Electric meter anomaly identification method and system based on MobileNetV3

Abstract

The application discloses a meter abnormality identification method and system based on MobileNetV3, comprising the following steps: acquiring image data of abnormal and normal electric energy meters, preprocessing the image data to obtain a sample data set, and randomly dividing the sample data set into a training set and a verification set according to a proportion; training an improved MobileNetV3 network based on the training set, verifying the improved MobileNetV3 network based on the verification set, and taking the trained improved MobileNetV3 network as an abnormal electric energy meter identification model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module of a reverse residual error; and image data of an electric energy meter to be identified is input into the abnormal electric energy meter identification model to output an electric energy meter abnormality identification result. The application improves the extraction capability of the appearance features of the electric meter, improves the abnormality identification accuracy, and can adaptively generate a special abnormality identification model for different types of electric energy meters.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electric energy metering device fault identification, in particular to an electric meter abnormality identification method and system based on MobileNetV3. BACKGROUND

[0002] Current smart meter abnormality identification mainly relies on traditional image processing methods and relatively basic deep learning networks. In the detection of smart meters, traditional methods detect the appearance of the electric meter through edge detection, color feature extraction, etc., combined with feature matching methods (such as SIFT or SURF). However, due to the influence of external conditions such as light and angle, the recognition effect is not stable, and the feature extraction time is long, which is difficult to meet the efficient requirements of the actual pipeline.

[0003] In addition, some existing deep learning-based models (such as VGG16, ResNet50, etc.) perform well in image classification, but their structures are relatively complex, and the number of model parameters is large, which makes it difficult to deploy them on devices with limited computing power. In particular in the abnormality detection scenario, the feature extraction effect of these networks is limited, resulting in low accuracy in identifying the appearance of the electric meter, and slow detection speed, which is not conducive to the real-time requirements in practical applications. Therefore, the existing technical solutions still have deficiencies in model lightweight and recognition accuracy.

[0004] Therefore, the present application is proposed. SUMMARY

[0005] The technical problem to be solved by the present application is that the existing technology has low accuracy in identifying the appearance of the electric meter, slow detection speed, and is not conducive to real-time requirements in practical applications, and cannot adapt to abnormality identification of different types of electric energy meters. The present application aims to provide an electric meter abnormality identification method and system based on MobileNetV3, which can automatically construct an electric energy meter appearance abnormality identification model through improved MobileNetV3 network, generate a dedicated identification model to adapt to different types of electric energy meters, and improve the generalization ability and adaptability of the model. The present application improves the extraction ability of the electric meter appearance features, significantly improves the accuracy of abnormality identification, can adaptively generate a dedicated abnormality identification model for different types of electric energy meters, effectively meets the identification requirements in different environments, and has fast detection speed.

[0006] The present application is implemented by the following technical solutions:

[0007] In a first aspect, the present application provides an electric meter abnormality identification method based on MobileNetV3, which comprises:

[0008] Image data of abnormal electric energy meters and normal electric energy meters is acquired, sample data sets are obtained by preprocessing the image data, and the sample data sets are randomly divided into a training set and a verification set according to a proportion;

[0009] The improved MobileNetV3 network is trained based on the training set, the accuracy of the improved MobileNetV3 network is verified based on the verification set, and the trained improved MobileNetV3 network is used as an abnormal electric energy meter identification model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module with a reverse residual error;

[0010] Image data of an electric energy meter to be identified is input into the abnormal electric energy meter identification model, and an electric energy meter abnormality identification result is output.

[0011] The present application is suitable for electric energy meter appearance abnormality identification under the condition of centralized detection and automatic production of electric energy meters, can adaptively generate a dedicated abnormality identification model for different types of electric energy meters, can effectively improve the accuracy of the model, and can also consider the identification rate under the condition of a flow line.

