Feed production information management system and method

Abstract

The invention relates to the field of feed production, and particularly discloses a feed production information management system and method, and the method comprises the steps: firstly obtaining a feed production monitoring video collected by a camera and order raw material information collected by a database, and then carrying out the feature extraction and correlation analysis of the two through a deep learning technology, and finally, a classification result is obtained through a classifier to judge whether feed production grinding is sufficient or not, so that dependence on manual monitoring is reduced, influences caused by human errors and operation fatigue are reduced, and then more objective adjustment suggestions and optimized production management decisions are provided.

Description

Technical Field

[0001] The present application relates to the field of feed production, and more specifically, to a feed production information management system and method. Background Art

[0002] Feed is a specialized food that provides nutritional support to animals, aiming to promote their health, growth and production performance. Feed usually includes a variety of ingredients, such as grains, protein sources, vitamins and minerals, to meet the nutritional needs of different animals. In feed production, the production process needs to be flexible because different orders have different feed formulation and production requirements.

[0003] However, existing technologies usually rely on preset production parameters and procedures, which are difficult to meet personalized production needs. Although manual monitoring by operators is a common solution, operators may make mistakes due to fatigue, negligence or lack of experience, resulting in untimely monitoring and adjustment, which in turn affects feed quality. In addition, manual monitoring has a lag and cannot respond to changes in the production process in real time.

[0004] Therefore, a feed production information management system and method is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a feed production information management system and method, which first obtains the feed production monitoring video collected by the camera and the order raw material information collected by the database, then uses deep learning technology to perform feature extraction and association analysis on the two, and finally obtains the classification result through the classifier to judge whether the feed production grinding is sufficient, thereby reducing the dependence on manual monitoring, reducing the impact of human errors and operator fatigue, and then providing more objective adjustment suggestions and optimizing production management decisions.

[0006] According to one aspect of the present application, there is provided a feed production information management system, comprising:

[0007] A feed production data acquisition module is used to acquire feed production monitoring videos collected by cameras and order raw material information collected by databases;

[0008] A feed production data extraction module, used to extract a multimodal feature vector of feed grinding uniformity and a feature vector of order raw material text understanding from the feed production monitoring video collected by the camera and the order raw material information collected by the database;

[0009] The feed production grinding judgment module is used to judge whether the feed production grinding is sufficient based on the multimodal feature vector of feed grinding uniformity and the feature vector of order raw material text understanding.

[0010] According to another aspect of the present application, a feed production information management method is provided, comprising:

[0011] Obtain feed production monitoring videos collected by cameras and order raw material information collected by databases;

[0012] Extracting a multimodal feature vector of feed grinding uniformity and a feature vector of order raw material text understanding from the feed production monitoring video collected by the camera and the order raw material information collected by the database;

[0013] Based on the multimodal feature vector of feed grinding uniformity and the feature vector of order raw material text understanding, it is judged whether the feed production grinding is sufficient.

[0014] Compared with the prior art, the present application provides a feed production information management system and method, which first obtains the feed production monitoring video collected by the camera and the order raw material information collected by the database, and then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally obtains the classification result through the classifier to determine whether the feed production is sufficiently ground, thereby reducing the dependence on manual monitoring, reducing the impact of human errors and operational fatigue, and then providing more objective adjustment suggestions and optimizing production management decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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.

[0016] Figure 1 It is a block diagram of a feed production information management system according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of a feed production data extraction module in a feed production information management system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of an order raw material feature extraction unit in a feed production information management system according to an embodiment of the present application.

[0019] Figure 4 It is a block diagram of a feed production grinding judgment module in a feed production information management system according to an embodiment of the present application.

[0020] Figure 5 It is a flowchart of the feed production information management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0022] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0023] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0024] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0025] Figure 1 Schematic diagram of a feed production information management system according to an embodiment of the present application. Figure 1 As shown, according to the feed production information management system 100 of the embodiment of the present application, it includes: a feed production data acquisition module 110, which is used to obtain the feed production monitoring video collected by the camera and the order raw material information collected by the database; a feed production data extraction module 120, which is used to extract the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector from the feed production monitoring video collected by the camera and the order raw material information collected by the database; a feed production grinding judgment module 130, which is used to judge whether the feed production grinding is sufficient based on the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector.

[0026] In the above-mentioned feed production information management system 100, the feed production data acquisition module 110 is used to obtain the feed production monitoring video collected by the camera and the order raw material information collected by the database. It should be understood that feed, as a food specifically for providing nutrition to animals, has the main goal of promoting the health, growth and production performance of animals. Generally, feed is composed of a variety of ingredients, including grains, protein sources, vitamins and minerals to meet the diverse nutritional needs of different animals. In the feed production process, since different orders have different requirements for formulas and production processes, the production process must have a certain degree of flexibility. However, current technology usually relies on fixed production parameters and procedures, which limits the adaptability of the production process to personalized needs. Although manual monitoring is a common solution in practice, the operator may make mistakes due to fatigue, negligence or lack of experience, resulting in the timeliness of monitoring and adjustment being affected, which may have a negative impact on the quality of feed. In addition, manual monitoring often has a certain lag and cannot respond to changes that may occur in the production process in real time. Therefore, in the technical solution of this application, the feed production monitoring video collected by the camera and the order raw material information collected by the database are obtained, and combined with deep learning technology, the reliance on manual monitoring is reduced by evaluating the grinding degree in feed production, thereby reducing the impact caused by human errors and operator fatigue. This method provides more objective adjustment suggestions and helps optimize production management decisions.

