Cooking flame form real-time monitoring and abnormity alarm system based on computer vision

Abstract

The invention relates to the technical field of cooking flame safety monitoring, and particularly discloses a cooking flame form real-time monitoring and abnormity alarm system based on computer vision. A flame stability evaluation module extracts a flicker difference frequency characteristic value, a combustion sufficiency evaluation module extracts a spectral distribution difference characteristic value, a flame combustion state evaluation module fuses the characteristics and inputs the fused characteristics into a random forest model, and a continuous flame state score is output and is judged whether to be abnormal or not; if the combustion state is abnormal, the abnormal alarm module starts a grading early warning mechanism which comprises sound-light alarm, user pushing, gas cutting off, remote notification and other operations, all-weather and high-precision monitoring of the combustion state of the gas flame in the kitchen is achieved, the accuracy and response speed of flame abnormal recognition are improved, and the working efficiency is improved. The safety protection effect and the application and popularization value are good.

Description

Technical Field

[0001] The present invention relates to the technical field of cooking flame safety monitoring, and particularly to a real-time monitoring and abnormal alarm system for cooking flame morphology based on computer vision. Background Art

[0002] With the rapid development of intelligent kitchen appliances, the safety and intelligent level of gas cookers have attracted increasing attention. The stability and sufficiency of the flame combustion state not only directly affect the cooking efficiency, but also are closely related to the safety of kitchen gas use. Traditional gas cookers mainly rely on manual observation or simple temperature and pressure sensors for abnormal judgment, and it is difficult to achieve real-time and accurate monitoring of the changes in the flame morphology. In recent years, flame detection technologies based on computer vision have gradually been applied to intelligent kitchen systems, and through image processing and pattern recognition methods, functions such as flame presence detection and color recognition have been initially realized. However, existing methods still have problems such as low recognition accuracy and response lag when facing complex environmental interferences (such as light changes and oil fume occlusion) and multiple types of abnormal flames (such as yellow flames, flashback, and blowout), and it is difficult to meet the application requirements in high-reliability scenarios.

[0003] The existing technologies have the following deficiencies: Most of the existing flame monitoring technologies only rely on a single feature (such as color, shape, or brightness) to judge the flame state, lacking the ability of multi-dimensional modeling of the dynamic change process of the flame, and it is difficult to accurately distinguish normal fluctuations from real abnormal situations. Most of the solutions do not deeply analyze the periodic stability of flame combustion and the characteristics of heat radiation distribution, resulting in easy misjudgment or missed judgment under complex working conditions. At the same time, existing systems generally lack an effective early warning grading mechanism and cannot implement differential alarm and intervention measures according to the degree of abnormality, which limits their actual application effects in household and commercial kitchens. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time monitoring and abnormal alarm system for cooking flame morphology based on computer vision to solve the problems in the above background.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A real-time monitoring and abnormal alarm system for cooking flame morphology based on computer vision, comprising: A data acquisition module, which acquires the flame image data of the gas cooker during the cooking process; A flame stability evaluation module, which performs time-domain differential processing on the flame image data and calculates the flame flicker differential frequency eigenvalue according to the degree of pixel gray change between adjacent frames for evaluating the periodic stability of the combustion process; Flame combustion sufficiency evaluation module, which evaluates the combustion sufficiency by analyzing the spectral energy distribution of flame image data in different bands and calculating the spectral distribution difference eigenvalue; Flame combustion state evaluation module, which inputs the flame flicker differential frequency eigenvalue and the spectral distribution difference eigenvalue into the flame state judgment model, and judges the combustion state of the flame according to the model output, including normal combustion state and abnormal combustion state; Abnormal alarm module, which performs early warning processing based on the abnormal combustion state of the flame.

[0006] As a further solution of the present invention: evaluating the periodic stability of the combustion process specifically includes: Real-time collect the flame image data of the gas stove during the cooking process, perform time-domain differential processing on the flame image data, calculate the flame flicker differential frequency eigenvalue according to the pixel gray difference between adjacent frames, and judge whether the flame flicker differential frequency eigenvalue is greater than or equal to the preset threshold. If so, the flame during the combustion process is unstable; if not, the flame during the combustion process is stable.

