A cooking robot, an intelligent controller and a monosodium glutamate feeding method

Abstract

The application discloses a cooking robot, an intelligent controller and a monosodium glutamate feeding method. The method comprises the following steps: feeding monosodium glutamate into a monosodium glutamate dissolving pool to form a monosodium glutamate aqueous solution; acquiring a monosodium glutamate aqueous solution image of the monosodium glutamate dissolving pool, and performing image analysis on the monosodium glutamate aqueous solution image in combination with a region clustering algorithm and a microscopic binary neural network; judging whether the monosodium glutamate inventory in the monosodium glutamate dissolving pool is up to standard based on the image analysis result; if it is judged that the monosodium glutamate inventory in the monosodium glutamate dissolving pool is not up to standard, continuously feeding monosodium glutamate into the monosodium glutamate dissolving pool until the monosodium glutamate inventory in the monosodium glutamate dissolving pool is up to standard; and conveying the monosodium glutamate aqueous solution in the monosodium glutamate dissolving pool into corresponding cookware. The monosodium glutamate aqueous solution image is quickly processed and accurately analyzed in real time by combining the region clustering algorithm and the microscopic binary neural network, the monosodium glutamate in the monosodium glutamate dissolving pool can be timely supplemented, and therefore, automatic blending and accurate feeding of monosodium glutamate in a cooking process can be realized.

Description

Technical Field

[0001] This invention relates to the field of automation technology, and in particular to a cooking robot, an intelligent controller, and a method for dispensing monosodium glutamate (MSG). Background Technology

[0002] Currently, cooking robots have become an important culinary aid in catering establishments. These devices feature built-in electronic recipes, automating the cooking process. They can perform complex cooking techniques such as shaking the pan, tossing, separating ingredients, and pouring, as well as various other cooking methods including stir-frying, quick-frying, sautéing, braising, and simmering. Users can enjoy a variety of authentic dishes with just a touch of a button. The cooking process is non-stick, non-burning, and non-overflowing, and is safe, energy-efficient, and smokeless. During cooking, the pan automatically seals to preserve freshness, ensuring the dishes are nutritious, delicious, and authentic.

[0003] Despite the numerous advantages mentioned above, the taste of dishes cooked by robotic chefs falls short of expectations in practical use. Specifically, while the recipes and programs of these robots are set according to the cooking processes of professional chefs, the taste of the dishes they produce is far from that of professional chefs, failing to meet the expected standards. One significant reason is the difficulty in precisely adding seasonings. Among many seasonings, monosodium glutamate (MSG) is an extremely important one, its dosage directly determining the taste of the dish. Therefore, accurately controlling the amount of MSG added is crucial.

[0004] Most existing cooking robots employ two methods for dispensing MSG: solid dispensing and liquid dispensing. Solid dispensing typically uses a spiral propulsion system to meter the MSG, controlling the amount dispensed by adjusting the spiral's rotation. Liquid dispensing involves preparing an aqueous solution of MSG and then dripping it into the pot. However, both methods struggle to achieve precise MSG dispensing. Specifically, with spiral propulsion systems, the dispensing device, positioned above the pot, is susceptible to steam, causing MSG to clump and potentially jamming the propulsion mechanism, resulting in inaccurate metering. Furthermore, the varying particle size of MSG makes precise dispensing difficult to control solely by adjusting the spiral's rotation. Regarding the method of adding MSG solution, if the prepared MSG solution is unsaturated, adding MSG is equivalent to adding more water to the pot, which will affect the taste of the cooked dish. If the prepared MSG solution is saturated, the solubility of MSG is affected by temperature, and the temperature cannot be controlled consistently each time a saturated solution is prepared. This means that the solubility of MSG solution will vary each time, making it difficult to achieve precise addition of MSG. In addition, the preparation of MSG solution is done manually, so when the cooking workload is large, it needs to be added frequently, which is time-consuming and laborious. Summary of the Invention

[0005] This invention provides a cooking robot, an intelligent controller, and a method for dispensing monosodium glutamate (MSG), aiming to achieve automatic mixing and precise dispensing of MSG during the cooking process.

[0006] This invention provides a method for dispensing monosodium glutamate (MSG) based on a cooking robot, comprising:

[0007] Monosodium glutamate (MSG) is added to the MSG dissolving tank to form an MSG aqueous solution;

[0008] An image of the monosodium glutamate (MSG) aqueous solution in a MSG dissolving tank is acquired, and the image of the MSG aqueous solution is analyzed by combining a region clustering algorithm and a microscopic binarization neural network.

[0009] Based on the image analysis results, determine whether the amount of monosodium glutamate in the monosodium glutamate dissolving tank meets the standard;

[0010] If it is determined that the amount of monosodium glutamate in the monosodium glutamate dissolving tank is not up to standard, then continue to add monosodium glutamate to the monosodium glutamate dissolving tank until the amount of monosodium glutamate in the dissolving tank reaches the standard.

[0011] The monosodium glutamate (MSG) solution in the MSG dissolving tank is transferred to the corresponding cookware.

[0012] Furthermore, the step of adding monosodium glutamate (MSG) to the MSG dissolving tank to form an MSG aqueous solution includes:

[0013] When the cooking robot is powered on, the drive motor pushes the stirrer to rotate slowly in the monosodium glutamate dissolving tank.

