Parameter regulation and control method, device and equipment of air frying oven and medium

Abstract

The invention discloses a parameter regulation and control method and device for an air frying oven, equipment and a medium, and relates to the technical field of plasmas. According to the scheme, multi-dimensional data in the cooking process are collected through a sensor, real-time analysis and prediction are conducted in combination with a preset deep neural network, and the accuracy of the parameter regulation and control is improved. The current plasma power, the current air circulation system parameters and other cooking parameters are automatically adjusted, and the heating efficiency and the intelligent degree of the air frying oven are improved; besides, the preset deep neural network monitors the cooking state of the food in real time, so that the accuracy of the cooking process can be further improved, the situation that the food is over-cooked or not cooked is avoided, and the user experience is improved; moreover, by constructing a preset deep neural network model composed of a timestamp sensing encoder, a Query encoder, a sensor coder-decoder, a sliding window Q-Former and a large language model layer, the collected data can be precisely processed, and the precision of parameter regulation and control of the air frying oven is further improved.

Description

Technical Field

[0001] The present application relates to the field of plasma technology, and in particular to a parameter control method, device, equipment and medium for an air fryer oven. Background Art

[0002] Air fryers are a common cooking tool in the current home appliance market, widely used in both home and commercial kitchens. Deep fryers utilize hot air circulation technology to achieve healthy cooking with minimal or no oil, and they quickly fry food. Ovens, on the other hand, heat food evenly and consistently using both radiation and convection.

[0003] However, despite their advantages, air fryers still have shortcomings in terms of heating efficiency, energy consumption, and operational precision. For example, traditional air fryers rely on hot air convection, which can cause food to dry out or unevenly distribute heat inside, and they also have low energy efficiency.

[0004] Therefore, how to improve the heating efficiency and intelligence of air fryer ovens is an urgent problem to be solved. Summary of the Invention

[0005] The present application aims to at least solve the technical problems existing in the prior art. To this end, in a first aspect, the present application proposes a parameter control method for an air fryer oven, the method comprising: Obtaining target sensor data and target image data collected in real time from the air fryer oven; wherein current plasma power and current air circulation system parameters corresponding to the air fryer oven are determined based on historical sensor data and historical image data; The target sensor data and target image data are input into a preset deep neural network model for processing to generate a prediction result of the state of the food in the air fryer. The preset deep neural network model is constructed by a timestamp-aware encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer. The current plasma power and the current air circulation system parameters are adjusted based on the state prediction results.

[0006] In one possible implementation, the target sensor data and the target image data are input into a preset deep neural network model for processing to generate a prediction result of the state of the food in the air fryer, including: Inputting the target sensor data and the target image data as input data into a preset deep neural network model, processing the input data through a timestamp-aware encoder of the preset deep neural network model to generate an image frame feature sequence; The image frame feature sequence is processed through the Query encoder to generate a processing result; The processing results are processed by the sensor codec to generate fused data feature representation and overall image feature representation; The fused data feature representation and the overall image feature representation are processed through the large language model layer to generate the status prediction results of the ingredients in the air fryer.

[0007] In one possible implementation, processing input data using a timestamp-aware encoder of a preset deep neural network model to generate an image frame feature sequence includes: The target image data in the input data is divided and processed by a timestamp-aware encoder to obtain multiple image blocks; For each image block, convolution processing is performed on the image block to obtain the initial visual vector; Add position code and timestamp code to the initial visual vector in sequence to obtain the current visual vector; Each current visual vector is spliced together to generate an image frame feature sequence.

[0008] In a possible implementation, processing the image frame feature sequence by a query encoder to generate a processing result includes: The query vector output by the previous layer is self-interacted through the self-attention layer in the Query encoder to generate the query matrix, key matrix and value matrix; Based on the query matrix, key matrix and value matrix, the self-attention output result is calculated; The cross-attention layer in the query encoder calculates the self-attention output and the image frame feature sequence to obtain the cross-attention output; The cross-attention output result is input into the feedforward network layer in the Query encoder for dimensionality transformation and nonlinear activation processing to generate the processing result.

