Electric appliance AI identification method and identification device based on matrix image analysis

Abstract

The invention relates to the technical field of AI intelligent identification, in particular to an electric appliance AI identification method and device based on matrix image analysis. Specifically, the identification method comprises the steps that S0, an image AI identification device arranged in a preset area carries out matrix type image identification on electric appliances in the preset area; if the image recognition result is pre-stored in the specific picture library, the step S1 is not carried out any more; otherwise, continuing to execute S1; the method comprises the following steps: S1, converting an obtained real-time electric signal of an electric appliance into a time domain feature and a frequency domain feature through Fourier transform; s2, predicting the safety coefficient of the current electric appliance based on a BP prediction model; and S3, performing type identification on the current electric appliance based on a safety coefficient threshold value and the obtained safety coefficient. Based on the method and the device, the use condition of the electric appliance in the preset area can be accurately and effectively identified, and the power utilization convenience and safety of the preset area are ensured by controlling the power on / off of the electric appliance.

Description

Technical Field

[0001] The present invention relates to the field of AI intelligent recognition technology, and in particular to an AI recognition method and a recognition device for electrical appliances based on matrix image analysis. Background Art

[0002] For school dormitories, which are large collective dormitories, electricity safety needs to be strictly controlled in order to reduce the risk of electricity use and protect the safety of teachers and students.

[0003] Electrical appliances that are explicitly prohibited from use in school dormitories include: purely resistive heating appliances (such as electric blankets, rice cookers, and electric ironing boards), resistive-inductive appliances (such as hair dryers and electric heaters that blow hot air), and resonant heating appliances (such as induction cookers); in response to the aforementioned "electrical appliances prohibited from use in school dormitories", schools usually install power-limiting modules, which analyze the waveform based on a dedicated chip; however, some students often use "anti-power-limiting sockets" to destroy the sine wave waveform and interfere with the judgment of the power-limiting module. Electrical appliances that are explicitly allowed to be used in school dormitories include: approximately resistive heating appliances (such as desktop computers), capacitive appliances (such as power adapters, laptops, mobile phone chargers, monitors), and inductive appliances (such as inkjet printers and small fans). Electrical appliances that are allowed to be used in some school dormitories include: purely resistive appliances (such as water dispensers) and capacitive appliances (such as refrigerated water dispensers, refrigeration air conditioners, and refrigerators).

[0004] Based on the foregoing, it can be known that for the control of the use of electrical appliances in student dormitories, it is necessary to allow some resistive electrical appliances to be used normally, and to disable some non-resistive electrical appliances; that is to say, it is difficult to deal with the on-off control problem of electrical appliances in student dormitories based on the simple classification of resistive, capacitive and inductive electrical appliances; in order to solve the aforementioned technical problems, it is urgent to propose an AI recognition method and recognition device for electrical appliances based on matrix image analysis. Summary of the invention

[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides an AI identification method and device for electrical appliances based on matrix image analysis, which solves the technical problem of "how to accurately and effectively identify the usage of electrical appliances in a preset area, and then ensure the convenience and safety of electricity use in the preset area by controlling the on / off power of the electrical appliances" in the current identification and control process of electrical appliances in student dormitories.

[0006] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0007] The present invention provides an AI recognition method for electrical appliances based on matrix image analysis. Specifically, the recognition method includes:

[0008] S0, an image AI recognition device is provided in a preset area, and performs matrix image recognition on electrical appliances in the preset area; if the image recognition result is pre-stored in a specific image library, S1 is not performed; otherwise, S1 is continued to be performed;

[0009] The specific image database is: a pre-established image database corresponding to dangerous electrical appliances;

[0010] S1. Convert the real-time electrical signals of electrical appliances obtained into time domain features and frequency domain features through Fourier transform; S2. Predict the safety factor of the current electrical appliances based on the BP prediction model; S3. Identify the type of the current electrical appliances based on the safety factor threshold and the obtained safety factor;

[0011] Among them, the initial training set of the BP prediction model is: the time domain characteristics, frequency domain characteristics and safety factors of pre-selected electrical appliances; the model input parameters are time domain characteristics and frequency domain characteristics, and the model output parameter is the safety factor; the types of electrical appliances include: safe electrical appliances and dangerous electrical appliances in the preset area.