[0014] Control the electric water tap to release water into the monosodium glutamate dissolving tank until the water level in the monosodium glutamate dissolving tank reaches the preset height, thereby obtaining a monosodium glutamate aqueous solution;

[0015] The temperature of the monosodium glutamate (MSG) solution in the MSG dissolving tank is monitored in real time by a temperature monitor installed around the drain outlet of the MSG dissolving tank. When the water temperature is detected to be lower than the preset first temperature, the heating rod is controlled to heat the MSG solution until the temperature of the MSG solution reaches the preset second temperature.

[0016] Furthermore, the step of acquiring an image of the monosodium glutamate (MSG) aqueous solution in the MSG dissolving pool, and performing image analysis on the MSG aqueous solution image using a region clustering algorithm and a microscopic binarization neural network, includes:

[0017] The monosodium glutamate aqueous solution image was denoised using a region clustering algorithm.

[0018] The image of the monosodium glutamate (MSG) aqueous solution after noise reduction is input into a micro-binarization neural network, and the corresponding MSG decision is output by the micro-binarization neural network. The MSG decision is then used as the image analysis result.

[0019] Furthermore, the denoising process for the monosodium glutamate aqueous solution image includes:

[0020] Iterate through every pixel of the monosodium glutamate aqueous solution image;

[0021] For each pixel of the monosodium glutamate aqueous solution image, a multi-level clustering task is performed to complete image segmentation;

[0022] Calculate the average pixel value for each block and initialize cluster labels for the pixels in each block;

[0023] For each pixel, a neighborhood is generated, and the median of the neighborhood is calculated. Then, the median is used to replace the pixel value at the corresponding position, thereby completing the noise reduction process.

[0024] Furthermore, the calculation of the average pixel value of each block and the initialization of cluster labels for the pixels of each block includes:

[0025] For each pixel, determine whether there exists a first target pixel that is adjacent to any pixel belonging to another cluster;

[0026] If it is determined that there is a first target pixel that is adjacent to any pixel belonging to another cluster, then calculate the first average distance between the first target pixel and its own cluster and the second average distance between the first target pixel and its neighboring clusters.

[0027] Compare the first average distance and the second average distance, and modify the cluster label of the first target pixel to the cluster label with the smallest distance.

[0028] Furthermore, the step of generating a neighborhood for each pixel, calculating the median of the neighborhood, and then replacing the pixel value at the corresponding position with the median to complete the noise reduction process includes:

[0029] For each pixel, arbitrarily select a second target pixel and determine whether the first neighboring pixel of the second target pixel belongs to the same cluster as the second target pixel;

[0030] If it is determined that the first adjacent pixel and the second target pixel belong to the same cluster, then the first adjacent pixel is added to a preset waiting queue;

[0031] Select the third target pixel that is most similar to the second target pixel from the first adjacent pixels in the waiting queue, and add the third target pixel to the preset result queue. At the same time, record the second adjacent pixels corresponding to the third target pixel in the preset set.

[0032] In the preset set, the second adjacent pixel with the same cluster label as the second target pixel is selected as the fourth target pixel, and the fourth target pixel is added to the waiting queue;

[0033] If the number of neighboring pixels generated at the lowest level is less than k, then pixels with the same cluster label from the previous level are added sequentially until the number of elements in the waiting queue is less than k.

[0034] Furthermore, the step of inputting the denoised MSG aqueous solution image into a micro-binarization neural network, and having the micro-binarization neural network output the corresponding MSG decision, includes:

[0035] The convolutional layers in the micro-binarization neural network are used to capture hidden features in the denoised monosodium glutamate aqueous solution image and generate feature maps.

[0036] The feature map is input into a value pooling layer to reduce the resolution of the feature map while preserving spatial invariance;

[0037] The MBC layer is used to quickly capture features from the feature map to obtain the corresponding target features.

[0038] The target features are nonlinearly fitted using a fully connected layer;

[0039] The target features processed by the fully connected layer are input into the softmax layer, and the softmax layer outputs the corresponding MSG decision.

[0040] This invention also provides an intelligent controller, comprising:

[0041] The first dispensing module is used to dispense monosodium glutamate (MSG) into the MSG dissolving tank to form an MSG aqueous solution.

[0042] The image analysis module is used to acquire images of the monosodium glutamate (MSG) aqueous solution in the MSG dissolving pool, and to perform image analysis on the MSG aqueous solution images by combining a region clustering algorithm and a microscopic binarization neural network.

[0043] The monosodium glutamate (MSG) quantity control module is used to determine whether the MSG content in the MSG dissolving tank meets the standard based on image analysis results.

[0044] The second dispensing module is used to continue dispensing monosodium glutamate (MSG) into the MSG dissolving tank if it is determined that the amount of MSG in the MSG dissolving tank is not up to standard, until the amount of MSG in the MSG dissolving tank reaches the standard.

[0045] The monosodium glutamate (MSG) delivery module is used to deliver the MSG aqueous solution from the MSG dissolving tank to the corresponding cookware.

[0046] Furthermore, the image analysis module includes:

[0047] The denoising unit is used to denoise the monosodium glutamate aqueous solution image using a region clustering algorithm;

[0048] The decision output unit is used to input the denoised monosodium glutamate (MSG) aqueous solution image into a micro-binarization neural network, and the micro-binarization neural network outputs the corresponding MSG decision, and the MSG decision is used as the image analysis result.

[0049] This invention also provides a cooking robot, including an intelligent controller, a stirrer disposed at the bottom of the monosodium glutamate dissolving tank, a drive motor connected to the stirrer, and an electrically controlled faucet, a temperature monitor, and a heating rod, all electrically connected to the intelligent controller, wherein:

[0050] The drive motor is used to push the stirrer to rotate slowly in the monosodium glutamate dissolving tank after the cooking robot is powered on.