[0009] In one possible implementation, the process of generating the fusion data feature representation includes: Perform self-attention encoding on the target sensor data in the input data through the sensor encoder-decoder to obtain the self-attention encoding result; The self-attention encoding result is used as the new query matrix, and the processing result is used as the new key matrix and the new value matrix, and the new cross-attention output result is calculated based on the new query matrix, the new key matrix and the new value matrix; The new cross-attention output result is input into the feedforward network layer in the Query encoder for dimension transformation and nonlinear activation processing to generate a fused data feature representation.

[0010] In one possible implementation, the process of generating the overall image feature representation includes: Using a preset window to segment the target image data in the input data to obtain multiple window image data; For each window image data, encoding is performed on the window image data by using a timestamp perceptual encoder to obtain a perceptual coding result; After averaging the perceptual coding results, the averaged result is input into the Query encoder for processing to generate the overall image feature representation.

[0011] In one possible implementation, the fused data feature representation and the overall image feature representation are processed by a large language model layer to generate a prediction result of the state of the ingredients in the air fryer, including: Perform linear transformation on the fusion data feature representation and the overall image feature representation to generate transformation results; The transformation results are input into the large language model layer for calculation to generate hidden states; The hidden state is processed to generate the state prediction results of the ingredients in the air fryer.

[0012] In a second aspect of the present application, a parameter control device for an air fryer oven is provided, the device comprising: an acquisition module for acquiring target sensor data and target image data collected in real time from the air fryer; wherein the current plasma power and current air circulation system parameters corresponding to the air fryer are determined based on the historical sensor data and historical image data; A generation module is used to input the target sensor data and target image data into a preset deep neural network model for processing to generate a prediction result of the state of the food in the air fryer; wherein the preset deep neural network model is constructed by a timestamp-aware encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer; The adjustment module is used to adjust the current plasma power and the current air circulation system parameters based on the state prediction result.

[0013] In a possible implementation, the generation module is specifically configured to: Inputting the target sensor data and the target image data as input data into a preset deep neural network model, processing the input data through a timestamp-aware encoder of the preset deep neural network model to generate an image frame feature sequence; The image frame feature sequence is processed through the Query encoder to generate a processing result; The processing results are processed by the sensor codec to generate fused data feature representation and overall image feature representation; The fused data feature representation and the overall image feature representation are processed through the large language model layer to generate the status prediction results of the ingredients in the air fryer.

[0014] In a possible implementation, the generation module is further configured to: The target image data in the input data is divided and processed by a timestamp-aware encoder to obtain multiple image blocks; For each image block, convolution processing is performed on the image block to obtain the initial visual vector; Add position code and timestamp code to the initial visual vector in sequence to obtain the current visual vector; Each current visual vector is spliced together to generate an image frame feature sequence.

[0015] In a possible implementation, the generation module is further configured to: The query vector output by the previous layer is self-interacted through the self-attention layer in the Query encoder to generate the query matrix, key matrix and value matrix; Based on the query matrix, key matrix and value matrix, the self-attention output result is calculated; The cross-attention layer in the query encoder calculates the self-attention output and the image frame feature sequence to obtain the cross-attention output; The cross-attention output result is input into the feedforward network layer in the Query encoder for dimensionality transformation and nonlinear activation processing to generate the processing result.

[0016] In a possible implementation, the generation module is further configured to: Perform self-attention encoding on the target sensor data in the input data through the sensor encoder-decoder to obtain the self-attention encoding result; The self-attention encoding result is used as the new query matrix, and the processing result is used as the new key matrix and the new value matrix, and the new cross-attention output result is calculated based on the new query matrix, the new key matrix and the new value matrix; The new cross-attention output result is input into the feedforward network layer in the Query encoder for dimension transformation and nonlinear activation processing to generate a fused data feature representation.

[0017] In a possible implementation, the generation module is further configured to: Using a preset window to segment the target image data in the input data to obtain multiple window image data; For each window image data, encoding is performed on the window image data by using a timestamp perceptual encoder to obtain a perceptual coding result; After averaging the perceptual coding results, the averaged result is input into the Query encoder for processing to generate the overall image feature representation.