[0007] As a further solution of the present invention: the acquisition process of the flame flicker differential frequency eigenvalue is as follows: Collect a continuous image sequence of the gas stove flame during the cooking process, perform gray-scale processing on the images, and extract the time-gray change information of the flame area; Within the preset flame region of interest, calculate the pixel gray difference between adjacent frames to obtain a differential image matrix; Perform average processing on each frame of the differential image within the flame region to obtain a one-dimensional time series signal for subsequent frequency-domain analysis; Use the fast Fourier transform algorithm to perform spectral analysis on the obtained one-dimensional time series signal, and extract the main frequency components of the flame flicker; Within the set normal flame flicker frequency range, calculate the total energy within the corresponding frequency band, and calculate the average energy within the corresponding frequency band through the mean calculation expression. Calculate the ratio of the average energy to the total energy to obtain the flame flicker differential frequency eigenvalue.

[0008] As a further solution of the present invention: evaluating the combustion sufficiency specifically includes: Real-time collect the flame image data of the gas stove during the cooking process, calculate the spectral distribution difference eigenvalue by analyzing the abnormal degree of the spectral energy distribution of the flame image data in different bands, and judge whether the spectral distribution difference eigenvalue is greater than or equal to the preset threshold. If so, the flame during the combustion process is insufficient; if not, the flame during the combustion process is sufficient.

[0009] As a further solution of the present invention: the process of obtaining the spectral distribution difference characteristic value is: The flame image data is collected by multi-spectral imaging technology, and the average spectral energy of each band is calculated within the selected band range to obtain the average spectral energy sequence of each band; Discrete wavelet transform is applied to the average spectral energy sequence of each band to obtain wavelet coefficients at different scales, and the spectral distribution difference characteristic values ​​are calculated based on the wavelet coefficients.

[0010] As a further solution of the present invention: the construction process of the flame state judgment model is: The flame flicker differential frequency eigenvalue and the spectral distribution difference eigenvalue are constructed into a comprehensive eigenvector as the input of the flame state judgment model to minimize the error between the predicted flame state score and the actual flame state score as the training target. The flame state score is output according to the trained flame state judgment model, and the flame state judgment model is a random forest model.

[0011] As a further solution of the present invention: the training process of the flame state judgment model is: The random forest algorithm is used to establish a flame state judgment model. The model consists of multiple decision trees. Each decision tree is trained based on randomly selected sample subsets and feature subsets to enhance the generalization ability of the model. During the training process, the actual flame state scores manually annotated are used as labels, the flame state scores predicted and output by the model are used as targets, and the minimization of the mean square error is used as the optimization target. After the model training is completed, the comprehensive feature vector extracted in real time is input into the trained random forest model to output the flame state score.

[0012] As a further solution of the present invention: judging the combustion state of the flame according to the model output specifically includes: It is determined whether the flame state score of the combustion process is greater than or equal to a preset threshold. If so, it is recorded as a normal combustion state; if not, it is recorded as an abnormal combustion state.

[0013] As a further solution of the present invention: the early warning process based on the abnormal combustion state of the flame specifically includes: The abnormal burning time of the flame is obtained, and the abnormal burning time of the flame is multiplied by the flame state score to obtain the grading score. If the grading score is less than or equal to the preset first-level abnormal score, it is recorded as a first-level abnormality, triggering the on-site sound and light alarm and simultaneously pushing the early warning information to the user terminal. If the grading score is greater than or equal to the first-level abnormality score, it is recorded as a second-level abnormality. In addition to executing the first-level abnormal operation, the gas valve automatic closing device is linked to control to cut off the fuel supply, and an alarm notification is sent to the remote management platform.

[0014] Beneficial effects of the present invention: (1) The present invention constructs a composite feature extraction mechanism for the recognition of the combustion state of cooking flames by integrating multimodal visual features in the time domain and the spectral domain. Specifically, in terms of flame stability assessment, based on the time domain difference processing and fast Fourier transform analysis of continuous image frames, the flame flicker differential frequency eigenvalue is extracted. This eigenvalue can reflect the periodic fluctuation characteristics of the flame brightness over time, and is highly sensitive to phenomena such as violent flame shaking and irregular flickering under abnormal combustion conditions; in terms of combustion sufficiency assessment, multispectral imaging technology is introduced to collect the radiation information of the flame in multiple bands, and the spectral distribution difference eigenvalue is extracted in combination with discrete wavelet transform, so as to achieve accurate quantification of the abnormal distribution of flame thermal radiation, and effectively identify non-ideal combustion states such as yellow flame and flashback caused by insufficient air mixing or incomplete combustion of fuel. By constructing the above two types of features into a comprehensive feature vector and inputting it into the flame state judgment model trained based on the random forest algorithm, the system can still maintain good discrimination ability in complex kitchen environments (such as strong light changes, oil smoke interference, and background clutter), significantly improving the accuracy, robustness and adaptability of flame state recognition. This method not only breaks through the limitations of traditional single feature analysis in dynamic flame monitoring, but also provides reliable technical support for the engineering application of smart kitchen equipment in gas safety early warning.

