A method and system for intelligent identification of distribution network circuit diagram

Abstract

The present application relates to the field of image processing technology, and in particular to a method and system for intelligently identifying distribution network line diagrams. The method comprises the following steps: obtaining a segmented image of a distribution network line diagram; the segmented image includes multiple segmented regions; for multiple segmented images formed by multiple segmentations, obtaining a segmentation score for each segmented image based on the gradient difference between each segmented region and the overall segmented image, thereby obtaining an optimal segmented image; for any segmented region in the optimal segmented image, calculating the detail density of each segmented region based on the gradient characteristics in the segmented region; adjusting the size of a preset block within the segmented region based on the detail density to obtain an adaptive block, wherein the side length of the adaptive block is inversely proportional to the detail density; and processing a historical scanned image using the CLAHE algorithm based on the adaptive block to obtain an enhanced image. The present application has the effect of improving the accuracy of distribution network line diagram recognition.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method and system for intelligently identifying distribution network circuit diagrams. Background Art

[0002] Distribution network diagrams are crucial reference materials for power system design, operation, and maintenance. They contain a large amount of power equipment symbols, connection relationships, and circuit information. Traditional distribution network diagram recognition relies primarily on manual interpretation, which is not only inefficient and labor-intensive, but also prone to recognition errors in scenarios with dense symbols and mixed graphics and text, severely hindering the digital transformation of distribution systems. In recent years, with the advancement of image processing and deep learning technologies, researchers have attempted to use artificial intelligence methods such as convolutional neural networks (CNNs), graph neural networks (GNNs), convolutional recurrent neural networks (CRNNs), and object detection algorithms such as YOLO to automatically recognize and analyze distribution network diagrams. For paper drawings, after capturing the image of the drawing, it is often necessary to process the image and then train a model based on the image to improve the subsequent recognition accuracy. The CLAHE (Contrast Limited Adaptive Histogram Equalization) algorithm is a common image enhancement method in image processing. Its core principle is to divide the image into sub-blocks of several sizes (such as 8×8 or 32×32 pixels), perform histogram equalization on each sub-block to enhance the local contrast, and then use bilinear interpolation to smooth the transition between blocks.

[0003] However, the density of text and symbols in different areas of the actual distribution network diagram is different. Figure 1 High-density areas contain a wealth of details and a mixture of text and symbols. This causes the traditional CLAHE algorithm to enhance image feature contrast, resulting in artifacts at block boundaries and local over-enhancement. For example, the "%" symbol in "69%~1" is broken, the "column" in "newly installed pole-mounted transformer" is blurred, and the contrast imbalance in the overlapping area between the dotted line and text in "newly built low-voltage three-phase four-wire overhead line" is unbalanced. This results in low OCR recognition accuracy, which in turn affects the accuracy of intelligent recognition of distribution network diagrams. Summary of the Invention

[0004] In order to solve the problem of low accuracy in traditional distribution network circuit diagram recognition, the present application provides a distribution network circuit diagram intelligent recognition method and system.

[0005] In a first aspect, the present application provides a method for intelligently identifying a distribution network diagram, which adopts the following technical solutions:

[0006] A method for intelligently identifying distribution network line diagrams comprises the following steps: scanning historical circuit drawings to obtain historical scanned images; performing multiple unequal segmentation on each historical image to form a segmentation map after segmentation; the segmented image comprises multiple segmented regions, and for the multiple segmented maps formed by the multiple segmentation, obtaining a segmentation score for each segmented map based on a gradient difference between each segmented region in the segmented map and the segmented map as a whole, thereby obtaining an optimal segmented image; for any segmented region in the optimal segmented image, calculating the detail density of each segmented region based on a gradient feature in the segmented region, adjusting the size of a preset block within the segmented region based on the detail density to obtain an adaptive block, wherein the side length of the adaptive block is inversely proportional to the detail density; processing the historical scanned image using the CLAHE algorithm based on the adaptive block to obtain an enhanced image; using the enhanced image to train a YOLO model to obtain a detection model, and using the detection model to identify the drawings.