[0018] In a possible implementation, the generation module is further configured to: Perform linear transformation on the fusion data feature representation and the overall image feature representation to generate transformation results; The transformation results are input into the large language model layer for calculation to generate hidden states; The hidden state is processed to generate the state prediction results of the ingredients in the air fryer.

[0019] In a third aspect, the present invention provides an apparatus for plasma heating, comprising: Controller; a memory for storing instructions executable by the controller; Wherein, the controller is configured to execute the instructions to implement any of the above-mentioned parameter control methods for the air fryer oven.

[0020] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement any of the above-mentioned parameter control methods for an air fryer oven.

[0021] The embodiments of the present application have the following beneficial effects: The embodiments of the present application provide a parameter control method, device, equipment, and medium for an air fryer oven. The method includes: obtaining target sensor data and target image data collected in real time from the air fryer oven, inputting the target sensor data and target image data into a preset deep neural network model for processing, and generating a state prediction result of the food in the air fryer oven, wherein the preset deep neural network model is constructed by a timestamp perception encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer, and adjusting the current plasma power and current air circulation system parameters based on the state prediction result. This solution uses sensors to collect multi-dimensional data during the cooking process, combines it with a preset deep neural network for real-time analysis and prediction, and automatically adjusts cooking parameters such as the current plasma power and the current air circulation system parameters, thereby improving the heating efficiency and intelligence of the air fryer oven. In addition, the preset deep neural network monitors the cooking status of food in real time, thereby further improving the accuracy of the cooking process, avoiding overcooked or undercooked food, and enhancing the user experience. Moreover, by constructing a preset deep neural network model consisting of a timestamp-aware encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer, the collected data can be accurately processed, further improving the accuracy of the parameter control of the air fryer oven. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart of the steps of a parameter control method for an air fryer provided in an embodiment of the present application; Figure 2 A flowchart of the steps for generating a prediction result of the state of food in an air fryer provided in an embodiment of the present application; Figure 3 A flowchart of the steps for generating an image frame feature sequence provided in an embodiment of the present application; Figure 4 A flowchart of the steps for generating a processing result provided in an embodiment of the present application; Figure 5 A flowchart of the steps for generating a fusion data feature representation provided in an embodiment of the present application; Figure 6 A flowchart of the steps for generating a fusion data feature representation provided in an embodiment of the present application; Figure 7 A flowchart of the steps for generating a state prediction result provided in an embodiment of the present application; Figure 8 This is a structural block diagram of a parameter control device for an air fryer provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0024] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, unless otherwise specified, "multiple" means two or more. In addition, the use of "based on" or "according to" means openness and inclusiveness, because the process, steps, calculations or other actions "based on" or "according to" one or more of the conditions or values may be based on additional conditions or values beyond the stated in practice.

[0025] Figure 1 This is a flow chart of the steps of a parameter control method of an air fryer provided in an embodiment of the present application. Figure 1 As shown, the method includes the following steps: Step 102: Acquire target sensor data and target image data collected in real time from the air fryer oven.

[0026] Among them, the various sensors built into the air fryer can be used to obtain cooking process data, including initial sensor data and the initial image data ,wherein, the initial sensor data may include but is not limited to various types of data such as temperature, humidity, and smoke concentration.

[0027] In some optional embodiments, the collected initial sensor data can be normalized to the interval [0, 1] to obtain the target sensor data , .in, , and are the minimum and maximum values of the initial sensor data during the investigation period.

[0028] In some optional embodiments, the collected initial image data may be normalized to the interval [0, 1] to obtain the target image data. .in, , 255 represents the maximum pixel value.

[0029] Step 104: Input the target sensor data and the target image data into a preset deep neural network model for processing to generate a state prediction result of the food in the air fryer.

[0030] Among them, the preset deep neural network model is constructed by timestamp perception encoder, query encoder, sensor encoder-decoder, sliding window Q-Former and large language model layer.