[0015] (2) The present invention adopts a flame state judgment model based on the random forest algorithm, fully exploits the nonlinear mapping relationship between flame image features and combustion state, and significantly improves the system's modeling and prediction capabilities for complex combustion behaviors. The model uses a comprehensive feature vector composed of flame flicker differential frequency eigenvalues ​​and spectral distribution difference eigenvalues ​​as input, and integrates multiple decision trees randomly divided in sample and feature space to achieve high-precision classification and scoring output of flame combustion state. Compared with traditional linear models or single discrimination methods, random forests have stronger anti-overfitting capabilities and generalization performance, and can maintain stable discrimination effects in actual kitchen scenes with noise interference, data fluctuations and environmental uncertainties. The continuous flame state score output by the model can dynamically reflect the health of the current combustion process, and combine the preset threshold to make a state judgment; further, the abnormal alarm module performs graded evaluation based on the score decline trend, abnormal duration and its change rate, and automatically triggers a multi-level early warning response mechanism. For level 1 abnormalities, the system will perform local sound and light alarms and push alarm information to the user terminal; for level 2 and above serious abnormalities, the gas valve closing device will be linked to cut off the fuel supply to prevent the accident from expanding, and the abnormal event log will be uploaded to the cloud server, and an alarm notification will be sent to the remote management platform to form a complete safety closed-loop control chain. This technical solution realizes the intelligent management of the entire process from flame state recognition to active intervention, effectively improving the safety of fire use, timely response, and intelligent operation and maintenance in home and commercial kitchen environments. Brief Description of the Drawings

[0016] The present invention will be further described below in conjunction with the accompanying drawings.

[0017] Figure 1 It is a flow block diagram of the real-time monitoring and abnormal alarm system for the cooking flame form based on computer vision of the present invention. Detailed Embodiments

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] Please refer to Figure 1 As shown, the present invention is a real-time monitoring and abnormal alarm system for the cooking flame form based on computer vision, including: A data acquisition module, which acquires the flame image data of the gas cooker during the cooking process; A flame stability evaluation module, which performs time-domain differential processing on the flame image data, calculates the flame flicker differential frequency eigenvalue according to the degree of pixel gray change between adjacent frames, and is used to evaluate the periodic stability of the combustion process; A flame combustion sufficiency evaluation module, which calculates the spectral distribution difference eigenvalue by analyzing the spectral energy distribution of the flame image data in different bands, and is used to evaluate the combustion sufficiency; A flame combustion state evaluation module, which inputs the flame flicker differential frequency eigenvalue and the spectral distribution difference eigenvalue into the flame state judgment model, and judges the combustion state of the flame according to the model output, including the normal combustion state and the abnormal combustion state; An abnormal alarm module, which performs early warning processing based on the abnormal combustion state of the flame.

[0020] In the data acquisition module, the flame image data of the gas cooker during the cooking process is acquired, specifically including: During the cooking process, high-definition cameras and multispectral imaging technology installed in the kitchen environment are used to capture real-time flame images during the combustion of gas stoves; the cameras are deployed above the side of the stove, avoiding areas with concentrated oil fumes, and equipped with optical filtering devices to enhance the recognition accuracy of the flame area; the image acquisition frequency is no greater than 30 frames per second to ensure the integrity of the flame dynamic change information; the collected original image data is sequentially subjected to white balance correction, color space conversion, and denoising processing, and then transmitted to the subsequent analysis module for extracting flame morphological features and evaluating the combustion state.