[0007] In this method, the historical scanned image is segmented multiple times based on different segmentation strategies, and the optimal segmented image is selected based on the segmentation score of the segmented image. In the optimal segmented image, the historical scanned image is accurately divided into areas with more details and areas with less details. In the optimal segmented image, the detail density of the segmented area is calculated, and the size of the detail density indicates the amount of detail in the segmented area. The size of the blocks in the segmented area is adjusted based on the amount of detail in the segmented area. Based on different historical scanned images, the side lengths of the adaptive blocks of each historical scanned image and each segmented area are dynamically adjusted, thereby improving the image processing effect and the subsequent accuracy of model training and the use of the detection model to recognize drawings.

[0008] Optionally, the step of obtaining the segmentation score of each segmentation image based on the gradient difference between each segmentation area in the segmentation image and the segmentation image as a whole includes: obtaining the gradient mean of the pixel points in the segmentation area in the segmentation image, and using the gradient mean as the local gradient level; using the mean of multiple local gradient levels as the overall gradient level; and using the mean of the absolute difference between each local gradient level and the overall gradient level as the segmentation score.

[0009] In this method, if a segmentation map can better distinguish between areas with more details and areas with less details, then the segmentation score of the segmentation map corresponding to the segmentation strategy is considered to be higher. When the mean gradient in a segmented image is high, it means that the area contains more details. If the mean gradient in a region is small, it means that the segmented area has less details. If a segmentation map can distinguish between the two well, then the score of the segmentation map will be higher. Therefore, this method calculates the mean of the absolute difference between the local gradient level and the global gradient level. The larger the mean, the higher the segmentation score.

[0010] Optionally, the segmentation image with the largest segmentation score is used as the optimal segmentation image.

[0011] Optionally, the step of calculating the detail density of each segmented region based on the gradient features in the segmented region includes: obtaining the gradient fluctuation of the segmented region; and taking the product of the gradient fluctuation and the local gradient level as the detail density of the segmented region.

[0012] The detail density in the segmented area is mainly based on the number of detail edges in the segmented area. If the gradient fluctuation of the pixel points in a segmented area is large, it means that the segmented area contains edge information, and then the size of the detail density in the segmented area is reflected based on this feature.

[0013] Optionally, the step of adjusting the size of the preset block within the segmented area according to the detail density to obtain the adaptive block includes: using the normalized result of the detail density as the adjustment index, using the difference between the custom constant and the adjustment index as the influence coefficient, and using the product of the influence coefficient and the side length of the preset block as the side length of the adaptive block.

[0014] Based on the principle of the CLAHE algorithm, smaller blocks are set for areas with high detail density to preserve more details. For areas with low detail density, larger blocks are used for enhancement to improve processing efficiency.

[0015] Optionally, the step of using the enhanced image to train the YOLO model to obtain the detection model includes: using convolution kernels of different scales to obtain the feature map of the enhanced image, and calculating the degree of detail retention of the feature map compared to the enhanced image; adjusting the spatial attention weight of each pixel point in the YOLO model based on the degree of detail retention to obtain a comprehensive attention weight, training the YOLO model based on the comprehensive attention weight, and obtaining the detection model after the training is completed.

[0016] The CBAM module in the traditional YOLO model uses fixed-size convolution kernels to obtain pixel attention weights, which can easily lose image details. Therefore, this method processes historical scanned images using convolution kernels of different scales. Based on the degree of detail retained in the processed images at different scales, it obtains comprehensive attention weights, further improving model training and the accuracy of subsequent distribution network line diagram recognition.

[0017] Optionally, the step of calculating the degree of detail retention of the feature map compared to the enhanced image includes: for the feature map convolved at any scale, obtaining the detail loss and pixel change of the segmented area; taking the product of the detail loss and the pixel change as the overall loss, and obtaining the degree of detail retention based on the overall loss, and the overall loss is proportional to the degree of detail retention.

[0018] After the enhanced image is processed by the convolution kernel, the image details will be lost to a certain extent. Here, the degree of detail retention when the historical scan image is processed by convolution kernels of different scales is determined based on the changes in the detail density and pixel values ​​of the historical scan image before and after convolution.