[0031] After obtaining the target sensor data and target image data, the target sensor data and target image data can be input into a preset deep neural network model for processing, thereby generating a state prediction result of the ingredients in the air fryer oven.

[0032] In some optional embodiments, such as Figure 2 As shown, Figure 2 A flowchart of the steps for generating a prediction result of the state of food in an air fryer provided in an embodiment of the present application includes: Step 202: Input the target sensor data and the target image data as input data into a preset deep neural network model, process the input data through the timestamp perception encoder of the preset deep neural network model, and generate an image frame feature sequence.

[0033] The sensor inside the cooking cavity captures image frames arranged in time sequence. These frames have specific timestamps that indicate their time interval or order. The timestamp-aware encoder generates a feature vector sequence by combining the spatial information and timestamp information of the target image data, providing visual features containing time sequence context to the subsequent encoder.

[0034] In some optional embodiments, such as Figure 3 As shown, Figure 3 A flowchart of the steps for generating an image frame feature sequence provided in an embodiment of the present application includes: Step 302: The target image data in the input data is divided and processed by a timestamp-aware encoder to obtain a plurality of image blocks.

[0035] Step 304: Perform convolution processing on each image block to obtain an initial visual vector.

[0036] Step 306: Add position code and time stamp code to the initial visual vector in sequence to obtain the current visual vector.

[0037] Step 308: perform splicing processing on each current visual vector to generate an image frame feature sequence.

[0038] The target image data in the input data is first divided into several small image blocks. Then, the image blocks can be convolved. Each image block is mapped to a feature vector through ViT convolution projection, thereby obtaining a set of initial visual vectors. This process can be recorded as ,Conv represents the convolution operation; is the number of image blocks after the image is divided.

[0039] Next, the initial visual vector Add position codes one by one , to retain the position information of each image block in the image, and obtain the visual vector after adding position encoding ,in, , It is The position encoding of the initial visual vector.

[0040] Timestamp t Encoded as a learnable timestamp vector , the timestamp vector and each visual vector after adding position encoding After addition and fusion, the current visual vector is obtained .in, , It is the first The current visual vector.

[0041] Finally, each current visual vector can be spliced to generate an image frame feature sequence .in, .

[0042] Step 204: Process the image frame feature sequence through the Query encoder to generate a processing result.

[0043] After the image frame feature sequence is obtained, the image frame feature sequence can be processed by a Query encoder to generate a processing result.

[0044] In some optional embodiments, such as Figure 4 As shown, Figure 4 A flowchart of steps for generating a processing result provided in an embodiment of the present application includes: Step 402: The query vector output by the previous layer is self-interacted through the self-attention layer in the Query encoder to generate a query matrix, a key matrix, and a value matrix.

[0045] Step 404: Based on the query matrix, the key matrix, and the value matrix, the self-attention output result is calculated.

[0046] Step 406: Calculate the self-attention output result and the image frame feature sequence through the cross-attention layer in the Query encoder to obtain the cross-attention output result.

[0047] Step 408: Input the cross-attention output result into the feedforward network layer in the Query encoder for dimensionality transformation and nonlinear activation processing to generate a processing result.

[0048] The query encoder uses a multi-layer Transformer structure. The Transformer structure is a deep learning model architecture that contains three sub-layers: self-attention layer, cross-attention layer, and feed-forward network layer. is initialized to A learnable vector is generated and updated as it is passed through multiple layers of Transformers. The following describes the computational process of a single-layer Transformer in detail.

[0049] In the In the layer, the query vector output by the previous layer It is fed into the self-attention mechanism for self-interaction. In this process, the query vector and its own attention mechanism are calculated to generate the query matrix through linear transformation. , key matrix Sum Matrix .in, , , , 、 and is a learnable projection matrix.

[0050] Then, the self-attention output can be calculated based on the query matrix, key matrix and value matrix , , is the dimension of the key matrix.