[0051] The electrically controlled faucet is used to add water to the monosodium glutamate dissolving tank until the water level in the monosodium glutamate dissolving tank reaches a preset height, thereby obtaining a monosodium glutamate aqueous solution;

[0052] The temperature monitor is used to monitor the temperature of the monosodium glutamate (MSG) aqueous solution in the MSG dissolving tank in real time.

[0053] The heating rod is used to heat the monosodium glutamate (MSG) aqueous solution when the water temperature is detected to be lower than a preset first temperature, until the temperature of the MSG aqueous solution reaches a preset second temperature.

[0054] This invention provides a cooking robot, an intelligent controller, and a method for dispensing monosodium glutamate (MSG). The method includes: adding MSG to a MSG dissolving tank to form an MSG aqueous solution; acquiring an image of the MSG aqueous solution in the dissolving tank and performing image analysis on the image using a region clustering algorithm and a microscopic binarization neural network; determining whether the MSG content in the dissolving tank meets the standard based on the image analysis results; if the MSG content in the dissolving tank is not determined to be sufficient, continuing to add MSG to the dissolving tank until the MSG content meets the standard; and conveying the MSG aqueous solution from the dissolving tank to the corresponding cookware. This invention, by combining a region clustering algorithm and a microscopic binarization neural network for rapid processing and real-time accurate analysis of the MSG aqueous solution image, enables timely replenishment of MSG in the dissolving tank, thereby achieving automatic mixing and precise dispensing of MSG during the cooking process. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 A flowchart illustrating a method for dispensing monosodium glutamate (MSG) based on a cooking robot, provided in an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the network structure of a monosodium glutamate (MSG) dispensing method based on a cooking robot, provided in an embodiment of the present invention.

[0058] Figure 3 A schematic block diagram of an intelligent controller provided in an embodiment of the present invention;

[0059] Figure 4 This is a structural schematic diagram of a cooking robot provided in an embodiment of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0062] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0063] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0064] Please see below. Figure 1 , Figure 1 The flowchart of a method for dispensing monosodium glutamate (MSG) based on a cooking robot provided in an embodiment of the present invention specifically includes steps S101 to S105.

[0065] S101. Add monosodium glutamate (MSG) to the MSG dissolving tank to form an MSG aqueous solution;

[0066] S102. Obtain an image of the monosodium glutamate (MSG) aqueous solution in the MSG dissolving pool, and perform image analysis on the MSG aqueous solution image using a region clustering algorithm and a microscopic binary neural network.

[0067] S103. Based on the image analysis results, determine whether the amount of monosodium glutamate in the monosodium glutamate dissolving tank meets the standard.

[0068] S104. If it is determined that the amount of monosodium glutamate in the monosodium glutamate dissolving tank does not meet the standard, then continue to add monosodium glutamate to the monosodium glutamate dissolving tank until the amount of monosodium glutamate in the monosodium glutamate dissolving tank meets the standard.

[0069] S105. Transfer the monosodium glutamate (MSG) solution in the MSG dissolving tank to the corresponding cookware.

[0070] In this embodiment, monosodium glutamate (MSG) is first added to a MSG dissolving tank containing a certain amount of aqueous solution to dissolve the MSG and obtain a corresponding MSG aqueous solution. Then, images of the MSG aqueous solution are acquired by taking photographs or other methods, and further image analysis is performed. Based on the image analysis results, it can be determined whether more MSG needs to be added, and once the requirements are met, the MSG aqueous solution is transferred to the cookware.

[0071] This embodiment combines a region clustering algorithm with a microscopic binary neural network to analyze captured photos in real time. This enables rapid processing and accurate analysis of real-time photos, allowing for timely replenishment of MSG in the dissolving tank. Furthermore, this embodiment automates the entire process from MSG addition to dissolution and delivery, requiring no manual intervention and further enhancing the automation level of the cooking robot. Moreover, the MSG delivery method provided in this embodiment allows for fully automated MSG addition, and by controlling the MSG temperature, the saturated solubility of MSG can be controlled, along with precise control of the amount of MSG solution delivered, achieving precise control over the MSG and thus promoting a higher level of taste in dishes cooked by the cooking robot.

[0072] In one embodiment, step S101 includes:

[0073] When the cooking robot is powered on, the drive motor pushes the stirrer to rotate slowly in the monosodium glutamate dissolving tank.

[0074] Control the electric water tap to release water into the monosodium glutamate dissolving tank until the water level in the monosodium glutamate dissolving tank reaches the preset height, thereby obtaining a monosodium glutamate aqueous solution;

[0075] The temperature of the monosodium glutamate (MSG) solution in the MSG dissolving tank is monitored in real time by a temperature monitor installed around the drain outlet of the MSG dissolving tank. When the water temperature is detected to be lower than the preset first temperature, the heating rod is controlled to heat the MSG solution until the temperature of the MSG solution reaches the preset second temperature.

