Automatic batching control system and method for feed production

Abstract

The invention relates to the field of intelligent feed production, and provides an automatic batching control system and method for feed production, which is characterized in that a monitoring key frame in a monitoring video in a mixing device is analyzed to know the mixing state of feed raw materials, and the opportunity of stopping stirring is judged based on the mixing state of the feed raw materials. Thus, the stirring time is adjusted according to actual needs, it can be ensured that all components are evenly mixed while energy resources are saved, and the device can be suitable for feed raw materials with different physical and chemical characteristics.

Description

Technical Field

[0001] The present application relates to the field of intelligent feed production, and more specifically, to an automatic batching control system and method for feed production. Background Art

[0002] Patent CN204707966U discloses a fully automatic batching and mixing system for feed production. In this fully automatic batching and mixing system, materials from various hoppers are added to the weighing hopper under the action of the screw feeder. When the set weight is reached, the control system instructs to stop adding materials. When each component is added in sufficient quantity, the control system opens the hopper valve to put the materials into the downstream mixing device. When the downstream mixing device is in operation, the operator presses the manual button switch, and the central processing unit controls the feed motor to cooperate with the feed hopper to feed materials and controls the mixing motor to stir and mix through the motor drive module. The indicator light flashes to show the working status of the feed motor and the mixing motor. Stirring and mixing can be timed, and the alarm prompts after the timing is completed.

[0003] Considering that different feed raw materials have different physical and chemical properties, such as density, viscosity, fluidity, etc., these properties will affect the mixing effect. This timed mixing method may not guarantee that all components can be evenly mixed, and the timing time depends on manual experience. For some feed raw materials, such a long mixing time may not be required, which will lead to unnecessary energy waste. For other feed raw materials, the mixing time may be too short, which will result in the failure to achieve the desired mixing effect. Manual intervention is required again, which increases the complexity of the operation.

[0004] Therefore, there is a need for an automatic batching control solution for feed production. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present application provides an automatic batching control system and method for feed production.

[0006] According to one aspect of the present application, there is provided an automatic batching control system for feed production, comprising:

[0007] A feed mixing monitoring data acquisition module is used to obtain monitoring videos in the mixing device for a predetermined period of time;

[0008] A feed mixing monitoring data sampling module, used for sampling the monitoring video in the mixing device in the predetermined time period to obtain a plurality of feed raw material mixing monitoring key frames;

[0009] A feed mixing monitoring key frame analysis module, used for performing key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector;

[0010] The hybrid motor control result generation module is used to obtain the hybrid motor control result based on the feed raw material mixing state characteristic vector.

[0011] According to another aspect of the present application, there is provided an automatic batching control method for feed production, comprising:

[0012] Acquiring surveillance video within the mixing device for a predetermined period of time;

[0013] Sampling the monitoring video in the mixing device during the predetermined time period to obtain a plurality of feed raw material mixing monitoring key frames;

[0014] Performing key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector;

[0015] Based on the feed raw material mixing state characteristic vector, a mixing motor control result is obtained.

[0016] This application has significant technical effects due to the adoption of the above technical solutions:

[0017] The automatic batching control system and method for feed production provided by the present application analyzes the monitoring key frames in the monitoring video in the mixing device to understand the mixing state of the feed raw materials, and determines the timing to stop stirring based on the mixing state of the feed raw materials. In this way, the stirring time can be adjusted according to actual needs, which can ensure that all components are mixed evenly while saving energy resources, and can be applied to feed raw materials with different physical and chemical properties. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 The present invention is a flowchart of an automatic batching control method for feed production according to an embodiment of the present application.

[0020] Figure 2 Schematic diagram of data flow of an automatic ingredient control method for feed production according to an embodiment of the present application.

[0021] Figure 3 The present invention is a flowchart of performing key frame content analysis on the multiple feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector in an automatic batching control method for feed production according to an embodiment of the present application.

[0022] Figure 4 The present invention is a flow chart of obtaining a mixing motor control result based on the feed raw material mixing state characteristic vector in the automatic batching control method for feed production according to an embodiment of the present application.

