A cooking equipment temperature control system based on machine vision

Abstract

The present invention relates to the field of heat control technology, specifically a heat control system for cooking equipment based on machine vision, including an image acquisition unit, an image partitioning unit, a heat control unit, a food recognition unit, a first degree of doneness recognition unit and a second degree of doneness recognition unit, and also includes: an image acquisition unit, the image acquisition unit is used to acquire images of food in a pot according to a preset sampling period based on an image acquisition device integrated in the cooking equipment to obtain a food cooking image, and transmit the food cooking image to the image partitioning unit. The present invention can realize accurate heat adjustment of food in each heating zone by being able to acquire images of food in the pot in real time, and based on image partitioning of different heating zones and identification of food type and degree of doneness, this refined control method can effectively avoid overcooking or undercooking of food during cooking.

Description

Technical Field

[0001] The invention relates to the technical field of temperature control, and in particular to a temperature control system for cooking equipment based on machine vision. Background Art

[0002] Cooking equipment refers to various tools and mechanical equipment used to prepare, cook, heat and process food. These equipment can help improve cooking efficiency, ensure the taste and nutrition of food, and make the cooking process more convenient.

[0003] Traditional systems usually rely on manual experience or simple timing control and lack the ability to monitor the status of food in real time. Cooks can only judge the heat by visual inspection, smelling, etc., which can easily result in food being overcooked, undercooked, or unevenly heated. Traditional systems rely on manual operation to adjust the heat, requiring cooks to constantly observe and adjust the heat throughout the process, which increases the difficulty of cooking, especially for people with insufficient cooking experience, who are prone to operating errors. Traditional pots usually lack fine heat zone control, which may result in uneven heating, with some food being overcooked or burnt while other parts may still be raw. This uneven heating will affect the overall taste and texture of the food. Traditional systems cannot recognize different types of food, nor can they make corresponding heat adjustments based on the characteristics of different ingredients. Cooks need to decide the heating time and temperature of different ingredients based on experience and manual settings, which may result in unsatisfactory cooking results. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the disadvantages of the above-mentioned prior art and provide a cooking equipment temperature control system based on machine vision.

[0005] The technical solution adopted to solve the above technical problems is: a cooking equipment temperature control system based on machine vision, including a temperature control unit, a food recognition unit, a first degree of doneness recognition unit and a second degree of doneness recognition unit, and also including:

[0006] An image acquisition unit, the image acquisition unit being used to acquire images of food in the cookware according to a preset sampling period based on an image acquisition device integrated in the cooking device, so as to obtain a food cooking image, and transmit the food cooking image to the image partitioning unit;

[0007] An image partitioning unit, wherein the image partitioning unit is used to receive the food cooking image transmitted by the image acquisition unit, and partition the food cooking image based on preset images of different heating zones of the cookware to obtain a plurality of food cooking sub-images, wherein the food cooking sub-image is a food cooking image within the heating zone, establish a partition mapping table between the plurality of food cooking sub-images and the different heating zones of the cookware, transmit the partition mapping table to the heat control unit, and transmit the plurality of food cooking sub-images to the food recognition unit.

[0008] Preferably, the food recognition unit is used to receive the multiple food cooking sub-images transmitted by the image partitioning unit, and perform food recognition on the food cooking sub-images based on a pre-trained food recognition model to obtain food type labels corresponding to the food cooking sub-images, wherein the food type labels include meat labels and plant labels, and the food cooking sub-images whose food type labels are meat labels are transmitted to the first doneness recognition unit, and the food cooking sub-images whose food type labels are plant labels are transmitted to the second doneness recognition unit;

[0009] The first doneness recognition unit is used to receive the food cooking sub-image with the food type label being a meat label transmitted by the food recognition unit, and perform feature extraction on the food cooking sub-image with the food type label being a meat label to obtain image features, wherein the image features include color features, shape features and texture features, and perform a first doneness recognition on the food cooking sub-image with the food type label being a meat label through the image features based on a pre-trained first doneness recognition model to obtain a first doneness, and transmit the first doneness to the heat control unit.