[0012] Further, the sample data sets are obtained by preprocessing the image data, including:

[0013] The photographed intelligent electric energy meters are classified into two categories by manual screening: the first category is normal electric meters, and the second category is abnormal electric meters; after classification, a classified sample data set is obtained;

[0014] The classification method is to save normal electric meter pictures in a folder and save abnormal electric meter pictures in another folder.

[0015] Further, the improved MobileNetV3 network includes an input layer, a convolution layer, an activation layer, a bottleneck module with a reverse residual error, a global pooling layer, an activation layer, a full connection layer, a judgment layer and an output layer connected in sequence;

[0016] The bottleneck module with a reverse residual error reduces the amount of calculation while maintaining performance through an expansion, depth convolution and compression structure, and obtains data after the bottleneck module with a reverse residual error.

[0017] The entire bottleneck module with a reverse residual error can be represented as:

[0018] X3=Bneck(X2)

[0019] Wherein X3 represents data after the bottleneck module with a reverse residual error, and Bneck() represents a combination of three operations of point-by-point convolution expansion, depth convolution and point-by-point convolution compression.

[0020] Further, the bottleneck module with reverse residual includes a point-wise convolution expansion unit, a depthwise convolution unit and a point-wise convolution compression unit.

[0021] The point-wise convolution expansion unit is configured to expand the data with a lower number of channels to obtain data with a higher number of channels as expanded data through a first point-wise convolution expansion operation.

[0022] The depthwise convolution unit is configured to perform independent depthwise convolution on each channel in the expanded data to obtain depthwise-convoluted data.

[0023] The point-wise convolution compression unit is configured to compress the depthwise-convoluted data to data with a lower number of channels through a second point-wise convolution.

[0024] Further, the expression of the first point-wise convolution is X expand = H_swish(Conv 1×1 (X)), where X expand is the data after point-wise convolution expansion, Conv 1×1 () represents a convolution operation using a 1x1 convolution kernel, and X is an independent variable input into the operation.

[0025] The expression of the second point-wise convolution is X compress = Conv 1×1 (X depthwise ), where X compress represents the data after point-wise convolution compression.

[0026] Further, the operation expression of the depthwise convolution is:

[0027] X depthwise = H_swish(DepthwiseConv 3×3 (X expand

[0028] where X depthwise is the data after depthwise convolution, and DepthwiseConv 3×3 () represents a depthwise convolution operation using a 3x3 convolution kernel, which is represented as:

[0029]

[0030] where Y h,w,c represents the value of the data after depthwise convolution at position (h, w) and channel c. X h+i-1,w+j-1,c represents the value of the data to be subjected to the depthwise convolution operation at position (h+i-1, w+j-1) and channel c; K i,j,c represents the weight (learnable) of the 3x3 convolution kernel on the cth channel; and b cbias (learnable) of the c-th channel.

[0031] Further, the loss function of the improved MobileNetV3 network is improved is:

[0032]

[0033] where i is a sample, y i is a true label, is a predicted label.

[0034] In a second aspect, the present application further provides a MobileNetV3-based electric meter anomaly identification system, which comprises:

[0035] an acquisition unit configured to acquire image data of abnormal electric meters and normal electric meters;

[0036] a preprocessing and division unit configured to preprocess the image data to obtain a sample data set, and randomly divide the sample data set into a training set and a verification set according to a proportion;

[0037] a model construction and training unit configured to train the improved MobileNetV3 network based on the training set, verify the accuracy of the improved MobileNetV3 network based on the verification set, and use the trained improved MobileNetV3 network as an abnormal electric meter identification model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module with a reverse residual;

[0038] an anomaly identification unit configured to input image data of an electric meter to be identified into the abnormal electric meter identification model, and output an electric meter anomaly identification result.

[0039] Further, the improved MobileNetV3 network comprises an input layer, a convolution layer, an activation layer, a bottleneck module with a reverse residual, a global pooling layer, an activation layer, a full connection layer, a judgment layer and an output layer connected in sequence;

[0040] The bottleneck module with a reverse residual reduces the amount of calculation while maintaining performance through the structure of expansion, deep convolution and compression, and obtains data after the bottleneck module with a reverse residual;

[0041] The bottleneck module with a reverse residual comprises a point-wise convolution expansion unit, a deep convolution unit and a point-wise convolution compression unit.