[0023] Figure 5 It is a system block diagram of an automatic batching control system for feed production according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] Patent CN204707966U discloses a fully automatic batching and mixing system in feed production. In this system, materials from various hoppers are added to the weighing hopper under the action of a screw feeder, and the feeding is stopped after reaching the set weight. When each component is sufficient, the hopper valve is opened to put the material into the downstream mixing device. During the operation of the mixing device, the operator presses the manual button switch, and the central processing unit controls the feeding motor to feed and the mixing motor to stir and mix through the motor drive module. The indicator light flashes to show the working status, and the stirring can be timed, and the alarm prompts when the timed completion. Considering that the physical and chemical properties of different feed raw materials (such as density, viscosity, fluidity, etc.) will affect the mixing effect, the timed stirring and mixing method is difficult to ensure that all components are evenly mixed, and the timing time depends on manual experience, which may cause energy waste or poor mixing effect, and manual intervention is required again, increasing the complexity of the operation. Therefore, an automatic batching control scheme for feed production is expected.

[0026] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for automatic ingredient control in feed production.

[0027] Figure 1 The present invention is a flowchart of an automatic batching control method for feed production according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the automatic batching control method for feed production according to an embodiment of the present application. Figure 1 and Figure 2As shown, the automatic ingredient control method for feed production according to the embodiment of the present application includes: S110, acquiring the monitoring video in the mixing device of a predetermined time period; S120, sampling the monitoring video in the mixing device of the predetermined time period to obtain a plurality of feed raw material mixing monitoring key frames; S130, performing key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector; S140, obtaining a mixing motor control result based on the feed raw material mixing state feature vector.

[0028] In step S110, the monitoring video in the mixing device of a predetermined time period is obtained. It should be understood that the monitoring video in the mixing device can observe the mixing process of the feed raw materials in the mixing device in real time, including the dispersion of the raw materials, the mixing uniformity, whether there are lumps or unevenly mixed areas, etc. It specifically shows the state of the feed raw materials inside the mixing device at different time points, including the distribution of the raw materials, flow conditions, color changes, etc. By analyzing the monitoring video in the mixing device, the mixing state can be judged more accurately and the control strategy can be adjusted in time, thereby improving the mixing quality, reducing energy waste and ensuring production safety. In particular, in a specific implementation method of the embodiment of the present application, the acquisition of monitoring video data in the mixing device can be carried out by an industrial camera.

[0029] In step S120, the monitoring video in the mixing device of the predetermined time period is sampled to obtain a plurality of feed raw material mixing monitoring key frames. Specifically, in an embodiment of the present application, the monitoring video in the mixing device of the predetermined time period is sampled to obtain a plurality of feed raw material mixing monitoring key frames, including: sampling the monitoring video in the mixing device of the predetermined time period at a predetermined sampling frequency to obtain the plurality of feed raw material mixing monitoring key frames. It should be understood that the amount of monitoring video data originally collected by the industrial camera is huge, and direct processing is not only time-consuming but also resource-intensive. In order to reduce the amount of calculated data and facilitate the model to quickly analyze and process the monitoring video, it is necessary to sample the monitoring video in the mixing device of the predetermined time period. During the mixing process, the mixing state of the raw materials will change over time. Through sampling, the key moments of these changes can be captured, such as a significant improvement in mixing uniformity, the occurrence of abnormal conditions, etc., which provides an important basis for the control of the hybrid motor.

[0030] In step S130, key frame content analysis is performed on the plurality of feed material mixing monitoring key frames to obtain a feed material mixing state feature vector. Specifically, Figure 3 The present invention is a flowchart of performing key frame content analysis on the multiple feed raw material mixing monitoring key frames to obtain the feed raw material mixing state feature vector in the automatic batching control method for feed production according to an embodiment of the present application. Figure 3As shown, the key frame content analysis is performed on the multiple feed raw material mixing monitoring key frames to obtain the feed raw material mixing state feature vector, including: S131, the multiple feed raw material mixing monitoring key frames are passed through a feed raw material mixing state change timing encoder to obtain a feed raw material mixing state timing change feature map; S132, the feed raw material mixing state timing change feature map is passed through a mixing state feature enhancer based on a double attention mechanism to obtain a feed raw material mixing state timing change enhanced feature map; S133, the feed raw material mixing state timing change feature map and the feed raw material mixing state timing change enhanced feature map are fused to obtain a feed raw material mixing state feature map; S134, the feed raw material mixing state feature map is subjected to mean pooling processing of each feature matrix along the channel dimension to obtain the feed raw material mixing state feature vector.