[0076] In this embodiment, combined with Figure 4 , Figure 4 is a schematic structural diagram of a stir-fry robot provided correspondingly. When putting monosodium glutamate into the monosodium glutamate dissolution tank S7 to form a monosodium glutamate aqueous solution, first, the stir-fry robot is turned on to energize the corresponding components. Then, the drive motor S18 runs to drive the stirrer S17 to slowly rotate in the monosodium glutamate dissolution tank S7. Next, the intelligent controller S11 controls the electronically controlled faucet S2 to open, and water is discharged into the monosodium glutamate dissolution tank S7, and the liquid level monitor S8 senses the water level height in the monosodium glutamate dissolution tank S7. When the water level reaches the preset height (h), the intelligent controller S11 controls the electronically controlled faucet S2 to close. At the same time, the temperature monitor S13 arranged near the drain outlet of the monosodium glutamate dissolution tank S7 monitors the temperature of the monosodium glutamate aqueous solution in the monosodium glutamate dissolution tank in real time, and when the temperature of the monosodium glutamate aqueous solution is lower than the preset first temperature t (for example, 70 °C < t < 90 °C, default is 85 °C), the intelligent controller S11 controls the heating rod S10 to heat the monosodium glutamate aqueous solution until the temperature of the monosodium glutamate aqueous solution reaches the preset second temperature.

[0077] Furthermore, a light source S12 (such as a flicker-free white light source) is used to continuously irradiate the bottom of the monosodium glutamate dissolution tank S7, and the four light sources are distributed around the monosodium glutamate dissolution tank. Preferably, the distances between the distributed light sources are equal to each other, and the light is directed at the center of the bottom of the monosodium glutamate dissolution tank S7. In addition, when the room temperature is relatively low, the fan S6 can be selectively turned on to blow away the mist emitted from the monosodium glutamate aqueous solution, so as to avoid affecting the imaging clarity of the imaging module S1.

[0078] After that, when the monosodium glutamate aqueous solution in the monosodium glutamate dissolution tank S7 is transported to the corresponding cookware S5 through the monosodium glutamate solution delivery pipe S15, the intelligent controller S11 controls the water pump S9 to run for a corresponding stroke, so that the monosodium glutamate aqueous solution is filtered through the filter screen S14 and then sprayed into the cookware S5 through the monosodium glutamate dosing nozzle S4. Preferably, to ensure the accuracy of the volume of the sprayed monosodium glutamate solution, the drive motor S18 of the water pump S9 is a linear motor, which has the characteristic of high positioning accuracy, and the water pump S9 can adopt a precision metering pump.

[0079] In one embodiment, the step S102 includes:

[0080] Adopting a region clustering algorithm to denoise the monosodium glutamate aqueous solution image;

[0081] Inputting the denoised monosodium glutamate aqueous solution image into a microscopic binary neural network, and the microscopic binary neural network outputs the corresponding monosodium glutamate decision, and taking the monosodium glutamate decision as the image analysis result.

[0082] In this embodiment, when performing image analysis on the monosodium glutamate (MSG) aqueous solution image, a region clustering algorithm is first used to denoise it. Then, a microscopic binary neural network is used to output corresponding MSG decisions for the denoised MSG aqueous solution image. These decisions can be, for example, whether the current MSG amount is sufficient, or whether the current MSG aqueous solution meets subsequent cooking requirements. The intelligent controller of the cooking robot can further determine whether to continue adding MSG to the dissolving tank based on these decisions. For example, when the microscopic binary neural network outputs 1, the intelligent controller controls the MSG supplementer S3 to slowly add MSG to the dissolving tank; when the microscopic binary neural network outputs 0, the intelligent controller does not output a control signal.

[0083] Furthermore, before performing image analysis on the monosodium glutamate (MSG) aqueous solution image, it is naturally necessary to first acquire the MSG aqueous solution image. Here, in this embodiment, a camera module S1 installed above the MSG dissolving tank S7 continuously takes photos from the upper part of the MSG dissolving tank S7 to obtain the MSG aqueous solution image, and then transmits the photos to the intelligent controller S11 for analysis.

[0084] Subsequently, the intelligent controller S11 determines the next operation based on the MSG decision. When the intelligent controller S11 determines that there is insufficient MSG at the bottom of the MSG dissolving tank S7, it controls the MSG replenisher S3 to slowly add MSG to the MSG dissolving tank until the camera module S1 captures obvious undissolved MSG particles at the bottom of the MSG dissolving tank for n consecutive seconds. At this point, the intelligent controller S11 can determine that there is enough MSG at the bottom of the MSG dissolving tank and controls the MSG replenisher S3 to stop adding MSG.

[0085] In one embodiment, the denoising process for the monosodium glutamate aqueous solution image includes:

[0086] Iterate through every pixel of the monosodium glutamate aqueous solution image;

[0087] For each pixel of the monosodium glutamate aqueous solution image, a multi-level clustering task is performed to complete image segmentation;

[0088] Calculate the average pixel value for each block and initialize cluster labels for the pixels in each block;

[0089] For each pixel, a neighborhood is generated, and the median of the neighborhood is calculated. Then, the median is used to replace the pixel value at the corresponding position, thereby completing the noise reduction process.

[0090] In this embodiment, a region clustering algorithm is used for noise reduction. First, a noisy monosodium glutamate (MSG) aqueous solution image is input. The MSG aqueous solution image can be represented by image[m][n][3], where m represents the length of the MSG aqueous solution image, n represents the width of the MSG aqueous solution image, and 3 represents the R, G, and B channels of the MSG aqueous solution image.

[0091] Then, each pixel of the MSG aqueous solution image is traversed to complete image segmentation. Image segmentation is a multi-level clustering task, with each level equivalent to a k-means clustering algorithm. The side length 'd' of each level's square is initialized, and the image is divided into squares with side length 'd'. Any extra space can be rectangles. The average pixel value of each square is calculated, and cluster labels are initialized for the pixels in each square.

[0092] Next, a neighborhood of each pixel is generated, and the median is calculated based on the pixels in the neighborhood. The median is then stored in the denoised MSG aqueous solution image, where the generated median will replace the pixel value at the corresponding position of the pixel.