[0010] Preferably, the second doneness recognition unit is used to receive the food cooking sub-image whose food type label is a plant label transmitted by the food recognition unit, and perform a second doneness recognition on the food cooking sub-image whose food type label is a plant label based on a pre-trained second doneness recognition model to obtain a second doneness, and transmit the second doneness to the heat control unit.

[0011] Preferably, the heat control unit is used to receive the partition mapping table transmitted by the image partitioning unit, the first doneness transmitted by the first doneness recognition unit and the second doneness transmitted by the second doneness recognition unit, and obtain the doneness corresponding to the multiple food cooking sub-images based on the first doneness and the second doneness, perform partition mapping on the multiple food cooking sub-images and the partition mapping table to obtain the heating zones corresponding to the multiple food cooking sub-images, obtain the doneness of the food in the corresponding heating zone based on the doneness corresponding to the multiple food cooking sub-images, input the doneness of the food in the heating zone into a pre-trained heat control model, output the heating power of the heating zone based on the heat control model, and perform heat control on the heating zone based on the heating power.

[0012] Preferably, the food cooking image is partitioned based on preset images of different heating zones of the cookware to obtain a plurality of food cooking sub-images, including:

[0013] Extracting corner points of the images of different heating zones of the cookware and the food cooking image based on the Harris corner point detection algorithm to obtain a first corner point set corresponding to the images of different heating zones of the cookware and a second corner point set corresponding to the food cooking image;

[0014] Calculating the distance between the first corner point in the first corner point set and the second corner point in the second corner point set based on the Euclidean distance, and if the distance is less than a preset distance threshold, taking the first corner point and the second corner point as a matching point pair, repeating the above operation until the first corner point set and the second corner point set are processed to obtain a matching point pair set;

[0015] Performing image registration on the images of different heating zones of the cookware and the food cooking image based on the set of matching point pairs using affine transformation to obtain a transformation matrix;

[0016] Different heating zone areas in the images of different heating zones of the cookware are mapped to the food cooking image according to the transformation matrix to obtain a plurality of food cooking sub-images.

[0017] Preferably, the food recognition model includes a feature extraction module and a feature fusion module, the feature extraction module includes a three-level feature extraction network, the three-level feature extraction network extracts the image features of the food cooking sub-image through a convolutional neural network, and then outputs the category and the probability value of each category, wherein the convolutional neural network adopts a pyramid structure, and the feature fusion module is used to use a 1×1 pixel size convolution kernel to perform feature fusion and pooling on the image features extracted by the three-level feature extraction network, and then use two 3×3 pixel size convolution kernels to perform convolution operation and activation function activation, and finally perform pooling to obtain 512 7×7 pixel feature maps, and classify the 512 7×7 pixel feature maps based on the fully connected layer to obtain the food type label corresponding to the food cooking sub-image.

[0018] Preferably, the categories and the probability values ​​of each category are expressed as follows:

[0019] ;

[0020] in, and Indicates The input and weight parameters of the convolutional neural network. Represents the convolution, pooling and activation operations performed during the convolution process. represents the feature extraction module, Indicates The output probability of each category of the convolutional neural network is Indicates The output category label of each category of the convolutional neural network.

[0021] Preferably, the first doneness recognition model adopts an improved support vector machine, and the support vector machine is optimized based on a swarm particle optimization algorithm.

[0022] Preferably, the support vector machine is optimized based on a swarm particle optimization algorithm, comprising:

[0023] Randomly generate the initial population and set the population size , initialize the fitness of each particle, where the fitness is the mean square error loss of the first maturity recognition model, and set the maximum number of iterations ;

[0024] Calculate each particle The fitness value of the particle is determined and the best particle in the current population is identified. , and its fitness is , calculate the other particles in the population and the optimal particle The distance between them is calculated as follows:

[0025] ;

[0026] in, Indicates the population The distance between each particle and the optimal particle;

[0027] The position is updated according to the position of the particle and the distance between the particles, wherein the position update formula is as follows:

[0028] ;

[0029] in, and Indicates that the coordinates in the population are The particle Second and The position in the d-dimensional solution space in the iteration, represents the first random number, represents the second random number, Indicates The optimal position of the current particle in the iteration, and represents the division coefficient;

[0030] Calculate the fitness of the new position. If the fitness of the new position is better than that of the old position, update the position and fitness of the particle. Repeat the above operation until the maximum number of iterations is reached. Or fitness convergence to obtain the particle with the best fitness in the current population and its fitness.