[0027] Specifically, the acquisition of feed production monitoring videos collected by cameras and order raw material information collected by databases aims to comprehensively monitor and optimize the feed production process. The monitoring videos collected by cameras provide the system with real-time visual data, enabling it to intuitively observe every detail on the production line. These video data can not only record the actual grinding process of the feed, but also help detect abnormalities in production. On the other hand, the order raw material information in the database contains detailed raw material data for each production batch, including raw material type, quantity and specifications. This information is crucial to understanding the specific raw materials used in the production process and their ratios, and can help the system evaluate the impact of raw materials on the final feed quality. By combining data from these sources, the system can achieve multi-dimensional analysis, identify the key factors affecting the uniformity of feed grinding, and thus perform precise quality control and production optimization. This data integration not only improves the transparency of production, but also enhances the traceability of the production process, providing a solid guarantee for the quality of the final product.

[0028] In the feed production information management system 100, the feed production data extraction module 120 is used to extract the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector from the feed production monitoring video collected by the camera and the order raw material information collected by the database. In this way, all aspects of feed production can be comprehensively evaluated to ensure that the grinding process meets the production standards and improve the overall quality of the feed, thereby discovering potential problems in the production process, proposing improvement measures, optimizing the production process, and improving the consistency and reliability of the product.

[0029] Figure 2 FIG. 1 is a block diagram of a feed production data extraction module in a feed production information management system according to an embodiment of the present application. Figure 2 As shown, in a specific embodiment of the present application, the feed production data extraction module 120 includes: a feed production data preprocessing unit 121, which is used to perform data preprocessing on the feed production monitoring video collected by the camera to obtain multiple feed grinding uniformity production pixel enhanced image key frames; a feed production target feature encoding unit 122, which is used to perform target feature encoding on the multiple feed grinding uniformity production pixel enhanced image key frames to obtain a feed grinding uniformity tracking feature vector; a feed production convolution encoding unit 123, which is used to perform convolution encoding on the multiple feed grinding uniformity production pixel enhanced image key frames to obtain a feed grinding uniformity production associated feature vector; a feed grinding uniformity feature association unit 124, which is used to associate the feed grinding uniformity tracking feature vector with the feed grinding uniformity production associated feature vector to obtain the feed grinding uniformity multimodal feature vector; an order raw material feature extraction unit 125, which is used to perform feature extraction on the order raw material information collected by the database to obtain the order raw material text understanding feature vector.

[0030] It should be understood that data preprocessing includes techniques such as denoising, smoothing, and image enhancement. These steps help eliminate visual noise introduced by camera acquisition and enhance image clarity and detail. By extracting key frames, the system can select representative and information-rich images from the entire production process. These images can not only show the changes in the operating status of the grinding equipment, but also capture the key moments that may affect the quality of feed. This data preprocessing and key frame extraction not only improves the efficiency and operability of the monitoring system, but also provides production managers with intuitive and decision-making visual information, helping them to promptly discover and solve potential production problems, and ultimately improve the quality and consistency of feed production.

[0031] Furthermore, target feature encoding is performed on multiple key frames of pixel-enhanced images of feed grinding uniformity production to obtain feed grinding uniformity tracking feature vectors, aiming to convert complex visual data into quantifiable and analyzable feature vectors through advanced image processing and computer vision technology, thereby achieving accurate tracking and analysis of changes in grinding uniformity during feed production. Among them, target feature encoding usually involves deep learning models or feature extraction algorithms, which can extract key visual features such as color distribution, texture information, and spatial layout from pixel-level enhanced images. Tracking using feature vectors can achieve real-time monitoring and analysis, and timely detect abnormalities or quality problems in the grinding process. This method not only improves the accuracy and response speed of the monitoring system, but also provides production managers with actionable data support to help them optimize production processes and improve product quality.

[0032] Furthermore, convolutional coding is performed on multiple keyframes of feed grinding uniformity production pixel-enhanced images, aiming to use the convolutional neural network (CNN) technology in deep learning to extract high-level features from complex visual data to obtain more accurate and informative feed grinding uniformity production-related feature vectors. The convolutional coding process first takes each keyframe as input, and gradually extracts the spatial features and hierarchical structure of the image through a series of convolutional layers and pooling layers. These hierarchical features capture important information such as particle distribution, color changes, and texture details during the grinding process, which helps to accurately describe and quantify the dynamic changes of feed grinding uniformity. The advantage of convolutional coding lies in its sensitivity to local image features and its ability to handle translation invariance, which makes it perform well in processing complex production monitoring images.