[0042] The point-wise convolution expansion unit is configured to expand data with a lower number of channels through a first point-wise convolution expansion operation to obtain data with a higher number of channels as expanded data.

[0043] The depth convolution unit is used to perform independent depth convolution on each channel in the expanded data to obtain the depth-convolved data;

[0044] The point-by-point convolution compression unit is used to compress the depth-wise convolved data back to data with a lower number of channels through a second point-by-point convolution.

[0045] In a third aspect, the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned electric meter anomaly identification method based on MobileNetV3.

[0046] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0047] 1. The present invention is based on the electric meter anomaly recognition method and system of MobileNetV3. The present invention is suitable for identifying electric meter appearance anomalies under the conditions of centralized detection and calibration of electric meters and automated production. It can adaptively generate exclusive anomaly recognition models for electric meters of different models, effectively improving the accuracy of the model while taking into account the recognition rate under assembly line conditions.

[0048] 2. Lightweight and efficient: Compared with traditional deep learning models (such as ResNet50, VGG16, etc.), this paper adopts an improved MobileNetV3 network to achieve a lightweight model design and reduce the amount of computation, enabling this proposal to run efficiently in a low-power environment and meet the real-time detection requirements of pipeline conditions.

[0049] 3. Higher recognition accuracy: Traditional methods are easily affected by the external environment and have unstable recognition results. This proposal improves the ability to extract meter appearance features through the combination of deep convolution and inverse residual bottleneck modules, significantly improving the accuracy of anomaly recognition and effectively addressing recognition needs in different environments.

[0050] 4. Good Adaptability: This proposal can adaptively model different types of meters and generate unique anomaly recognition models. Compared with traditional methods, this solution has stronger model generalization capabilities and can adaptively adjust the model structure based on the appearance characteristics of different meters, ensuring high recognition accuracy across various meter models.

[0051] 5. Deployment Ease: Our MobileNetV3-based recognition method offers excellent deployment advantages. Traditional models have large parameter counts and are unsuitable for embedded devices. However, our proposed method can run efficiently on edge devices, facilitating widespread adoption in real-world scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:

[0053] Figure 1 Flow chart of the electric meter abnormality identification method based on MobileNetV3 of the present application;

[0054] Figure 2 Schematic diagram of the improved MobileNetV3 network structure of the present application;

[0055] Figure 3 Effect diagram of loss comparison between the present application and other existing methods;

[0056] Figure 4 Effect diagram of accuracy rate comparison between the present application and other existing methods;

[0057] Figure 5 Structure block diagram of the electric meter abnormality identification system based on MobileNetV3 of the present application. DETAILED DESCRIPTION

[0058] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be given below in combination with embodiments and drawings, and the illustrative embodiments of the present application and their descriptions are only used to explain the present application and do not limit the present application.

[0059] Embodiment 1

[0060] As shown in Figure 1 , the electric meter abnormality identification method based on MobileNetV3 of the present application comprises:

[0061] Step 1, obtaining image data of abnormal electric energy meters and normal electric energy meters, pre-processing the image data to obtain a sample data set, and randomly dividing the sample data set into a training set and a validation set according to a proportion;

[0062] In this embodiment, the image data is pre-processed to obtain a sample data set, which comprises:

[0063] The photographed intelligent electric energy meters are classified into two categories by artificial screening: the first category is normal electric meters, and the second category is abnormal electric meters; after the classification is completed, a classified sample data set D is obtained;

[0064] The classification method is to save the normal electric meter pictures in one folder and save the abnormal electric meter pictures in another folder.