[0012] Optionally, the identification method further includes:

[0013] S4. If the time domain characteristics and frequency domain characteristics corresponding to the identified electrical appliances do not exist in the current BP prediction model training set, the corresponding time domain characteristics, frequency domain characteristics and safety factors are added to the current BP prediction model training set for model retraining to obtain a new BP prediction model.

[0014] Optionally, the matrix image recognition method includes:

[0015] S0-1. Acquire the image of the electrical appliance based on the image AI recognition device, acquire the pixel data corresponding to the image of the electrical appliance based on matrix, and pre-process the image of the electrical appliance based on matrix analysis;

[0016] Among them, matrix analysis includes addition, subtraction, multiplication and inverse matrix and rank of matrix pixel data matrix; preprocessing includes grayscale transformation and filtering processing;

[0017] S0-2, performing secondary processing on the pre-processed electrical appliance image based on matrix analysis;

[0018] Among them, secondary processing includes: edge detection, texture analysis and color feature extraction;

[0019] S0-3. Based on matrix analysis, the images of electrical appliances that have completed secondary processing are subjected to image recognition through convolutional neural networks.

[0020] Optionally, S4 also includes:

[0021] If the time domain features and frequency domain features corresponding to the identified dangerous electrical appliances do not exist in the current BP prediction model training set, the image data of the corresponding electrical appliances are obtained through the image AI recognition device, and the image data is stored in a specific image library.

[0022] Optionally, after S1 and before S2, it also includes:

[0023] S1-1, if the time domain features and frequency domain features obtained by transformation exist in the specific feature data set, S2 to S4 are no longer performed; otherwise, S2 is continued to be performed;

[0024] Among them, the specific feature data set includes: time domain characteristics and frequency domain characteristics corresponding to dangerous electrical appliances allowed to be used in a preset area.

[0025] Optionally, the S1 includes:

[0026] The power distribution device obtains the electrical signals of all electrical appliances in the preset area;

[0027] The acquired electrical signal is converted into the corresponding amplitude spectrum and phase spectrum based on Fourier transform; and the corresponding time domain features and frequency domain features are acquired based on the amplitude spectrum and phase spectrum.

[0028] Optionally, the safety factor is: a value set based on the safe use of electrical appliances in a preset area, combined with corresponding time domain characteristics and frequency domain characteristics.

[0029] Optionally, the safety factor threshold is: a threshold value set based on the safety factor included in the current BP prediction model training set.

[0030] Optionally, the S3 includes:

[0031] If the obtained safety factor is within the safety factor threshold range, the current electrical appliance is regarded as a safe electrical appliance in the preset area;

[0032] Otherwise, the current electrical appliance is regarded as a dangerous electrical appliance in the preset area.

[0033] The present invention also provides an identification device based on the aforementioned AI identification method for electrical appliances. Specifically, the identification device is a combined AI electronic communication device integrated and packaged based on the aforementioned AI identification method for electrical appliances.

[0034] The beneficial effects of this application are:

[0035] Compared with the traditional methods of "simply dividing resistive, capacitive and inductive electrical appliances and installing power limiting modules", this application adopts the method of "solid-state characteristic analysis of electrical appliances combined with matrix image AI recognition". First, based on the pre-stored specific image library, matrix image AI recognition is performed on electrical appliances to quickly identify dangerous electrical appliances in real time; secondly, the real-time electrical signals of electrical appliances are converted into time domain features and frequency domain features based on Fourier transform, and the safety factor of the current electrical appliances is predicted based on the BP prediction model. Finally, the type of the current electrical appliances is identified based on the safety factor threshold and the obtained safety factor, so as to achieve: accurate and effective identification of the use of electrical appliances in the preset area;

[0036] When the time domain features and frequency domain features corresponding to the identified safety electrical appliances do not exist in the current BP prediction model training set, on the one hand, the present application updates the BP prediction model training set to ensure the prediction accuracy of the BP prediction model; on the other hand, the present application obtains image data of the electrical appliances through an image AI recognition device, and stores the image data in a specific image library to ensure the recognition accuracy of the image AI recognition device; thereby greatly ensuring the accurate recognition of electrical appliances;