[0033] In particular, the feed grinding uniformity tracking feature vector is associated with the feed grinding uniformity production-related feature vector to obtain the feed grinding uniformity multimodal feature vector, in order to comprehensively utilize the feature information from different sources, thereby providing a more comprehensive and accurate assessment of feed grinding quality. Specifically, the feed grinding uniformity tracking feature vector usually comes from real-time image data analysis, reflecting the immediate changes and grinding quality in the production process. These feature vectors are based on real-time monitoring data and can capture dynamic changes and short-term fluctuations in the grinding process. On the other hand, the feed grinding uniformity production-related feature vector is extracted from multiple key frames through convolutional coding, representing the feature data accumulated over a long period of time in the production process, reflecting the overall grinding quality and the long-term performance of production equipment. Associating these two types of feature vectors helps to comprehensively evaluate feed grinding uniformity at two different time scales.

[0034] More specifically, feature extraction is performed on the order raw material information collected from the database to obtain the order raw material text understanding feature vector, aiming to convert the original text data into machine-processable structured features to support deeper analysis and decision-making. Order raw material information usually includes a large amount of descriptive data, such as raw material name, specification, quantity, supplier information and related attributes, which are often unstructured text in their original form and difficult to directly analyze effectively. The feature extraction process uses natural language processing (NLP) technology to convert text information into numerical feature vectors. This process first reduces noise and extracts core information through text preprocessing steps such as removing stop words, tokenization, and stemming. Next, word embedding technology (such as Word2Vec, GloVe) or context embedding technology (such as BERT) is used to map words or phrases to a continuous vector space. These embedding technologies can capture the semantic relationship and context information between words, so that the extracted feature vectors can effectively represent the important features and patterns in the raw material information.

[0035] In a specific embodiment of the present application, the feed production data preprocessing unit 121 includes: extracting multiple feed grinding uniformity production image key frames from the feed production monitoring video collected by the camera; passing the multiple feed grinding uniformity production image key frames through a noise reduction module to obtain the multiple feed grinding uniformity production pixel enhanced image key frames.

[0036] It should be understood that surveillance video provides a continuous visual data stream, containing real-time production scenes and process details. By extracting keyframes, representative time points or events in the video can be selected, such as specific frames per second or key moments of key operations, to extract static representations of the image. These keyframes are significant in both time and space: they capture visual information at key moments in the production process, such as grinding equipment operation, raw material delivery, and particle movement. Compared with continuous video data, keyframes provide a more compact and efficient data representation, which can reduce storage costs and simplify data transmission processes, especially when applied in large-scale monitoring systems.

[0037] Furthermore, multiple feed grinding uniformity production image keyframes are passed through a noise reduction module to obtain pixel enhanced image keyframes, aiming to optimize image quality and enhance useful information in the image, so as to more accurately analyze and evaluate the uniformity in the grinding process. In production monitoring, images may be affected by various factors, such as poor lighting conditions, equipment vibration or lens blur, resulting in noise or unclear details in the image. The role of the noise reduction module is to eliminate or reduce these noises through signal processing technology, so that the image is visually clearer and more accurate. The noise reduction module can apply various image processing techniques, such as smoothing filtering, wavelet transform or neural network model based on deep learning, and select according to the specific noise type and characteristics in the image. These techniques can effectively filter out random noise from sensors or the environment while retaining key features and structural information in the image. Through noise reduction processing, the visual quality of the image is improved, which helps to more accurately analyze and measure the particle distribution and uniformity in the grinding process. Specifically, the multiple feed grinding uniformity production image key frames are input into the encoder of the noise reduction module, wherein the encoder uses a convolution layer to perform explicit spatial encoding on the multiple feed grinding uniformity production image key frames to obtain image features; and the image features are input into the decoder of the noise reduction module, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the image features to obtain the multiple feed grinding uniformity production pixel enhanced image key frames.

[0038] In a specific embodiment of the present application, the feed production target feature encoding unit 122 includes: passing the multiple feed grinding uniformity production pixel enhanced image key frames through a target detection network to obtain multiple feed grinding uniformity regions of interest; passing the multiple feed grinding uniformity regions of interest through a feed grinding uniformity significant target detector to obtain the feed grinding uniformity tracking feature vector.