[0031] In the mixing process of feed raw materials, the state change of feed raw materials is a process that changes continuously over time. In order to capture the continuity and dependence of the state of feed raw materials in time, it is necessary to pass the multiple feed raw material mixing monitoring key frames through the feed raw material mixing state change timing encoder to obtain the feed raw material mixing state time series change feature map. Through the feed raw material mixing state change timing encoder, features related to the mixing state change, such as mixing uniformity, mixing speed, raw material distribution, etc., can be extracted from the monitoring key frames. In this application, the feed raw material mixing state change timing encoder is a convolutional neural network model using a three-dimensional convolution kernel. The three-dimensional convolutional neural network adds the calculation of the channel dimension (the channel dimension in this application specifically refers to the time dimension) on the basis of the two-dimensional convolutional neural network, so that it can simultaneously extract the spatial features and time features in the video sequence, which is conducive to analyzing the spatiotemporal changes of raw materials in the mixing process. It is also worth mentioning that in the feed production process, the environment in the mixing device may be more complex, with problems such as illumination changes and occlusion. The three-dimensional convolutional neural network can better adapt to these complex scenes and improve the robustness of the system through its powerful feature extraction ability.

[0032] In an embodiment of the present application, a possible implementation method of passing the multiple feed raw material mixing monitoring key frames through a feed raw material mixing state change timing encoder to obtain a feed raw material mixing state time series change feature map can be: using each layer of the feed raw material mixing state change timing encoder to perform the following on the input data in the forward transmission of the layer: using the convolution units of each layer of the feed raw material mixing state change timing encoder to perform convolution processing based on a three-dimensional convolution kernel on the input data to obtain a convolution feature map; using the pooling units of each layer of the feed raw material mixing state change timing encoder to perform pooling processing based on a local feature matrix on the convolution feature map to obtain a pooling feature map; using the activation units of each layer of the feed raw material mixing state change timing encoder to perform nonlinear activation on the feature values ​​of each position in the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the feed raw material mixing state change timing encoder is the feed raw material mixing state time series change feature map.

[0033] Not all features at all positions in the original feed raw material mixing state time series change feature map are effective for the final mixed motor control. In order to enhance the expression of features in the original feature map, it is necessary to pass the feed raw material mixing state time series change feature map through a mixed state feature enhancer based on a dual attention mechanism to obtain a feed raw material mixing state time series change enhanced feature map. The mixed state feature enhancer based on the dual attention mechanism in this application is a convolutional neural network model including a spatial attention mechanism and a channel attention mechanism. Those of ordinary skill in the art should understand that the spatial attention mechanism can guide the model to pay attention to the area in the feature map that is most relevant to the mixed state change, and help to more accurately locate the location of key information, that is, by weighting the spatial information in the original feature map, the model can better capture the subtle changes in the mixing process; and the channel attention mechanism can automatically evaluate the importance of features on different channels to the current task, and selectively enhance or suppress these features, thereby improving the generalization ability and recognition accuracy of the model, that is, through the integration and optimization of channel information, the model can more comprehensively understand the changes in the mixed state, and then optimize the control of the mixed motor. In general, the spatial attention mechanism and channel attention mechanism can focus on the important information in the spatial position and channel dimension of the feature map respectively. Through weighted processing, important features are made more prominent and unimportant features are suppressed, thereby achieving feature enhancement.

[0034] In an embodiment of the present application, a possible implementation method of passing the temporal change feature map of the mixing state of the feed raw materials through a mixed state feature enhancer based on a dual attention mechanism to obtain an enhanced feature map of the temporal change of the mixing state of the feed raw materials can be: using each layer of the mixed state feature enhancer based on the dual attention mechanism to perform the following on the input data in the forward pass of the layer: performing convolution processing on the input data based on the convolution kernel to obtain a convolution feature map; performing global mean pooling on each feature matrix along the channel dimension of the convolution feature map to obtain a channel feature vector; passing the channel feature vector through a Softmax function to obtain a normalized channel feature vector; using the eigenvalues ​​of each position in the normalized channel feature vector as weights to weight the feature matrix along the channel dimension of the convolution feature map to obtain a channel attention map; performing average pooling and maximum pooling on the channel attention map along the channel dimension to obtain an average feature matrix and maximum feature matrix; cascading and channel-adjusting the average feature matrix and the maximum feature matrix to obtain a channel feature matrix; using the convolution layer of the spatial attention module to convolutionally encode the channel feature matrix to obtain a convolution feature matrix; passing the convolution feature matrix through the Softmax function to obtain a spatial attention score matrix; multiplying the spatial attention score matrix with each feature matrix of the channel attention map along the channel dimension by position point to obtain the channel-spatial attention map; pooling the channel-spatial attention map based on the local feature matrix to obtain a pooled feature map; performing nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the input of the first layer of the mixed state feature enhancer is the temporal change feature map of the mixed state of the feed raw materials, and the output of the last layer of the mixed state feature enhancer is the enhanced feature map of the temporal change of the mixed state of the feed raw materials.

