Process and device for producing prefabricated spicy beef tallow sauce
Through intelligent control solutions, deep learning technology is used to monitor and analyze the stir-frying process in real time, and automatically determine the timing of turning off the fire, solving the problem of poor product consistency and stability in the production process of pre-made spicy butter sauces in the existing technology, realizing the More efficient intelligent fire-off control improves product quality.
Patent Information
- Application Number
- CN202411027225.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing pre-made spicy butter sauce production process has poor product consistency and stability due to manual control, and the garlic paste stir-frying process is a complex process involving multivariable participation. Single variable fire-off control cannot improve product consistency and stability. .
Using an intelligent control scheme, the frying state video is obtained through the camera, keyframe sampling and image semantic feature extraction, the frying state prototype characterization feature vector is determined by using the convolutional neural network and the prototype feature extraction network, and the frying state prototype characterization feature vector is automatically judged, and the timing of turning off fire is realized. Intelligent fire turn-off control.
It improves the product consistency and stability of pre-made spicy butter sauce, ensures that the garlic paste is fully discolored and avoids scorching, and improves production efficiency and product quality.
Smart Images

Figure CN118902059A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent production, and more specifically, to a production process and device for prefabricated spicy butter sauce. Background Art
[0002] Pre-made spicy beef sauce is a convenient condiment that can be used to make various spicy dishes, such as spicy hot pot, spicy hot pot, spicy hot pot, etc. The main raw materials of pre-made spicy beef sauce include chili peppers, peppercorns, spices, cooking oil, etc. Through fine processing technology, the raw materials are mixed, fried, filled, sealed and other steps to make a sauce with rich aroma and spicy taste.
[0003] In the production process of pre-made spicy butter sauce, since the production process is mostly controlled manually based on experience, the taste and quality consistency of each batch of sauce is poor, and manual operation and contact will also introduce unnecessary pollution and errors. Therefore, a more optimized production process of pre-made spicy butter sauce is expected, especially to improve its intelligence and automation. Summary of the invention
[0004] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiment of the present application provides a prefabricated spicy beef butter sauce production process and device, which uses an intelligent control scheme to construct an intelligent fire-off control scheme to improve the product consistency and stability of the prefabricated spicy beef butter sauce. However, in the technical scheme of the present application, judging the timing of turning off the fire is a very complex technical problem. If a threshold-based fire-off control is performed based on a single variable such as temperature or smell, this cannot improve the consistency and stability of the product. The reason is that the process of stir-frying mashed garlic is a process involving multiple variables, and the relationship between each variable is complex. In recent years, the development of deep learning, especially artificial intelligence, has provided new solutions and solutions for intelligent fire-off control.
[0005] According to one aspect of the present application, a process for producing a prefabricated spicy butter sauce is provided, comprising:
[0006] Mix and grind the spices in proportion to obtain fine spice powder;
[0007] Process the garlic into garlic paste;
[0008] After heating the butter to a predetermined temperature, pour the heated butter into a wok, add appropriate amounts of salt and sugar, stir-fry evenly, then add the spice powder, and continue to stir-fry;
[0009] Pour the mashed garlic into the wok, stir-fry until the mashed garlic changes color, and turn off the heat to obtain the spicy butter sauce;
[0010] The spicy slurry is cooled and then put into a sealed container to obtain a pre-made spicy butter sauce.
[0011] According to another aspect of the present application, a prefabricated spicy butter sauce production device is provided, comprising:
[0012] A monitoring video acquisition module, used to acquire the stir-frying status monitoring video collected by the camera;
[0013] A sampling module, used for performing key frame sampling on the stir-frying state monitoring video to obtain a sequence of stir-frying state monitoring key frames;
[0014] An image semantic feature extraction module, used for extracting the semantic features of the stir-frying state image from the sequence of the stir-frying state monitoring key frames to obtain a sequence of the stir-frying state image semantic feature graphs;
[0015] a stir-fry state prototype representation feature extraction module, used to determine a stir-fry state prototype representation feature vector based on a sequence of the stir-fry state image semantic feature graphs; and
[0016] The control result generation module is used to determine whether to automatically turn off the fire based on the stir-fry state prototype characterization feature vector.