[0031] Preferably, the second maturity recognition model adopts a ResNet50 model.

[0032] The beneficial effects of the present invention are as follows: (1) The present invention can collect images of food in the cooker in real time, and based on the image partitioning of different heating zones and the identification of the type and degree of doneness of the food, it can achieve precise heat adjustment of the food in each heating zone. This refined control method can effectively avoid the situation where the food is overcooked or undercooked during the cooking process. Through the machine learning food recognition model, different types of food (such as meat and plants) are automatically distinguished, and different degree of doneness recognition models are used based on the food type. In this way, more reasonable heat control can be performed according to the characteristics of different ingredients. (2) The present invention finely divides the different heating zones in the cooker through the image partitioning unit. The system can adjust the temperature of the food according to the actual conditions of each heating zone. The system adjusts the heat according to the situation to ensure that the food in each area is cooked at the appropriate heat, thereby avoiding overheating or uneven heating. By using image processing technology and machine learning, the system can automatically identify the cooking status of the food and adjust the heating power in real time, reducing manual intervention and improving cooking efficiency and accuracy. (3) The present invention uses different types of ingredients to have different requirements for heat. The system can adapt to the different cooking requirements of meat and plant foods through different doneness recognition units and trained models, optimize the cooking effect, and improve the taste and quality of the food. By accurately controlling the heat of the heating zone, overheating of the food is avoided, energy is saved, waste is reduced, and the energy efficiency of the cooking equipment is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of the system architecture of an overall system in an embodiment of the present invention.

[0034] Figure numerals: 1. Image acquisition unit; 2. Image partitioning unit; 3. Heat control unit; 4. Food recognition unit; 5. First degree of doneness recognition unit; 6. Second degree of doneness recognition unit. DETAILED DESCRIPTION

[0035] Embodiment 1, as Figure 1 As shown, a cooking equipment temperature control system based on machine vision proposed by the present invention includes a temperature control unit 3, a food recognition unit 4, a first degree of doneness recognition unit 5 and a second degree of doneness recognition unit 6, and also includes:

[0036] An image acquisition unit 1, the image acquisition unit 1 is used to acquire images of food in the pot according to a preset sampling period based on an image acquisition device integrated in the cooking device to obtain a food cooking image, and transmit the food cooking image to the image partitioning unit 2;

[0037] The image partitioning unit 2 is used to receive the food cooking image transmitted by the image acquisition unit 1, and partition the food cooking image based on the preset images of different heating zones of the cookware to obtain multiple food cooking sub-images, wherein the food cooking sub-image is the food cooking image within the heating zone, establish a partition mapping table between the multiple food cooking sub-images and the different heating zones of the cookware, transmit the partition mapping table to the heat control unit 3, and transmit the multiple food cooking sub-images to the food recognition unit 4.

[0038] In the present invention, the image acquisition device is a device for taking images during the cooking process, usually a camera or other image sensor, which is installed on the cooking device and can capture the image of food in the pot in real time; the sampling period refers to the time interval at which the image acquisition device acquires images of food at a predetermined time interval during the cooking process, for example, an image of food is acquired every 5 seconds to capture changes in food; the partition mapping table records the relationship between different heating zones of the pot and the food cooking sub-images, for example, the food image in the first zone corresponds to the left heating zone of the pot, the second zone corresponds to the right heating zone, etc. The table is used to guide subsequent image processing and analysis to ensure that each sub-image corresponds to a specific heating zone.

[0039] In an optional embodiment, the food recognition unit 4 is used to receive multiple food cooking sub-images transmitted by the image partitioning unit 2, and perform food recognition on the food cooking sub-images based on a pre-trained food recognition model to obtain food type labels corresponding to the food cooking sub-images, wherein the food type labels include meat labels and plant labels, and the food cooking sub-images with the food type labels as meat labels are transmitted to the first doneness recognition unit 5, and the food cooking sub-images with the food type labels as plant labels are transmitted to the second doneness recognition unit 6.