[0065] Step 2, training the improved MobileNetV3 network based on the training set, verifying the accuracy of the improved MobileNetV3 network based on the validation set, and taking the trained improved MobileNetV3 network as an abnormal electric energy meter identification model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module with a reverse residual;

[0066] In this embodiment, the classified pictures are input into the improved MobileNetV3 neural network, and the improved MobileNetV3 neural network structure is as shown in Figure 2 The input is an image X, X is a labeled picture in a data set D, where H is the height of the image, W is the width of the image, and C is the number of channels of the image. The optimization method used by the network is Adam optimization.

[0067] As shown in Figure 2 The improved MobileNetV3 network includes an input layer, a convolutional layer, an activation layer, a bottleneck module with a reverse residual, a global pooling layer, an activation layer, a fully connected layer, a judgment layer and an output layer connected in sequence. The main layer structure is described as follows:

[0068] (1) Convolutional layer

[0069] First, set the convolution kernel: where 3 is the height and width of the convolution kernel, 3 is the input channel number, 16 is the output channel number, and the convolution stride is 1. The operation can be represented as:

[0070] X1=Conv 3×3 (X,K1)=X*K1+b

[0071] where X1 is the picture after convolution, Conv 3×3 () is the convolution operation, and b is a trainable offset term.

[0072] (2) Activation layer

[0073] The activation layer uses the H_swish activation function to replace the ReLU activation function. Compared with the traditional ReLU activation, H_swish retains the negative value information and has higher efficiency in lightweight networks. The operation can be represented as:

[0074]

[0075] where X2 is the activated data, ReLU6 is a function, and the expression is:

[0076] ReLU6(x)=min(max(0,x),6)

[0077] (3) Bottleneck module with reverse residual

[0078] The bottleneck module with reverse residual is a key structure used in the improved MobileNetV3 network, which enhances the feature extraction capability of the network while maintaining a lightweight design. The core idea of this module is to reduce the computational complexity while maintaining performance through the structure of expansion, depth convolution, and compression, and to obtain data after the bottleneck module with reverse residual.

[0079] Specifically, the bottleneck module with reverse residual includes a pointwise convolution expansion unit, a depth convolution unit, and a pointwise convolution compression unit.

[0080] 1) The pointwise convolution expansion unit is used to expand the data with a lower number of channels through a first pointwise convolution expansion operation to obtain data with a higher number of channels as expanded data; it has a higher number of channels to improve the representation ability of features in the data; the operation expression of the first pointwise convolution is:

[0081] X expand =H_swish(Conv 1×1 (X))

[0082] where X expand is the data after pointwise convolution expansion, Conv 1×1 () represents a convolution operation using a 1×1 convolution kernel, and X is the independent variable input into the operation;

[0083] 2) The depth convolution unit is used to perform independent depth convolution on each channel in the expanded data to obtain depth-convolved data; depth convolution has low computational complexity and can perform spatial feature extraction while maintaining computational efficiency. The operation expression of depth convolution is:

[0084] X depthwise =H_swish(DepthwiseConv 3×3 (X expand

[0085] where X depthwise is the data after depth convolution, DepthwiseConv 3×3 () represents a depth convolution operation using a 3×3 convolution kernel, which is represented as:

[0086]

[0087] where Y h,w,c represents the value of the data after depth convolution at position (h, w), channel c. X h+i-1,w+j-1,cIndicates that the data that needs to be convolved is at position (h+i-1,w+j-1), the value of channel c; K i,j,c represents the weight of the 3×3 convolution kernel on the cth channel (learnable); b c Represents the bias of the c-th channel (learnable).

[0088] 3) Point-by-point convolution compression unit, used to compress the depthwise convolution data back to data with a lower number of channels through the second point-by-point convolution. This operation speeds up subsequent calculations while ensuring that the feature information is not overly distorted during the compression process. The expression of the second point-by-point convolution is:

[0089] X compress =Conv 1×1 (X depthwise )

[0090] where X compress Represents the data after point-by-point convolution compression.

