Electric power fitting shielding image identification method and system

Abstract

The invention relates to the technical field of image recognition, and provides an electric power fitting shielding image recognition method and system, and the method comprises the steps: obtaining the historical image data of an electric power fitting; preprocessing the acquired historical image data of the electric power fittings, and constructing a first recognition model and a historical image data set; establishing a second identification model; inputting a historical image data set in the second identification model for training, and outputting the category and the shielding condition of the electric power fitting; real-time digital image data of the electric power fittings are acquired and preprocessed through image acquisition equipment, and the electric power fittings are automatically identified by using an electric power fitting shielding identification model. According to the method, through cooperative work of the double models, the identification accuracy of the electric power fitting is effectively improved, in addition, the shielding condition can be identified, the installation position and angle of the electric power fitting can be conveniently inferred subsequently, and therefore more accurate data support is provided for maintenance and management of an electric power system.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for recognizing an image of electric hardware obstruction. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In the daily operation and maintenance of the power system, power fittings play a vital role. They are iron or aluminum metal accessories widely used in power transmission lines. They can protect the wires from the influence of the external environment; provide reliable electrical contact to ensure the flow of current and the transmission of electric energy; prevent the wires from loosening or falling off, so that the wires maintain the correct position and stability during operation. Power fittings can also adjust the length and angle of the wires to make them more suitable for the needs of the power system. In addition, power fittings can distribute electricity to different wires, thereby improving the efficiency of the power system.

[0004] Power fittings are widely used in power systems. In addition to being used in power lines, they are also often used in substations, etc. However, these fittings are often located in complex environments. They are relatively small in size and often block each other due to dense installation, resulting in low recognition accuracy and frequent missed detections, which poses a great challenge to image detection and recognition. Therefore, being able to accurately identify the type of small-sized power fittings and determine the occlusion situation is an urgent problem to be solved. Summary of the invention

[0005] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a power grid fault monitoring method and system, the purpose of which is to solve the problem that the existing image recognition method has low classification accuracy for small-sized power fittings and cannot accurately identify occlusion conditions.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of the present invention provides a method for recognizing an image of an electric power fitting obstruction, comprising:

[0008] Obtain historical image data of power fittings;

[0009] Preprocessing the acquired historical image data of power fittings to construct a first recognition model and a historical image data set;

[0010] Establish a second recognition model; input the historical image data set into the second recognition model for training, and output the category and occlusion of the power fittings;

[0011] The real-time digital image data of the power fittings is collected and preprocessed through the image acquisition equipment, and the power fittings are automatically identified by applying the power fittings occlusion recognition model.

[0012] As a further limitation, the preprocessing includes grayscale processing, normalization processing, noise reduction processing and enhancement processing; wherein, the image is grayscaled using a weighted average method; the image size is determined and the image size is normalized; the image is denoised using a median filter method; and the enhancement processing includes adjusting the image's chroma, brightness, sharpness and contrast.

[0013] As a further limitation, the establishing of the first recognition model is specifically:

[0014] The preprocessed historical image data of power fittings is input into the convolution layer to extract the specific features of the nodes and generate the first feature map; it passes through the activation layer, pooling layer, and fully connected layer in sequence, and finally inputs the softmax layer to output the probability that a node is a power fitting based on the specific features.

[0015] As a further limitation, the specific features include the shape, size, outline, and texture of the node; the nodes include power fittings, conductors, transformers, and insulators.

[0016] As a further limitation, the construction of the historical image dataset is specifically:

[0017] Marking edge contour points of the electric hardware in the feature map output by the first recognition model to obtain a contour point map;

[0018] According to the output of the first recognition model, marking the power fittings labels on the contour point diagram;

[0019] The historical image dataset is constructed using the annotated contour point map.

[0020] As a further limitation, the edge contour points of the power fittings are marked in the feature map output by the first recognition model, specifically:

[0021] Using the improved Canny edge detection, all gray values ​​in the image are represented as {0, 1, 2, ..., Y-1}, a threshold k is selected, and the pixels in the image whose gray values ​​are in the range of [0, k] and [k+1, Y-1] are grouped into q 0 and q 1 ; Use the following formula to make the objective function The maximum threshold k is the optimal threshold:

[0022]

[0023] ω 0 and ω 1 Respectively represent the pixel points belong to q0 and q 1 The probability of 0 and μ 1 Respectively represent q 0 and q 1 The average gray value.

