Material matching method for power distribution network engineering

Abstract

The invention discloses a material matching method for a power distribution network project, and relates to the technical field of material matching, and the method comprises the following steps: collecting a material image of a completed part in a construction process of the power distribution network project in real time as an initial image; performing preprocessing and target extraction on the initial image to obtain a feature image; training a material matching model based on a support vector machine according to the feature image and a preset power distribution network material drawing, and generating a material matching degree; the material matching degree is compared with a preset material matching threshold value, and whether material matching of the completed power distribution network project is abnormal or not is judged; and when the material matching is abnormal, reworking is performed on the completed power distribution network project, and early warning and adjustment are performed on the power distribution network project to be performed. According to the method, the abnormal condition of material matching of the completed power distribution network is judged, so that material matching errors in the subsequent construction process are reduced, extra cost caused by large rework or repair is reduced, resource utilization is optimized, and budget is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of material matching, and more specifically, to a material matching method for a distribution network project. Background Art

[0002] In 2019, State Grid Corporation of China proposed the goal of "Three Types and Two Networks, World-Class", of which the "Three Types" refers to the State Grid Corporation of China building a shared, platform-based, and hub-based power grid enterprise, and the "Two Networks" refers to the operation and construction of a ubiquitous power Internet of Things and a strong power grid. The construction of a ubiquitous power Internet of Things was first proposed by State Grid Corporation of China, with the aim of making full use of big data technology, making good use of the big data information within the power grid company, and sharing it with the society and power customers, keeping pace with the times, and making greater contributions to power grid construction and people's better lives.

[0003] The so-called ubiquitous power Internet of Things is to combine the current power grid with information technology and communication technology such as artificial intelligence and mobile Internet, so as to realize human-machine interaction and the interconnection of all things, and thus develop the power grid into an intelligent system that can flexibly perceive information, efficiently process information, and flexibly apply information. Therefore, it can be foreseen that using machines to replace manual labor to achieve refined work management and streamline labor costs will become the work direction of the State Grid Corporation of China for some time to come, and it is also the key goal of the power grid in the field of control research for some time to come.

[0004] The foundation for building a ubiquitous power Internet of Things is supported by power grid data. Therefore, the collection and organization of basic power grid data is particularly important. Due to the large power supply area of ​​my country's power grid and the scattered power supply population, most distribution network lines are still single-power branch networks. Therefore, the amount of data to be collected is huge and the data collection is difficult. Traditional distribution network information collection often uses experienced employees to hand-draw the distribution network on foot, and then re-draw the drawings in front of the computer using AutoCAD or other drawing software. After that, the distribution network line management personnel will analyze the electronic circuit diagram information one by one, count each distribution network component, and then use the on-site work experience to calculate the distribution network assets, that is, equipment information, which greatly wastes labor costs. At the same time, due to the lack of experience of line management personnel, calculation errors and other reasons, major errors in the final line data collection are caused, which often cause errors in information upload, and shortages of materials in the construction process caused by inaccurate data during line transformation, as well as the failure of the final settlement of the project.

[0005] In the process of modern industry and urbanization, the power distribution network plays a vital role. It is responsible for transmitting electrical energy from power stations to end users. As a key component of the power system, the distribution network engineering covers the entire process from high-voltage transmission to low-voltage distribution. Its design and construction are crucial to ensure the stability, safety and efficiency of power supply.

[0006] The above disclosed technical solutions have at least the following technical problems: the traditional distribution network construction cannot adjust the engineering construction by real-time material matching, which leads to the subsequent readjustment of the distribution network engineering, thereby increasing the total cost of the project and reducing the construction efficiency.

[0007] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0008] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a material matching method for a distribution network project, which solves the problems of increased cost and reduced efficiency through material matching.

[0009] To achieve the above object, the present invention provides the following technical solutions: A material matching method for a distribution network project comprises the following steps: real-time acquisition of a material image of a completed part in the construction process of the distribution network project as an initial image; preprocessing the initial image and extracting a target to obtain a feature image; training a material matching model based on a support vector machine according to the feature image and a preset distribution network material drawing to generate a material matching degree; comparing the material matching degree with a preset material matching threshold to determine whether the material matching of the completed distribution network project is abnormal; when the material matching is abnormal, reworking the completed distribution network project, and giving an early warning and adjustment to the upcoming distribution network project.