[0035] Considering that the time series change feature map of the mixing state of feed raw materials mainly captures the characteristics of the mixing process of feed raw materials that change over time, such as the changing trend of mixing uniformity, and the time series change enhanced feature map of the mixing state of feed raw materials further optimizes and enhances the time series change feature map through the dual attention mechanism, highlights the key information and suppresses unimportant information. In order to make full use of the advantages of the two feature maps and form a more comprehensive and accurate description of the mixing state, it is necessary to fuse the time series change feature map of the mixing state of feed raw materials and the time series change enhanced feature map of the mixing state of feed raw materials in the technical solution of the present application to obtain the feed raw material mixing state feature map. It should be understood that the use of the time series change feature map or the enhanced feature map alone may lead to a decrease in recognition accuracy due to incomplete information or insufficient feature expression. By fusing these two feature maps, their respective shortcomings can be compensated and the recognition accuracy of the mixing state of feed raw materials can be improved.

[0036] In an embodiment of the present application, a possible implementation method of fusing the feed raw material mixing state time series change characteristic diagram and the feed raw material mixing state time series change enhanced characteristic diagram to obtain the feed raw material mixing state characteristic diagram may be: fusing the feed raw material mixing state time series change characteristic diagram and the feed raw material mixing state time series change enhanced characteristic diagram to obtain the feed raw material mixing state characteristic diagram using the following fusion formula; wherein the fusion formula is:

[0037]

[0038] Among them, F c is the feed raw material mixing state characteristic diagram, F a is the time series variation characteristic diagram of the mixing state of the feed raw materials, F b is an enhanced characteristic diagram of the time series change of the mixing state of the feed raw materials, It represents the addition of the elements at corresponding positions of the feed raw material mixing state time series change characteristic diagram and the feed raw material mixing state time series change enhanced characteristic diagram, and α and β are weighted parameters for controlling the balance between the feed raw material mixing state time series change characteristic diagram and the feed raw material mixing state time series change enhanced characteristic diagram in the feed raw material mixing state characteristic diagram.

[0039] Next, the feed raw material mixed state characteristic graph is subjected to mean pooling processing of each characteristic matrix along the channel dimension to obtain the feed raw material mixed state characteristic vector. It should be understood that in the automatic batching control for feed production, real-time performance and computational efficiency are crucial, and the characteristic graph dimension is high, and the operation consumes more time and computing resources. In order to improve the response speed and stability of the control, it is necessary to perform a pooling operation on the feed raw material mixed state characteristic graph. Specifically, the pooling operation reduces the spatial size of the feed raw material mixed state characteristic graph by aggregating the features of the local area, thereby reducing the amount of data for subsequent processing. Mean pooling retains the main feature information of the area by averaging the characteristic values ​​in the local area. The feed raw material mixed state characteristic vector obtained by mean pooling is more refined and rich in important information, making the identification of the mixed state more accurate. Based on this accurate information, the start time of the mixing motor can be adjusted more accurately, thereby improving the accuracy and efficiency of mixing.

[0040] In step S140, a mixing motor control result is obtained based on the feed material mixing state feature vector. Specifically, Figure 4 The flowchart of the automatic batching control method for feed production according to the embodiment of the present application is to obtain the control result of the mixing motor based on the feed raw material mixing state feature vector. Figure 4As shown, based on the feed raw material mixing state feature vector, a hybrid motor control result is obtained, including: S141, performing feature dynamic mask mixing adjustment based on gated synthesis on the feed raw material mixing state feature vector to obtain an optimized feed raw material mixing state feature vector; S142, passing the optimized feed raw material mixing state feature vector through a classifier-based hybrid motor control result generator to obtain a classification result, and the classification result is used to indicate whether the hybrid motor needs to be turned off.

[0041] In particular, in the present application, the feed raw material mixing state feature vector is obtained by processing the feed raw material mixing state time series change feature graph, and this feature graph itself is extracted from multiple feed raw material mixing monitoring key frames obtained through sampling. Even after sampling, there is still a large amount of repeated information in the multiple feed raw material mixing monitoring key frames. In the subsequent feature encoding and feature enhancement process, it is difficult to completely remove these repeated information, resulting in redundant features in the feed raw material mixing state feature vector. The redundant features in the feed raw material mixing state feature vector will interfere with the classifier's recognition and weight allocation of key features, which makes it difficult for the classifier to accurately focus on the features that can truly distinguish different categories, which will affect the final hybrid motor control result. Therefore, in the technical solution of the present application, it is necessary to perform a gated synthesis-based feature dynamic mask mixing adjustment on the feed raw material mixing state feature vector to obtain an optimized feed raw material mixing state feature vector.