[0093] In one embodiment, calculating the average pixel value of each block and initializing cluster labels for the pixels of each block includes:

[0094] For each pixel, determine whether there exists a first target pixel that is adjacent to any pixel belonging to another cluster;

[0095] If it is determined that there is a first target pixel that is adjacent to any pixel belonging to another cluster, then calculate the first average distance between the first target pixel and its own cluster and the second average distance between the first target pixel and its neighboring clusters.

[0096] Compare the first average distance and the second average distance, and modify the cluster label of the first target pixel to the cluster label with the smallest distance.

[0097] In this embodiment, if a pixel (i.e., the first target pixel) is adjacent to another cluster (adjacent to a pixel belonging to another cluster), the average distance between the first target pixel and its own cluster and the average distance between the first target pixel and its neighboring clusters are compared. The cluster label of the first target pixel is then modified to the label of the cluster with the smallest distance. The above steps are performed for each pixel and repeated until the cluster labels of all pixels no longer change. Then, image segmentation and partitioning at different levels are completed according to different square side lengths d.

[0098] In one embodiment, the step of generating a neighborhood for each pixel, calculating the median of the neighborhood, and then replacing the pixel value at the corresponding position with the median to complete the noise reduction process includes:

[0099] For each pixel, arbitrarily select a second target pixel and determine whether the first neighboring pixel of the second target pixel belongs to the same cluster as the second target pixel;

[0100] If it is determined that the first adjacent pixel and the second target pixel belong to the same cluster, then the first adjacent pixel is added to a preset waiting queue;

[0101] Select the third target pixel that is most similar to the second target pixel from the first adjacent pixels in the waiting queue, and add the third target pixel to the preset result queue. At the same time, record the second adjacent pixels corresponding to the third target pixel in the preset set.

[0102] In the preset set, the second adjacent pixel with the same cluster label as the second target pixel is selected as the fourth target pixel, and the fourth target pixel is added to the waiting queue;

[0103] If the number of neighboring pixels generated at the lowest level is less than k, then pixels with the same cluster label from the previous level are added sequentially until the number of elements in the waiting queue is less than k.

[0104] In this embodiment, for each pixel in the monosodium glutamate aqueous solution image, the following steps are performed:

[0105] (1) Generate the neighborhood of each pixel. This step involves two queues: a waiting queue (each element of which is a pixel, denoted as WL) and a result queue (denoted as RL). First, consider the four pixels above, below, left, and right of the target pixel (i.e., the second target pixel). If they belong to the same cluster (lowest level) as the second target pixel, add them to WL. Additionally, initialize RL to an empty set.

[0106] (2) Complete the generation of the neighborhood. If the number of elements in RL is less than the set neighborhood size (number of pixels k, where k is a parameter of the algorithm), then perform the following operations in a loop: Select the pixel in WL that is most similar to the second target pixel (i.e., the third target pixel, denoted as ps), add ps to RL, and denote the set of the four pixels above, below, left, and right of ps as S. Add the pixels in S whose cluster labels are the same as the cluster labels of the second target pixel (i.e., the fourth target pixel) to WL. If the number of neighborhood pixels generated at the lowest level is less than k, then add the pixels with the same cluster labels from the previous level in sequence. After this step, WL contains k pixels.

[0107] (3) Calculate the median of the neighborhood of the second target pixel. In the previous steps, the algorithm generated the neighborhood of the second target pixel, i.e., its result queue RL. This step will calculate the median based on the pixels in the neighborhood and store the median in the denoised image. In the denoised image, the generated median will replace the pixel value at the corresponding position of the target pixel.

[0108] In one embodiment, the step of inputting the denoised monosodium glutamate (MSG) aqueous solution image into a micro-binarization neural network, and having the micro-binarization neural network output the corresponding MSG decision, includes:

[0109] The convolutional layers in the micro-binarization neural network are used to capture hidden features in the denoised monosodium glutamate aqueous solution image and generate feature maps.

[0110] The feature map is input into a value pooling layer to reduce the resolution of the feature map while preserving spatial invariance;

[0111] The MBC layer is used to quickly capture features from the feature map to obtain the corresponding target features.

[0112] The target features are nonlinearly fitted using a fully connected layer;

[0113] The target features processed by the fully connected layer are input into the softmax layer, and the softmax layer outputs the corresponding MSG decision.

[0114] In this embodiment, the micro-binarization neural network consists of five key components: convolutional layers, pooling layers, micro-binary convolutional (MBC) layers, fully connected layers, and softmax layers. First, the denoised MSG aqueous solution image is input into the convolutional layers. The convolutional kernels automatically capture hidden features to generate new feature maps, which are then input into the pooling layers. The pooling layers reduce the resolution of the feature maps while preserving spatial invariance. Next, the MBC layers are used for fast feature capture of the feature maps. The fully connected layers are used for non-linear fitting of the extracted features. Finally, the softmax layers are used to obtain the final decision result. Figure 3 As shown in the schematic diagram of a micro-binarized neural network structure provided in this embodiment, it can be seen that... Figure 3 The process involves more than just one convolutional layer and one pooling layer; the fully connected layer is the same. Therefore, it is understandable that in practical applications, the MSG aqueous solution image is processed sequentially through a convolutional layer, a pooling layer, another convolutional layer, a pooling layer, an MBC layer, a fully connected layer 1, a fully connected layer 2, and a softmax layer before the corresponding MSG decision is output.