[0021] In the flame stability evaluation module, time-domain differential processing is performed on the flame image data, and according to the degree of pixel gray-scale change between adjacent frames, the flame flicker differential frequency eigenvalue is calculated to evaluate the periodic stability of the combustion process, specifically including: Real-time collect the flame image data of the gas stove during the cooking process, perform time-domain differential processing on the flame image data, calculate the flame flicker differential frequency eigenvalue according to the pixel gray-scale difference between adjacent frames, and determine whether the flame flicker differential frequency eigenvalue is greater than or equal to the preset threshold. If so, the flame during the combustion process is unstable; if not, the flame during the combustion process is stable; The process of obtaining the flame flicker differential frequency eigenvalue is as follows: Collect a continuous image sequence of the gas stove flame during the cooking process, perform gray-scale processing on the images, and extract the time-gray-scale change information of the flame area; Within the preset region of interest of the flame, calculate the pixel gray-scale difference between adjacent frames to obtain a differential image matrix; Perform average processing on each frame of the differential image within the flame area to obtain a one-dimensional time series signal for subsequent frequency-domain analysis; Use the fast Fourier transform algorithm to perform spectral analysis on the obtained one-dimensional time series signal to extract the main frequency components of the flame flicker; Among them, the frequency-domain energy calculation formula is: ; In the formula, represents the complex spectral coefficient corresponding to the frequency ; represents the number of frequencies, is the one-dimensional time series signal obtained from the difference between adjacent frames, represents the fast Fourier transform, represents the acquisition time; Within the set normal flame flicker frequency range, calculate the total energy within the corresponding frequency band, and calculate the average energy within the corresponding frequency band through the mean calculation expression. Calculate the ratio of the average energy to the total energy to obtain the flame flicker differential frequency eigenvalue. Among them, the calculation expression for the total energy within the corresponding frequency band is: ; In the formula, represents the maximum frequency, represents the minimum frequency, represents the total energy.

[0022] In the flame combustion sufficiency evaluation module, by analyzing the spectral energy distribution of the flame image data in different bands, the spectral distribution difference eigenvalue is calculated to evaluate the combustion sufficiency, specifically including: Real-time collect the flame image data of the gas stove during the cooking process. By analyzing the abnormal degree of the spectral energy distribution of the flame image data in different bands, calculate the spectral distribution difference eigenvalue, and judge whether the spectral distribution difference eigenvalue is greater than or equal to the preset threshold. If so, the flame during the combustion process is insufficient; if not, the flame during the combustion process is sufficient.

[0023] The process of obtaining the spectral distribution difference eigenvalue is as follows: Collect the flame image data through multispectral imaging technology. In the selected band range, calculate the average spectral energy of each band to obtain the average spectral energy sequence of each band. Among them, the calculation expression of the average spectral energy is: ; In the formula, represents the average spectral energy, represents the position coordinates of the image, represents the pixel gray value of the th band, represents the preset area, represents the number of bands; ; In the formula, represents the spectral distribution difference eigenvalue, represents the number of discrete wavelet decomposition layers, represents the position of the discrete wavelet decomposition, represents the wavelet coefficient.

[0024] In the flame combustion state evaluation module, input the flame flicker differential frequency eigenvalue and the spectral distribution difference eigenvalue into the flame state judgment model. According to the model output, judge the combustion state of the flame, including the normal combustion state and the abnormal combustion state, specifically including: Construct the differential frequency eigenvalue of flame flicker and the spectral distribution difference eigenvalue into a comprehensive feature vector as the input of the flame state judgment model, and minimize the error between the predicted flame state score and the actual flame state score as the training objective. According to the trained flame state judgment model, output the flame state score, and the flame state judgment model is a random forest model.

[0025] It should be noted that: by extracting the differential frequency eigenvalue of flame flicker and the spectral distribution difference eigenvalue, the combustion state of the flame is quantitatively analyzed in multiple dimensions from the time domain and the spectral domain, significantly improving the accuracy and robustness of flame anomaly recognition. In the flame stability assessment, based on the fast Fourier transform, the spectral analysis of the flame grayscale time series signal is carried out, the energy distribution within the set frequency range is calculated, and the ratio of the average energy to the total energy is introduced as the differential frequency eigenvalue of flame flicker to effectively capture the abnormal trend of the periodic fluctuation of the flame; in the combustion sufficiency assessment, the multi-spectral imaging technology is used to obtain the flame radiation information in different bands, the average spectral energy sequence is constructed and combined with wavelet transform to extract the spectral distribution difference eigenvalue, so as to reflect the combustion sufficiency and thermal efficiency state of the fuel. The above method not only realizes the high-sensitivity monitoring of the dynamic changes of the flame morphology, but also enhances the system's ability to identify abnormal flame states under complex working conditions by fusing multi-modal features, providing an intelligent and real-time monitoring means for kitchen gas safety.