[0019] Optionally, the step of obtaining the detail loss amount and the pixel change amount of the segmented area includes: obtaining the detail density of each segmented area after convolution as the retained density; and taking the difference between the detail density of the segmented area before convolution and the retained density as the detail loss amount;

[0020] The mean pixel value of all pixels in the segmented area of ​​the feature map is taken as the pixel level in the segmented area; the mean of the absolute difference between each pixel and the pixel level is taken as the pixel change.

[0021] Optionally, the step of adjusting the spatial attention weight of each pixel in the YOLO model based on the degree of detail preservation to obtain a comprehensive attention weight includes: for any pixel, obtaining the degree of detail preservation of the segmented area corresponding to the pixel; and taking the product of the normalized result of the detail preservation degree and the spatial attention weight as the total attention weight of the pixel.

[0022] In a second aspect, the present application provides a distribution network circuit diagram intelligent recognition system, which adopts the following technical solutions:

[0023] A distribution network circuit diagram intelligent identification system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the distribution network circuit diagram intelligent identification method described above is implemented.

[0024] The above-mentioned method for intelligent identification of distribution network line diagram is generated into a computer program and stored in a memory so as to be loaded and executed by a processor. Thus, a system is made based on the memory and the processor for easy use.

[0025] This application has the following technical effects:

[0026] According to the detail density in different segmented areas, adaptive blocks of different sizes are set for each segmented area, so that more details of the historical scanned images can be retained during the enhancement process, thereby improving the image enhancement effect; ultimately, the accuracy of the subsequent YOLO model in recognizing drawings is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is the distribution network circuit diagram in the background technology of this application.

[0028] Figure 2 This is a method flow chart of a method for intelligent identification of distribution network circuit diagrams in this application.

[0029] Figure 3 This is an image segmentation reference image of the intelligent recognition method for distribution network line diagrams in this application (mainly used to show the effect after image segmentation in step S2). DETAILED DESCRIPTION

[0030] The embodiment of the present application discloses a method for intelligent recognition of distribution network line diagrams, which scans circuit drawings and obtains historical scanned images. Different segmentation strategies are preset, and each historical scanned image is segmented to obtain multiple segmentation maps corresponding to the historical images. The optimal segmentation image is obtained based on the segmentation score of the segmentation map, and then adaptive blocks of different sizes are set according to the detail density of different segmented areas in the optimal segmentation image, thereby improving the effect of enhancing the historical scanned images. Compared with traditional image enhancement methods, the method in the present application can divide blocks of different sizes based on the detail density (the amount of details) of different areas in the image, effectively retain image details, reduce contrast imbalance and other problems, and improve OCR recognition accuracy.

[0031] Reference Figure 2 A method for intelligently identifying a distribution network diagram includes steps S1 to S5.

[0032] S1: Scan historical circuit drawings to obtain historical scan images.

[0033] A high-precision scanner was used to scan various historical distribution network line drawings to obtain historical scanned images. Non-local mean denoising was used to remove stains and wrinkles from the scanned images. The Hough transform was used to detect the line diagram borders, calculate the rotation angle, and correct the image to obtain the historical scanned images.

[0034] In this embodiment, the distribution network line diagram includes two types: one is the entire line diagram, which is a drawing used as a reference during the historical construction process; the other is the image of the symbols in the line diagram, in which one symbol corresponds to one image.

[0035] A historical distribution network diagram corresponds to a historical scanned image, and a distribution network diagram to be identified corresponds to a scanned image to be identified. The historical scanned images contain a variety of mixed distribution network symbols, and each distribution network symbol image corresponds to a symbol.

[0036] S2: Perform multiple unequal segmentations on each historical image to form a segmentation map. The segmentation image includes multiple segmentation regions. For the multiple segmentation maps formed by the multiple segmentation processes, obtain a segmentation score for each segmentation map based on the gradient difference between each segmentation region in the segmentation map and the overall segmentation map to obtain the optimal segmentation image.

[0037] For any historical scanned image, first divide it into rectangular regions of different sizes. An example segmentation method is as follows: Figure 3 , each historical scan image is divided into multiple rectangular areas, and each rectangular area is used as a segmentation area. At the same time, each historical scan image is segmented multiple times, which can also be understood as using different segmentation strategies to segment the image. The optimal segmentation image is selected based on the segmentation score of the segmentation map formed after segmentation. In this process, a large number of segmentation maps generated by segmentation strategies will be traversed, and the optimal segmentation image will be selected based on the segmentation scores of different segmentation maps. The size of the segmentation score indicates the quality of the segmentation map generated by the segmentation strategy. In order to facilitate calculation and processing, the number of rectangular areas (segmentation areas) formed after segmentation can be limited in this process, for example, the number of segmentation areas in the segmentation map is not greater than 20.