[0091] After the above triple operations of point-by-point convolution expansion → depth-wise convolution → point-by-point convolution compression, the entire bottleneck module with reverse residual can be expressed as:

[0092] X3=Bneck(X2)

[0093] Where X3 represents the data after passing through the bottleneck module with reverse residual, and Bneck() represents the combination of the triple operations of point-by-point convolution expansion → depth-wise convolution → point-by-point convolution compression.

[0094] (4) Global pooling layer

[0095] Global average pooling compresses the number of features in all spatial dimensions into the average value in the channel dimension. This operation can be expressed as:

[0096]

[0097] where X pool Represents the data after pooling, c represents the number of channels, here is the number of channels of data X3.

[0098] (5) Fully connected layer and judgment layer

[0099] The output after global average pooling is mapped to two categories through the fully connected layer. Let the weight matrix of the fully connected layer be With bias b, the output is:

[0100]

[0101] Softmax is used to normalize the output to a probability distribution of two classes.

[0102] Specifically, the model is optimized using a cross-entropy loss function. For each sample i, the true label is y i , and the predicted label is The loss function is: the loss function of the improved MobileNetV3 network is:

[0103]

[0104] Step 3: input the image data of the electric energy meter to be identified into the abnormal electric energy meter identification model, and output the electric energy meter abnormality identification result.

[0105] In specific implementation, the electric energy meter image data collected on-site in a batch of single-phase electric energy meter calibration assembly line is used as sample data for implementation.

[0106] S1: first, the photographed smart meters are classified into two categories by manual screening, one category is normal electric meters (positive samples), and the other category is abnormal electric meters (negative samples). The classification method is to manually save the normal electric meter pictures in a folder and save the abnormal electric meter pictures in another folder. After classification, the classified data set D is obtained, and the specific composition of the data D is shown in Table 1.

[0107] Table 1

[0108] Training set Test set Normal meters 2000 500 Abnormal meters 2000 500 Total 4000 1000

[0109] S2: build an improved MobileNetV3 network model, bring the data set D into the model for training, and test the model performance with the test set after training, and the training is iterated for 10 times, and the specific training and test effects are shown in Figure 3 and Figure 4 .

[0110] S3: in order to further reflect the effect of the electric energy meter abnormality identification of the present application, the method adopted by the present application is compared with VGG16 and ResNet50, and the test set of the data set D is used for comparative experiment, and the comparison effect is shown in Table 2.

[0111] Table 2

[0112]

[0113] The present application (1) is based on the lightweight model design of MobileNetV3: the present application adopts an improved MobileNetV3 network structure, has efficient feature extraction capability and lightweight design, is suitable for deployment on devices with limited computing resources, and meets the needs of real-time detection of the pipeline. (2) Construction of an automatic electric meter anomaly identification model: the present application realizes the automatic construction of an electric meter appearance anomaly identification model by improving the MobileNetV3 network, can generate a dedicated identification model to adapt to different models of electric meters, and improves the generalization ability and adaptability of the model. (3) Data-driven adaptive learning strategy: based on the division of the training set and the validation set, the adaptive learning is realized through the training of MobileNetV3, which effectively improves the accuracy of anomaly identification while retaining the identification rate under the pipeline condition.

[0114] The present application is suitable for electric meter appearance anomaly identification under the condition of automatic production of electric meter centralized detection and testing, can adaptively generate a dedicated anomaly identification model for different models of electric meters, and can effectively improve the accuracy of the model while taking into account the identification rate under the pipeline condition.

[0115] Embodiment 2

[0116] As shown in Figure 5 , the difference between the present embodiment and embodiment 1 is that the present embodiment provides an electric meter anomaly identification system based on MobileNetV3, which is one-to-one corresponding to the function of the electric meter anomaly identification method based on MobileNetV3 of embodiment 1; the system comprises:

[0117] An acquisition unit is configured to acquire image data of abnormal electric meters and normal electric meters;

[0118] A preprocessing and division unit is configured to preprocess the image data to obtain a sample data set, and randomly divide the sample data set into a training set and a validation set according to a proportion;

[0119] A model construction and training unit is configured to train an improved MobileNetV3 network based on the training set, verify the accuracy of the judgment of the improved MobileNetV3 network based on the validation set, and use the trained improved MobileNetV3 network as an abnormal electric meter identification model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module with a reverse residual;

[0120] An anomaly identification unit is configured to input image data of an electric meter to be identified into the abnormal electric meter identification model, and output an electric meter anomaly identification result.

