal of reducing the loss value, by continuously adjusting the parameters of the recognition model, the recognition performance of the recognition model can be optimized to improve the accuracy of the recognition model.

[0064] In addition, the trained recognition model can be tested based on a test data set (images containing the nozzle atomization effect different from the training images). By comparing the difference between the predicted recognition angle output by the recognition model and the true spraying angle corresponding to the test image in the test sample, evaluation metrics such as the accuracy rate, recall rate, and mean square error of the recognition model can be evaluated to assess the performance of the recognition model. When each of the above evaluation metrics of the recognition model is lower than the preset evaluation threshold, the recognition model can be continuously adjusted and optimized. For example, adjusting the recognition model parameters, changing the model structure, adding data augmentation methods, or adopting other optimization strategies, etc., to improve the recognition performance and generalization ability of the recognition model.

[0065] In addition, during the training process of the above recognition model, reference can be made to Figure 7 , Figure 7 which is an application schematic diagram of the graphics processing unit provided in the embodiment of the present application. The training process of the recognition model can be processed in parallel by a graphics processing unit (GPU), that is, the training process of the recognition model can be synchronously processed by multiple parallel processing units, thereby improving the training efficiency. At the same time, the GPU can accelerate the recognition process of the nozzle atomization image through floating-point operations. Compared with the Central Processing Unit (CPU), the GPU has higher computing power and parallel processing ability. The spraying angle recognition process of the nozzle atomization image can also be processed in parallel by the GPU, that is, the spraying angle recognition process of the nozzle atomization image can be synchronously processed by multiple parallel processing units, thereby improving the recognition efficiency.

[0066] S203: Compare the nozzle atomization image with the preset atomization image corresponding to the recognition angle to obtain a comparison result.

[0067] The comparison result is used to indicate whether the nozzle system that generates the current nozzle atomization image needs to be adjusted. If the comparison result is that the nozzle atomization image is different from the preset atomization image corresponding to the recognition angle, it is considered that the nozzle system that generates the nozzle atomization image needs to be adjusted; if the comparison result is that the nozzle atomization image is the same as the preset atomization image corresponding to the recognition angle, it is considered that the nozzle system that generates the nozzle atomization image does not need to be adjusted. Among them, the preset atomization image is used to indicate the nozzle atomization image generated under the ideal spraying state.

[0068] Among them, the determination method of the preset atomization image corresponding to the recognition angle can refer to the preset atomization image determined by the spray gun based on the data shown in Table 1 in S202. Taking Figure 5c the preset atomization image shown as an example, the obtained nozzle atomization image can be compared with the preset atomization image corresponding to the recognition angle in the nozzle atomization image to obtain a comparison result.

[0069] S204: Determine the nozzle atomization recognition result based on the comparison result.

[0070] Finally, after determining the comparison result, the nozzle atomization recognition result can be determined based on the comparison result. The nozzle atomization recognition result is used to indicate whether to end the recognition process. As an example, reference can be made to Figure 8 , Figure 8 which is a schematic diagram of the nozzle atomization recognition method provided in the embodiments of this application. First, a nozzle atomization image can be obtained. Then, the nozzle atomization image can be input into an identification model for jet angle recognition, that is, operations such as image preprocessing, feature extraction, feature fusion, target segmentation recognition, and edge detection can be performed on the nozzle atomization image, and the recognition angle of the nozzle atomization image output by the recognition model can be obtained. After determining the recognition angle of the nozzle atomization image, this recognition angle can be fed back to the coating machine control system, so that the coating machine control system can determine the corresponding comparison result based on the recognition angle. When the comparison result is that the nozzle atomization image is the same as the preset atomization image corresponding to the recognition angle, it can be considered that this situation is not an abnormal situation, and it is considered that the nozzle atomization result is to end the recognition process, that is, it indicates that the atomization effect indicated by the nozzle atomization image belongs to the atomization effect generated under the ideal spraying state, and the coating process can be ended. When the comparison result is that the nozzle atomization image is different from the preset atomization image corresponding to the recognition angle, it can be considered that this situation is an abnormal situation, and it is considered that the nozzle atomization recognition result is to adjust the nozzle system, that is, it indicates that the atomization effect indicated by the nozzle atomization image does not belong to the atomization effect generated under the ideal spraying state. The atomization effect indicated by the nozzle atomization image can be adjusted by calling and adjusting the nozzle system through the coating machine control system until the atomization effect under the ideal spraying state is obtained, so as to realize the recognition of the nozzle atomization effect generated by different slurry types in different process flows by the coating machine control system.