[0038] For multiple segmentation images formed by multiple segmentation, the segmentation score of each segmentation image is obtained according to the gradient difference between each segmentation area in the segmentation image and the segmentation image as a whole, and the optimal segmentation image is obtained.

[0039] S3: Obtain the gradient mean of the pixel points in the segmentation area in the segmentation map, and use the gradient mean as the local gradient level; use the mean of multiple local gradient levels as the overall gradient level; and use the mean of the absolute differences between each local gradient level and the overall gradient level as the segmentation score.

[0040] Specifically, the calculation formula of the segmentation score of the segmentation map can be expressed as:

[0041] Where, represents the segmentation score of the segmentation map, Represents the number of segmented regions in the segmentation map, Indicates the The local gradient level of the segmented region, Indicates the overall gradient level.

[0042] The formula is mainly used to determine whether the areas with more details and the areas with less details in the historical scanned image are effectively segmented. If the gradient level in a segmented area is higher, it means that the segmented area contains more edge details. It represents the difference between the gradient amplitude of the segmented area and the gradient amplitude of the whole image. The larger this part is, the better the image segmentation effect is. For example, for a segmented area with a high gradient amplitude, it means that the segmented area is rich in details. At the same time, if the gradient amplitude of a segmented area is low, it means that the segmented area has few details and may belong to the background area of ​​the image. The larger the value, the more effective and accurate the segmentation.

[0043] The same historical scanned image forms different segmentation maps based on different segmentation strategies, calculates the segmentation score of each segmentation map, and takes the segmentation map with the largest segmentation score as the optimal segmentation image.

[0044] S4: For any segmented area in the optimal segmentation image, the detail density of each segmented area is calculated according to the gradient features in the segmented area; the size of the preset blocks inside the segmented area is adjusted according to the detail density to obtain adaptive blocks, and the side length of the adaptive blocks is inversely proportional to the detail density.

[0045] Obtain the gradient fluctuation of the segmented region; and use the product of the gradient fluctuation and the local gradient level as the detail density of the segmented region.

[0046] In this embodiment, the standard deviation of the gradient amplitude of each pixel in the segmented region is used as the gradient fluctuation of the segmented region. When the variation of the gradient amplitude between each pixel in the segmented region is large, it means that the segmented region contains more edge details, and thus the detail density of the region is greater.

[0047] The normalized result of detail density is used as the adjustment index, the difference between the custom constant and the adjustment index is used as the influence coefficient, and the product of the influence coefficient and the side length of the preset block is used as the side length of the adaptive block.

[0048] Specifically, the calculation formula of the side length of the adaptive block can be expressed as:

[0049] Where, Indicates the The adaptive block side length of the segmented area (in the calculation process, if the value is not an integer, it will be rounded down), Indicates a preset block side length, an example The value can be 32. Indicates the The detail density of the segmented area, It represents the maximum value of the detail density of the segmented area in all segmentation maps. It is mainly used to normalize the detail density corresponding to the segmented area to facilitate subsequent calculations. Indicates a custom constant. In this embodiment, it is a custom constant The main function of 1 is to make the normalized result of detail density inversely proportional to the side length of the final adaptive block.

[0050] Since the CLAHE algorithm reconstructs pixel values ​​through the pixel values ​​and their distribution within the block, the area with greater detail density should have smaller blocks to avoid artifacts or blurring caused by interference of details by neighboring pixels; the area with lower detail density should have larger blocks to accelerate image feature enhancement due to fewer details.

[0051] refer to Figure 3 The sizes of the adaptive blocks corresponding to different segmented areas in the optimal segmented image are different, so that the historical scanned image can be enhanced more accurately according to the detailed features of the image. Based on the adaptive block, the CLAHE algorithm is used to enhance the features and contrast of the historical scanned image to obtain an enhanced image.