[0037] When the time domain features and frequency domain features obtained through transformation exist in a specific feature data set, the corresponding electrical appliances will not be identified, providing convenient conditions for the use of specific electrical appliances such as water dispensers, air conditioners, refrigerators, etc. in school dormitories, making humanized intelligent management and control more reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flow chart of an electrical appliance AI identification method based on matrix image analysis provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0040] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0041] Embodiment 1

[0042] This embodiment proposes an AI recognition method for electrical appliances based on matrix image analysis;

[0043] This application needs to explain that: image recognition involves converting image data into digital information, using computer algorithms to perform object recognition and information recognition based on digital information, realizing image data analysis and processing, and effectively recognizing images. Matrix analysis involves matrix operations, solving equations, and solving linear systems. Images can be regarded as a representation of matrices. Through matrix analysis, image rotation, translation, scaling, image feature extraction, and image recognition can be achieved.

[0044] like Figure 1 As shown, the recognition method includes: S0, an image AI recognition device is set in a preset area, and matrix image recognition is performed on the electrical appliances in the preset area; if the image recognition result is pre-stored in a specific image library, S1 is no longer performed; otherwise, S1 is continued.

[0045] For the aforementioned S0, the image AI recognition device may be: a combined device in which a camera device arranged in multiple directions is connected to a background analysis device;

[0046] The specific image library is a pre-established image database corresponding to dangerous electrical appliances; specifically, the photos or videos of various types of dangerous electrical appliances are taken in advance by a camera device, and the photos or videos are converted into data and stored in a specific image library. Therefore, when the data of electrical appliances recognized by the image AI recognition device in the preset area is the same as any set of data in the specific image library, it indicates that the current electrical appliance is a dangerous electrical appliance.

[0047] In this embodiment, the matrix image recognition method includes:

[0048] S0-1. Acquire the image of the electrical appliance based on the image AI recognition device, acquire the pixel data corresponding to the image of the electrical appliance based on matrix formation, and pre-process the image of the electrical appliance based on matrix analysis.

[0049] Among them, matrix analysis includes addition, subtraction, multiplication and inverse matrix and rank of matrix pixel data matrix; preprocessing includes grayscale transformation and filtering processing;

[0050] For the aforementioned S0-1, the acquired color image can be converted into a grayscale image through grayscale transformation to simplify the subsequent data processing process; the calculation formula of grayscale transformation is shown in (1):

[0051] g(x,y)=0.299R(x,y)+0.587G(x,y)+0.114B(x,y) (1)

[0052] Among them, R(x,y), G(x,y), and B(x,y) represent the red, green, and blue components of the color image, respectively, and g(x,y) represents the converted grayscale image.

[0053] For the aforementioned S0-1, the image data is smoothed by filtering, which can eliminate noise and improve image quality. The filtering method used in this application is median filtering, and the calculation method of median filtering is: sorting the current pixel value and its surrounding pixel values, and then taking the middle value; specifically:

[0054] Pixel value sorting: f(xp,y)…f(x+p,y)…f(x,yq)…f(x,y+q); where p and q represent the radius of the filter kernel; the middle value is the filtered pixel value f(x,y).

[0055] S0-2. Perform secondary processing on the pre-processed electrical appliance image based on matrix analysis.

[0056] Among them, secondary processing includes: edge detection, texture analysis and color feature extraction;

[0057] For the aforementioned S0-2, edge detection is used to extract edge information of the image data to extract information about the object in the image. The edge detection method used in this application is Canny edge detection. The steps of Canny edge detection include:

[0058] 1) Calculate the gradient of the image; 2) Eliminate noise through double threshold filtering; 3) Eliminate edge interference through non-maximum suppression; 4) Extract edges based on double thresholds.