[0115] In a convolutional layer, the input data is convolved by a set of convolutional kernels to generate new feature maps. This process can be described as follows:

[0116]

[0117] in This is the output of the l-th convolutional layer, representing the newly obtained feature map, which is used as input to the next layer (the (l+1)-th layer). m' = 1, 2, ..., M' are the indices of the output feature map, while m = 1, 2, ..., M are the indices of the input feature maps. The index is M = 1 if the layer is the first layer of the model. This represents the weight matrix of the m-th convolutional kernel. This is the bias matrix for this layer. * indicates a convolution operation. f(·) represents the ReLU nonlinear activation function.

[0118] Pooling layers primarily perform pooling operations to reduce the size of the feature map, optimize the number of neurons, accelerate convergence, and prevent overfitting. The pooling operation divides the feature map into many non-overlapping rectangular regions through a pooling window, and then obtains the fused features of each region through pooling, represented as:

[0119]

[0120] in express The [i,j]th element. p and q are the length and width of the pooling window, respectively. The maximum value in the pooling window is selected as the element value of the new feature map.

[0121] In the MBC layer, the micro binary convolution kernel (MBP) performs operations using an overlapping scanning method similar to that of traditional convolution kernels. Multiple micro binary convolution kernels are used in the MBC layer. Input feature mapping A convolution operation is performed, where i = 1, 2, ..., K, generating K feature matrices. These feature matrices are then computed using a non-linear activation function. In the MBC layer, the ReLU activation function replaces the Heaviside step function in traditional LBP. Finally, the K feature matrices are combined with K learnable weights. (i = 1, 2, ..., K) are linearly combined to generate the output of the final MBC layer. This process can be described as follows:

[0122]

[0123] in is the output feature map of the nth layer, and m and m' are the indices of the input and output feature maps, respectively. f(·) represents the ReLU activation function. This represents the microscopic binary convolution operation.

[0124] Fully connected layers primarily perform nonlinear transformations on feature maps, a process described as follows:

[0125] F = f(W) fc X fc +b fc )

[0126] Where X fc It is the input feature map, W fc and b fc represents the weights and biases of the fully connected layer, and F represents the output of the fully connected layer.

[0127] The Softmax layer is used for the output. Given a dataset containing z samples... in The corresponding tags are Where y (i) ∈{0,1}. The Softmax output is as follows:

[0128]

[0129] Where θ = [θ1, θ2] represents the parameters of the Softmax classifier. If P(y (i) =0|f (i) ;θ)≥P(y (i) =1|f (i) If θ), the micro-binarized neural network outputs 0; otherwise, it outputs 1.

[0130] Figure 3 This invention provides a schematic block diagram of an intelligent controller S11, which includes:

[0131] The first dispensing module 301 is used to dispense monosodium glutamate (MSG) into the MSG dissolving tank to form an MSG aqueous solution.

[0132] Image analysis module 302 is used to acquire images of monosodium glutamate (MSG) aqueous solution in a MSG dissolving pool, and to perform image analysis on the MSG aqueous solution images by combining a region clustering algorithm and a microscopic binarization neural network.

[0133] The monosodium glutamate (MSG) quantity control module 303 is used to determine whether the MSG content in the MSG dissolving tank meets the standard based on the image analysis results.

[0134] The second dispensing module 304 is used to continue dispensing monosodium glutamate (MSG) into the MSG dissolving tank if it is determined that the amount of MSG in the MSG dissolving tank is not up to standard, until the amount of MSG in the MSG dissolving tank reaches the standard.

[0135] The monosodium glutamate (MSG) delivery module 305 is used to deliver the MSG aqueous solution in the MSG dissolving tank to the corresponding cookware.

[0136] In one embodiment, the image analysis module 302 includes:

[0137] The denoising unit is used to denoise the monosodium glutamate aqueous solution image using a region clustering algorithm;

[0138] The decision output unit is used to input the denoised monosodium glutamate (MSG) aqueous solution image into a micro-binarization neural network, and the micro-binarization neural network outputs the corresponding MSG decision, and the MSG decision is used as the image analysis result.

[0139] In one embodiment, the noise reduction unit includes:

[0140] The first traversal unit is used to traverse each pixel of the monosodium glutamate aqueous solution image;

[0141] The image segmentation unit is used to perform a multi-level clustering task for each pixel of the monosodium glutamate aqueous solution image to complete image segmentation;

[0142] The label initialization unit is used to calculate the average pixel value of each block and initialize cluster labels for the pixels of each block.

[0143] The neighborhood generation unit is used to generate a neighborhood for each pixel, calculate the median of the neighborhood, and then replace the pixel value at the corresponding position with the median, thereby completing the noise reduction process.

[0144] In one embodiment, the tag initialization unit includes:

[0145] The first judgment unit is used to determine, for each pixel, whether there exists a first target pixel that is adjacent to any pixel belonging to another cluster;

[0146] The first determination unit is used to calculate the first average distance between the first target pixel and its own cluster and the second average distance between the first target pixel and the adjacent cluster if it is determined that there is a first target pixel adjacent to any pixel point belonging to another cluster.

[0147] The label modification unit is used to compare the first average distance and the second average distance, and modify the cluster label of the first target pixel to the cluster label with the smallest distance.

[0148] In one embodiment, the neighborhood generation unit includes:

[0149] The second judgment unit is used to arbitrarily select a second target pixel for each pixel and determine whether the first adjacent pixel of the second target pixel belongs to the same cluster as the second target pixel.