[0040] In an optional embodiment, the first doneness recognition unit 5 is used to receive the food cooking sub-image with the food type label of meat transmitted by the food recognition unit 4, and perform feature extraction on the food cooking sub-image with the food type label of meat to obtain image features, wherein the image features include color features, shape features and texture features, and perform a first doneness recognition on the food cooking sub-image with the food type label of meat through image features based on a pre-trained first doneness recognition model to obtain a first doneness, and transmit the first doneness to the heat control unit 3.

[0041] It should be noted that color features describe the characteristics of color distribution in an image, such as the color average, color contrast, etc.; shape features describe the characteristics of the shape of an object in an image, such as edges, contours, and area; texture features describe the characteristics of the image surface texture, such as the roughness and texture direction in the image.

[0042] In an optional embodiment, the second doneness recognition unit 6 is used to receive the food cooking sub-image whose food type label is a plant label transmitted by the food recognition unit 4, and perform a second doneness recognition on the food cooking sub-image whose food type label is a plant label based on a pre-trained second doneness recognition model to obtain a second doneness, and transmit the second doneness to the heat control unit 3.

[0043] In an optional embodiment, the temperature control unit 3 is used to receive the partition mapping table transmitted by the image partitioning unit 2, the first degree of doneness transmitted by the first degree of doneness recognition unit 5 and the second degree of doneness transmitted by the second degree of doneness recognition unit 6, and obtain the doneness corresponding to the multiple food cooking sub-images based on the first degree of doneness and the second degree of doneness, partition-map the multiple food cooking sub-images with the partition mapping table to obtain the heating zones corresponding to the multiple food cooking sub-images, obtain the doneness of the food in the corresponding heating zone based on the doneness corresponding to the multiple food cooking sub-images, input the doneness of the food in the heating zone into a pre-trained temperature control model, output the heating power of the heating zone based on the temperature control model, and perform temperature control on the heating zone based on the heating power.

[0044] It should be noted that the heat control model is an intelligent model trained based on machine learning or deep learning. It can predict and control the power output of the heating equipment according to the doneness of food in different heating zones. The heat control model takes into account the current doneness of the food, the state of the heating zone and other factors (such as temperature, time, etc.), so as to optimize the heating process and avoid overcooking or undercooking of food. The heat control model can adopt a BP neural network, and by learning the nonlinear relationship between the doneness of food and the state of the heating zone, it can achieve efficient and accurate heat control. By inputting multiple features (such as doneness, temperature, time, etc.), the neural network outputs the appropriate heating power or food doneness prediction, thereby optimizing the cooking process and achieving precise temperature control and heating power adjustment.

[0045] Embodiment 2, a cooking equipment temperature control system based on machine vision proposed by the present invention, compared with embodiment 1, this embodiment further includes: partitioning the food cooking image based on preset images of different heating zones of the pot to obtain multiple food cooking sub-images, including:

[0046] Corner point extraction is performed on the images of different heating zones of the pot and the food cooking image based on the Harris corner point detection algorithm to obtain a first corner point set corresponding to the images of different heating zones of the pot and a second corner point set corresponding to the food cooking image;

[0047] The distance between the first corner point in the first corner point set and the second corner point in the second corner point set is calculated based on the Euclidean distance. If the distance is less than a preset distance threshold, the first corner point and the second corner point are taken as a matching point pair. The above operation is repeated until the first corner point set and the second corner point set are processed to obtain a matching point pair set.

[0048] Using affine transformation to register the images of different heating zones of the pot and the food cooking images based on the matching point pair set, so as to obtain a transformation matrix;

[0049] Different heating zone areas in the images of different heating zones of the cookware are mapped to the food cooking image according to the transformation matrix to obtain a plurality of food cooking sub-images.