[0059] For the aforementioned S0-2, the fine structures in the image are analyzed by texture analysis to extract the feature information in the image; the texture analysis method in this application is natural field analysis (Gabor), and the calculation formula corresponding to the natural field analysis is shown in (2):

[0060] h(x,y)=exp(-((xx 0 ) 2 +(yy 0 ) 2 ) / (2×sigma 2 ))×cos(2×pix / lambda+phi) (2)

[0061] Where h(x,y) represents the response value of the Gabor filter at the coordinate (x,y); 0 ,y 0 represents the center position of the Gabor filter; sigma represents the standard deviation of the filter; lambda represents the wavelength of the Gabor filter; phi represents the phase of the Gabor filter.

[0062] For the aforementioned S0-2, the color information in the image is extracted by color feature extraction; the method adopted in this application is color similarity calculation, and the formula for color similarity calculation is shown in (3):

[0063] D=sqrt((H 1 -H 2 ) 2 +(S 1 -S 2 ) 2 +(V 1 -V 2 ) 2 ) (3)

[0064] Where D represents the distance between two colors; H 1 , S 1 、V 1 Indicates the hue, saturation and brightness of the first color; H 2 , S 2 、V 2 Indicates the hue, saturation, and lightness of the second color.

[0065] S0-3. Based on matrix analysis, the images of electrical appliances that have completed secondary processing are subjected to image recognition through convolutional neural networks.

[0066] For the above S0-3, the feature information of the image is automatically learned through the convolutional neural network to achieve image recognition; the main structure of the convolutional neural network is:

[0067] 1) Convolutional layer, used to filter the image through the convolution kernel and extract the feature information of the image;

[0068] 2) Pooling layer, which is used to compress the feature information of the image through downsampling method to reduce the amount of calculation;

[0069] 3) Fully connected layer, which is used to connect the feature information of the convolutional layer and the pooling layer to achieve image classification and recognition.

[0070] What needs to be explained in this application is that: the electrical structures and internal components of different types of electrical appliances are different, which in turn determines that the electrical characteristics of different electrical appliances are also different; since the electrical characteristics of electrical appliances are fixed when the design is completed, the electrical characteristics remain unchanged during the use of the appliances; therefore, this application adopts the method of "solid-state characteristic analysis of electrical appliances" to perform AI identification of electrical appliances.

[0071] like Figure 1 As shown, the identification method further includes: S1, converting the acquired real-time electrical signals of the electrical appliances into time domain features and frequency domain features through Fourier transform.

[0072] For the aforementioned S1, the power distribution device obtains the electrical signals of all electrical appliances in the preset area, converts the obtained electrical signals into corresponding amplitude spectrum and phase spectrum based on Fourier transform, and obtains corresponding time domain features and frequency domain features based on the amplitude spectrum and phase spectrum;

[0073] Among them, the power distribution device can be installed in the school dormitory distribution room, and the preset area is the school dormitory area.

[0074] For the aforementioned S1, the time domain features include: maximum value, maximum absolute value, minimum value, mean value, peak-to-peak value, absolute mean value, root mean square value, root amplitude, standard deviation, kurtosis, skewness, margin index, waveform index, pulse index, peak index; the frequency domain features include: center of gravity frequency, mean square frequency, root mean square frequency, frequency variance, frequency standard deviation;

[0075] In this embodiment, the time domain features and frequency domain features can be acquired, analyzed and displayed based on corresponding simulation computing software.

[0076] In this embodiment, if the time domain features and frequency domain features obtained through transformation exist in the specific feature data set, no subsequent operations are performed;

[0077] Among them, the specific feature data set includes: time domain features and frequency domain features corresponding to dangerous electrical appliances allowed to be used in the preset area;

[0078] It should be noted that in order to improve the flexibility of the usage rules of electrical appliances in the preset area, some electrical appliances are directly set as usable electrical appliances. This application sets up a special interface in the simulation calculation software for obtaining time domain characteristics and frequency domain characteristics, so that some electrical appliances do not undergo identification operations.

[0079] like Figure 1 As shown, the identification method further includes: S2, predicting the safety factor of the current electrical appliance based on the BP prediction model.