[0052] S5: The historical scanned image is processed using the CLAHE algorithm based on adaptive block segmentation to obtain an enhanced image; the enhanced image is used to train the YOLO model to obtain a detection model, and the detection model is used to recognize the drawing.

[0053] The CBAM (Convolutional Block Attention Module, attention mechanism) module in the traditional Yolo model adopts The fixed-size convolution kernel obtains the spatial attention weight of each pixel in the enhanced image, resulting in the local features being diluted by the weighted average of the surrounding background pixels when the distribution network symbol is smaller than the convolution kernel receptive field. The attention weight will be dispersed when the convolution kernel receptive field cannot cover the complete structure of the distribution network symbol or line. In local areas, the spatial attention weight of pixels is too small, the weight is empty, and key details are ignored. Therefore, this application uses convolution kernels of different sizes to obtain feature maps of different scales for historical scanned images, calculates the importance of each scale based on the difference in detail density changes before and after convolution of each segmented area, and calculates the comprehensive spatial weight through the importance and the spatial attention weight of each pixel at different scales.

[0054] Specifically, first, convolution kernels of different scales are used to obtain the feature map of the enhanced image, and the degree of detail preservation of the feature map compared to the enhanced image is calculated.

[0055] For the feature map of convolution at any scale, the detail loss and pixel change of the segmented area are obtained; the product of the detail loss and pixel change is taken as the overall loss, and the degree of detail retention is obtained based on the overall loss. The overall loss is proportional to the degree of detail retention.

[0056] Obtain the detail density of each segmented area after convolution as the retained density; the difference between the detail density of the segmented area before convolution and the retained density is taken as the detail loss amount;

[0057] The mean pixel value of all pixels in the segmented area of ​​the feature map is taken as the pixel level in the segmented area; the mean of the absolute difference between each pixel and the pixel level is taken as the pixel change.

[0058] Specifically, the calculation formula for the degree of detail preservation before and after image convolution can be expressed as:

[0059] ; Representation scale The convolution kernel of the feature map after convolution The degree of detail preservation in a rectangular area, Indicates the The detail density of the segmented area, Represents the first feature map after convolution The detail density of each segmented region; Indicates the number of the historical scanned images The pixel change of each segmented area.

[0060] Where, It represents the difference between the detail density of a segmented area in the historical scan image and the detail density of the same area in the convolved feature map. Reflects the amount of detail loss after convolution. The smaller this part is, the more image details the convolution kernel of this scale can retain. Similarly, This represents the change in pixel values ​​before and after convolution in a segmented region, reflecting the degree of detail loss. The greater the degree of detail preservation, the greater the influence of surrounding pixels on the pixel after convolution kernel processing of that scale, and thus the lower the confidence level of spatial attention obtained by the CBAM in the traditional YOLO model.

[0061] Based on the degree of detail preservation, the spatial attention weight of each pixel in the YOLO model is adjusted to obtain the comprehensive attention weight.

[0062] For any pixel point, obtain the detail retention degree of the segmented area corresponding to the pixel point; take the product of the normalized result of the detail retention degree and the spatial attention weight as the comprehensive attention weight of the pixel point.

[0063] Specifically, the calculation formula of the comprehensive attention weight can be expressed as: Where, Represents the comprehensive attention weight of the pixel; Indicates the number of convolution kernels; Indicates that the convolution kernel is scale the degree of detail preservation when Indicates that the pixel point obtained by the CBAM module is at scale The spatial attention weights on .

[0064] The YOLO model is trained based on the comprehensive attention weights, and a detection model is obtained after training. A historical scanned image is input, and the target box and the name of the symbol within the target box are output. After a preset number of training cycles, the detection model is obtained. The distribution line diagram to be identified is scanned to obtain the scanned image to be identified. The scanned image to be identified is enhanced to obtain an enhanced image to be identified. This enhanced image to be identified is input into the detection model to obtain the target box and name of the distribution network symbol in the enhanced image to be identified. The output results can be manually verified for secondary verification to achieve intelligent and accurate recognition.

[0065] An embodiment of the present application also discloses a distribution network circuit diagram intelligent identification system, including a processor and a memory, the memory storing computer program instructions, and when the computer program instructions are executed by the processor, a distribution network circuit diagram intelligent identification method according to the present application is implemented.