[0071] As an example, when the nozzle atomization recognition result is to adjust the nozzle system, the atomization system in the nozzle system can be adjusted, or the peristaltic pump flow rate in the nozzle system can also be adjusted, or an alarm can be given to notify the operation and maintenance personnel for manual adjustment. Specifically, the adjustment process can include: the programmable logic controller in the nozzle system can exchange data with the industrial PC through Ethernet technology. When the programmable logic controller receives the peristaltic pump adjustment signal, it can send this peristaltic pump adjustment signal to the frequency converter, and then the peristaltic pump flow rate in the nozzle system can be adjusted by the frequency converter based on this peristaltic pump adjustment signal, that is, precise control of the peristaltic pump flow rate can be achieved by the frequency converter based on this peristaltic pump adjustment signal. For example, after the frequency converter receives the peristaltic pump adjustment signal, it can stop the flow rate of the spraying peristaltic pump and execute the pulse gun cleaning function. When dripping occurs during the stop of spraying, the spray needle pulse can be closed to perform atomization operation to prevent liquid droplets from falling onto the tablets, etc.

[0072] Among them, the programmable logic controller can control the frequency converter through communication based on the automation bus standard PROFINET. When the programmable logic controller receives the atomization system adjustment signal, it can adjust the atomization system in the nozzle system through analog output, that is, it can control the atomization pressure and the fog width pressure through analog output, and then control the atomization effect of the nozzle.

[0073] In the nozzle atomization recognition method provided in the embodiment of the present application, the method includes obtaining a nozzle atomization image, where the nozzle atomization image is used to indicate the image of the nozzle during operation, that is, first, the image of the nozzle during operation can be obtained to complete automatic recognition based on the nozzle atomization image. Then, the nozzle atomization image can be input into the recognition model for jet angle recognition to determine the angle of the atomization effect generated by the nozzle during operation, and obtain the recognition angle of the nozzle atomization image. After determining the recognition angle of the nozzle atomization image, the nozzle atomization image can be compared with the preset atomization image corresponding to the determined recognition angle to obtain a comparison result. After determining the comparison result, the nozzle atomization recognition result can be determined, that is, based on the comparison result between the nozzle atomization image and the preset atomization image corresponding to the recognition angle, it can be determined whether the nozzle system that generates the current nozzle atomization image needs to be adjusted. It can be seen that through the above method, an automated nozzle atomization recognition solution is provided. By inputting the nozzle atomization image into the recognition model, the recognition angle of the nozzle atomization image can be determined, and then the nozzle atomization recognition result can be further determined based on this recognition angle, without manual detection of the nozzle atomization effect, which not only improves the recognition efficiency of nozzle atomization recognition, but also can realize real-time detection of the atomization effect generated by the nozzle.

[0074] The above are some specific implementation manners of the nozzle atomization recognition method provided in the embodiment of the present application. Based on this, the present application also provides a corresponding device. Next, the device provided in the embodiment of the present application will be introduced from the perspective of functional modularization.

[0075] See Figure 9 , which is a schematic structural diagram of a nozzle atomization recognition device 900 provided in the embodiment of the present application. The device 900 may include:

[0076] An acquisition module 901, configured to acquire a nozzle atomization image, where the nozzle atomization image is used to indicate the image of the nozzle during operation;

[0077] A recognition module 902, configured to input the nozzle atomization image into a recognition model for jet angle recognition to obtain the recognition angle of the nozzle atomization image;

[0078] A comparison module 903, configured to compare the nozzle atomization image with a preset atomization image corresponding to the recognition angle to obtain a comparison result;

[0079] A determination module 904, configured to determine a nozzle atomization recognition result based on the comparison result.

[0080] Optionally, the device includes the following units for training an identification model:

[0081] An annotation unit, configured to annotate the injection angle in the training image to obtain an angle label corresponding to the training image;

[0082] A feature extraction unit, configured to extract features from the training image to obtain image features;

[0083] A fusion unit, configured to perform feature fusion on the image features output by each convolutional layer in the identification model to obtain fusion features;

[0084] A segmentation and recognition unit, configured to perform target segmentation and recognition on the fusion features to obtain a segmentation result;

[0085] A detection unit, configured to perform edge detection on the segmentation result to obtain a predicted recognition angle;

[0086] A training unit, configured to train the identification model based on the difference between the predicted recognition angle and the angle label corresponding to the training image.

[0087] Optionally, the device further includes:

[0088] A parallel processing unit, configured to parallel-process the training process of the identification model through a graphics processor.

[0089] Optionally, the comparison module 904 includes:

[0090] A first determination unit, configured to determine that the nozzle atomization recognition result is to end the recognition process when the comparison result is that the nozzle atomization image is the same as the preset atomization image corresponding to the recognition angle;

[0091] A second determination unit, configured to determine that the nozzle atomization recognition result is to adjust the nozzle system when the comparison result is that the nozzle atomization image is different from the preset atomization image corresponding to the recognition angle.

[0092] Optionally, the device further includes:

[0093] A first adjustment unit, configured to adjust the atomization system in the nozzle system or adjust the peristaltic pump flow rate in the nozzle system when the nozzle atomization recognition result is to adjust the nozzle system.

[0094] Optionally, a programmable logic controller and a frequency converter are further included in the nozzle system. The first adjustment unit includes:

[0095] A second adjustment unit, configured to send the peristaltic pump adjustment signal to the frequency converter when the programmable logic controller receives the peristaltic pump adjustment signal, so that the frequency converter adjusts the flow rate of the peristaltic pump in the nozzle system based on the peristaltic pump adjustment signal;

[0096] A third adjustment unit, configured to adjust the atomization system in the nozzle system through analog quantity output when the programmable logic controller receives the atomization system adjustment signal.