[0024] As a further limitation, a second recognition model is established, specifically,

[0025] S1: Input the historical image dataset into the feature engineering unit to generate a feature map;

[0026] S1.1: Input the historical image data set into a convolution encoder, extract environmental features, and generate a second feature map; the convolution encoder structure is: depth-separable convolution layer-pooling layer-downsampling layer;

[0027] S1.2: A convolution decoder is skip-connected after the convolution encoder to gradually restore the detailed information of the image. The convolution decoder structure is: upsampling-deconvolution layer-activation layer;

[0028] S1.3: Two fully connected layers are connected after the convolution decoder to convert high-dimensional features to low-dimensional features;

[0029] S1.4: Input the Softmax layer, classify the power fittings according to the environmental characteristics, further output the probability that the power fittings belong to the wire fittings, grounding fittings, insulation fittings, protection fittings and bracket fittings, and label them;

[0030] S2: Input the annotated feature map into the candidate region generation unit to generate multiple region anchor boxes; use the softmax activation function to determine whether each anchor box belongs to the foreground, and then use the non-maximum suppression method to filter and generate the corrected anchor box;

[0031] S3: After correcting the anchor frame, input the ROI pooling unit. During the pooling process, each ROI is divided into a fixed number of small areas, and the features in each small area are aggregated to generate a third feature map with a uniform size;

[0032] S4: input the third feature map into the occlusion recognition unit, and determine the occlusion situation according to the coordinate information of the intersection area between the power hardware anchor frames;

[0033] Specifically, two anchor frames of power fittings a and b are selected, and the shielding situation is obtained according to the following formula:

[0034]

[0035] Among them, x a 、x b is the central horizontal coordinate of the anchor frame of power fittings a and b, y a ,yb is the central ordinate of the anchor frames of power fittings a and b, w a 、w b is the width of the anchor frame of power fittings a and b, l a , l b is the length of the anchor frames of power fittings a and b;

[0036] When the anchor frames of power fittings a and b satisfy the above formula, it means that there is no obstruction between power fittings a and b;

[0037]

[0038] When the anchor frames of power fittings a and b satisfy the above formula, it means that there is slight obstruction between power fittings a and b;

[0039]

[0040] When the anchor frames of power fittings a and b satisfy the above formula, it means that there is serious obstruction between power fittings a and b;

[0041] S5: Perform secondary recognition based on the output of the first recognition model, and output the category and occlusion status of the electrical fittings after the secondary recognition.

[0042] A second aspect of the present invention provides an electric hardware occlusion image recognition system, comprising:

[0043] The data acquisition module is configured to: acquire historical image data of power fittings;

[0044] The initial recognition module is configured to: pre-process the acquired historical image data of the power fittings to construct a first recognition model and a historical image data set;

[0045] The secondary recognition module is configured to: establish a second recognition model; input a historical image data set into the second recognition model for training, and output the category and occlusion of the power fittings;

[0046] The real-time identification module for electric fittings is configured to: collect real-time digital image data of electric fittings through image acquisition equipment and perform preprocessing, and automatically identify the electric fittings using the first recognition model and the second recognition model.

[0047] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for recognizing an image of electric power fitting obstruction as described in the first aspect of the present invention.

[0048] The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for recognizing an image of electrical fitting obstruction as described in the first aspect of the present invention are implemented.

[0049] One or more of the above technical solutions have the following beneficial effects:

[0050] The present invention realizes accurate judgment of the obstruction of power fittings by sequentially passing through two recognition models. The first recognition model focuses on identifying the basic characteristics of power fittings, and the second recognition model further analyzes the environmental characteristics of image data. Through the collaborative work of the two models, the recognition accuracy of the power fitting obstruction image can be effectively improved. The improved neural network simplifies the calculation process and can maintain stable working performance even under complex field conditions.

[0051] The first recognition model of the present invention focuses on identifying basic features such as the shape and outline of the node, while the second recognition model focuses on analyzing the usage categories of power fittings in different environments. In addition, the system can also facilitate the subsequent inference of the installation position and angle of the power fittings based on the occlusion situation, thereby providing more accurate data support for the maintenance and management of the power system.