[0051] Thus, the self-attention output result and the image frame feature sequence can be calculated through the cross attention layer in the Query encoder to obtain the cross attention output result. As the query vector, the image frame feature sequence As the key matrix and value matrix, calculate the cross attention, that is, , , , and the cross attention output can be obtained by calculating the attention distribution and weighting the value matrix , .

[0052] Finally, the cross-attention output can be input into the feedforward network layer in the Query encoder for dimension transformation and nonlinear activation processing to generate the processing result. FFN represents the feed-forward network layer. After L layers of Transformer modules, the final query vector is represented for: , which is the processing result of the feedforward network layer.

[0053] Step 206: Process the processing result through the sensor codec to generate a fusion data feature representation and an overall image feature representation.

[0054] Among them, the sensor codec consists of an encoder and a decoder. Similarly, the target sensor data It can also combine timestamp information with the processing results from the Query encoder Attention interaction.

[0055] In some optional embodiments, such as Figure 5 As shown, Figure 5 A flowchart of the steps for generating a fusion data feature representation provided in an embodiment of the present application includes: Step 502: Perform self-attention encoding on the target sensor data in the input data through the sensor encoder-decoder to obtain a self-attention encoding result.

[0056] Step 504: Use the self-attention encoding result as a new query matrix, and use the processing result as a new key matrix and a new value matrix, and calculate a new cross-attention output result based on the new query matrix, the new key matrix and the new value matrix.

[0057] Step 506: Input the new cross-attention output result into the feedforward network layer in the Query encoder for dimensionality transformation and nonlinear activation processing to generate a fused data feature representation.

[0058] Among them, first of all, Target sensor data of the layer Perform self-attention encoding to obtain self-attention encoding results .in, , Represents the self-attention operation function, which can help capture the interactive relationship between various features of the target sensor data.

[0059] Then, the self-attention encoding result can be as a new query matrix and process the results As the new key matrix and the new value matrix, the new cross attention output result is calculated based on the new query matrix, the new key matrix and the new value matrix .in, , Represents the cross-operation function. Through cross-attention, the target sensor data can be fused with the target image data to obtain an updated sensor representation.

[0060] Finally, the new cross attention output result can be The feedforward network layer input to the Query encoder performs dimension transformation and nonlinear activation processing to generate fused data feature representation Among them, the new cross attention output result It is sent to the feedforward network layer for dimension transformation and nonlinear activation processing to obtain the Output of the layer , Finally, after stacking multiple layers of Transformer, the fused data feature representation is obtained. .

[0061] In some optional embodiments, while the air fryer is continuously operating, the sensor captures target image data, which changes over time. To avoid information loss or over-compression due to long-term image input, a sliding window strategy can be used to segment multiple frames of target image data. This method can perform relatively fine feature extraction on the target image data for each time period.

[0062] like Figure 6 As shown, Figure 6 A flowchart of the steps for generating a fusion data feature representation provided in an embodiment of the present application includes: Step 602: Use a preset window to segment the target image data in the input data to obtain a plurality of window image data.

[0063] Step 604: For each window image data, the window image data is encoded by a timestamp perceptual encoder to obtain a perceptual coding result.

[0064] Step 606: After averaging the perceptual coding results, the averaged result is input into the Query encoder for processing to generate an overall image feature representation.

[0065] The target image data in the input data is segmented using a preset window to obtain multiple window image data. Include The target image data of the frame, where .

[0066] For each window image data, the window image data can be encoded by a timestamp perceptual encoder to obtain a perceptual encoding result. The timestamp perceptual encoder takes the timing information of each frame into consideration through the timestamp information to obtain the representation of each frame. , which is the perceptual coding result.

[0067] Then, after averaging the perceptual coding results, the average processing result is input into the Query encoder for processing to generate the overall image feature representation. Among them, after inputting the average processing result into the Query encoder for processing, the query representation of this window can be obtained. .

[0068] in, , Represents the average operation. For queries in different windows , and concatenate them. In this way, the feature information of multiple windows is effectively summarized. Finally, FFN is used to adjust the feature dimension required by the large language model layer, that is, the overall image feature representation is obtained. .in, , Represents a splicing operation.