[0050] In this embodiment, the Harris corner detection algorithm is an image processing technology, which is often used for feature extraction of images. The goal of the algorithm is to detect points with significant features from the image, which are called corner points. Corner points are usually high-contrast areas in the image, which can provide rich local information for subsequent image matching and registration; Euclidean distance is a common method for calculating the straight-line distance between two points. In image processing, Euclidean distance can be used to measure the similarity between image feature points; affine transformation is an image transformation technology that can be used to rotate, translate, scale and shear an image without changing the parallel line characteristics of the image. Affine transformation is represented by a transformation matrix, which can map the position of a point in the image to another position. Through affine transformation, the features of the pot image and the food image can be aligned to provide the correct spatial relationship for subsequent analysis; the transformation matrix is ​​a matrix used to describe the spatial coordinate transformation in the affine transformation. It contains the parameters required for the image to perform operations such as translation, rotation, and scaling. By matching the point pair set, the affine transformation algorithm can calculate a transformation matrix, which can map the corner points in the pot image to the corner points in the food image, thereby realizing image registration.

[0051] In an optional embodiment, the food recognition model includes a feature extraction module and a feature fusion module, the feature extraction module includes a three-level feature extraction network, the three-level feature extraction network extracts image features of a food cooking sub-image through a convolutional neural network, and then outputs categories and probability values ​​of each category, wherein the convolutional neural network adopts a pyramid structure, and the feature fusion module is used to use a 1×1 pixel size convolution kernel to perform feature fusion and pooling on the image features extracted by the three-level feature extraction network, and then use two 3×3 pixel size convolution kernels to perform convolution operations and activation function activation, and finally perform pooling to obtain 512 7×7 pixel feature maps, and classify the 512 7×7 pixel feature maps based on the fully connected layer to obtain the food type label corresponding to the food cooking sub-image.

[0052] It should be noted that the three-level feature extraction network refers to a feature extraction structure composed of three layers of networks. Usually, each layer of the network will extract features at different levels. The first layer may extract some simple low-level features (such as edges and textures), while the second and third layers will extract more complex high-level features. Convolutional neural network (CNN) is a deep learning model commonly used in image processing. It gradually extracts features from images through structures such as convolutional layers, pooling layers, and fully connected layers. It is particularly suitable for tasks such as image classification, object detection, and image recognition. A pyramid structure usually refers to a hierarchical structure in which the size of the feature map of each layer gradually decreases and the degree of abstraction gradually increases. The pyramid structure is often used in convolutional neural networks to capture feature information of different scales, which is suitable for processing objects of different scales and sizes.

[0053] In an optional embodiment, the expressions of the categories and the probability values ​​of each category are as follows:

[0054] ;

[0055] in, and Indicates The input and weight parameters of the convolutional neural network. Represents the convolution, pooling and activation operations performed during the convolution process. represents the feature extraction module, Indicates The output probability of each category of the convolutional neural network is Indicates The output category label of each category of the convolutional neural network.

[0056] In an optional embodiment, the first doneness recognition model adopts an improved support vector machine, and the support vector machine is optimized based on a swarm particle optimization algorithm.

[0057] It should be noted that Support Vector Machine (SVM) is a commonly used supervised learning model for classification and regression analysis. The main idea of ​​SVM is to find an optimal hyperplane (a straight line in two-dimensional space and a plane in three-dimensional space) to separate data points of different categories. The goal of SVM is to maximize the classification interval, that is, to find a hyperplane that maximizes the distance from the support vector (the point closest to the hyperplane) to the hyperplane; Particle Swarm Optimization (PSO) is a global optimization algorithm that simulates the foraging behavior of bird flocks in nature. PSO searches for the optimal solution in the solution space through a group of "particles". Each particle represents a potential solution, and the particle adjusts its position based on its own experience and the experience of other particles in the group.

[0058] In an optional embodiment, the support vector machine is optimized based on a swarm particle optimization algorithm, including:

[0059] Randomly generate the initial population and set the population size , initialize the fitness of each particle, where the fitness is the mean square error loss of the first maturity recognition model, and set the maximum number of iterations ;

[0060] Calculate each particle The fitness value of the particle is determined and the best particle in the current population is identified. , and its fitness is , calculate the other particles in the population and the optimal particle The distance between them is calculated as follows:

[0061] ;

[0062] in, Indicates the population The distance between each particle and the optimal particle;

[0063] The position is updated according to the position of the particle and the distance between the particles, where the position update formula is as follows:

[0064] ;

[0065] in, and Indicates that the coordinates in the population are The particle Second and The position in the d-dimensional solution space in the iteration, represents the first random number, represents the second random number, Indicates The optimal position of the current particle in the iteration, and represents the division coefficient;

[0066] Calculate the fitness of the new position. If the fitness of the new position is better than that of the old position, update the position and fitness of the particle. Repeat the above operation until the maximum number of iterations is reached. Or fitness convergence to obtain the particle with the best fitness in the current population and its fitness.