[0066] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0067] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for intelligently identifying a distribution network diagram, characterized in that: The method comprises the following steps: scanning historical circuit drawings to obtain historical scanned images; performing multiple unequal segmentation on each historical image to form a segmentation map after segmentation; the segmented image comprises multiple segmentation regions, and for the multiple segmentation maps formed by the multiple segmentation, obtaining a segmentation score for each segmentation map based on the gradient difference between each segmentation region in the segmentation map and the overall segmentation map, thereby obtaining an optimal segmentation image; for any segmentation region in the optimal segmentation image, calculating the detail density of each segmentation region based on the gradient features in the segmentation region, including: obtaining the gradient mean of the pixel points in the segmentation region in the segmentation map, and using the gradient mean as the local gradient level; using the standard deviation of the gradient amplitude of each pixel point in the segmentation region as the gradient fluctuation of the segmentation region; and using the product of the gradient fluctuation and the local gradient level as the detail density of the segmentation region; The size of the preset blocks within the segmented area is adjusted according to the detail density to obtain adaptive blocks, and the side length of the adaptive blocks is inversely proportional to the detail density; the historical scanned image is processed using the CLAHE algorithm based on the adaptive blocks to obtain an enhanced image; the enhanced image is used to train the YOLO model to obtain a detection model, including: using convolution kernels of different scales to obtain the feature map of the enhanced image, and calculating the degree of detail retention of the feature map compared to the enhanced image; based on the degree of detail retention, the spatial attention weight of each pixel point in the YOLO model is adjusted to obtain a comprehensive attention weight, and the YOLO model is trained based on the comprehensive attention weight. After the training is completed, the detection model is obtained; and the detection model is used to recognize the drawings.

2. A method for intelligently identifying a distribution network circuit diagram according to claim 1, characterized in that: The step of obtaining the segmentation score of each segmentation image according to the gradient difference between each segmentation area in the segmentation image and the overall segmentation image includes: taking the average of multiple local gradient levels as the overall gradient level; and taking the average of the absolute differences between each local gradient level and the overall gradient level as the segmentation score.

3. The method for intelligently identifying a distribution network diagram according to claim 1, wherein: The segmentation image with the largest segmentation score is taken as the optimal segmentation image.

4. A method for intelligently identifying a distribution network diagram according to claim 1, characterized in that: The steps of adjusting the size of the preset blocks within the segmented area according to the detail density to obtain adaptive blocks include: using the normalized result of the detail density as the adjustment index, using the difference between the custom constant and the adjustment index as the influence coefficient, and using the product of the influence coefficient and the side length of the preset block as the side length of the adaptive block.

5. The method for intelligently identifying a distribution network diagram according to claim 1, wherein: The steps of calculating the degree of detail preservation of the feature map compared to the enhanced image include: for the feature map convolved at any scale, obtaining the detail loss and pixel change of the segmented area; taking the product of the detail loss and the pixel change as the overall loss, and obtaining the degree of detail preservation based on the overall loss, and the overall loss is proportional to the degree of detail preservation.

6. A method for intelligently identifying a distribution network diagram according to claim 5, characterized in that: The step of obtaining the detail loss amount and the pixel change amount of the segmented area includes: obtaining the detail density of each segmented area after convolution as the retained density; and taking the difference between the detail density of the segmented area before convolution and the retained density as the detail loss amount; The mean pixel value of all pixels in the segmented area of ​​the feature map is taken as the pixel level in the segmented area; the mean of the absolute difference between each pixel and the pixel level is taken as the pixel change.

7. The method for intelligently identifying a distribution network diagram according to claim 5, characterized in that: The steps of adjusting the spatial attention weight of each pixel in the YOLO model based on the degree of detail preservation to obtain the comprehensive attention weight include: for any pixel, obtaining the degree of detail preservation of the segmented area corresponding to the pixel; and taking the product of the normalized result of the detail preservation degree and the spatial attention weight as the total attention weight of the pixel.

8. A distribution network line diagram intelligent recognition system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for intelligently identifying a distribution network circuit diagram according to any one of claims 1 to 7 is implemented.

Images