[0052] In the second recognition model of the present invention, the cooperation of the encoder, decoder and fully connected layer can effectively capture key environmental information while reducing unnecessary redundant features. The dimensionality reduction operation not only helps to improve computing efficiency, but also ensures that the model does not ignore the recognition of smaller electrical fittings.

[0053] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0055] Figure 1 is a flow chart of the method of embodiment 1;

[0056] Figure 2 This is a structural diagram of the second recognition model in Example 1. DETAILED DESCRIPTION

[0057] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0058] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0059] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0060] Embodiment 1

[0061] This embodiment discloses a method for recognizing an image of an electric hardware obstruction. Figure 1 As shown, the following steps are included:

[0062] Step 1: Obtain historical image data of power fittings;

[0063] Step 2: pre-process the acquired historical image data of power fittings to construct a first recognition model and a historical image data set;

[0064] Step 3: Establish a second recognition model; input the historical image data set into the second recognition model for training, and output the category and occlusion of the power fittings;

[0065] Step 4: Use image acquisition equipment to collect real-time digital image data of power fittings and perform preprocessing, and apply the power fitting occlusion recognition model to automatically identify the power fittings.

[0066] Specifically,

[0067] In step 1, historical image data of power fittings collected by a camera or other image acquisition equipment is obtained in the power system database.

[0068] In step 2,

[0069] Step 2.1: Preprocess the acquired historical image data of power fittings, including grayscale processing, normalization processing, noise reduction processing and enhancement processing.

[0070] Among them, the weighted average method is used to grayscale the image;

[0071] The images collected by the camera or other imaging devices are RGB images. In order to improve the efficiency of image processing, the weighted average method is used to convert the RGB image into a grayscale image. The formula is as follows:

[0072] Y i =r 0 R i +r1 G i +r 2 B i

[0073] Among them, i represents the pixel point, Y i Represents the gray value of the pixel after processing, R i , G i , B i Respectively represent the R, G, and B values ​​of the pixel, r 0 、r 1 、r 2 Respectively represent the weights of R, G, and B. In this embodiment, r 0 、r 1 、r 2 The values ​​are 0.299, 0.587, and 0.114 respectively.

[0074] The original image size is relatively large, and a fixed size needs to be determined to normalize the image size. In this embodiment, the size is determined to be 3000*2500.

[0075] Median filtering is used to reduce image noise. Median filtering is a nonlinear filtering method that sorts the pixel values ​​in the neighborhood of each pixel in the image by size and takes the middle value to replace the original pixel value. It has a good noise suppression effect, and it can also retain edge information and reduce edge blur.

[0076] The enhancement processing includes adjusting the chroma, brightness, sharpness and contrast of the image. In this embodiment, the opencv computer vision library is used to enhance the image data, and the provided function can effectively improve the clarity of the image. Specifically, the Color function is used to increase the chroma of the image, the Brighter function is used to increase the brightness of the image, the Sharpness function is used to increase the sharpness of the image, and the Contrast function is used to increase the contrast of the image.

[0077] The preprocessed image data can ensure the best image quality and provide a clear visual basis for subsequent analysis.

[0078] Step 2.2: Establish the first recognition model, specifically:

[0079] The first recognition model adopts a convolutional neural network model, which passes through the input layer, convolution layer, activation layer, pooling layer, fully connected layer in sequence, and finally inputs the Softmax layer.

[0080] Among them, the preprocessed historical image data of power fittings is input into the input layer, and the specific features of the nodes are extracted in the convolution layer to generate the first feature map; after the feature map is generated, the activation layer uses a nonlinear activation function, such as the ReLU function, to increase the nonlinear expression ability, so that the model can capture more complex features; the pooling layer reduces the amount of calculation by reducing the spatial dimension of the feature map and maintains the invariance of the features; the fully connected layer integrates the features extracted by the previous layer to form a feature vector; the Softmax layer converts these feature vectors into probability distributions, and outputs the probability that a node is a power fitting according to specific features to achieve node classification.

[0081] Nodes include electrical fittings, conductors, transformers, insulators, etc.; specific features include the shape, size, outline, texture, etc. of the nodes.