[0067] It should be noted that the population refers to the collection of all particles in the particle swarm optimization algorithm. In the PSO algorithm, the population consists of multiple "particles", each particle represents a combination of the penalty coefficient and kernel function of a support vector machine; fitness is an indicator to measure the quality of the particle solution. In machine learning tasks, fitness is usually related to the performance of the model (such as error); mean square error loss (MSE Loss) is a way to measure the gap between the model prediction and the true value.

[0068] In an optional embodiment, the second maturity recognition model adopts the ResNet50 model.

[0069] It should be noted that ResNet (Residual Network) is a deep neural network architecture that solves the "vanishing gradient" problem that deep neural networks may encounter during training. ResNet skips some layers by introducing residual connections, so that information can be passed directly from one layer to another, thereby greatly reducing the difficulty of training deep networks; 50 means that there are 50 layers in the ResNet model (i.e. 50 convolutional layers, fully connected layers, etc.). It is a deeper network that can learn more complex data features.

[0070] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A cooking equipment temperature control system based on machine vision, comprising a temperature control unit (3), a food recognition unit (4), a first degree of doneness recognition unit (5) and a second degree of doneness recognition unit (6), characterized in that: An image acquisition unit (1), the image acquisition unit (1) being used to acquire images of food in the cookware according to a preset sampling period based on an image acquisition device integrated in the cooking device, so as to obtain a food cooking image, and transmit the food cooking image to the image partitioning unit (2); An image partitioning unit (2), the image partitioning unit (2) being used to receive the food cooking image transmitted by the image acquisition unit (1), and to partition the food cooking image based on preset images of different heating zones of the cookware, so as to obtain a plurality of food cooking sub-images, wherein the food cooking sub-images are food cooking images within the heating zones, to establish a partition mapping table between the plurality of food cooking sub-images and the different heating zones of the cookware, to transmit the partition mapping table to the heat control unit (3), and to transmit the plurality of food cooking sub-images to the food recognition unit (4); The food cooking image is partitioned based on preset images of different heating zones of the cookware to obtain a plurality of food cooking sub-images, including: Extracting corner points of the images of different heating zones of the cookware and the food cooking image based on the Harris corner point detection algorithm to obtain a first corner point set corresponding to the images of different heating zones of the cookware and a second corner point set corresponding to the food cooking image; Calculating the distance between the first corner point in the first corner point set and the second corner point in the second corner point set based on the Euclidean distance, and if the distance is less than a preset distance threshold, taking the first corner point and the second corner point as a matching point pair, repeating the above operation until the first corner point set and the second corner point set are processed to obtain a matching point pair set; Performing image registration on the images of different heating zones of the cookware and the food cooking image based on the set of matching point pairs using affine transformation to obtain a transformation matrix; Different heating zone areas in the images of different heating zones of the cookware are mapped to the food cooking image according to the transformation matrix to obtain a plurality of food cooking sub-images.

2. The cooking equipment temperature control system based on machine vision according to claim 1, characterized in that: The food recognition unit (4) is used to receive the multiple food cooking sub-images transmitted by the image partitioning unit (2), and perform food recognition on the food cooking sub-images based on a pre-trained food recognition model to obtain food type labels corresponding to the food cooking sub-images, wherein the food type labels include meat labels and plant labels, and transmit the food cooking sub-images whose food type labels are meat labels to the first doneness recognition unit (5), and transmit the food cooking sub-images whose food type labels are plant labels to the second doneness recognition unit (6); The first doneness recognition unit (5) is used to receive the food cooking sub-image with the food type label being a meat label transmitted by the food recognition unit (4), and perform feature extraction on the food cooking sub-image with the food type label being a meat label to obtain image features, wherein the image features include color features, shape features and texture features, and perform first doneness recognition on the food cooking sub-image with the food type label being a meat label through the image features based on a pre-trained first doneness recognition model to obtain a first doneness, and transmit the first doneness to the heat control unit (3).