[0017] Compared with the prior art, the present application provides a prefabricated spicy butter sauce production process and device, which uses an intelligent control scheme to construct an intelligent fire-off control scheme to improve the product consistency and stability of the prefabricated spicy butter sauce. However, in the technical solution of the present application, judging the timing of turning off the fire is a very complex technical problem. If a threshold-based fire-off control is performed based on a single variable such as temperature or smell, this cannot improve the consistency and stability of the product. The reason is that the process of stir-frying mashed garlic is a multi-variable process, and the relationship between each variable is complex. In recent years, the development of deep learning, especially artificial intelligence, has provided new solutions and solutions for intelligent fire-off control. 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 This is a flow chart of a production process for prefabricated spicy butter sauce according to an embodiment of the present application;
[0020] Figure 2 This is a system architecture diagram of a prefabricated spicy butter sauce production process according to an embodiment of the present application;
[0021] Figure 3 This is a flow chart of sub-step S4 of the production process of pre-made spicy butter sauce according to an embodiment of the present application;
[0022] Figure 4 It is a flowchart of sub-step S42 of the production process of prefabricated spicy butter sauce according to an embodiment of the present application;
[0023] Figure 5 This is a block diagram of a prefabricated spicy beef sauce production device 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] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.
[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.
[0028] 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.
[0029] In a specific example, after the butter is heated to a predetermined temperature of 180°C, the heated butter is poured into a wok, and an appropriate amount of salt and sugar are added, stir-fried evenly, and then the spice powder is added and stir-fried continuously. Then, the mashed garlic is poured into the wok, stir-fried until the mashed garlic changes color, and the heat is turned off to obtain the spicy butter sauce. Finally, the spicy slurry is cooled and placed in a sealed container to obtain the pre-made spicy butter sauce.
[0030] In the above-mentioned pre-made spicy butter sauce production process, it has been found through research that one of the key steps to improve the consistency of taste and quality of each batch of sauce is to determine the timing of turning off the fire. The reason is that if the fire is turned off too early, the garlic paste will not change color sufficiently, and the aroma of the pre-made spicy butter sauce will not come out. If the fire is turned off too late, the garlic paste may be burnt, which will introduce new flavors and affect the taste of the pre-made spicy butter sauce. The traditional fire-off control is manually controlled, which makes it difficult to ensure the product consistency and stability of the pre-made spicy butter sauce.
[0031] In the technical solution of the present application, a production process for prefabricated spicy butter sauce is proposed. Figure 1 Flow chart of the production process of prefabricated spicy butter sauce according to an embodiment of the present application. Figure 1 As shown, the production process of the prefabricated spicy butter sauce according to the embodiment of the present application includes the following steps: S1, mixing spices in proportion and grinding them into fine powder to obtain spice fine powder; S2, processing garlic into garlic paste; S3, after heating the butter to a predetermined temperature, pouring the heated butter into a wok, adding appropriate amounts of salt and sugar, stir-frying evenly, then adding the spice fine powder, and continuing to stir-fry; S4, pouring the garlic paste into the wok, stir-frying until the garlic paste changes color, turning off the heat to obtain the spicy butter sauce; S5, cooling the spicy slurry and putting it into a sealed container to obtain the prefabricated spicy butter sauce.
[0032] In particular, in S1, spices are mixed in proportion and ground into fine powder to obtain spice fine powder. It should be understood that grinding spices into fine powder can release more fragrance and taste. Fine powdered spices are easier to mix with other ingredients, so that dishes can better absorb the flavor of spices, improving the taste and flavor of dishes.
[0033] In particular, the step S2 is to process the garlic into mashed garlic. It should be understood that processing the garlic into mashed garlic can enhance the flavor of the dishes, facilitate use and storage, and reduce the irritation of the garlic. This makes mashed garlic one of the commonly used seasonings in cooking.
[0034] In particular, in S3, after the butter is heated to a predetermined temperature, the heated butter is poured into a wok, and appropriate amounts of salt and sugar are added, stir-fried evenly, and then the spice powder is added, and stir-fried continuously. It should be understood that the above steps can improve the flavor and taste of the food.