[0097] The embodiment of the present application further provides a corresponding device and a computer storage medium for implementing the solution provided by the embodiment of the present application.

[0098] Wherein, the device includes a memory and a processor. The memory is used to store a computer program, and the processor is used to execute the computer program so that the device executes the nozzle atomization recognition method described in any embodiment of the present application.

[0099] The computer storage medium stores a computer program. When the code runs, the device running the computer program implements the nozzle atomization recognition method described in any embodiment of the present application.

[0100] In the embodiment of the present application, the "first", "second" (if any) in the names such as "first" and "second" are only used as name identifiers and do not represent the first and second in sequence.

[0101] From the description of the above embodiments, those skilled in the art can clearly understand that all or part of the steps in the above embodiment methods can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as read-only memory (ROM) / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0102] It should be noted that the various embodiments in this specification are described in a progressive manner. For the identical or similar parts among the embodiments, reference can be made to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the description of the method embodiments. The apparatus embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components referred to as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative efforts.

[0103] As described above, this is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A nozzle atomization identification method, characterized in that: The method comprises: Acquire a nozzle atomization image, where the nozzle atomization image is used to indicate an image of the nozzle when it is working; Inputting the nozzle atomization image into a recognition model to perform spray angle recognition, thereby obtaining a recognition angle of the nozzle atomization image; Comparing the nozzle atomization image with a preset atomization image corresponding to the recognition angle to obtain a comparison result; Based on the comparison result, a nozzle atomization identification result is determined.

2. The method according to claim 1, characterized in that The recognition model is trained by the following steps: Annotate the injection angles in the training images to obtain angle labels corresponding to the training images; Performing feature extraction on the training image to obtain image features; Performing feature fusion on the image features output by each convolutional layer in the recognition model to obtain fused features; Performing target segmentation and recognition on the fused features to obtain a segmentation result; Performing edge detection on the segmentation result to obtain a predicted recognition angle; The recognition model is trained based on the difference between the predicted recognition angle and the angle label corresponding to the training image.

3. The method according to claim 2, characterized in that The method further comprises: The training process of the recognition model is processed in parallel by a graphics processor.

4. The method according to claim 1, characterized in that: Determining the nozzle atomization recognition result based on the comparison result includes: When the comparison result is that the nozzle atomization image is identical to the preset atomization image corresponding to the recognition angle, the nozzle atomization recognition result is determined to end the recognition process; When the comparison result is that the nozzle atomization image is different from the preset atomization image corresponding to the recognition angle, it is determined that the nozzle atomization recognition result is to adjust the nozzle system.

5. The method according to claim 4, characterized in that The method further comprises: When the nozzle atomization identification result is to adjust the nozzle system, the atomization system in the nozzle system is adjusted or the peristaltic pump flow in the nozzle system is adjusted.

6. The method according to claim 5, characterized in that The nozzle system also includes a programmable logic controller and a frequency converter, and the adjustment of the atomization system in the nozzle system or the adjustment of the peristaltic pump flow in the nozzle system includes: When the programmable logic controller receives the peristaltic pump adjustment signal, it sends the peristaltic pump adjustment signal to the frequency converter, so that the frequency converter adjusts the peristaltic pump flow in the nozzle system based on the peristaltic pump adjustment signal; When the programmable logic controller receives the atomization system adjustment signal, it adjusts the atomization system in the nozzle system through analog output.

7. A nozzle atomization identification device, characterized in that: The device comprises: An acquisition module, used for acquiring a nozzle atomization image, wherein the nozzle atomization image is used for indicating an image of the nozzle when in operation; A recognition module, used for inputting the nozzle atomization image into a recognition model to perform spray angle recognition, and obtain a recognition angle of the nozzle atomization image; A comparison module, used for comparing the nozzle atomization image with a preset atomization image corresponding to the recognition angle to obtain a comparison result; A determination module is used to determine a nozzle atomization recognition result based on the comparison result.

8. The device according to claim 7, characterized in that The device includes the following units for training the recognition model: A labeling unit, used for labeling the injection angle in the training image to obtain an angle label corresponding to the training image; A feature extraction unit, used to extract features from the training image to obtain image features; A fusion unit, used for fusing the image features output by each convolutional layer in the recognition model to obtain a fusion feature; A segmentation and recognition unit, used to perform target segmentation and recognition on the fusion features to obtain a segmentation result; A detection unit, configured to perform edge detection on the segmentation result to obtain a predicted recognition angle; A training unit is used to train the recognition model based on the difference between the predicted recognition angle and the angle label corresponding to the training image.

9. A nozzle atomization identification device, characterized in that: The device comprises: Memory for storing computer programs; A processor is used to execute the computer program so that the device performs the nozzle atomization identification method as described in any one of claims 1 to 6.

10. A computer storage medium, characterized in that: The computer storage medium stores a computer program, and when the computer program is executed by a processor, the nozzle atomization identification method according to any one of claims 1 to 6 is implemented.

Images