[0042] Specifically, in an embodiment of the present application, the feed raw material mixing state feature vector is subjected to feature dynamic mask mixing adjustment based on gated synthesis to obtain an optimized feed raw material mixing state feature vector, including: extracting the feed raw material covariance prior basis matrix; performing core prior information low-rank feature distillation on the feed raw material covariance prior basis matrix to obtain a set of core prior information sparse dictionary atomic vectors of the feed raw material model; constructing a feed raw material model core prior information kernel induced latent space mapping matrix between the feed raw material mixing state feature vector and each feed raw material model core prior information sparse dictionary atomic vector in the set of the feed raw material model core prior information sparse dictionary atomic vectors to obtain the feed raw material model core prior information kernel induced latent space mapping matrix. A set of spatial mapping matrices; calculating the feed raw material prior information response feature coupling strength symbol of each feed raw material model core prior information kernel induced latent space mapping matrix in the set of the feed raw material model core prior information kernel induced latent space mapping matrices to obtain a set of feed raw material prior information response feature coupling strength symbols; based on the set of feed raw material prior information response feature coupling strength symbols, performing soft mask feature mixing on the set of feed raw material model core prior information kernel induced latent space mapping matrices to obtain a feed raw material model prior information response projection coding matrix; mapping the feed raw material mixing state feature vector to the feature space of the feed raw material model prior information response projection coding matrix to obtain the optimized feed raw material mixing state feature vector.

[0043] The feed raw material mixing state feature vector is subjected to feature dynamic mask mixing adjustment based on gated synthesis to obtain an optimized feed raw material mixing state feature vector, including:

[0044] First, extract the covariance prior basis matrix of feed raw materials. It should be understood that extracting the covariance prior basis matrix of feed raw materials is not just about obtaining information, but is actually a key operation to structure and calculate prior knowledge. The principle is to condense domain expertise, model design concepts, or macro laws inverted in data into the mathematical form of the covariance prior basis matrix of feed raw materials for effective use of subsequent algorithms.

[0045] Then, the core prior information low-rank feature distillation is performed on the covariance prior basis matrix of the feed raw material to obtain a set of core prior information sparse dictionary atomic vectors of the feed raw material model, which is expressed as the core prior information low-rank feature distillation formula:

[0046]

[0047] Among them, U represents the set of atomic vectors of the sparse dictionary of core prior information of the feed raw material model, v 1 、v 2 、vm Respectively represent the first, second, and mth core prior information sparse dictionary atomic vectors of the feed raw material model, T represents the transposition operation, CoreExtraction represents the core prior information low-rank feature distillation, and M p represents the prior basis matrix of the covariance of feed raw materials, Λ represents the diagonal matrix, λ 1 , m Respectively represent the values ​​of the first and mth positions on the diagonal of the diagonal matrix. It should be understood that the covariance prior basis matrix of feed raw materials may contain redundant or high-dimensional information, and direct application may lead to excessive computational burden and interference of non-critical information on the optimization process. Therefore, the principle of low-rank feature distillation of core prior information is to reduce noise and refine prior knowledge, just like the filtering process in signal processing, retaining the main components and filtering out redundancy.

[0048] Next, a feed raw material model core prior information core induced latent space mapping matrix is ​​constructed between the feed raw material mixing state feature vector and each feed raw material model core prior information sparse dictionary atomic vector in the set of the feed raw material model core prior information sparse dictionary atomic vector to obtain a set of feed raw material model core prior information core induced latent space mapping matrices, which is expressed as a core induced latent space mapping formula:

[0049]

[0050] Among them, x o represents the characteristic vector of the mixed state of feed raw materials, l i (x o ) represents the o After linear transformation, the feature vector has the same characteristic scale as the corresponding feed raw material model core prior information sparse dictionary atomic vector, v i represents the core prior information sparse dictionary atomic vector of the i-th feed raw material model, represents matrix multiplication, L represents the length of the atomic vector of the sparse dictionary of the core prior information of the feed raw material model, MR i Represents the core prior information kernel induced latent space mapping matrix of the ith feed raw material model. It should be understood that the core of this step is to construct the interactive relationship between the feed raw material mixing state feature vector and the refined prior knowledge. Specifically, through latent space mapping, the nonlinear response pattern of the feed raw material mixing state feature vector to different aspects of prior knowledge is learned. In essence, the core prior information kernel induced latent space mapping matrix of the feed raw material model is the re-encoding of the feed raw material mixing state feature vector from the perspective of prior knowledge, which incorporates the interpretation and processing of knowledge. Its function exceeds information association and realizes the directional enhancement of feature representation and the extraction of multi-perspective feature information.