[0035] In particular, in S4, pouring the mashed garlic into the wok, stir-frying until the mashed garlic changes color, and turning off the heat to obtain the spicy butter sauce. In particular, in a specific example of the present application, Figure 3 As shown, the S4 includes: S41, deploying a camera toward the wok; and S42, judging whether to automatically turn off the fire through machine vision.
[0036] Specifically, in S41, a camera is deployed toward the wok. It should be understood that the situation in the wok can be observed in real time through the camera, including the state of the ingredients, the control of the fire power, etc. This can help the chef or operator to adjust the cooking process in time to ensure the quality and taste of the food.
[0037] Specifically, the S42 is to determine whether to automatically turn off the fire through machine vision. In particular, in a specific example of the present application, Figure 2 This is a system architecture diagram of a prefabricated spicy butter sauce production process according to an embodiment of the present application. Figure 4 FIG. 4 is a flow chart of sub-step S42 of the production process of prefabricated spicy butter sauce according to an embodiment of the present application. Figure 2 and Figure 4 As shown, the S42 includes: S421, acquiring the stir-frying state monitoring video collected by the camera; S422, performing key frame sampling on the stir-frying state monitoring video to obtain a sequence of stir-frying state monitoring key frames; S423, extracting the semantic features of the stir-frying state image from the sequence of the stir-frying state monitoring key frames to obtain a sequence of stir-frying state image semantic feature maps; S424, determining the stir-frying state prototype representation feature vector based on the sequence of the stir-frying state image semantic feature maps; and, S425, determining whether to automatically turn off the fire based on the stir-frying state prototype representation feature vector.
[0038] More specifically, the S421, obtains the stir-fry state monitoring video collected by the camera. In the technical solution of the present application, the purpose of obtaining the stir-fry state monitoring video collected by the camera is to monitor the stir-fry process of the spicy butter sauce in real time, and extract relevant information through the video for subsequent processing. Here, the video collected by the camera can observe the stir-fry state in real time and obtain information in the stir-fry process in time. In this way, it can be ensured that the judgment of the time to turn off the fire is based on the latest stir-fry situation, avoiding inaccurate judgment due to time delay. In addition, the video contains rich visual information, such as color changes in the stir-fry process, the condition of mashed garlic, etc. By analyzing and processing the video, key frames and stir-fry state images can be extracted, and semantic features can be further extracted for judging the time to turn off the fire. It is worth mentioning that by recording the stir-fry process of each batch of sauces, a historical data set can be established, data analysis and mining can be performed, and the production process and taste quality of the spicy butter sauce can be further optimized.
[0039] More specifically, in S422, the stir-fry state monitoring video is sampled for key frames to obtain a sequence of stir-fry state monitoring key frames. That is, the stir-fry state monitoring video is sampled for key frames to obtain a sequence of stir-fry state monitoring key frames. Here, the stir-fry state monitoring video is sampled for key frames to obtain a sequence of stir-fry state monitoring key frames in order to reduce the amount of data and extract key information in the stir-fry process. Specifically, the stir-fry state monitoring video is usually a continuous frame sequence, containing a large amount of redundant information. By sampling the video for key frames, the amount of data can be reduced, and only the key frames are retained, thereby reducing the computational cost and storage requirements of subsequent processing. Here, the key frames are frames containing important information in the video, which usually represent key state changes in the stir-fry process. By sampling the key frames, key information such as garlic mash color change and stir-fry state transition can be captured for subsequent stir-fry state judgment and control instruction generation.
[0040] More specifically, the S423 extracts the semantic features of the stir-fry state image from the sequence of the stir-fry state monitoring key frames to obtain the sequence of the semantic feature graphs of the stir-fry state image. That is, in the technical solution of the present application, the sequence of the stir-fry state monitoring key frames is passed through the stir-fry state image semantic feature extractor based on the convolutional neural network model to obtain the sequence of the semantic feature graphs of the stir-fry state image; the sequence of the stir-fry state monitoring key frames is passed through the stir-fry state image semantic feature extractor based on the convolutional neural network model to obtain the sequence of the semantic feature graphs of the stir-fry state image in order to convert the image information into semantic features to support the subsequent judgment of the timing of turning off the fire and the generation of control instructions. Specifically, through the convolutional neural network model, the visual information in the stir-fry state image can be converted into semantic features. The convolutional neural network can learn the local and global features of the image and extract the key information in the image, such as color, texture, shape, etc. These semantic features can better represent the changes and features of the stir-fry state for subsequent judgment of the timing of turning off the fire. That is, compared with the original image, the semantic feature graph can more accurately capture the features of the stir-fry state, such as color changes, the condition of the mashed garlic, etc. This can improve the accuracy and stability of the judgment of the time to turn off the fire. More specifically, each layer of the stir-fry state image semantic feature extractor based on the convolutional neural network model performs the following operations on the input data in the forward transfer of the layer: convolution processing on the input data to obtain a convolution feature map; pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the stir-fry state image semantic feature extractor based on the convolutional neural network model is a sequence of the stir-fry state image semantic feature maps, and the input of the first layer of the stir-fry state image semantic feature extractor based on the convolutional neural network model is a sequence of the stir-fry state monitoring key frames.