[0082] Step 2.3: Construct a historical image dataset, specifically:

[0083] Mark the edge contour points of the power fittings in the feature map output by the first recognition model to obtain a contour point map; specifically:

[0084] Using the improved Canny edge detection, all gray values ​​in the image are represented as {0, 1, 2, ..., Y-1}, a threshold k is selected, and the pixels in the image whose gray values ​​are in the range of [0, k] and [k+1, Y-1] are grouped into q 0 and q 1 Among them, the probability of gray value y in all pixels is defined as p y :

[0085]

[0086] Where n y is the number of pixels with gray value y, and n is the total number of pixels in the image. ;

[0087] The objective function is made as follows The maximum k is the optimal threshold:

[0088]

[0089] Among them, ω 0 and ω 1 Respectively represent the pixel points belong to q 0 and q 1 Probability:

[0090]

[0091]

[0092] μ 0 and μ1 Respectively represent q 0 and q 1 The average gray value of:

[0093]

[0094]

[0095] The threshold of the Canny edge detection algorithm before the improvement is fixed and easily affected by noise, resulting in false edges in the detection results. The improved Canny edge detection algorithm can calculate the optimal threshold according to the size and grayscale value of different images. The obtained contour point map will be more accurate, providing more accurate edge information for subsequent electrical hardware identification.

[0096] According to the output of the first recognition model, marking the power fittings labels on the contour point diagram;

[0097] The historical image dataset is constructed using the annotated contour point map.

[0098] The first identification model can comprehensively cover the key node equipment in the power system, ensuring that it can accurately and comprehensively identify the node type.

[0099] In step 3, a second recognition model is established, specifically,

[0100] Step 3.1: Input the historical image dataset into the feature engineering unit to generate a feature map;

[0101] Step 3.1.1: Input the historical image dataset into the convolution encoder, extract the environmental features, and generate the second feature map; the convolution encoder structure is: depthwise separable convolution layer-pooling layer-downsampling layer. Among them, the depthwise separable convolution layer is used to replace the ordinary convolution layer for feature extraction; the depthwise separable convolution layer uses 1 depthwise separable convolution with a convolution kernel size of 3×3 and 1 ordinary convolution with a convolution kernel size of 1×1; the pooling layer uses average pooling; the convolution kernel size of the downsampling layer is 3×3 and the step size is 1;

[0102] Step 3.1.2: After the convolution encoder, the convolution decoder is skipped to gradually restore the image details. The convolution decoder structure is: upsampling-deconvolution layer-activation layer. The upsampling layer convolution kernel size is 2×2 and the step size is 1; the deconvolution layer transposed convolution kernel size is 3×3 and the step size is 2; the activation layer uses the ReLU function.

[0103] Step 3.1.3: The convolutional decoder is followed by two fully connected layers to convert high-dimensional features to low-dimensional features.

[0104] The number of neurons in the first fully connected layer is 512, and the number of neurons in the second fully connected layer is the number of categories of electrical fittings, in order to achieve the conversion from high-dimensional features to low-dimensional features. In this way, the network can effectively capture key information while reducing unnecessary redundant features. This dimensionality reduction operation not only helps to improve computational efficiency, but also ensures that the model does not ignore the recognition of smaller electrical fittings. The use of deep separable convolutional layers and deconvolutional layers greatly reduces the number of model parameters, improves computational efficiency, and ensures the accuracy of feature extraction.

[0105] In addition, the convolutional encoder is responsible for extracting environmental features, while the convolutional decoder is used to gradually restore the detailed information of the image. In this process, the jump connection between the encoder and the decoder allows the convolutional decoder to directly access the feature maps of different levels in the convolutional encoder, thereby restoring image details while retaining important structural information and improving recognition accuracy.

[0106] Step 3.1.4: Input the Softmax layer, classify the power fittings according to the environmental characteristics, further output the probability that the power fittings belong to wire fittings, grounding fittings, insulating fittings, protective fittings and bracket fittings, and label them.

[0107] Step 3.2: Input the annotated feature map into the candidate region generation unit to generate multiple region anchor boxes; use the Softmax activation function to determine whether each anchor box belongs to the foreground, and then use the non-maximum suppression method to filter and generate the corrected anchor box.

[0108] Anchor frame correction can ensure that the boundaries of each anchor frame are more precise and more in line with the actual shape and size of electrical fittings, thereby improving the accuracy of subsequent classification.

[0109] Step 3.3: After correcting the anchor frame, input the ROI pooling unit. During the pooling process, each ROI is divided into a fixed number of small areas, and the features in each small area are aggregated to generate a third feature map with a uniform size.

[0110] This feature map of uniform size can effectively improve the recognition accuracy of the model, enable features at different levels to be better combined, and further enhance the model's ability to process complex environmental data.