[0051] Next, the feed raw material prior information response characteristic coupling strength symbol of each feed raw material model core prior information kernel induced latent space mapping matrix in the set of the feed raw material model core prior information kernel induced latent space mapping matrix is ​​calculated to obtain a set of feed raw material prior information response characteristic coupling strength symbols, which is expressed as the characteristic coupling strength formula:

[0052] S i =||MR i || F

[0053] Among them, ||·|| F represents the F norm of the matrix, S i Represents the coupling strength symbol of the feature response of the prior information of the i-th feed raw material. It should be understood that the principle of this step is to extract key information and simplify the feature representation of the core prior information kernel-induced latent space mapping matrix of each feed raw material model, just like generating an information summary, and extracting the summary or feature vector that best represents the core information. The coupling strength symbol of the feature response of the prior information of the feed raw material must be representative and discriminative. It contains the ideas of feature selection and feature aggregation, selects the most informative part to retain, and aggregates it into a concise feature coupling strength symbol to achieve deeper compression and refinement. The function of the coupling strength symbol of the feature response of the prior information of the feed raw material is not only to compress information, but also to improve the efficiency and robustness of the subsequent fusion process, reduce the data dimension, and reduce the computational burden, especially in high-dimensional matrices.

[0054] Subsequently, based on the set of coupling strength symbols of the feed raw material prior information response characteristics, the set of the feed raw material model core prior information kernel induced latent space mapping matrices is soft-masked feature mixed to obtain the feed raw material model prior information response projection coding matrix, which is expressed as a soft-masked feature mixing formula:

[0055]

[0056] Among them, softmax represents the normalized exponential function, and P represents the projection coding matrix of the prior information response of the feed raw material model. It should be understood that the core principle of the soft mask feature mixing is to emphasize the adaptive and selective prior information integration strategy. Sparsity reflects that not all prior information responses are equally important. The fusion should be selective, focusing on more important responses, weakening or ignoring unimportant responses, improving feature selection and generalization capabilities, and avoiding overfitting. Dynamicity means that the weights or methods of soft mask feature mixing are not fixed, but are adaptively adjusted according to the input data or model state to improve the flexibility and adaptability of the model. The essence of soft mask feature mixing of the set of core prior information kernel induced latent space mapping matrices of the feed raw material model is to optimally combine the response information from different prior knowledge perspectives to form a feed raw material model prior information response projection coding matrix that comprehensively reflects the model prior response.

[0057] Finally, the feed raw material mixing state feature vector is mapped to the feature space of the feed raw material model prior information response projection coding matrix to obtain the optimized feed raw material mixing state feature vector, which is expressed as a mapping formula:

[0058]

[0059] Among them, x opt Represents the optimized feed raw material mixed state feature vector. It should be understood that the entire adaptation process is finally completed by mapping the feed raw material mixed state feature vector to the feature space of the feed raw material model prior information response projection coding matrix to obtain the optimized feed raw material mixed state feature vector. The principle of this step is to use the feature space defined by the feed raw material model prior information response projection coding matrix to project the feed raw material mixed state feature vector into the space. In essence, the feed raw material model prior information response projection coding matrix acts as a transformation matrix, and linearly or nonlinearly maps the feed raw material mixed state feature vector to embed it into the feature space of the fusion model prior information. The optimized feed raw material mixed state feature vector better fits the model prior constraints and guidance, and realizes boundary adaptation on the feature manifold. Compared with the feed raw material mixed state feature vector, the optimized feed raw material mixed state feature vector is usually significantly improved in terms of expressiveness, distinguishability and generalization, providing a better feature representation for subsequent machine learning tasks.