[0039] It should be understood that through the target detection network, each frame in the pixel enhanced image keyframe is first fully scanned and analyzed. The network detects and identifies features or objects related to feed grinding uniformity that may exist in the image, such as particles, equipment parts or specific operating areas. These regions of interest can be classified based on predefined labels or categories, such as areas with good uniformity, areas with abnormal particle distribution, etc. The identified regions of interest facilitate more detailed data analysis and quality control. For example, further feature extraction, such as size, shape and color distribution, can be performed on the particles in each area to evaluate the specific parameters of grinding uniformity. The target detection network provides a highly automated solution, reducing the possibility of manual intervention and errors, while improving the efficiency and accuracy of the production monitoring system. Specifically, the target detection network is used to process the multiple feed grinding uniformity production pixel enhanced image keyframes using the following target detection formula to obtain the multiple feed grinding uniformity regions of interest;

[0040] Among them, the target detection formula is:

[0041] ROI=(cls(ψ det ,B),Regr(ψ det ,B))

[0042] Among them, ψ det Generate pixel-enhanced image keyframes for multiple feed grinding uniformity, B is the anchor frame, ROI is the region of interest for multiple feed grinding uniformity, cls(ψ det ,B) indicates classification, Regr(ψ det ,B) indicates regression.

[0043] Furthermore, multiple feed grinding uniformity regions of interest are processed through a feed grinding uniformity salient object detector, which aims to further accurately track and quantify uniformity features in the grinding process, thereby achieving more refined monitoring and analysis of production quality. The feed grinding uniformity salient object detector is a specially designed algorithm or model that can efficiently identify and track important uniformity-related features or objects in an image, such as the shape, color distribution of particles, and their movement trajectory within the equipment. Through this process, representative uniformity feature points or regions are first selected from the region of interest, which may reflect key details and changes in the grinding process. The object detector uses advanced computer vision techniques, such as deep learning convolutional neural networks (CNNs), to identify and extract these features and map them into numerical feature vectors. The feed grinding uniformity tracking feature vectors also provide important data support for real-time monitoring and feedback systems. Based on these feature vectors, an early warning mechanism or an automated adjustment system can be established to respond to problems or fluctuations in the grinding process in a timely manner to ensure the stability and consistency of production quality. Specifically, each layer of the feed grinding uniformity significant target detector is used to perform the following operations on the input data in the forward pass of the layer: convolution processing is performed on the input data using a first convolution kernel to obtain a first convolution feature map; convolution processing is performed on the first convolution feature map using a second convolution kernel to obtain a second convolution feature map, wherein the size of the first convolution kernel is larger than the size of the second convolution kernel; and, pooling processing is performed on the second convolution feature map to obtain a pooling feature map; and, activation processing is performed on the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the feed grinding uniformity significant target detector is the feed grinding uniformity tracking feature vector, and the input of the first layer of the feed grinding uniformity significant target detector is the multiple feed grinding uniformity regions of interest.

[0044] In a specific embodiment of the present application, the feed production convolutional encoding unit 123 includes: arranging the multiple feed grinding uniformity production pixel enhanced image key frames into a feed grinding uniformity production input tensor; passing the feed grinding uniformity production input tensor through a feed grinding uniformity production three-dimensional convolutional neural network to obtain the feed grinding uniformity production associated feature vector.

[0045] It should be understood that the process of arranging these image frames into an input tensor involves converting each image into a numerical data structure so that the computer can understand and process it. This data format not only allows the processing of information at multiple time points at the same time, but also preserves the spatial and temporal dimensional features in the image. The construction process of the input tensor usually includes normalizing each image, such as resizing, normalizing brightness, or color space conversion, to ensure data consistency and comparability. Then, these processed images are organized into tensors in chronological order or specific time windows, forming a multi-dimensional data structure, in which each dimension corresponds to a specific attribute or feature of the image. The main purpose of this data arrangement is to perform end-to-end learning and prediction through a deep learning model. By using such a data structure, real-time monitoring and prediction of grinding uniformity in the production process can be achieved. The deep learning model can learn the correlation and change pattern between images at different time points during the training phase, so as to predict possible future grinding uniformity problems or change trends in real-time production.

[0046] Furthermore, the feed grinding uniformity production input tensor is processed by a three-dimensional convolutional neural network (3DCNN) to obtain the feed grinding uniformity production-related feature vector, in order to use deep learning technology to efficiently capture and analyze the complex spatial and temporal features in the image data. The 3D convolutional neural network is a model that specializes in processing three-dimensional data structures, which can simultaneously consider the spatial dimension (width and height) and temporal dimension (frame sequence) of the image, so as to more comprehensively understand and analyze the dynamic changes and temporal characteristics in the data. In the feed grinding uniformity production monitoring, the input tensor usually consists of image frames at multiple time points, and each frame of the image represents the state of the grinding process at a different time point. The 3D convolutional neural network can extract and learn local features and global patterns in the image data by performing convolution operations in these three dimensions. This method allows the model to combine temporal and spatial information when processing data, and mine complex features in the image sequence, such as the change trend of the grinding particles, the temporal fluctuation of uniformity, and the long-term effect of equipment operation. By training the 3D convolutional neural network, the model can automatically learn the deep-level features related to the feed grinding uniformity and encode these features into feature vectors. These feature vectors provide a comprehensive representation of the state of the production process, which can help further analysis and decision-making. Specifically, each layer of the three-dimensional convolutional neural network for feed grinding uniformity production performs the following operations on the input data in the forward pass of the layer: using the convolution units of each layer of the three-dimensional convolutional neural network for feed grinding uniformity production to perform convolution processing based on the convolution kernel on the input data to obtain a convolution feature map; using the pooling units of each layer of the three-dimensional convolutional neural network for feed grinding uniformity production to perform pooling processing on the convolution feature map along the channel dimension to obtain a pooling feature map; and, using the activation units of each layer of the three-dimensional convolutional neural network for feed grinding uniformity production 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 three-dimensional convolutional neural network for feed grinding uniformity production is the associated feature vector of feed grinding uniformity production.