[0026] The random forest algorithm is used to establish the flame state judgment model. The model consists of multiple decision trees. Each decision tree is trained based on a randomly selected sample subset and feature subset to enhance the generalization ability of the model. During the training process, the actual score of the manually labeled flame state is used as the label, the flame state score predicted by the model is used as the target, and minimizing the mean square error is used as the optimization objective. After the model training is completed, the real-time extracted comprehensive feature vector is input into the trained random forest model to output the flame state score.

[0027] Judging the combustion state of the flame according to the model output specifically includes: Judge whether the flame state score during the combustion process is greater than or equal to the preset threshold. If so, it is recorded as the normal combustion state; if not, it is recorded as the abnormal combustion state.

[0028] In the abnormal alarm module, based on the abnormal combustion state of the flame, early warning processing is carried out, specifically including: The abnormal burning time of the flame is obtained, and the abnormal burning time of the flame is multiplied by the flame state score to obtain the grading score. If the grading score is less than or equal to the preset first-level abnormal score, it is recorded as a first-level abnormality, triggering the on-site sound and light alarm and simultaneously pushing the early warning information to the user terminal. If the grading score is greater than or equal to the first-level abnormality score, it is recorded as a second-level abnormality. In addition to executing the first-level abnormal operation, the gas valve automatic closing device is linked to control to cut off the fuel supply, and an alarm notification is sent to the remote management platform.

[0029] The working principle of the present invention is to realize high-precision and all-weather monitoring of the flame combustion state of kitchen gas stoves through multimodal feature fusion and intelligent analysis technology. The system includes a data acquisition module, a flame stability evaluation module, a flame combustion sufficiency evaluation module, a flame combustion state evaluation module and an abnormal alarm module. Among them, the data acquisition module obtains the flame image through a high-definition camera and a multi-spectral imaging device, and performs preprocessing to improve the image quality; the flame stability evaluation module extracts the flame flicker differential frequency eigenvalue based on time domain difference and fast Fourier transform to quantify the stability of the periodic fluctuation of the flame; the flame combustion sufficiency evaluation module uses multi-spectral imaging combined with wavelet transform to calculate the spectral distribution difference eigenvalue, so as to evaluate whether the fuel is fully burned. The above two types of features are constructed into a comprehensive feature vector and input into the flame state judgment model trained based on the random forest algorithm. The model realizes accurate recognition of the flame state by minimizing the mean square error between the predicted score and the manually labeled score, outputs a continuous combustion state score and determines the normal or abnormal state according to the threshold. The abnormal alarm module calculates the classification score based on the product of the abnormal duration and the state score, implements a multi-level early warning mechanism, and supports functions such as sound and light alarm, user push, gas cut-off and remote alarm. The present invention realizes closed-loop management from image acquisition to intelligent decision-making to safety response, significantly improving the safety and intelligence level of kitchen fire, and has good application prospects and engineering promotion value.

[0030] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0031] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. 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 by wired or wireless (such as 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 a data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0032] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context before and after.

[0033] It should be understood that in various embodiments of the present application, the order numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0034] The above has described in detail one embodiment of the present invention, but the content described is only the preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A cooking flame morphology real-time monitoring and abnormal alarm system based on computer vision, characterized in that: include: A data acquisition module, wherein the data acquisition module acquires flame image data of the gas cooker during cooking; A flame stability evaluation module, which performs time-domain differential processing on flame image data and calculates flame flicker differential frequency characteristic values ​​according to the degree of pixel grayscale change between adjacent frames, so as to evaluate the periodic stability of the combustion process; A flame combustion sufficiency evaluation module, wherein the flame combustion sufficiency evaluation module calculates a spectral distribution difference characteristic value by analyzing the spectral energy distribution of the flame image data in different bands, so as to evaluate the combustion sufficiency; A flame combustion state evaluation module, wherein the flame combustion state evaluation module inputs the flame flicker differential frequency characteristic value and the spectral distribution difference characteristic value into a flame state judgment model, and judges the combustion state of the flame according to the model output, including a normal combustion state and an abnormal combustion state; The abnormal alarm module performs early warning processing based on the abnormal combustion state of the flame.

2. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 1 is characterized in that: The evaluation of the periodic stability of the combustion process specifically includes: The flame image data of the gas stove during the cooking process is collected in real time, and the flame image data is processed in the time domain difference. The flame flicker differential frequency eigenvalue is calculated according to the pixel grayscale difference between adjacent frames, and it is determined whether the flame flicker differential frequency eigenvalue is greater than or equal to a preset threshold. If so, the flame of the combustion process is unstable, if not, the flame of the combustion process is stable.

3. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 2 is characterized in that: The process of obtaining the flame flicker differential frequency characteristic value is as follows: Collect continuous image sequences of the gas stove flame during cooking, convert the images into grayscale, and extract the temporal grayscale change information of the flame area; In the preset flame region of interest, the pixel grayscale difference between adjacent frames is calculated to obtain a differential image matrix; Each frame of differential image is averaged in the flame area to obtain a one-dimensional time series signal for subsequent frequency domain analysis; The fast Fourier transform algorithm is used to perform spectrum analysis on the obtained one-dimensional time series signal to extract the main frequency components of the flame flicker. Within the set normal flame flicker frequency range, the total energy in the corresponding frequency band is calculated, and the average energy in the corresponding frequency band is calculated by the mean calculation expression. The ratio of the average energy to the total energy is calculated to obtain the flame flicker differential frequency characteristic value.

4. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 1 is characterized in that: The evaluation of combustion sufficiency specifically includes: The flame image data of the gas stove during the cooking process is collected in real time. The abnormal degree of spectral energy distribution of the flame image data in different bands is analyzed, the spectral distribution difference characteristic value is calculated, and it is determined whether the spectral distribution difference characteristic value is greater than or equal to the preset threshold. If so, the flame of the combustion process is insufficient, if not, the flame of the combustion process is sufficient.

5. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 4 is characterized in that: The process of obtaining the spectral distribution difference characteristic value is as follows: The flame image data is collected by multi-spectral imaging technology, and the average spectral energy of each band is calculated within the selected band range to obtain the average spectral energy sequence of each band; Discrete wavelet transform is applied to the average spectral energy sequence of each band to obtain wavelet coefficients at different scales, and the spectral distribution difference characteristic values ​​are calculated based on the wavelet coefficients.

6. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 1 is characterized in that: The construction process of the flame state judgment model is as follows: The flame flicker differential frequency eigenvalue and the spectral distribution difference eigenvalue are constructed into a comprehensive eigenvector as the input of the flame state judgment model to minimize the error between the predicted flame state score and the actual flame state score as the training target. The flame state score is output according to the trained flame state judgment model, and the flame state judgment model is a random forest model.

7. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 6 is characterized in that: The training process of the flame state judgment model is as follows: The random forest algorithm is used to establish a flame state judgment model. The model consists of multiple decision trees. Each decision tree is trained based on randomly selected sample subsets and feature subsets to enhance the generalization ability of the model. During the training process, the actual flame state scores manually annotated are used as labels, the flame state scores predicted and output by the model are used as targets, and the minimization of the mean square error is used as the optimization target. After the model training is completed, the comprehensive feature vector extracted in real time is input into the trained random forest model to output the flame state score.

8. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 1 is characterized in that: The step of judging the combustion state of the flame according to the model output specifically includes: It is determined whether the flame state score of the combustion process is greater than or equal to a preset threshold. If so, it is recorded as a normal combustion state; if not, it is recorded as an abnormal combustion state.

9. The computer vision-based cooking flame morphology real-time monitoring and abnormal alarm system according to claim 1, characterized in that: The abnormal combustion state of the flame is used to perform early warning processing, which specifically includes: The abnormal burning time of the flame is obtained, and the abnormal burning time of the flame is multiplied by the flame state score to obtain the grading score. If the grading score is less than or equal to the preset first-level abnormal score, it is recorded as a first-level abnormality, triggering the on-site sound and light alarm and simultaneously pushing the early warning information to the user terminal. If the grading score is greater than or equal to the first-level abnormality score, it is recorded as a second-level abnormality. In addition to executing the first-level abnormal operation, the gas valve automatic closing device is linked to control to cut off the fuel supply, and an alarm notification is sent to the remote management platform.

Images