[0080] For the aforementioned S2, the initial training set of the BP prediction model is: the time domain characteristics, frequency domain characteristics and safety factors of the pre-selected electrical appliances;

[0081] The present application ensures the prediction accuracy of the BP prediction model by improving the applicability; since different types of electrical appliances need to be identified, the pre-selected electrical appliances include: capacitive electrical appliances, resistive electrical appliances, inductive electrical appliances and comprehensive load electrical appliances.

[0082] In this application, the safety factor is: a value set based on the safe use of electrical appliances in a preset area, combined with corresponding time domain characteristics and frequency domain characteristics; the safety factor is a judgment indicator for determining whether the current electrical appliance can be used in the preset area.

[0083] In this application, the model input parameters of the BP prediction model are time domain characteristics and frequency domain characteristics, and the model output parameter is the safety factor; the BP prediction model can be embodied as: based on a simulation calculation software with an interactive interface; the simulation calculation software corresponding to the BP prediction model is docked with the simulation calculation software for obtaining time domain characteristics and frequency domain characteristics, so that the time domain characteristics and frequency domain characteristics obtained in real time can be imported into the BP prediction model, and the corresponding safety factor can be displayed based on the visual interface.

[0084] like Figure 1 As shown, the identification method further includes: S3, identifying the type of the current electrical appliance based on the safety factor threshold and the acquired safety factor.

[0085] For the aforementioned S3, the safety factor threshold is: a threshold value divided and set based on the safety factor included in the current BP prediction model training set; the types of electrical appliances include: safe electrical appliances and dangerous electrical appliances in a preset area.

[0086] In this embodiment, when the safety factor predicted based on S2 is within the safety factor threshold range, it indicates that the current electrical appliance can be used and is a safe electrical appliance; otherwise, it cannot be used and is a dangerous electrical appliance.

[0087] In this embodiment, the identification result of the type of the current electrical appliance can also be interactively displayed based on the simulation calculation software corresponding to the aforementioned BP prediction model.

[0088] In this embodiment, the identification method also includes: S4. If the time domain characteristics and frequency domain characteristics corresponding to the identified electrical appliances do not exist in the current BP prediction model training set, the corresponding time domain characteristics, frequency domain characteristics and safety factors are added to the current BP prediction model training set for model retraining to obtain a new BP prediction model.

[0089] For the aforementioned S4, in order to improve the accuracy of electrical appliance identification,

[0090] The present application improves the applicability of the BP prediction model by timely adding and updating the BP prediction model training set.

[0091] It should be noted that as the BP prediction model training set is updated, the safety factor threshold is also updated accordingly.

[0092] In this embodiment, it is also necessary to explain that: if the time domain features and frequency domain features corresponding to the identified dangerous electrical appliances do not exist in the current BP prediction model training set, it means that the image data of the current dangerous electrical appliances have not been pre-stored in the specific image library, and then the image data of the corresponding electrical appliances are obtained through the image AI recognition device, and the image data is stored in the specific image library, so as to incrementally update the image data of the specific image library, thereby improving the accuracy of the matrix image recognition performed by the image AI recognition device on the electrical appliances in the preset area.

[0093] Based on the AI ​​identification method of electrical appliances based on matrix image analysis described in Example 1, compared with the traditional methods of "simply dividing resistive, capacitive and inductive electrical appliances and installing power limiting modules", this application adopts the method of "combining solid-state characteristic analysis of electrical appliances with matrix image AI identification". First, based on the pre-stored specific image library, matrix image AI identification is performed on electrical appliances to quickly identify dangerous electrical appliances in real time; secondly, the real-time electrical signals of electrical appliances obtained are converted into time domain features and frequency domain features based on Fourier transform, and the safety factor of the current electrical appliance is predicted based on the BP prediction model. Finally, the type of the current electrical appliance is identified based on the safety factor threshold and the obtained safety factor, so as to achieve: accurate and effective identification of the use of electrical appliances in a preset area;

[0094] Based on the AI ​​identification method for electrical appliances based on matrix image analysis described in Example 1, when the time domain features and frequency domain features corresponding to the identified safe electrical appliances do not exist in the current BP prediction model training set, on the one hand, the present application updates the BP prediction model training set to ensure the prediction accuracy of the BP prediction model; on the other hand, the present application obtains image data of the electrical appliances through an image AI recognition device, and stores the image data in a specific image library to ensure the recognition accuracy of the image AI recognition device; thereby greatly ensuring the accurate recognition of electrical appliances;