[0111] Step 3.4: Input the third feature map into the occlusion recognition unit, and determine the occlusion situation according to the coordinate information of the intersection area between the power fitting anchor frames.

[0112] Specifically, two anchor frames of power fittings a and b are selected, and the shielding situation is obtained according to the following formula:

[0113]

[0114] Among them, x a 、x b is the central horizontal coordinate of the anchor frame of power fittings a and b, y a ,y b is the central ordinate of the anchor frames of power fittings a and b, w a 、w b is the width of the anchor frame of power fittings a and b, l a , l b is the length of the anchor frames of power fittings a and b;

[0115] When the anchor frames of power fittings a and b satisfy the above formula, it means that there is no obstruction between power fittings a and b;

[0116]

[0117] When the anchor frames of power fittings a and b satisfy the above formula, it means that there is slight obstruction between power fittings a and b;

[0118]

[0119] When the anchor frames of power fittings a and b satisfy the above formula, it means that there is serious obstruction between power fittings a and b;

[0120] Step 3.5: Perform secondary recognition based on the output of the first recognition model, and output the category and occlusion status of the electrical fittings after secondary recognition.

[0121] Step 4: Use image acquisition equipment to collect real-time digital image data of power fittings and perform preprocessing, and apply the power fitting occlusion recognition model to automatically identify the power fittings.

[0122] The first recognition model is used to preliminarily identify the power fittings in various nodes, and the second recognition model is used to identify different categories of power fittings in different usage environments. The collaborative work of the two models can effectively improve the recognition accuracy of the occluded images of power fittings. The first recognition model is mainly responsible for identifying the basic features of the power fittings, such as shape, size, and contour, while the second recognition model focuses on analyzing the categories of power fittings in different environments. In addition, the system can also infer the installation position and angle of the power fittings based on the occlusion situation, thereby providing more accurate data support for the maintenance and management of the power system.

[0123] Embodiment 2

[0124] This embodiment discloses a power fitting occlusion image recognition system, which specifically includes:

[0125] The data acquisition module is configured to: acquire historical image data of power fittings;

[0126] The initial recognition module is configured to: pre-process the acquired historical image data of the power fittings, and construct a first recognition model and a historical image data set;

[0127] The secondary recognition module is configured to: establish a second recognition model; input a historical image data set into the second recognition model for training, and output the category and occlusion of the power fittings;

[0128] The real-time identification module for electric fittings is configured to: collect real-time digital image data of electric fittings through image acquisition equipment and perform preprocessing, and automatically identify the electric fittings using the first recognition model and the second recognition model.

[0129] Embodiment 3

[0130] The purpose of this embodiment is to provide a computer-readable storage medium.

[0131] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in a method for recognizing an image of electric fitting obstruction as described in the first embodiment of the present disclosure.

[0132] Embodiment 4

[0133] The purpose of this embodiment is to provide an electronic device.

[0134] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for recognizing an image of electric fitting obstruction as described in the first embodiment of the present disclosure are implemented.

[0135] The steps involved in the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0136] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0137] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for recognizing an image of an electric power fitting occlusion, characterized in that: include: Obtain historical image data of power fittings; Preprocessing the acquired historical image data of power fittings to construct a first recognition model and a historical image data set; establishing a second recognition model; The historical image data set is input into the second recognition model for training, and the category and occlusion of the electrical fittings are output; The real-time digital image data of the electric fittings is collected and preprocessed through the image acquisition equipment, and the electric fittings are automatically identified by applying the electric fittings occlusion recognition model.

2. The method for recognizing an image of an electric power fitting obstruction according to claim 1, characterized in that: The preprocessing includes grayscale processing, normalization processing, noise reduction processing and enhancement processing; wherein, the image is grayscaled by using a weighted average method; the image size is determined and the image size is normalized; the image is denoised by using a median filter method; and the enhancement processing includes adjusting the chroma, brightness, sharpness and contrast of the image.

3. The method for recognizing an image of an electric power fitting obstruction according to claim 1, characterized in that: The establishing of the first recognition model is specifically as follows: The preprocessed historical image data of power fittings is input into the convolution layer to extract the specific features of the nodes and generate the first feature map; it passes through the activation layer, pooling layer, and fully connected layer in sequence, and finally inputs the softmax layer to output the probability that a node is a power fitting based on the specific features.