[0060] It should be understood that in order to perform efficient and accurate pattern recognition based on the input feature vector (i.e., the optimized feed raw material mixing state feature vector), it is necessary to train the optimized feed raw material mixing state feature vector through a deep learning algorithm to obtain a classifier, and use the output of the classifier as the basis for the control decision of the mixing motor, that is, by real-time monitoring of the mixing state and adjusting the operating state of the mixing motor accordingly, automatic batching control for feed production can be achieved. By accurately controlling the mixing process, the mixing uniformity and quality stability of the feed raw materials can be ensured, thereby improving the overall quality of the product. At the same time, it avoids waste of resources and product quality problems caused by excessive or insufficient mixing, and improves production efficiency.

[0061] In an embodiment of the present application, a possible implementation method of passing the optimized feed raw material mixing state feature vector through a classifier-based hybrid motor control result generator to obtain a classification result indicating whether the hybrid motor needs to be shut down may be: using the fully connected layer of the classifier to fully connect encode the optimized feed raw material mixing state feature vector to obtain a feed raw material mixing state encoded classification feature vector; and inputting the feed raw material mixing state encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0062] In summary, the automatic batching control method for feed production based on the embodiment of the present application is explained, which analyzes the monitoring key frames in the monitoring video in the mixing device to understand the mixing state of the feed raw materials, and judges the timing of stopping stirring based on the mixing state of the feed raw materials. In this way, the stirring time is adjusted according to actual needs, which can ensure that all components are mixed evenly while saving energy resources, and can be applied to feed raw materials with different physical and chemical properties.

[0063] Figure 5 1 is a system block diagram of an automatic batching control system for feed production according to an embodiment of the present application. Figure 5 As shown, according to the embodiment of the present application, the automatic batching control system 100 for feed production includes: a feed mixing monitoring data acquisition module 110, which is used to obtain the monitoring video in the mixing device of a predetermined time period; a feed mixing monitoring data sampling module 120, which is used to sample the monitoring video in the mixing device of the predetermined time period to obtain a plurality of feed raw material mixing monitoring key frames; a feed mixing monitoring key frame analysis module 130, which is used to perform key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector; a mixing motor control result generation module 140, which is used to obtain a mixing motor control result based on the feed raw material mixing state feature vector.

[0064] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the automatic batching control system 100 for feed production have been described in detail above. Figures 1 to 4 The invention has been described in detail in the description of the automatic ingredient control method for feed production, and therefore, its repeated description will be omitted.

[0065] In summary, the automatic batching control system 100 for feed production based on the embodiment of the present application is explained, which analyzes the monitoring key frames in the monitoring video in the mixing device to understand the mixing state of the feed raw materials, and determines the timing to stop stirring based on the mixing state of the feed raw materials. In this way, the stirring time is adjusted according to actual needs, which can ensure that all components are mixed evenly while saving energy resources, and can be applicable to feed raw materials with different physical and chemical properties.

[0066] As described above, the automatic batching control system 100 for feed production according to the embodiment of the present application can be implemented in various wireless terminals, such as a server for automatic batching control of feed production, etc. In one example, the automatic batching control system 100 for feed production according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the automatic batching control system 100 for feed production can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the automatic batching control system 100 for feed production can also be one of the many hardware modules of the wireless terminal.

[0067] Alternatively, in another example, the automatic ingredient control system 100 for feed production and the wireless terminal may also be separate devices, and the automatic ingredient control system 100 for feed production may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

Claims

1. An automatic batching control method for feed production, characterized in that: include: Acquiring surveillance video within the mixing device for a predetermined period of time; Sampling the monitoring video in the mixing device during the predetermined time period to obtain a plurality of feed raw material mixing monitoring key frames; Performing key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector; Based on the feed raw material mixing state characteristic vector, a mixing motor control result is obtained.

2. The automatic batching control method for feed production according to claim 1, characterized in that: The monitoring video in the mixing device in the predetermined time period is sampled to obtain a plurality of feed raw material mixing monitoring key frames, including: sampling the monitoring video in the mixing device in the predetermined time period at a predetermined sampling frequency to obtain the plurality of feed raw material mixing monitoring key frames.

3. The automatic batching control method for feed production according to claim 2, characterized in that: Performing key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector includes: Passing the plurality of feed raw material mixing monitoring key frames through a feed raw material mixing state change time series encoder to obtain a feed raw material mixing state time series change characteristic diagram; The feed raw material mixing state time series change characteristic graph is passed through a mixing state feature enhancer based on a dual attention mechanism to obtain a feed raw material mixing state time series change enhanced characteristic graph; The feed raw material mixing state time series change characteristic diagram and the feed raw material mixing state time series change enhanced characteristic diagram are merged to obtain the feed raw material mixing state characteristic diagram; The feed raw material mixing state characteristic graph is subjected to mean pooling processing of each characteristic matrix along the channel dimension to obtain the feed raw material mixing state characteristic vector.