[0010] In a preferred embodiment, the preprocessing includes image grayscale, image noise reduction and edge detection.

[0011] In a preferred embodiment, the specific steps of the edge detection are as follows: preprocessing the initial image to convert the initial image into a grayscale image; using an edge filter to smooth the grayscale image to remove noise to obtain a filtered image; using threshold processing to extract the highlight edges in the filtered image and calculate the gradient amplitude and direction of the highlight edges based on the Canny edge detection operator; suppressing non-edge pixels along the gradient direction and retaining the local maximum of the highlight edge gradient amplitude; classifying edge points based on a preset threshold, and processing the extracted edges according to the detection needs, the processing including edge refinement, contour generation, merging non-continuous edges and separating backgrounds.

[0012] In a preferred embodiment, the method for acquiring the feature image is as follows: outlining the edge of the image after edge detection, and fitting the outlined curve into a number of regular fitting circles; after generating the fitting circles, determining the coordinates of the center of the circle and the radius of the circle, and using the coordinates of the circle as the center of the circle and a number of pixel points as the side length, the preprocessed image is cropped to obtain the feature image.

[0013] In a preferred embodiment, the method for obtaining the fitting circle is as follows: based on three random known points in the image after edge detection, any two perpendicular bisectors corresponding to the three points are obtained, and the fitting circle is obtained based on the intersection of the two perpendicular bisectors, and circles with diameters close to the known image are screened out, while circles with diameters that are not close are discarded.

[0014] In a preferred embodiment, the specific method for obtaining the material matching model is as follows: extracting features from the feature image and the preset distribution network material drawing and performing normalization processing; identifying the features in the feature image that are consistent with the preset material drawing as a match, otherwise it is a mismatch; constructing a data set based on the feature image and the preset distribution network material drawing, the data set includes positive samples and negative samples, the positive samples represent instances where the feature image and the material drawing match, and the negative samples represent instances where the feature image and the material drawing do not match; training the material matching model based on the data set using a kernel function and adjusting the model parameters in real time during the training process to obtain the optimal matching algorithm; applying the optimal matching algorithm to the matching of the initial image; the material matching model outputs the predicted matching result, i.e., the material matching degree.

[0015] In a preferred embodiment, the specific method for obtaining the optimal matching algorithm is as follows: mapping the training samples in the training set to a high-order feature space through a kernel function; in the high-order feature space, introducing a non-negative relaxation coefficient to determine the linear separability of the training samples, and obtaining a decision function; performing a quadratic programming operation on the feature space based on the decision function, and solving the dual problem of the quadratic programming by introducing Lagrange multipliers, thereby obtaining the optimal matching algorithm; the quadratic programming operation is specifically: in the high-dimensional feature space, the SVM searches for the plane that maximizes the distance to the nearest training sample.

[0016] Technical effects and advantages of a material matching method for distribution network engineering of the present invention: 1. The present invention greatly reduces the waste of labor costs by training the material matching model. Accurate material matching ensures a smooth construction process, improves project progress and efficiency, ensures the use of materials that meet specifications, effectively reduces safety hazards caused by material mismatch, and improves the operational safety of the distribution network.

[0017] 2. The present invention reduces material matching errors in subsequent construction processes by judging abnormal conditions of material matching in completed distribution networks, reduces additional costs caused by extensive rework or repair, optimizes resource utilization, and saves budget. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The present invention is a structural schematic diagram of a material matching method for a distribution network project. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Embodiment 1, Figure 1 The present invention provides a material matching method for a distribution network project, and the steps are as follows: S1 collects material images of the completed parts of the distribution network project in real time as initial images.

[0021] The specific steps of collection are as follows: Determine the time and location for collection: after a section of line is completed at the construction site.

[0022] Prepare shooting equipment: Use a digital camera with sufficient resolution and quality.