[0069] Step 208: Process the fused data feature representation and the overall image feature representation through the large language model layer to generate a prediction result of the state of the ingredients in the air fryer.

[0070] Among them, in some optional embodiments, such as Figure 7 As shown, Figure 7 A flowchart of the steps for generating a state prediction result provided in an embodiment of the present application includes: Step 702: Perform linear transformation processing on the fused data feature representation and the overall image feature representation respectively to generate transformation results.

[0071] Step 704: Input the transformation result into the large language model layer for calculation to generate a hidden state.

[0072] Step 706: Process the hidden state to generate a state prediction result of the food in the air fryer.

[0073] Among them, the large language model layer can come from the overall image feature representation of the sliding window Q-Former Mapped to the large language model word vector space. Specifically, a linear transformation can be used to transform Convert to the internal embedding space of the large language model to generate the transformation result corresponding to the overall image feature representation .in, , It is the internal embedding dimension of the large language model. This mapping converts image information into a vector representation compatible with the language model.

[0074] Fusion data feature representation It also needs to be mapped to the same dimension as the image representation to obtain the transformation result corresponding to the fused data feature representation.

[0075] Then, all the transformation results can be concatenated to form the input sequence of the large language model Then it is input into the large language model layer for forward calculation. The large language model uses a multi-layer self-attention mechanism to propagate and process information and finally outputs the hidden state. , which contains a comprehensive representation of all input information.

[0076] Finally, the hidden state can be processed to generate the state prediction results of the ingredients in the air fryer. Optionally, for generation tasks, such as the next wind speed setting or ingredient status, a language decoder can be added on top of the large language model. The language decoder can be used to predict the state of the ingredients in the air fryer according to the hidden state. Output instructions or decisions in text format.

[0077] For regression tasks, such as predicting remaining cooking time or equipment control parameters, the output hidden state of the large language model Converted to numerical output after linear mapping ,in, , MLP represents multi-layer perceptron; Represents the result of the regression task.

[0078] Step 106: Adjust the current plasma power and current air circulation system parameters based on the state prediction result.

[0079] The current plasma power and current air circulation system parameters corresponding to the air fryer oven are determined based on historical sensor data and historical image data.

[0080] Specifically, during the cooking process, the sensor will continuously collect real-time data and continuously input this data into the above-mentioned preset deep neural network model. This model will predict the real-time status of the ingredients in real time and obtain the status prediction results. , so that the results can be predicted according to the state of the ingredients and whether the ingredients have reached the expected degree of doneness.

[0081] If the preset deep neural network model predicts that the temperature distribution, food color, or other parameters deviate from the expected state, the model recalculates the error between the predicted food state and the target state, and adjusts the current plasma power and air circulation system parameters to ensure that the system parameters always meet the actual needs of the food, thereby achieving the best cooking results. Through this closed-loop feedback mechanism, the cooking process can be optimized in real time to ensure that the food is heated to its optimal state.

[0082] To optimize the oven's heating effect and efficiency during cooking, precise control of the plasma power and air circulation system parameters is required. Specifically, state predictions can be used to determine the plasma generator's power requirements. The voltage and current are then adjusted to meet these requirements. This process ensures efficient heating, avoids overheating or underheating, and ensures uniform heating across the food surface.

[0083] Additionally, depending on the power requirements of the air fryer, you can adjust the current air circulation system parameters, specifically the fan speed and air flow, to optimize air circulation within the oven. By adjusting these parameters, you can ensure even distribution of hot air and enhance heat exchange efficiency across the food surface, achieving optimal heating results.

[0084] In some optional embodiments, after cooking is completed, key indicators for judging the state of the food can be collected and analyzed through error. To evaluate the cooking effect, , Indicates the target state of the food. According to the comparison results, if there is a significant deviation, the current model parameters Fine-tuning can be performed based on the error, and the optimization process can be achieved through the gradient descent method to obtain new model parameters. ,in, , represents the learning rate; Represents the gradient of the error function with respect to the model parameters.