[0047] Figure 3 FIG. 1 is a block diagram of an order raw material feature extraction unit in a feed production information management system according to an embodiment of the present application. Figure 3As shown, in a specific embodiment of the present application, the order raw material feature extraction unit 125 includes: an order raw material information word embedding subunit 1251, which is used to pass the order raw material information collected from the database through an order raw material information word embedding model to obtain a sequence of order raw material information embedding vectors; an order raw material information construction subunit 1252, which is used to construct the sequence of order raw material information embedding vectors into an order raw material semantic feature matrix; an order raw material information semantic convolution subunit 1253, which is used to pass the order raw material semantic feature matrix through an order raw material semantic feature extractor to obtain an order raw material text understanding feature vector.

[0048] It should be understood that the word embedding model of order raw material information is a deep learning technology that can map each word or phrase to a vector in a high-dimensional space, so that semantically similar words are close in the vector space. Through such an embedding process, the semantics and contextual information of the order raw material information can be effectively encoded and expressed. Among them, the order raw material information in the database may include product names, specifications, supplier information, etc. After preprocessing and standardization, this information is input into the word embedding model for processing. The model generates a corresponding high-dimensional embedding vector for each order raw material information by learning the contextual relationship and semantic information in the text. These vectors not only retain the semantic information of the original text, but also capture the association and similarity between different raw materials. Specifically, the order raw material information collected from the database is segmented to obtain an order raw material word sequence; the embedding layer of the order raw material information word embedding model is used to map each order raw material word in the order raw material word sequence to a word embedding vector to obtain a sequence of order raw material information embedding vectors.

[0049] Furthermore, the sequence of embedding vectors of order raw material information is constructed into an order raw material semantic feature matrix, aiming to express and analyze the semantic features of order raw materials in a more in-depth manner through integration and structured methods. The order raw material semantic feature matrix is ​​a two-dimensional data structure formed by arranging each order raw material information embedding vector according to certain rules or strategies, where each row represents an order raw material and each column corresponds to a specific semantic feature dimension. The embedding vector sequence is usually arranged in the order of orders or time series. By arranging these vectors into a matrix in a predetermined standardized manner, efficient organization and management of order raw material information can be achieved.

[0050] Furthermore, the order raw material semantic feature matrix is ​​passed through the order raw material semantic feature extractor to obtain the order raw material text understanding feature vector, in order to further enhance the in-depth understanding and analysis capabilities of the order raw material information. The order raw material semantic feature extractor is a specially designed algorithm or model that can extract higher-level abstract feature representations from the structured semantic feature matrix. These feature vectors are more explanatory and applicable. The order raw material semantic feature matrix contains multi-dimensional semantic feature representations of each order raw material. These features may cover information on product attributes, supplier relationships, market trends, etc. Through the order raw material semantic feature extractor, various technologies (such as deep learning, natural language processing, or statistical analysis methods) can be used to extract potential high-level semantic features from these feature matrices, such as potential demand patterns, price sensitivity, supply chain risks, etc. of order raw materials. Specifically, each layer of the order raw material semantic feature extractor is used to perform the following operations on the input data in the forward pass of the layer: convolution processing is performed on the input data based on the convolution kernel to generate a convolution feature map; global mean pooling processing is performed on the convolution feature map based on the feature matrix to generate a pooling feature map; and non-linear activation is performed on the feature values ​​at each position in the pooling feature map to generate an activation feature addition map; wherein the output of the last layer of the order raw material semantic feature extractor is the order raw material text understanding feature vector, the input from the second layer to the last layer of the order raw material semantic feature extractor is the output of the previous layer, and the input of the order raw material semantic feature extractor is the order raw material semantic feature matrix.

[0051] In the above-mentioned feed production information management system 100, the feed production grinding judgment module 130 is used to judge whether the feed production grinding is sufficient based on the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector. It should be understood that by combining the information of these two feature vectors, a comprehensive evaluation model can be established to judge whether the feed production grinding is sufficient. For example, by analyzing the uniformity of particle distribution and the dispersion of particle size in the grinding uniformity feature vector, combined with the raw material fineness requirements and physical properties before grinding in the order raw material feature vector, the satisfaction degree in the grinding process can be quantitatively evaluated.