[0023] Take images of materials; take pictures of completed parts to ensure that key material details and locations are captured, such as cable joints, installed transformers or switchgear.

[0024] Record relevant information: The captured images should at least include the shooting time, location, construction stage, and material model.

[0025] Upload and store images: Upload the captured material images to a designated storage location or project management system.

[0026] In this embodiment, the distribution network project is to assemble the distribution network using the same construction method. The distribution network project refers to the project invested by the power grid company to build or renovate 10kV and below lines and equipment. The 10kV distribution network is very important in the entire power grid and is the foundation of the power grid. The construction of the distribution network project can improve the 10kV power grid structure and make the distribution network structure more reasonable, thereby improving the power supply reliability and the safe operation level of the distribution network.

[0027] The section of the power system from the step-down distribution substation (high-voltage distribution substation) to the user end is called the distribution system. The distribution system is a power network system composed of a variety of distribution equipment (or components) and distribution facilities that transforms voltage and directly distributes electric energy to end users.

[0028] The distribution network consists of overhead lines, poles, cables, distribution transformers, switchgear, reactive power compensation capacitors and other distribution equipment and ancillary facilities. Its main function in the power grid is to distribute electrical energy. From the perspective of the nature of the distribution network, the distribution network equipment also includes the distribution equipment of the substation.

[0029] S2 preprocesses the initial image and extracts the target to obtain a feature image.

[0030] Preprocessing includes image grayscale, image denoising and edge detection.

[0031] Image grayscale conversion is the process of converting a color image into a grayscale image. In a color image, each pixel is usually composed of three color channels: red, green, and blue (RGB), and the value range of each channel is 0 to 255. A grayscale image has only one channel, and the value range of each pixel is usually 0 to 255, indicating the depth of the grayscale.

[0032] In this embodiment, the initial image with a length of N and a width of M is represented in the form of a grayscale matrix:

[0033] In the grayscale matrix, each pixel is Indicates that It represents the color information of this pixel, where i and j represent the horizontal and vertical coordinate points in the image respectively. The color information also includes three channel information of red (R), green (G), and blue (B).

[0034] In image analysis, in order to reduce the storage space of the image and retain the color information and position information of the image, the image is converted into a grayscale image. Transformed into , The range is 0~255, and the formula for converting an image into a grayscale image is as follows.

[0035]

[0036] Using the color information of each pixel , as well as the position information i and j, we can get the M*N matrix of the final grayscale image, as shown below.

[0037]

[0038] In the field of image processing, the common way to reduce noise in images is to use blur processing. Blur processing is also called smoothing processing. Blur processing is a very common and very simple filtering and noise reduction method that can effectively reduce noise and distortion in images.

[0039] The principle of image filtering is that the main information of an image is mostly retained in the mid-frequency band and the low-frequency band, while the high-frequency band is usually disturbed by noise. Therefore, generally speaking, reducing the interference of the high-frequency band can reduce the impact of noise on the image, thereby obtaining a clearer result. Image denoising can reduce the noise of the digitized image to facilitate subsequent image processing, while strengthening the key points of image features to facilitate subsequent feature extraction.

[0040] At present, image denoising approaches can be divided into linear filtering and nonlinear filtering. Linear filtering can be specifically divided into mean filtering, Gaussian filtering and square filtering, while nonlinear filtering can be specifically divided into bilateral filtering and median filtering.

[0041] It should be noted that this embodiment uses Gaussian filtering to reduce noise on the image.

[0042] Gaussian filtering is the most common filtering method currently used. With this filtering mode, it can effectively remove noise while retaining the details of the image. Compared with mean filtering, Gaussian filtering also uses the information around the original pixel to replace the original pixel information, but the difference is that the pixel points of this kernel do not take the absolute average value, but use the weighted average value to replace the original pixel points. From a mathematical point of view, the process of Gaussian filtering is to perform a convolution algorithm on the image and the normal distribution. Since the normal distribution is also called Gaussian distribution in mathematics, it is also called Gaussian filtering, which is equivalent to a low-pass filtering of the image. The image processed by Gaussian filtering looks like the original image through a frosted glass to the human eye. Unlike the shadow effect and out-of-focus bokeh effect, it will generate an accurate out-of-focus imaging. Since Gaussian filtering is the convolution of the image and the normal distribution, Gaussian filtering has a very good suppression effect on the noise under the normal distribution.