[0041] Convolutional Neural Network (CNN) is a deep learning model that is particularly suitable for processing data with grid structures, such as images and videos. Below is the general structure and step-by-step expansion of the convolutional neural network model: Input layer: Accepts input data, usually images, audio or text, etc.; Convolution layer: The convolution layer is one of the core components of CNN. It extracts local features from the input data by applying a series of filters (also called convolution kernels). The convolution operation multiplies the filter with the input data element by element and sums them to generate a feature map; Activation function: After the convolution layer, a nonlinear activation function such as ReLU is usually applied to introduce nonlinear features; Pooling layer: The pooling layer is used to reduce the spatial size of the feature map and retain the most important features. Common pooling operations include maximum pooling and average pooling; Fully connected layer: The fully connected layer connects the output of the pooling layer to one or more fully connected layers to map the features to the final output category or regression value. Each neuron in the fully connected layer is connected to all neurons in the previous layer; Output layer: The output layer selects appropriate activation functions according to different tasks, such as softmax function for multi-classification tasks and linear activation function for regression tasks; Loss function: Select appropriate loss functions according to different tasks, such as cross entropy loss function for classification tasks and mean square error loss function for regression tasks; Back propagation and optimization: Through the back propagation algorithm and gradient descent optimization algorithm, the gradient of the model parameters is calculated according to the loss function, and the parameters are updated to minimize the loss function.
[0042] More specifically, the S424 determines the prototype representation feature vector of the stir-fry state based on the sequence of the semantic feature graphs of the stir-fry state image. In particular, in a specific example of the present application, the S424 includes: using a prototype feature extraction network to process the sequence of the semantic feature graphs of the stir-fry state image to obtain the prototype representation feature vector of the stir-fry state. That is, the sequence of the semantic feature graphs of the stir-fry state image is processed using a prototype feature extraction network to obtain the prototype representation feature vector of the stir-fry state. Here, the purpose of using the prototype feature extraction network to process the sequence of the semantic feature graphs of the stir-fry state image is to aggregate and summarize the semantic features in the sequence, so as to obtain a more compact and representative feature vector. Specifically, each feature graph in the sequence of the semantic feature graphs of the stir-fry state image contains information about the stir-fry state. Through the prototype feature extraction network, these feature graphs can be aggregated, and multiple feature graphs in the sequence can be merged into one feature vector. This can reduce the dimension of the feature and improve the compactness and representation ability of the feature. At the same time, the prototype feature extraction network can summarize the feature maps in the sequence by learning the commonalities and differences of the stir-frying state to obtain the prototype representation of the stir-frying state. In this way, the feature vector can better represent the key features of the entire stir-frying process and capture the main patterns and changes of the stir-frying state. Specifically, the sequence of the semantic feature maps of the stir-frying state image is processed using the prototype feature extraction network to obtain the prototype representation feature vector of the stir-frying state, including: using the prototype feature extraction network to process the sequence of the semantic feature maps of the stir-frying state image with the following formula to obtain the prototype representation feature vector of the stir-frying state; wherein the prototype formula is:
[0043]
[0044] Among them, F i and F j are the i-th and j-th optimized stir-fry state image semantic feature maps in the sequence of the stir-fry state image semantic feature maps, F k is the sequence of the semantic feature graphs of the stir-fry state image, ‖·‖1 is the norm of the vector, M is the number of graphs in the sequence of the semantic feature graphs of the stir-fry state image - 1, D i is the eigenvalue of each position in the eigenvector representing the prototype of the stir-frying state, N s is the length of the characteristic vector representing the prototype of the stir-frying state, P k is the prototype characteristic vector of the stir-frying state, and exp(·) is an exponential operation.