[0150] The second determination unit is used to add the first adjacent pixel to a preset waiting queue if it is determined that the first adjacent pixel and the second target pixel belong to the same cluster.

[0151] The first adding unit is used to select the third target pixel that is most similar to the second target pixel from the first adjacent pixels in the waiting queue, add the third target pixel to the preset result queue, and record the second adjacent pixels corresponding to the third target pixel in the preset set.

[0152] The second adding unit is used to select a second adjacent pixel with the same cluster label as the second target pixel from a preset set as the fourth target pixel, and add the fourth target pixel to the waiting queue;

[0153] The pixel addition unit is used to add pixels with the same cluster label from the previous level in sequence if the number of neighboring pixels generated at the lowest level is less than k, until the number of elements in the waiting queue is less than k.

[0154] In one embodiment, the decision output unit includes:

[0155] The first feature capture unit is used to capture hidden features of the denoised monosodium glutamate aqueous solution image through the convolutional layer in the micro-binarization neural network and generate feature maps.

[0156] Pooling units are used to input the feature map values ​​into a pooling layer to reduce the resolution of the feature map while maintaining spatial invariance;

[0157] The second feature capture unit is used to quickly capture features from the feature map using the MBC layer to obtain the corresponding target features.

[0158] A fitting unit is used to perform nonlinear fitting of the target features using a fully connected layer;

[0159] The feature input unit is used to input the target features processed by the fully connected layer into the softmax layer, and the softmax layer outputs the corresponding MSG decision.

[0160] Figure 4 This is a schematic diagram of a cooking robot provided in an embodiment of the present invention. The cooking robot includes an intelligent controller S11, a stirrer S17 disposed at the bottom of the monosodium glutamate dissolving tank, a drive motor S18 connected to the stirrer S17, and an electrically controlled water tap S2, a temperature monitor S13, and a heating rod S10, all electrically connected to the intelligent controller S11.

[0161] The drive motor S18 is used to push the stirrer S17 to rotate slowly in the monosodium glutamate dissolving tank after the cooking robot is powered on.

[0162] The electrically controlled faucet S2 is used to add water to the monosodium glutamate dissolving tank until the water level in the monosodium glutamate dissolving tank reaches a preset height, thereby obtaining a monosodium glutamate aqueous solution.

[0163] The temperature monitor S13 is used to monitor the temperature of the monosodium glutamate aqueous solution in the monosodium glutamate dissolving tank in real time.

[0164] The heating rod S10 is used to heat the monosodium glutamate aqueous solution when the water temperature is detected to be lower than the preset first temperature, until the temperature of the monosodium glutamate aqueous solution reaches the preset second temperature.

[0165] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0166] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0167] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for dispensing monosodium glutamate (MSG) based on a cooking robot, characterized in that, include: Monosodium glutamate (MSG) is added to the MSG dissolving tank to form an MSG aqueous solution; An image of the monosodium glutamate (MSG) aqueous solution in a MSG dissolving tank is acquired, and the image of the MSG aqueous solution is analyzed by combining a region clustering algorithm and a microscopic binarization neural network. Based on the image analysis results, determine whether the amount of monosodium glutamate in the monosodium glutamate dissolving tank meets the standard; If it is determined that the amount of monosodium glutamate in the monosodium glutamate dissolving tank is not up to standard, then continue to add monosodium glutamate to the monosodium glutamate dissolving tank until the amount of monosodium glutamate in the dissolving tank reaches the standard. The monosodium glutamate (MSG) solution in the MSG dissolving tank is transferred to the corresponding cookware; The process of acquiring an image of the monosodium glutamate (MSG) aqueous solution in a MSG dissolving tank, and performing image analysis on the MSG aqueous solution image using a region clustering algorithm and a microscopic binary neural network, includes: The monosodium glutamate aqueous solution image was denoised using a region clustering algorithm. The image of the monosodium glutamate (MSG) aqueous solution after noise reduction is input into a micro-binarization neural network, and the corresponding MSG decision is output by the micro-binarization neural network. The MSG decision is then used as the image analysis result. The denoising process for the monosodium glutamate aqueous solution image includes: Iterate through every pixel of the monosodium glutamate aqueous solution image; For each pixel of the monosodium glutamate aqueous solution image, a multi-level clustering task is performed to complete image segmentation; Calculate the average pixel value for each block and initialize cluster labels for the pixels in each block; For each pixel, a neighborhood is generated, and the median of the neighborhood is calculated. Then, the median is used to replace the pixel value at the corresponding position, thereby completing the noise reduction process. The process of inputting the denoised MSG aqueous solution image into a micro-binarization neural network and having the micro-binarization neural network output the corresponding MSG decision includes: The convolutional layers in the micro-binarization neural network are used to capture hidden features in the denoised monosodium glutamate aqueous solution image and generate feature maps. The feature map is input into a value pooling layer to reduce the resolution of the feature map while preserving spatial invariance; The MBC layer is used to quickly capture features from the feature map to obtain the corresponding target features. The target features are nonlinearly fitted using a fully connected layer; The target features processed by the fully connected layer are input into the softmax layer, and the softmax layer outputs the corresponding MSG decision.

2. The method for dispensing monosodium glutamate (MSG) based on a cooking robot according to claim 1, characterized in that, The step of adding monosodium glutamate (MSG) to the MSG dissolving tank to form an MSG aqueous solution includes: When the cooking robot is powered on, the drive motor pushes the stirrer to rotate slowly in the monosodium glutamate dissolving tank. Control the electric water tap to release water into the monosodium glutamate dissolving tank until the water level in the monosodium glutamate dissolving tank reaches the preset height, thereby obtaining a monosodium glutamate aqueous solution; The temperature of the monosodium glutamate (MSG) solution in the MSG dissolving tank is monitored in real time by a temperature monitor installed around the drain outlet of the MSG dissolving tank. When the water temperature is detected to be lower than the preset first temperature, the heating rod is controlled to heat the MSG solution until the temperature of the MSG solution reaches the preset second temperature.