[0095] Based on the AI ​​identification method for electrical appliances based on matrix image analysis described in Example 1, when the time domain features and frequency domain features obtained through transformation exist in a specific feature data set, the corresponding electrical appliances are not identified, thereby providing convenient conditions for the use of specific electrical appliances such as water dispensers, air conditioners, refrigerators, etc. in school dormitories, and humanized intelligent management and control is more reasonable.

[0096] Embodiment 2

[0097] This embodiment proposes an identification device based on the AI ​​identification method for electrical appliances based on matrix image analysis described in the first embodiment above. The identification device is a combined AI electronic communication device integrated and packaged based on the AI ​​identification method for electrical appliances described in the first embodiment above.

[0098] Embodiment 3

[0099] This embodiment proposes an application example of the AI ​​identification method for electrical appliances based on matrix image analysis described in the first embodiment, which is as follows:

[0100] For the construction of the BP prediction model, the initial training set and the validation set are obtained; specifically, Table 1 is the data table of the initial training set and the validation set of the BP prediction model. Combined with Table 1, there are n electrical appliances in total from electrical appliances 1 to electrical appliances n, and the n electrical appliances include: resistive electrical appliances, container electrical appliances, inductive electrical appliances, and comprehensive electrical appliances of multiple categories;

[0101] Each electrical appliance has 15 time domain feature data, 5 frequency domain feature data and 1 safety factor data; the 15 time domain feature data correspond to the aforementioned "maximum value, maximum absolute value, minimum value, mean value, peak-to-peak value, absolute mean value, root mean square value, root amplitude, standard deviation, kurtosis, skewness, margin index, waveform index, pulse index, peak index", and the 5 frequency domain feature data correspond to the aforementioned "center of gravity frequency, mean square frequency, root mean square frequency, frequency variance, frequency standard deviation"; optionally, if n = 1200, the first 10 of the n electrical appliances are The 15 time domain feature data, 5 frequency domain feature data and 1 safety factor data corresponding to the 100 electrical appliances are used as the training set of the BP prediction model. The 15 time domain feature data, 5 frequency domain feature data and 1 safety factor data corresponding to the last 200 electrical appliances among the n electrical appliances are used as the verification set of the BP prediction model. After training, a preliminary BP prediction model is obtained to construct the corresponding relationship between the time domain features, the frequency domain features and the safety factor. The specific corresponding relationship is as shown in (4), and then the safety factor is predicted based on the time domain features, the frequency domain features and the safety factor.

[0102] Z = f(X 1 , X 2 ...X 15 , Y 1 , Y 2 ...Y 5 ) (4)

[0103] Combined with the above Table 1, the safety factor threshold is based on the safety factor set {Z 1 , Z 2 , Z 3 ...Z n Specifically, the safety factor threshold is determined based on the safety factor set of electrical appliances 1 to n belonging to the preset area safety electrical appliances.

[0104] In this embodiment, Table 2 is a data table of a specific feature data set. In combination with Table 2, if the time domain features and frequency domain features converted by Fourier transform are the same as the data of an electrical appliance in Table 2, it means that the current electrical appliance is directly allowed to be used in school dormitories and no type identification is required.

[0105] In this embodiment, Table 3 is a feature data set and safety factor data table acquired in real time. In combination with Table 1 and Table 3, if the 15 time domain feature data and 5 frequency domain feature data corresponding to any one or more electrical appliances in Table 3 are the same as the 15 time domain feature data and 5 frequency domain feature data corresponding to any electrical appliance in Table 1, then the corresponding data of one or more electrical appliances in Table 3 are supplemented to Table 1, so as to realize continuous updating of the BP prediction model training set and continuous model training to gradually improve the prediction accuracy of the BP prediction model.