[0085] The fine-tuned model is used to adjust the control strategy. This optimized control strategy influences the control variables during the cooking process, thereby improving subsequent cooking quality. Through this self-learning approach, the entire process involves calculating errors, adjusting the model, and continuously optimizing the control strategy, ultimately achieving improved cooking quality.

[0086] The present application provides a parameter control method for an air fryer oven, the method comprising: obtaining target sensor data and target image data collected in real time from the air fryer oven, inputting the target sensor data and target image data into a preset deep neural network model for processing, and generating a state prediction result of food in the air fryer oven, wherein the preset deep neural network model is constructed by a timestamp perception encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer, and adjusting the current plasma power and current air circulation system parameters based on the state prediction result. This solution uses sensors to collect multi-dimensional data during the cooking process, combines it with a preset deep neural network for real-time analysis and prediction, and automatically adjusts cooking parameters such as the current plasma power and the current air circulation system parameters, thereby improving the heating efficiency and intelligence of the air fryer oven. In addition, the preset deep neural network monitors the cooking status of food in real time, thereby further improving the accuracy of the cooking process, avoiding overcooked or undercooked food, and enhancing the user experience. Moreover, by constructing a preset deep neural network model consisting of a timestamp-aware encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer, the collected data can be accurately processed, further improving the accuracy of the parameter control of the air fryer oven.

[0087] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0088] Figure 8 This is a structural block diagram of a parameter control device for an air fryer provided in an embodiment of the present application.

[0089] like Figure 8 As shown, the parameter control device 800 of the air fryer oven includes: The acquisition module 802 is used to obtain target sensor data and target image data collected in real time from the air fryer oven; wherein the current plasma power and current air circulation system parameters corresponding to the air fryer oven are determined based on the historical sensor data and historical image data.

[0090] Generation module 804 is used to input the target sensor data and target image data into a preset deep neural network model for processing to generate a state prediction result of the food in the air fryer; wherein the preset deep neural network model is constructed by a timestamp perception encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former and a large language model layer.

[0091] The adjustment module 806 is used to adjust the current plasma power and the current air circulation system parameters based on the state prediction result.

[0092] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method and will not be elaborated on here. The various modules in the above graph-based retrieval enhancement generation apparatus can be implemented in whole or in part by software, hardware, or a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations of the above modules.

[0093] In one embodiment of the present application, there is provided a device for plasma heating, comprising: Controller; a memory for storing instructions executable by the controller; Wherein, the controller is configured to execute the instructions to implement any of the above-mentioned parameter control methods for the air fryer oven.

[0094] The device using plasma heating provided in the embodiment of the present application has similar implementation principles and technical effects as those of the above-mentioned method embodiment, and will not be described in detail here.

[0095] In one embodiment of the present application, a computer-readable storage medium is provided, wherein the storage medium stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the parameter control method of the air fryer oven in the embodiment of the present invention.

[0096] The computer-readable storage medium provided in this embodiment has similar implementation principles and technical effects to those of the above-mentioned method embodiment, and will not be described in detail here.

[0097] In the above embodiments, all or part of the embodiments can be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0098] It is easy to understand that those skilled in the art can combine, split, reorganize, etc. the embodiments of the present application based on the several embodiments provided in the present application to obtain other embodiments, and these embodiments do not exceed the scope of protection of the present application.

[0099] The above specific implementation methods further explain in detail the purpose, technical solutions and beneficial effects of the embodiments of the present application. It should be understood that the above are only specific implementation methods of the embodiments of the present application and are not intended to limit the scope of protection of the embodiments of the present application. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application should be included in the scope of protection of the embodiments of the present application.

Claims

1. A parameter control method for an air fryer oven, characterized in that: The method comprises: Obtaining target sensor data and target image data collected in real time from the air fryer oven; wherein current plasma power and current air circulation system parameters corresponding to the air fryer oven are determined based on historical sensor data and historical image data; Inputting the target sensor data and the target image data into a preset deep neural network model for processing to generate a prediction result of the state of the food in the air fryer; wherein the preset deep neural network model is constructed by a timestamp-aware encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer; The current plasma power and the current air circulation system parameters are adjusted based on the state prediction result.