[0121] As a further implementation, the improved MobileNetV3 network comprises an input layer, a convolution layer, an activation layer, a bottleneck module with reverse residual, a global pooling layer, an activation layer, a full connection layer, a judgment layer and an output layer connected in sequence;

[0122] The bottleneck module with reverse residual reduces the amount of calculation while maintaining performance through the structure of expansion, deep convolution and compression, and obtains data after the bottleneck module with reverse residual;

[0123] The bottleneck module with reverse residual includes a point-wise convolution expansion unit, a deep convolution unit and a point-wise convolution compression unit.

[0124] The point-wise convolution expansion unit is configured to expand data with a lower number of channels through a first point-wise convolution expansion operation to obtain data with a higher number of channels as expanded data.

[0125] The deep convolution unit is configured to perform independent deep convolution on each channel in the expanded data to obtain deep convolution data.

[0126] The point-wise convolution compression unit is configured to compress the deep convolution data back to data with a lower number of channels through a second point-wise convolution.

[0127] The execution process of each unit can be performed according to the execution process of the MobileNetV3-based electric meter anomaly identification method of Embodiment 1, and will not be repeated here.

[0128] Meanwhile, the application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the MobileNetV3-based electric meter anomaly identification method.

[0129] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0130] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0131] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0132] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0133] The above description is only specific implementation of the present application, and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. The electric meter anomaly recognition method based on MobileNetV3 is characterized by: The method includes: Obtain image data of abnormal electric energy meters and normal electric energy meters, preprocess the image data to obtain a sample data set, and randomly divide the data into a training set and a validation set according to a proportion; An improved MobileNetV3 network is trained based on a training set, and the improved MobileNetV3 network is verified based on a validation set, and the trained improved MobileNetV3 network is used as an abnormal electricity meter recognition model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module with a reverse residual; Inputting the image data of the electric energy meter to be identified into the abnormal electric energy meter identification model, and outputting the electric energy meter abnormality identification result; The improved MobileNetV3 network includes an input layer, a convolutional layer, an activation layer, a bottleneck module with a reverse residual, a global pooling layer, an activation layer, a fully connected layer, a judgment layer, and an output layer connected in sequence; The bottleneck module with reverse residual reduces the amount of calculation while maintaining performance through the structure of expansion, depth convolution and compression, and obtains data after passing through the bottleneck module with reverse residual; The bottleneck module with reverse residual includes a point-by-point convolution expansion unit, a depth-wise convolution unit and a point-by-point convolution compression unit; The point-by-point convolution expansion unit is configured to expand the data with a low number of channels through a first point-by-point convolution expansion operation to obtain data with a high number of channels as expanded data; The depthwise convolution unit is configured to perform independent depthwise convolution on each channel in the expanded data to obtain depthwise convolved data; The point-by-point convolution compression unit is used to compress the depth-convolved data back into data with a low number of channels through a second point-by-point convolution; The expression of the first point-by-point convolution is: expand =H_swish(Conv 1×1 (X)), where X expand It is the data after point-by-point convolution expansion, Conv 1×1 () indicates that a convolution operation is performed using a 1×1 convolution kernel, and X is the independent variable input into the operation; The expression of the second point-by-point convolution is: compress =Conv 1×1 (X depthwise ), where X compress Represents data compressed by point-by-point convolution; The operational expression of the depth convolution is: X depthwise =H_swish(DepthwiseConv 3×3 (X expand )) where X depthwise DepthwiseConv is the data after deep convolution. 3×3 () indicates that the 3×3 convolution kernel performs a depthwise convolution operation, which is expressed as: where Y h,w,c Indicates the value of channel c at position (h, w) after depth convolution; X h+i-1,w+j-1,c Indicates that the data that needs to be convolved is at position (h+i-1,w+j-1), the value of channel c; K i,j,c represents the weight of the 3×3 convolution kernel on the cth channel; b c Indicates the offset of the c-th channel.