3. The cooking equipment temperature control system based on machine vision according to claim 2, characterized in that: The second doneness recognition unit (6) is used to receive the food cooking sub-image whose food type label is a plant-type label transmitted by the food recognition unit (4), and perform a second doneness recognition on the food cooking sub-image whose food type label is a plant-type label based on a pre-trained second doneness recognition model to obtain a second doneness, and transmit the second doneness to the heat control unit (3).

4. The cooking equipment temperature control system based on machine vision according to claim 3, characterized in that: The temperature control unit (3) is used to receive the partition mapping table transmitted by the image partition unit (2), the first degree of doneness transmitted by the first degree of doneness recognition unit (5), and the second degree of doneness transmitted by the second degree of doneness recognition unit (6), and obtain the degree of doneness corresponding to the multiple food cooking sub-images based on the first degree of doneness and the second degree of doneness, perform partition mapping on the multiple food cooking sub-images and the partition mapping table to obtain the heating zones corresponding to the multiple food cooking sub-images, obtain the degree of doneness of food in the corresponding heating zones based on the degree of doneness corresponding to the multiple food cooking sub-images, input the degree of doneness of food in the heating zones into a pre-trained temperature control model, output the heating power of the heating zones based on the temperature control model, and perform temperature control on the heating zones based on the heating power.

5. The cooking equipment temperature control system based on machine vision according to claim 4, characterized in that: The food recognition model includes a feature extraction module and a feature fusion module. The feature extraction module includes a three-level feature extraction network. The three-level feature extraction network extracts image features of the food cooking sub-image through a convolutional neural network, and then outputs categories and probability values ​​of each category. The convolutional neural network adopts a pyramid structure. The feature fusion module is used to use a 1×1 pixel size convolution kernel to perform feature fusion and pooling on the image features extracted by the three-level feature extraction network, and then use two 3×3 pixel size convolution kernels to perform convolution operations and activation function activation, and finally perform pooling to obtain 512 7×7 pixel feature maps. The 512 7×7 pixel feature maps are classified based on the fully connected layer to obtain the food type label corresponding to the food cooking sub-image.

6. The cooking equipment temperature control system based on machine vision according to claim 5, characterized in that: The expressions for the categories and the probability values ​​of each category are as follows: ； in, and Indicates The input and weight parameters of the convolutional neural network. Represents the convolution, pooling and activation operations performed during the convolution process. represents the feature extraction module, Indicates The output probability of each category of the convolutional neural network is Indicates The output category label of each category of the convolutional neural network.

7. The cooking equipment temperature control system based on machine vision according to claim 6, characterized in that: The first maturity recognition model adopts an improved support vector machine, and the support vector machine is optimized based on a swarm particle optimization algorithm.

8. The cooking equipment temperature control system based on machine vision according to claim 7, characterized in that: The support vector machine is optimized based on a swarm particle optimization algorithm, including: Randomly generate the initial population and set the population size , initialize the fitness of each particle, where the fitness is the mean square error loss of the first maturity recognition model, and set the maximum number of iterations ; Calculate each particle The fitness value of the particle is determined and the best particle in the current population is identified. , and its fitness is , calculate the other particles in the population and the optimal particle The distance between them is calculated as follows: ； in, Indicates the population The distance between each particle and the optimal particle; The position is updated according to the position of the particle and the distance between the particles, wherein the position update formula is as follows: ； in, and Indicates that the coordinates in the population are The particle Second and The position in the d-dimensional solution space in the iteration, represents the first random number, represents the second random number, Indicates The optimal position of the current particle in the iteration, and represents the division coefficient; Calculate the fitness of the new position. If the fitness of the new position is better than that of the old position, update the position and fitness of the particle. Repeat the above operation until the maximum number of iterations is reached. Or fitness convergence to obtain the particle with the best fitness in the current population and its fitness.

9. The cooking equipment temperature control system based on machine vision according to claim 8, characterized in that: The second maturity recognition model adopts the ResNet50 model.

Images