2. The method according to claim 1, characterized in that The step of inputting the target sensor data and the target image data into a preset deep neural network model for processing to generate a state prediction result of the food in the air fryer includes: Inputting the target sensor data and the target image data as input data into a preset deep neural network model, processing the input data through a timestamp-aware encoder of the preset deep neural network model to generate an image frame feature sequence; Processing the image frame feature sequence by the Query encoder to generate a processing result; Processing the processing result by the sensor codec to generate a fused data feature representation and an overall image feature representation; The fused data feature representation and the overall image feature representation are processed by the large language model layer to generate a state prediction result of the ingredients in the air fryer.

3. The method according to claim 2, characterized in that The processing of the input data by the timestamp-aware encoder of the preset deep neural network model to generate an image frame feature sequence includes: dividing the target image data in the input data by the timestamp-aware encoder to obtain a plurality of image blocks; For each of the image blocks, performing convolution processing on the image block to obtain an initial visual vector; Adding position code and timestamp code to the initial visual vector in sequence to obtain a current visual vector; The current visual vectors are spliced to generate the image frame feature sequence.

4. The method according to claim 2, characterized in that The processing of the image frame feature sequence by the Query encoder to generate a processing result includes: The query vector output by the previous layer is self-interacted by the self-attention layer in the Query encoder to generate a query matrix, a key matrix and a value matrix; Calculating a self-attention output result based on the query matrix, the key matrix, and the value matrix; Calculating the self-attention output result and the image frame feature sequence through the cross-attention layer in the Query encoder to obtain a cross-attention output result; The cross-attention output result is input into the feedforward network layer in the Query encoder for dimensional transformation and nonlinear activation processing to generate the processing result.

5. The method according to claim 2, characterized in that The generation process of the fused data feature representation includes: Performing self-attention encoding on the target sensor data in the input data by the sensor encoder-decoder to obtain a self-attention encoding result; Using the self-attention encoding result as a new query matrix, and using the processing result as a new key matrix and a new value matrix, and calculating a new cross-attention output result based on the new query matrix, the new key matrix and the new value matrix; The new cross-attention output result is input into the feedforward network layer in the Query encoder for dimensionality transformation and nonlinear activation processing to generate the fused data feature representation.

6. The method according to claim 2, characterized in that The generation process of the overall image feature representation includes: Using a preset window to segment the target image data in the input data to obtain a plurality of window image data; For each of the window image data, encoding is performed on the window image data by using a timestamp perceptual encoder to obtain a perceptual coding result; After averaging the perceptual coding results, the averaged result is input into the Query encoder for processing to generate the overall image feature representation.

7. The method according to claim 2, characterized in that The processing of the fused data feature representation and the overall image feature representation by the large language model layer to generate a prediction result of the state of the food in the air fryer includes: Performing linear transformation processing on the fused data feature representation and the overall image feature representation respectively to generate transformation results; Inputting the transformation result into the large language model layer for calculation to generate the hidden state; The hidden state is processed to generate a state prediction result of the food in the air fryer.

8. A parameter control device for an air fryer oven, characterized in that: The device comprises: an acquisition module for acquiring target sensor data and target image data collected in real time from the air fryer; wherein the current plasma power and current air circulation system parameters corresponding to the air fryer are determined based on the historical sensor data and historical image data; a generation module, configured to input the target sensor data and the target image data into a preset deep neural network model for processing, and generate a prediction result of the state of the food in the air fryer; wherein the preset deep neural network model is constructed by a timestamp-aware encoder, a query encoder, a sensor encoder-decoder, a sliding window Q-Former, and a large language model layer; An adjustment module is used to adjust the current plasma power and the current air circulation system parameters based on the state prediction result.

9. A device using plasma heating, characterized in that include: Controller; a memory for storing instructions executable by the controller; The controller is configured to execute the instructions to implement the parameter control method of the air fryer according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the parameter control method of the air fryer according to any one of claims 1 to 7.

Images