4. The method for recognizing an image of an electric power fitting obstruction according to claim 3, characterized in that: The specific features include the shape, size, outline, and texture of the nodes; the nodes include power fittings, conductors, transformers, and insulators.

5. The method for recognizing an image of an electric power fitting obstruction according to claim 1, characterized in that: The construction of the historical image dataset is specifically as follows: Marking edge contour points of the electric hardware in the feature map output by the first recognition model to obtain a contour point map; According to the output of the first recognition model, marking the power fittings labels on the contour point diagram; The historical image dataset is constructed using the annotated contour point map.

6. The method for recognizing an image of an electric power fitting obstruction according to claim 5, characterized in that: The edge contour points of the power fittings are marked in the feature map output by the first recognition model, specifically: Using the improved Canny edge detection, all gray values ​​in the image are represented as {0, 1, 2, ..., Y-1}, a threshold k is selected, and the pixels in the image whose gray values ​​are in the range of [0, k] and [k+1, Y-1] are grouped into q0 and q1; the objective function is The maximum threshold k is the optimal threshold: ω0 and ω1 represent the probability that the pixel belongs to q0 and q1 respectively, μ0 and μ1 represent the average grayscale values ​​of q0 and q1 respectively.

7. The method for recognizing an image of an electric power fitting obstruction according to claim 1, characterized in that: Establish a second recognition model, specifically, S1: Input the historical image dataset into the feature engineering unit to generate a feature map; S1.1: Input the historical image data set into a convolution encoder, extract environmental features, and generate a second feature map; the convolution encoder structure is: depth-separable convolution layer-pooling layer-downsampling layer; S1.2: A convolution decoder is skip-connected after the convolution encoder to gradually restore the detailed information of the image. The convolution decoder structure is: upsampling-deconvolution layer-activation layer; S1.3: Two fully connected layers are connected after the convolution decoder to convert high-dimensional features to low-dimensional features; S1.4: Input the Softmax layer, classify the power fittings according to the environmental characteristics, further output the probability that the power fittings belong to the wire fittings, grounding fittings, insulation fittings, protection fittings and bracket fittings, and label them; S2: Input the annotated feature map into the candidate region generation unit to generate multiple region anchor boxes; The softmax activation function is used to determine whether each anchor box belongs to the foreground, and then the non-maximum suppression method is used to filter it to generate the corrected anchor box; S3: After correcting the anchor frame, input the ROI pooling unit. During the pooling process, each ROI is divided into a fixed number of small regions, and the features in each small region are aggregated to generate a third feature map with a uniform size; S4: input the third feature map into the occlusion recognition unit, and determine the occlusion situation according to the coordinate information of the intersection area between the power hardware anchor frames; Specifically, two anchor frames of power fittings a and b are selected, and the shielding situation is obtained according to the following formula: Among them, x a 、x b is the central horizontal coordinate of the anchor frame of power fittings a and b, y a ,y b is the central ordinate of the anchor frames of power fittings a and b, w a 、w b is the width of the anchor frame of power fittings a and b, l a , l b is the length of the anchor frames of power fittings a and b; When the anchor frames of power fittings a and b satisfy the above formula, it means that there is no obstruction between power fittings a and b; When the anchor frames of power fittings a and b satisfy the above formula, it means that there is slight obstruction between power fittings a and b; When the anchor frames of power fittings a and b satisfy the above formula, it means that there is serious obstruction between power fittings a and b; S5: Perform secondary recognition in combination with the output of the first recognition model, and output the category and occlusion status of the electrical fittings after the secondary recognition.

8. An electric power fitting occlusion image recognition system, characterized in that: include: The data acquisition module is configured to: acquire historical image data of power fittings; The initial recognition module is configured to: pre-process the acquired historical image data of the power fittings, and construct a first recognition model and a historical image data set; The secondary recognition module is configured to: establish a second recognition model; input a historical image data set into the second recognition model for training, and output the category and occlusion of the power fittings; The real-time identification module for electric fittings is configured to: collect real-time digital image data of electric fittings through image acquisition equipment and perform preprocessing, and automatically identify the electric fittings using the first recognition model and the second recognition model.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for recognizing an image of electric fitting obstruction as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for recognizing an image of electrical fitting obstruction as described in any one of claims 1 to 7 are implemented.

Images