[0045] It is worth mentioning that in other specific examples of the present application, the prototype representation feature vector of the stir-fry state can also be determined based on the sequence of the semantic feature map of the stir-fry state image in other ways, for example: input the sequence of the semantic feature map of the stir-fry state image; extract the feature matrix from each image through the convolutional neural network model; in order to obtain the prototype representation of the stir-fry state of each feature matrix, perform a spatial average pooling operation on each feature matrix; organize each feature vector into a sequence in order; for the entire sequence, a pooling operation can be applied to obtain a global prototype representation feature vector of the stir-fry state. Common pooling operations include average pooling and maximum pooling, which obtain the prototype representation feature vector of the stir-fry state by pooling the feature vector of each time step in the sequence.
[0046] More specifically, the S425 determines whether to automatically turn off the fire based on the prototype characterization feature vector of the stir-fry state. That is, in the technical solution of the present application, the prototype characterization feature vector of the stir-fry state is passed through a fire-off controller based on a classifier to obtain a control instruction, and the control instruction is used to indicate whether to turn off the fire. Here, the feature vector of the stir-fry state can be classified by the fire-off controller based on the classifier. The classifier can learn the relationship between different feature vectors and the timing of turning off the fire, and divide the feature vector into two categories: turning off the fire and not turning off the fire. In this way, it can be judged whether the current stir-fry state needs to turn off the fire according to the characteristics of the feature vector. The control instruction generated by the controller can indicate whether to turn off the fire. If the classifier determines that the current stir-fry state needs to turn off the fire, the controller will generate a corresponding fire-off instruction to control the closing of the firepower. In this way, the timing of turning off the fire can be automatically controlled to improve production efficiency and product consistency. In a specific example, the prototype characterization feature vector of the stir-fry state is fully connected and encoded using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and the encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.
[0047] That is, in the technical solution of the present application, the labels of the classifier include turning off the fire (first label), and not turning off the fire (second label), wherein the classifier determines which classification label the stir-fry state prototype characterization feature vector belongs to through a soft maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "whether to turn off the fire". It just has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether to turn off the fire is actually converted into a binary class probability distribution that conforms to the laws of nature through the classification label. In essence, the physical meaning of the natural probability distribution of the label is used, rather than the linguistic text meaning of "whether to turn off the fire".
[0048] A classifier is a machine learning model or algorithm that is used to classify input data into different categories or labels. Classifiers are part of supervised learning and perform classification tasks by learning the mapping relationship from input data to output categories.
[0049] A fully connected layer is a common type of layer in neural networks. In a fully connected layer, each neuron is connected to all neurons in the previous layer, and each connection has a weight. This means that each neuron in the fully connected layer receives inputs from all neurons in the previous layer, and performs a weighted sum of these inputs through the weights, and then passes the result to the next layer.
[0050] The Softmax classification function is a commonly used activation function for multi-classification problems. It converts each element of the input vector into a probability value between 0 and 1, and the sum of these probability values is equal to 1. The Softmax function is often used in the output layer of a neural network, especially for multi-classification problems, because it can map the network output to the probability distribution of each category. During the training process, the output of the Softmax function can be used to calculate the loss function and update the network parameters through the back propagation algorithm. It is worth noting that the output of the Softmax function does not change the relative size relationship between the elements, but only normalizes them. Therefore, the Softmax function does not change the characteristics of the input vector, but only converts it into a probability distribution form.