[0052] Figure 4 FIG. 1 is a block diagram of a feed production grinding judgment module in a feed production information management system according to an embodiment of the present application. Figure 4As shown, in a specific embodiment of the present application, the feed production and grinding judgment module 130 includes: a feed production feature fusion unit 131, which is used to fuse the multimodal feature vector of feed grinding uniformity and the feature vector of order raw material text understanding to obtain a feed production and grinding sufficient judgment feature vector; a feed production feature optimization unit 132, which is used to perform local disturbance correction on the feed production and grinding sufficient judgment feature vector based on backpropagation shift compensation to obtain an optimized feed production and grinding sufficient judgment feature vector; a feed production and grinding sufficient judgment classification unit 133, which is used to pass the optimized feed production and grinding sufficient judgment feature vector through a classifier to obtain a classification result, and the classification result is used to judge whether the feed production and grinding is sufficient.

[0053] It should be understood that by integrating these two eigenvectors, a comprehensive evaluation model can be established, namely the feed production grinding full judgment eigenvector. This eigenvector combines the physical characteristics of grinding uniformity and the multidimensional data of raw material characteristics, and can quantitatively evaluate the effect of the grinding process. For example, the particle distribution and raw material characteristics in the eigenvector can be used to predict whether the uniformity and fineness of the particles after grinding meet the production standards, and then adjust the parameters of the grinding equipment or the raw material ratio to optimize the production process and product quality.

[0054] In particular, in the technical solution of the present application, the feed production grinding sufficient judgment feature vector is obtained by feature processing of data from different data sources, wherein the data from different data sources include feed production monitoring videos collected by cameras and order raw material information collected by databases. In the process of image denoising and target detection in feed production monitoring, local features may be distorted due to incomplete processing or error accumulation, thereby affecting the final multimodal feature vector. The text understanding feature vector of order raw material information may also be affected by factors such as input text quality and embedding model accuracy, resulting in inaccurate extraction of semantic features. The disturbances generated by the noise and variant data of these features in the fused feed production grinding sufficient judgment feature vector will be further amplified, so that the classifier cannot accurately judge the adequacy of feed production grinding. That is, the work of the classifier is based on the input feed production grinding sufficient judgment feature vector. If the feed production grinding sufficient judgment feature vector is unstable due to disturbances in the data processing process, the classifier will have insufficient classification accuracy. Therefore, in the technical solution of the present application, the feed production grinding sufficient judgment feature vector is corrected for local disturbances based on back-propagation shift compensation to obtain an optimized feed production grinding sufficient judgment feature vector.

[0055] Among them, the feed production and grinding sufficient judgment feature vector is subjected to local disturbance correction based on back-propagation shift compensation to obtain an optimized feed production and grinding sufficient judgment feature vector, including: determining a first feed production and grinding sufficient judgment weight matrix and a second feed production and grinding sufficient judgment weight matrix of the classifier before and after each iterative update; matrix multiplying the first feed production and grinding sufficient judgment weight matrix and the second feed production and grinding sufficient judgment weight matrix with the feed production and grinding sufficient judgment feature vector respectively to obtain a first feed production and grinding sufficient judgment classification modulation feature vector and a second feed production and grinding sufficient judgment classification modulation feature vector; calculating the position difference between the first feed production and grinding sufficient judgment classification modulation feature vector and the second feed production and grinding sufficient judgment classification modulation feature vector to obtain a feed production and grinding sufficient judgment shift information representation vector; calculate the F norm of the position mean vector between the first feed production and grinding sufficient judgment classification modulation feature vector and the second feed production and grinding sufficient judgment classification modulation feature vector as the feed production and grinding sufficient judgment shift compensation scaling factor; linearly scale the feed production and grinding sufficient judgment shift information representation vector with the feed production and grinding sufficient judgment shift compensation scaling factor to obtain the scaled feed production and grinding sufficient judgment shift information representation vector, and input the scaled feed production and grinding sufficient judgment shift information representation vector into the Sigmoid activation function to obtain the feed production and grinding sufficient judgment back-propagation shift compensation representation vector; calculate the position point multiplication of the feed production and grinding sufficient judgment back-propagation shift compensation representation vector and the feed production and grinding sufficient judgment feature vector to obtain the optimized feed production and grinding sufficient judgment feature vector.

[0056] The optimization steps are specifically expressed as follows:

[0057]

[0058] Wherein, V represents the feed production and grinding sufficient judgment feature vector, S represents the feed production and grinding sufficient judgment shift compensation scaling factor, M 2 represents the second feed production grinding sufficiency judgment weight matrix, M 1 represents the first feed production grinding sufficient judgment weight matrix, sigmoid represents the activation function, ⊙ represents the position point multiplication, represents matrix multiplication, represents the difference by position, represents addition by position, ||·||F represents the Frobenius norm, and V' represents the optimized feed production grinding sufficiency judgment feature vector.