2. The electric meter anomaly recognition method based on MobileNetV3 according to claim 1 is characterized in that: Preprocessing the image data to obtain a sample data set includes: Through manual screening, the photographed smart energy meters are classified into two categories: the first category is normal meters, and the second category is abnormal meters. After the classification is completed, a classified sample data set is obtained; The classification method is to save the normal meter pictures in one folder and the abnormal meter pictures in another folder.

3. The electric meter anomaly recognition method based on MobileNetV3 according to claim 1 is characterized in that: The improved loss function of the MobileNetV3 network for: Where i is the sample, y i is the true label, is the predicted label.

4. The electric meter anomaly recognition system based on MobileNetV3 is characterized by: The system includes: an acquisition unit, configured to acquire image data of abnormal electric energy meters and normal electric energy meters; A preprocessing and division unit, configured to preprocess the image data to obtain a sample data set, and randomly divide the data into a training set and a validation set according to a proportion; A model building and training unit is used to train an improved MobileNetV3 network based on a training set, verify the improved MobileNetV3 network based on a validation set, and use the trained improved MobileNetV3 network as an abnormal electricity meter recognition model; the improved MobileNetV3 network is a MobileNetV3 network with a bottleneck module with a reverse residual; an abnormality identification unit, configured to input the image data of the electric energy meter to be identified into the abnormal electric energy meter identification model and output an abnormality identification result of the electric energy meter; The improved MobileNetV3 network includes an input layer, a convolutional layer, an activation layer, a bottleneck module with a reverse residual, a global pooling layer, an activation layer, a fully connected layer, a judgment layer, and an output layer connected in sequence; The bottleneck module with reverse residual reduces the amount of calculation while maintaining performance through the structure of expansion, depth convolution and compression, and obtains data after passing through the bottleneck module with reverse residual; The bottleneck module with reverse residual includes a point-by-point convolution expansion unit, a depth-wise convolution unit and a point-by-point convolution compression unit; The point-by-point convolution expansion unit is configured to expand the data with a low number of channels through a first point-by-point convolution expansion operation to obtain data with a high number of channels as expanded data; The depthwise convolution unit is configured to perform independent depthwise convolution on each channel in the expanded data to obtain depthwise convolved data; The point-by-point convolution compression unit is configured to compress the depth-convolved data back into data with a low number of channels through a second point-by-point convolution; The expression of the first point-by-point convolution is: expand =H_swish(Conv 1×1 (X)), where X expand It is the data after point-by-point convolution expansion, Conv 1×1 () indicates that a convolution operation is performed using a 1×1 convolution kernel, and X is the independent variable input into the operation; The expression of the second point-by-point convolution is: compress =Conv 1×1 (X depthwise ), where X compress Represents data compressed by point-by-point convolution; The operational expression of the depth convolution is: X depthwise =H_swish(DepthwiseConv 3×3 (X expand )) where X depthwise DepthwiseConv is the data after deep convolution. 3×3 () indicates that the 3×3 convolution kernel performs a depthwise convolution operation, which is expressed as: where Y h,w,c Indicates the value of channel c at position (h, w) after depth convolution; X h+i-1,w+j-1,c Indicates that the data that needs to be convolved is at position (h+i-1,w+j-1), the value of channel c; K i,j,c represents the weight of the 3×3 convolution kernel on the cth channel; b c Indicates the offset of the c-th channel.

5. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the electric meter abnormality identification method based on MobileNetV3 as described in any one of claims 1 to 3 is implemented.

Images