[0051] In particular, in a specific example of the present application, the prefabricated spicy butter sauce production process also includes: a training step for training the stir-fry state image semantic feature extractor based on the convolutional neural network model, the prototype feature extraction network and the classifier-based fire-off controller; wherein the training step includes: acquiring training data, the training data including: training stir-fry state monitoring video; performing key frame sampling on the training stir-fry state monitoring video to obtain a sequence of training stir-fry state monitoring key frames; passing the sequence of training stir-fry state monitoring key frames through the stir-fry state image semantic feature extractor based on the convolutional neural network model to obtain a training stir-fry state monitoring key frame sequence; A sequence of semantic feature maps of stir-fry state images; using a prototype feature extraction network to process the sequence of training semantic feature maps of stir-fry state images to obtain a training prototype representation feature vector of stir-fry state; calculating a predetermined loss function value for the sequence of training semantic feature maps of stir-fry state images; passing the training prototype representation feature vector of stir-fry state through a classifier-based fire-off controller to obtain a classification loss function value; using the weighted sum of the classification loss function value and the predetermined loss function value as the loss function value to train the stir-fry state image semantic feature extractor based on the convolutional neural network model, the prototype feature extraction network and the classifier-based fire-off controller.
[0052] In the above technical scheme, for the sequence of training stir-fry state image semantic feature maps obtained by passing the sequence of training stir-fry state monitoring key frames through a stir-fry state image semantic feature extractor based on a convolutional neural network model, since the training stir-fry state image semantic feature maps express the image semantic features of the training stir-fry state monitoring key frames, considering the semantic differences of source images between frames of the training stir-fry state monitoring key frames and the differences in the expression of channel image semantic features of the convolutional neural network model, when the sequence of stir-fry state image semantic feature maps is processed in units of the overall image semantic distribution of the feature matrix using a prototype feature extraction network, it is expected to enhance the accuracy of the classification results by targeting the complex feature distribution expression characteristics of the sequence of training stir-fry state image semantic feature maps while maintaining the complexity of the feature matrix image semantic expression of the sequence of training stir-fry state image semantic feature maps in the channel dimension.
[0053] Based on this, the present application further introduces a loss function for promoting classification understanding by expressing complex features of the sequence of semantic feature maps of the training stir-fry state images, that is, calculating a predetermined loss function value for the sequence of semantic feature maps of the training stir-fry state images, including: first expanding the sequence of semantic feature maps of the training stir-fry state images and cascading them into a training stir-fry state image semantic joint feature vector, calculating the mean and variance between the eigenvalues at any two positions of the training stir-fry state image semantic joint feature vector to obtain a mean weight matrix and a variance weight matrix, then performing query-style multiplication of the training stir-fry state image semantic joint feature vector with the mean weight matrix and the variance weight matrix respectively to obtain a mean feature vector and a variance feature vector, then calculating the Frobenius norm of the autocorrelation matrix of the obtained mean feature vector and variance feature vector, and subtracting the product of the Frobenius norm of the autocorrelation matrix of the training stir-fry state image semantic joint feature vector and the weight as a hyperparameter to obtain the predetermined loss function value.
[0054] For example, represented as:
[0055]
[0056] Wherein, V is the semantic joint feature vector of the training stir-fry state image, M μ is the mean weight matrix, M σ is the variance weight matrix, represents matrix multiplication operation, α is a predetermined weight hyperparameter, T represents transpose, ‖·‖ F represents the Frobenius norm of the matrix, and Loss represents the predetermined loss function value;
[0057] And as mentioned above:
[0058]
[0059] Among them, v i represents the i-th eigenvalue of the semantic joint feature vector of the training stir-fry state image, v j represents the jth eigenvalue of the semantic joint feature vector of the training stir-fry state image, M μ (i, j) represents the mean of the i-th eigenvalue and the j-th eigenvalue of the semantic joint feature vector of the training stir-fry state image, M σ (i, j) represents half of the absolute value of the difference between the i-th eigenvalue and the j-th eigenvalue of the semantic joint feature vector of the training stir-frying state image.
[0060] That is, the predetermined loss function value, on the basis of determining the overall distributed structural bottleneck of the training stir-fry state image semantic joint feature vector through the multi-dimensional interaction of the detail group of the training stir-fry state image semantic joint feature vector, simulates the correlation between the structural details of the training stir-fry state image semantic joint feature vector and the macro-structured feature representation behavior through the low-rank structured reasoning of the training stir-fry state image semantic joint feature vector. In this way, by training the classifier with the loss function, the structural detail dependency relative to the macro-complexity can be modeled in the classification scenario, thereby promoting the understanding of the macro-complex feature expression based on the structural details of the sequence of the training stir-fry state image semantic feature graphs, and improving the accuracy of the classification results.