[0106] Table 1 BP prediction model initial training set and validation set data table

[0107]

[0108] Table 2 Data table of specific feature datasets

[0109]

[0110] Table 3. Feature data set and safety factor data table acquired in real time

[0111]

[0112] It should be noted that in the claims, any reference numerals placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The use of the words first, second, third, etc., is only for convenience of expression and does not indicate any order. These words may be understood as part of the component name.

[0113] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0114] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments after knowing the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0115] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention should also include these modifications and variations.

Claims

1. An AI recognition method for electrical appliances based on matrix image analysis, characterized in that: The identification method comprises: S0, an image AI recognition device is provided in a preset area, and performs matrix image recognition on electrical appliances in the preset area; if the image recognition result is pre-stored in a specific image library, S1 is not performed; otherwise, S1 is continued to be performed; The specific image database is: a pre-established image database corresponding to dangerous electrical appliances; S1. Convert the real-time electrical signals of electrical appliances obtained into time domain features and frequency domain features through Fourier transform; S2. Predict the safety factor of the current electrical appliances based on the BP prediction model; S3. Identify the type of the current electrical appliances based on the safety factor threshold and the obtained safety factor; Among them, the initial training set of the BP prediction model is: the time domain characteristics, frequency domain characteristics and safety factors of pre-selected electrical appliances; the model input parameters are time domain characteristics and frequency domain characteristics, and the model output parameter is the safety factor; the types of electrical appliances include: safe electrical appliances and dangerous electrical appliances in the preset area.

2. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: The identification method further comprises: S4. If the time domain characteristics and frequency domain characteristics corresponding to the identified electrical appliances do not exist in the current BP prediction model training set, the corresponding time domain characteristics, frequency domain characteristics and safety factors are added to the current BP prediction model training set for model retraining to obtain a new BP prediction model.

3. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: Matrix image recognition methods include: S0-1. Acquire the image of the electrical appliance based on the image AI recognition device, acquire the pixel data corresponding to the image of the electrical appliance based on matrix, and pre-process the image of the electrical appliance based on matrix analysis; Among them, matrix analysis includes addition, subtraction, multiplication and inverse matrix and rank of matrix pixel data matrix; preprocessing includes grayscale transformation and filtering processing; S0-2, performing secondary processing on the pre-processed electrical appliance image based on matrix analysis; Among them, secondary processing includes: edge detection, texture analysis and color feature extraction; S0-3. Based on matrix analysis, the images of electrical appliances that have completed secondary processing are subjected to image recognition through convolutional neural networks.

4. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: S4 also includes: If the time domain features and frequency domain features corresponding to the identified dangerous electrical appliances do not exist in the current BP prediction model training set, the image data of the corresponding electrical appliances are obtained through the image AI recognition device, and the image data is stored in a specific image library.

5. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: After S1 and before S2, it also includes: S1-1, if the time domain features and frequency domain features obtained by transformation exist in the specific feature data set, S2 to S4 are no longer performed; otherwise, S2 is continued to be performed; Among them, the specific feature data set includes: time domain characteristics and frequency domain characteristics corresponding to dangerous electrical appliances allowed to be used in a preset area.

6. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: The S1 includes: The power distribution device obtains the electrical signals of all electrical appliances in the preset area; The acquired electrical signal is converted into the corresponding amplitude spectrum and phase spectrum based on Fourier transform; the corresponding time domain features and frequency domain features are acquired based on the amplitude spectrum and phase spectrum.

7. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: The safety factor is a value set based on the safe use of electrical appliances in a preset area, combined with the corresponding time domain characteristics and frequency domain characteristics.

8. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: The safety factor threshold is: a threshold value set based on the safety factor included in the current BP prediction model training set.

9. The AI ​​identification method for electrical appliances according to claim 1, characterized in that: The S3 includes: If the obtained safety factor is within the safety factor threshold range, the current electrical appliance is regarded as a safe electrical appliance in the preset area; Otherwise, the current electrical appliance is regarded as a dangerous electrical appliance in the preset area.

10. An identification device based on the electrical appliance AI identification method according to any one of claims 1 to 9, characterized in that: The identification device is a combined AI electronic communication device integrated and packaged based on the AI ​​identification method of electrical appliances.

Images