[0059] In the technical solution of the present application, in order to enhance the robustness of the classifier to the local disturbance in the input feature vector, the feed production and grinding sufficient judgment feature vector is corrected by the local disturbance based on the back propagation shift compensation. Among them, the key to the local disturbance correction based on the back propagation shift compensation of the feed production and grinding sufficient judgment feature vector is to capture the information of the weight matrix changing with the training process, and use this information to generate the feed production and grinding sufficient judgment shift information representation vector, and then generate the feed production and grinding sufficient judgment back propagation shift compensation representation vector through scaling and nonlinear transformation (Sigmoid function). In this way, the feed production and grinding sufficient judgment feature vector is supported by the distribution difference of different classification scenarios of the classifier during the training process to describe the local displacement of the feature vector, so that not only the displacement of the feature vector caused by the change of the weight matrix can be quantified, but also a compensation vector can be generated according to the displacement information. The role of the compensation vector is to weight and adjust each element in the feature vector according to the degree of influence of the disturbance, thereby reducing the influence of the disturbance on the classification result. In this way, the classifier's robustness to local disturbances in the input data can be significantly improved, the classifier's stability and generalization ability can be enhanced, and by reducing the impact of local disturbances on the classification results, the classifier can still maintain a high classification accuracy when facing noisy or variant data.

[0060] Furthermore, the optimized feed production grinding sufficient judgment feature vector is obtained by fusing the multimodal features of feed grinding uniformity and the order raw material text understanding features. These feature vectors reflect key information such as the distribution uniformity of particles, particle size and raw material characteristics during the grinding process. Through the classifier, the optimized feed production grinding sufficient judgment feature vector can be classified and learned using the existing training data set, and a classification model can be established to distinguish whether the grinding meets the expected requirements. The application of classification results is not limited to real-time monitoring and feedback. This data-driven approach helps to reduce production costs, improve product quality, and enhance the competitiveness of enterprises in the highly competitive market.

[0061] In summary, the embodiment of the present application first obtains the feed production monitoring video collected by the camera and the order raw material information collected by the database, and then uses deep learning technology to perform feature extraction and association analysis on the two, and finally obtains the classification result through the classifier to determine whether the feed production is sufficiently ground, thereby reducing the dependence on manual monitoring, reducing the impact of human errors and operational fatigue, and providing more objective adjustment suggestions to optimize production management decisions.

[0062] As described above, the feed production information management system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the feed production information management system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the feed production information management system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the feed production information management system 100 can also be one of the many hardware modules of the terminal device.

[0063] Alternatively, in another example, the feed production information management system 100 and the terminal device may also be separate devices, and the feed production information management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0064] Figure 5 Flow chart of the feed production information management method according to an embodiment of the present application. Figure 5 As shown, according to the feed production information management method of the embodiment of the present application, it includes: S110, acquiring the feed production monitoring video collected by the camera and the order raw material information collected by the database; S120, extracting the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector from the feed production monitoring video collected by the camera and the order raw material information collected by the database; S130, judging whether the feed production grinding is sufficient based on the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector.

[0065] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned feed production information management method have been described in detail above. Figures 1 to 4 The description of the feed production information management system has been described in detail, and therefore, its repeated description will be omitted.

[0066] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0067] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0068] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0069] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0070] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0071] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit of the technical solution of the present invention.

Claims

1. A feed production information management system, characterized in that: include: A feed production data acquisition module is used to acquire feed production monitoring videos collected by cameras and order raw material information collected by databases; A feed production data extraction module, used to extract a multimodal feature vector of feed grinding uniformity and a feature vector of order raw material text understanding from the feed production monitoring video collected by the camera and the order raw material information collected by the database; The feed production grinding judgment module is used to judge whether the feed production grinding is sufficient based on the multimodal feature vector of feed grinding uniformity and the feature vector of order raw material text understanding.

2. The feed production information management system according to claim 1, characterized in that: The feed production data extraction module comprises: A feed production data preprocessing unit, used for performing data preprocessing on the feed production monitoring video collected by the camera to obtain a plurality of feed grinding uniformity production pixel enhanced image key frames; A feed production target feature encoding unit, used for performing target feature encoding on the plurality of feed grinding uniformity production pixel enhancement image key frames to obtain a feed grinding uniformity tracking feature vector; A feed production convolutional coding unit, configured to perform convolutional coding on the plurality of feed grinding uniformity production pixel enhancement image key frames to obtain a feed grinding uniformity production associated feature vector; A feed grinding uniformity feature association unit, used for associating the feed grinding uniformity tracking feature vector with the feed grinding uniformity production association feature vector to obtain the feed grinding uniformity multimodal feature vector; The order material feature extraction unit is used to extract features from the order material information collected from the database to obtain the order material text understanding feature vector.

3. The feed production information management system according to claim 2, characterized in that: The feed production data preprocessing unit comprises: Extracting a plurality of feed grinding uniformity production image key frames from the feed production monitoring video collected by the camera; The multiple feed grinding uniformity production image key frames are passed through a noise reduction module to obtain the multiple feed grinding uniformity production pixel enhanced image key frames.