[0061] In particular, in S5, the spicy slurry is cooled and then placed in a sealed container to obtain a prefabricated spicy butter sauce. It should be understood that after the spicy slurry is placed in a sealed container, its freshness and taste can be effectively maintained. The sealed container can prevent the entry of air and moisture, slow down the oxidation and corruption process of the food, and extend the shelf life of the sauce.
[0062] In summary, the production process of the prefabricated spicy butter sauce according to the embodiment of the present application is explained, and an intelligent control scheme is used to construct an intelligent fire-off control scheme to improve the product consistency and stability of the prefabricated spicy butter sauce. Specifically, by real-time monitoring of the stir-frying process of the spicy butter sauce and extracting relevant features, it is determined whether the current stir-frying state needs to be turned off according to different features, thereby realizing adaptive fire control. In this way, the timing of turning off the fire can be automatically controlled to improve production efficiency and product consistency.
[0063] Furthermore, a device for producing a prefabricated spicy butter sauce is also provided.
[0064] Figure 5 FIG. 1 is a block diagram of a device for producing a prefabricated spicy butter sauce according to an embodiment of the present application. Figure 5 As shown, according to the prefabricated spicy beef sauce production device 300 of the embodiment of the present application, it includes: a monitoring video acquisition module 310, which is used to obtain the stir-frying state monitoring video collected by the camera; a sampling module 320, which is used to perform key frame sampling on the stir-frying state monitoring video to obtain a sequence of stir-frying state monitoring key frames; an image semantic feature extraction module 330, which is used to extract stir-frying state image semantic features from the sequence of stir-frying state monitoring key frames to obtain a sequence of stir-frying state image semantic feature maps; a stir-frying state prototype representation feature extraction module 340, which is used to determine the stir-frying state prototype representation feature vector based on the sequence of stir-frying state image semantic feature maps; and a control result generation module 350, which is used to determine whether to automatically turn off the fire based on the stir-frying state prototype representation feature vector.
[0065] As described above, the prefabricated spicy butter sauce production device 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a prefabricated spicy butter sauce production algorithm. In a possible implementation, the prefabricated spicy butter sauce production device 300 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 prefabricated spicy butter sauce production device 300 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 prefabricated spicy butter sauce production device 300 can also be one of the many hardware modules of the wireless terminal.
[0066] Alternatively, in another example, the prefabricated spicy beef sauce production device system 300 and the wireless terminal may be separate devices, and the prefabricated spicy beef sauce production device 300 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.
[0067] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.
Claims
1. A production process for prefabricated spicy butter sauce, characterized in that: include: Mix and grind the spices in proportion to obtain fine spice powder; Process the garlic into garlic paste; After heating the butter to a predetermined temperature, pour the heated butter into a wok, add appropriate amounts of salt and sugar, stir-fry evenly, then add the spice powder, and continue to stir-fry; Pour the mashed garlic into the wok, stir-fry until the mashed garlic changes color, and turn off the heat to obtain the spicy butter sauce; The spicy slurry is cooled and then put into a sealed container to obtain a pre-made spicy butter sauce.
2. The process for producing the prefabricated spicy butter sauce according to claim 1, characterized in that: The spices include dried chili peppers, prickly ash, star anise, cinnamon bark, bay leaves, tsaoko and coriander seeds.
3. The process for producing the prefabricated spicy butter sauce according to claim 2, characterized in that: The predetermined temperature is 180°C.
4. The process for producing the prefabricated spicy butter sauce according to claim 3, characterized in that: include: The garlic paste is poured into the wok, and stir-fried until the garlic paste changes color, and then the heat is turned off to obtain a spicy butter sauce, comprising: deploying a camera toward the wok; as well as Use machine vision to determine whether to automatically turn off the fire.
5. The process for producing the prefabricated spicy butter sauce according to claim 4, characterized in that: The method of judging whether to automatically turn off the fire by machine vision includes: Acquire a monitoring video of the stir-frying state collected by the camera; Performing key frame sampling on the stir-frying state monitoring video to obtain a sequence of stir-frying state monitoring key frames; Extracting the semantic features of the stir-frying state images from the sequence of the stir-frying state monitoring key frames to obtain a sequence of the semantic feature graphs of the stir-frying state images; Determining a prototype representation feature vector of the stir-fry state based on the sequence of the semantic feature graphs of the stir-fry state image; and Based on the prototype characterization feature vector of the stir-frying state, it is determined whether to automatically turn off the fire.