3. The method for dispensing monosodium glutamate (MSG) based on a cooking robot according to claim 1, characterized in that, The calculation of the average pixel value for each block and the initialization of cluster labels for the pixels in each block include: For each pixel, determine whether there exists a first target pixel that is adjacent to any pixel belonging to another cluster; If it is determined that there is a first target pixel that is adjacent to any pixel belonging to another cluster, then calculate the first average distance between the first target pixel and its own cluster and the second average distance between the first target pixel and its neighboring clusters. Compare the first average distance and the second average distance, and modify the cluster label of the first target pixel to the cluster label with the smallest distance.

4. The method for dispensing monosodium glutamate (MSG) based on a cooking robot according to claim 3, characterized in that, The step of generating a neighborhood for each pixel, calculating the median of the neighborhood, and then replacing the pixel value at the corresponding position with the median to complete the noise reduction process includes: For each pixel, arbitrarily select a second target pixel and determine whether the first neighboring pixel of the second target pixel belongs to the same cluster as the second target pixel; If it is determined that the first adjacent pixel and the second target pixel belong to the same cluster, then the first adjacent pixel is added to a preset waiting queue; Select the third target pixel that is most similar to the second target pixel from the first adjacent pixels in the waiting queue, and add the third target pixel to the preset result queue. At the same time, record the second adjacent pixels corresponding to the third target pixel in the preset set. In the preset set, the second adjacent pixel with the same cluster label as the second target pixel is selected as the fourth target pixel, and the fourth target pixel is added to the waiting queue; If the number of neighboring pixels generated at the lowest level is less than k, then pixels with the same cluster label from the previous level are added sequentially until the number of elements in the waiting queue is less than k.

5. An intelligent controller, characterized in that, The intelligent controller includes: The first dispensing module is used to dispense monosodium glutamate (MSG) into the MSG dissolving tank to form an MSG aqueous solution. The image analysis module is used to acquire images of the monosodium glutamate (MSG) aqueous solution in the MSG dissolving pool, and to perform image analysis on the MSG aqueous solution images by combining a region clustering algorithm and a microscopic binarization neural network. The monosodium glutamate (MSG) quantity control module is used to determine whether the MSG content in the MSG dissolving tank meets the standard based on image analysis results. The second dispensing module is used to continue dispensing monosodium glutamate (MSG) into the MSG dissolving tank if it is determined that the amount of MSG in the MSG dissolving tank is not up to standard, until the amount of MSG in the MSG dissolving tank reaches the standard. The monosodium glutamate (MSG) conveying module is used to convey the MSG aqueous solution in the MSG dissolving tank to the corresponding cookware; The image analysis module includes: The denoising unit is used to denoise the monosodium glutamate aqueous solution image using a region clustering algorithm; The decision output unit is used to input the denoised monosodium glutamate (MSG) aqueous solution image into a micro-binarization neural network, and the micro-binarization neural network outputs the corresponding MSG decision, and uses the MSG decision as the image analysis result. The noise reduction unit includes: The first traversal unit is used to traverse each pixel of the monosodium glutamate aqueous solution image; The image segmentation unit is used to perform a multi-level clustering task for each pixel of the monosodium glutamate aqueous solution image to complete image segmentation; The label initialization unit is used to calculate the average pixel value of each block and initialize cluster labels for the pixels of each block. The neighborhood generation unit is used to generate a neighborhood for each pixel, calculate the median of the neighborhood, and then replace the pixel value at the corresponding position with the median, thereby completing the noise reduction process. The decision output unit includes: The first feature capture unit is used to capture hidden features of the denoised monosodium glutamate aqueous solution image through the convolutional layer in the micro-binarization neural network and generate feature maps. Pooling units are used to input the feature map values ​​into a pooling layer to reduce the resolution of the feature map while maintaining spatial invariance; The second feature capture unit is used to quickly capture features from the feature map using the MBC layer to obtain the corresponding target features. A fitting unit is used to perform nonlinear fitting of the target features using a fully connected layer; The feature input unit is used to input the target features processed by the fully connected layer into the softmax layer, and the softmax layer outputs the corresponding MSG decision.

6. A cooking robot, characterized in that, Including the intelligent controller as described in claim 5, it further includes a stirrer disposed at the bottom of the monosodium glutamate dissolving tank, a drive motor connected to the stirrer, and an electrically controlled faucet, a temperature monitor, and a heating rod respectively electrically connected to the intelligent controller, wherein: The drive motor is used to push the stirrer to rotate slowly in the monosodium glutamate dissolving tank after the cooking robot is powered on. The electrically controlled faucet is used to add water to the monosodium glutamate dissolving tank until the water level in the monosodium glutamate dissolving tank reaches a preset height, thereby obtaining a monosodium glutamate aqueous solution; The temperature monitor is used to monitor the temperature of the monosodium glutamate (MSG) aqueous solution in the MSG dissolving tank in real time. The heating rod is used to heat the monosodium glutamate (MSG) aqueous solution when the water temperature is detected to be lower than a preset first temperature, until the temperature of the MSG aqueous solution reaches a preset second temperature.

Images