4. The feed production information management system according to claim 3, characterized in that: The feed production target feature coding unit comprises: Passing the plurality of feed grinding uniformity production pixel enhanced image key frames through an object detection network to obtain a plurality of feed grinding uniformity interest regions; The multiple feed grinding uniformity regions of interest are passed through a feed grinding uniformity salient object detector to obtain the feed grinding uniformity tracking feature vector.

5. The feed production information management system according to claim 4, characterized in that: The feed production convolutional encoding unit comprises: Arranging the plurality of feed grinding uniformity production pixel enhanced image key frames into a feed grinding uniformity production input tensor; The feed grinding uniformity production input tensor is passed through a feed grinding uniformity production three-dimensional convolutional neural network to obtain the feed grinding uniformity production associated feature vector.

6. The feed production information management system according to claim 5, characterized in that: The order raw material feature extraction unit comprises: An order raw material information word embedding subunit, used for passing the order raw material information collected from the database through an order raw material information word embedding model to obtain a sequence of order raw material information embedding vectors; An order raw material information construction subunit, used to construct the sequence of the order raw material information embedding vectors into an order raw material semantic feature matrix; The order material information semantic convolution subunit is used to pass the order material semantic feature matrix through the order material semantic feature extractor to obtain the order material text understanding feature vector.

7. The feed production information management system according to claim 6, characterized in that: The feed production and grinding judgment module comprises: A feed production feature fusion unit, used for fusing the feed grinding uniformity multimodal feature vector and the order raw material text understanding feature vector to obtain a feed production grinding sufficiency judgment feature vector; A feed production feature optimization unit, used for performing local disturbance correction based on back-propagation shift compensation on the feed production grinding sufficiency judgment feature vector to obtain an optimized feed production grinding sufficiency judgment feature vector; The feed production and grinding sufficiency judgment classification unit is used to pass the optimized feed production and grinding sufficiency judgment feature vector through a classifier to obtain a classification result, and the classification result is used to judge whether the feed production and grinding are sufficient.

8. The feed production information management system according to claim 7, characterized in that: The feed production characteristic optimization unit comprises: Determine a first feed production and grinding sufficiency judgment weight matrix and a second feed production and grinding sufficiency judgment weight matrix before and after each iterative update of the classifier; The first feed production and grinding sufficient judgment weight matrix and the second feed production and grinding sufficient judgment weight matrix are respectively matrix-multiplied with the feed production and grinding sufficient judgment feature vector to obtain a first feed production and grinding sufficient judgment classification modulation feature vector and a second feed production and grinding sufficient judgment classification modulation feature vector; Calculating the position difference between the first feed production and grinding sufficiency judgment classification modulation feature vector and the second feed production and grinding sufficiency judgment classification modulation feature vector to obtain a feed production and grinding sufficiency judgment shift information representation vector; Calculate the F norm of the position mean vector between the first feed production and grinding sufficiency judgment classification modulation feature vector and the second feed production and grinding sufficiency judgment classification modulation feature vector as the feed production and grinding sufficiency judgment shift compensation scaling factor; Linearly scaling the feed production and grinding sufficient judgment shift information representation vector by the feed production and grinding sufficient judgment shift compensation scaling factor to obtain a scaled feed production and grinding sufficient judgment shift information representation vector, and inputting the scaled feed production and grinding sufficient judgment shift information representation vector into a Sigmoid activation function to obtain a feed production and grinding sufficient judgment back-propagation shift compensation representation vector; The optimized feed production and grinding adequacy judgment feature vector is obtained by calculating the point-by-point multiplication of the back-propagation shift compensation representation vector of the feed production and grinding adequacy judgment and the feature vector of the feed production and grinding adequacy judgment.

9. A method for managing feed production information, characterized in that: include: Obtain feed production monitoring videos collected by cameras and order raw material information collected by databases; Extracting a multimodal feature vector of feed grinding uniformity and a feature vector of order raw material text understanding from the feed production monitoring video collected by the camera and the order raw material information collected by the database; Based on the multimodal feature vector of feed grinding uniformity and the feature vector of order raw material text understanding, it is judged whether the feed production grinding is sufficient.

10. The feed production information management method according to claim 9, characterized in that: Extracting a multimodal feature vector of feed grinding uniformity and a feature vector of order raw material text understanding from the feed production monitoring video collected by the camera and the order raw material information collected by the database, including: Performing data preprocessing on the feed production monitoring video collected by the camera to obtain a plurality of feed grinding uniformity production pixel enhanced image key frames; Performing target feature encoding on the plurality of feed grinding uniformity production pixel enhanced image key frames to obtain a feed grinding uniformity tracking feature vector; Convolution encoding is performed on the plurality of feed grinding uniformity production pixel enhancement image key frames to obtain a feed grinding uniformity production associated feature vector; Associating the feed grinding uniformity tracking feature vector with the feed grinding uniformity production association feature vector to obtain the feed grinding uniformity multimodal feature vector; Feature extraction is performed on the order raw material information collected from the database to obtain the order raw material text understanding feature vector.

Images