6. The process for producing prefabricated spicy butter sauce according to claim 5, characterized in that: The step of extracting the semantic features of the stir-frying state images from the sequence of the stir-frying state monitoring key frames to obtain a sequence of the semantic feature graphs of the stir-frying state images comprises: The sequence of the stir-frying state monitoring key frames is passed through a stir-frying state image semantic feature extractor based on a convolutional neural network model to obtain a sequence of the stir-frying state image semantic feature maps.
7. The process for producing prefabricated spicy butter sauce according to claim 6, characterized in that: The step of determining the prototype representation feature vector of the stir-fry state based on the sequence of the semantic feature graphs of the stir-fry state image comprises: Using the prototype feature extraction network to process the sequence of the stir-fry state image semantic feature graphs using the following formula to obtain the stir-fry state prototype representation feature vector; Among them, the prototype formula is: Among them, F i and F j are the i-th and j-th optimized stir-fry state image semantic feature maps in the sequence of the stir-fry state image semantic feature maps, F k is the sequence of the semantic feature graphs of the stir-fry state image, ‖·‖1 is the norm of the vector, M is the number of graphs in the sequence of the semantic feature graphs of the stir-fry state image - 1, D i is the eigenvalue of each position in the eigenvector representing the prototype of the stir-frying state, N s is the length of the characteristic vector representing the prototype of the stir-frying state, P k is the prototype characteristic vector of the stir-frying state, and exp(·) is an exponential operation.
8. The process for producing prefabricated spicy butter sauce according to claim 7, characterized in that: The determining whether to automatically turn off the fire based on the prototype characterization feature vector of the stir-frying state includes: passing the prototype characterization feature vector of the stir-frying state through a fire-off controller based on a classifier to obtain a control instruction, wherein the control instruction is used to indicate whether to turn off the fire.
9. The process for producing prefabricated spicy butter sauce according to claim 8, characterized in that: Also includes: A training step for training the stir-fry state image semantic feature extractor based on the convolutional neural network model, the prototype feature extraction network and the classifier-based fire-off controller; Wherein, the training step includes: Acquire training data, the training data comprising: a monitoring video of a training stir-frying state; Performing key frame sampling on the training stir-frying state monitoring video to obtain a sequence of training stir-frying state monitoring key frames; The sequence of training stir-frying state monitoring key frames is passed through a stir-frying state image semantic feature extractor based on a convolutional neural network model to obtain a sequence of training stir-frying state image semantic feature maps; Using a prototype feature extraction network to process the sequence of semantic feature maps of the training stir-fry state images to obtain a prototype representation feature vector of the training stir-fry state; Calculating a predetermined loss function value for the sequence of semantic feature maps of the training stir-frying state images; Passing the training stir-frying state prototype characterization feature vector through a classifier-based fire-off controller to obtain a classification loss function value; The weighted sum of the classification loss function value and the predetermined loss function value is used as the loss function value to train the stir-frying state image semantic feature extractor based on the convolutional neural network model, the prototype feature extraction network and the classifier-based fire-off controller.
10. A device for producing prefabricated spicy butter sauce, characterized in that: include: A monitoring video acquisition module, used to acquire the stir-frying status monitoring video collected by the camera; A sampling module, used for performing key frame sampling on the stir-frying state monitoring video to obtain a sequence of stir-frying state monitoring key frames; An image semantic feature extraction module, used for extracting the semantic features of the stir-frying state image from the sequence of the stir-frying state monitoring key frames to obtain a sequence of the stir-frying state image semantic feature graphs; a stir-fry state prototype representation feature extraction module, used to determine a stir-fry state prototype representation feature vector based on a sequence of the stir-fry state image semantic feature graphs; and The control result generation module is used to determine whether to automatically turn off the fire based on the stir-fry state prototype characterization feature vector.
Citation Information
Patent Citations
Low-salt hot pot seasoning and preparation process thereof
CN109329853A
Medium-spicy beef tallow hotpot condiment
CN113907312A
Garlic sauce and processing method thereof
CN115944071A