4. The automatic batching control method for feed production according to claim 3, characterized in that: The feed raw material mixing state change time series encoder is a convolutional neural network model using a three-dimensional convolution kernel, and the mixed state feature enhancer based on a dual attention mechanism is a convolutional neural network model including a spatial attention mechanism and a channel attention mechanism.

5. The automatic batching control method for feed production according to claim 4, characterized in that: Based on the feed raw material mixing state characteristic vector, a mixing motor control result is obtained, including: Performing a feature dynamic mask mixing adjustment based on gated synthesis on the feed raw material mixing state feature vector to obtain an optimized feed raw material mixing state feature vector; The optimized feed raw material mixing state feature vector is passed through a classifier-based mixing motor control result generator to obtain a classification result, and the classification result is used to indicate whether the mixing motor needs to be shut down.

6. The automatic batching control method for feed production according to claim 5, characterized in that: The feed raw material mixing state feature vector is subjected to feature dynamic mask mixing adjustment based on gated synthesis to obtain an optimized feed raw material mixing state feature vector, including: Extract the covariance prior basis matrix of feed ingredients; Performing low-rank feature distillation of core prior information on the feed raw material covariance prior basis matrix to obtain a set of sparse dictionary atomic vectors of core prior information of the feed raw material model; Constructing a feed raw material model core prior information kernel induced latent space mapping matrix between the feed raw material mixing state feature vector and each feed raw material model core prior information sparse dictionary atomic vector in the set of the feed raw material model core prior information sparse dictionary atomic vectors to obtain a set of feed raw material model core prior information kernel induced latent space mapping matrices; Calculating the feed raw material prior information response characteristic coupling strength symbol of each feed raw material model core prior information kernel induced latent space mapping matrix in the set of the feed raw material model core prior information kernel induced latent space mapping matrix to obtain a set of feed raw material prior information response characteristic coupling strength symbols; Based on the set of the feed raw material prior information response feature coupling strength symbols, a set of the feed raw material model core prior information kernel induced latent space mapping matrices is soft-masked feature mixed to obtain a feed raw material model prior information response projection coding matrix; The feed raw material mixing state feature vector is mapped to the feature space of the feed raw material model prior information response projection coding matrix to obtain the optimized feed raw material mixing state feature vector.

7. An automatic batching control system for feed production, characterized in that: include: A feed mixing monitoring data acquisition module is used to obtain monitoring videos in the mixing device for a predetermined period of time; A feed mixing monitoring data sampling module, used for sampling the monitoring video in the mixing device in the predetermined time period to obtain a plurality of feed raw material mixing monitoring key frames; A feed mixing monitoring key frame analysis module, used for performing key frame content analysis on the plurality of feed raw material mixing monitoring key frames to obtain a feed raw material mixing state feature vector; The hybrid motor control result generation module is used to obtain the hybrid motor control result based on the feed raw material mixing state characteristic vector.

8. The automatic batching control system for feed production according to claim 7, characterized in that: The feed mixing monitoring key frame analysis module includes: A feed raw material mixing state time series change feature extraction unit, used for passing the plurality of feed raw material mixing monitoring key frames through a feed raw material mixing state change time series encoder to obtain a feed raw material mixing state time series change feature diagram; A feed raw material mixing state temporal change feature enhancement unit, used for passing the feed raw material mixing state temporal change feature graph through a mixing state feature enhancer based on a double attention mechanism to obtain a feed raw material mixing state temporal change enhanced feature graph; A feature fusion unit, used for fusing the feed raw material mixing state time series change feature graph and the feed raw material mixing state time series change enhanced feature graph to obtain the feed raw material mixing state feature graph; The feed raw material mixing state characteristic graph is subjected to mean pooling processing of each characteristic matrix along the channel dimension to obtain the feed raw material mixing state characteristic vector.

9. The automatic batching control system for feed production according to claim 8, characterized in that: The hybrid motor control result generating module comprises: A feed raw material mixed state feature optimization unit, used for performing feature dynamic mask mixing adjustment based on gated synthesis on the feed raw material mixed state feature vector to obtain an optimized feed raw material mixed state feature vector; The optimized feed raw material mixing state feature analysis unit is used to pass the optimized feed raw material mixing state feature vector through a classifier-based mixing motor control result generator to obtain a classification result, and the classification result is used to indicate whether the mixing motor needs to be turned off.

Images