[0043] The expressions of one-dimensional Gaussian function and two-dimensional Gaussian function are listed below.

[0044]

[0045] Among them, x and y are independent variables, representing the input values ​​of the function; is the mean of the Gaussian distribution, indicating the center position of the function on the horizontal axis; is the standard deviation, which controls the distribution width of the function in the horizontal direction; is a constant used to ensure the normalization of the two-dimensional Gaussian function.

[0046] After removing image noise, some of the original image features are still not obvious. If the edges are directly extracted, it may cause recognition errors. Therefore, in the process of image processing, a method that can extract the edges of objects is needed, which is called edge detection technology. The edge of an object is formed by the orientation of the object or its own material. There are often differences in color, sensitivity, and texture at the edge. People distinguish these differences to outline the outer contour of the object. Edge detection technology can collect useful information to the greatest extent and remove a large amount of useless information. It is the key to the image processing process. It has been widely used in texture analysis, image extraction, image segmentation and other professions.

[0047] Edge detection is a basic tool in graphics and image processing, computer vision, and machine vision. It is usually used for feature extraction and feature detection. It aims to detect edges with obvious changes or discontinuous areas in a digital image. In one-dimensional space, similar operations are called step detection. An edge is a boundary line between different regions in an image. Usually an edge image is a binary image. The purpose of edge detection is to capture areas where brightness changes sharply. Color edge detection methods use edge filters. These filters extract edges by finding pixels at the boundary of brighter and darker areas. The filters look for parts of the image where gradients change significantly. These gradients are generally described as the amplitude and direction of the edge. All pixels with high edge amplitudes are selected to complete the edge contour extraction of the area.

[0048] The three most common operators are: The Sobel operator combines Gaussian smoothing and differential derivation. It is a first-order derivative edge detection operator that uses a convolution kernel to perform convolution and operation on each pixel in the image, and then uses a suitable threshold to extract the edge. The Sobel operator has two convolution kernels, corresponding to the x and y directions respectively. The calculation process is as follows: Derivatives are taken in the x and y directions respectively, and at each pixel in the image, the approximate gradient is obtained by combining the above two results.

[0049] The Laplace operator is a second-order derivative operator. In the edge area of ​​the image, the pixel values ​​will change significantly, and the derivative of these pixels will produce extreme values. At these extreme value locations, the second-order derivative is 0, so the second-order derivative is used to detect the edge of the image.

[0050] The basic idea of ​​the Canny operator is to find the local maximum of the gradient. First, a Gaussian smoothing filter is used for convolution to reduce noise, then a pair of convolution arrays are used to calculate the edge gradient and direction, then non-maximum suppression is used to remove non-edge lines, and finally a hysteresis threshold (high threshold and low threshold) is used to detect and connect edges.

[0051] The general processing flow of edge detection is as follows: Preprocess the initial image and convert it into a grayscale image; The grayscale image is smoothed and filtered using an edge filter to remove noise to obtain a filtered image; Using threshold processing, the highlight edges in the filtered image are extracted and the gradient amplitude and direction of the highlight edges are calculated based on the Canny edge detection operator; Suppress non-edge pixels along the gradient direction and retain the local maximum value of the gradient amplitude of the highlight edge; The edge points are classified based on a preset threshold value, and the extracted edges are processed according to the detection requirements, including edge thinning, contour generation, merging discontinuous edges and separating backgrounds.

[0052] In this example, the Canny edge detection operator extracts the image edge according to three criteria: one-to-one correspondence between the detected edge pixels and the actual edge pixels, minimum deviation between the detected edge pixels and the actual positions, large output signal-to-noise ratio, low false recognition rate, and high recognition completeness.

[0053] The derivative of the second-order Gaussian function on the x-axis is:

[0054] The derivative on the y-axis is:

[0055] Performing a first-order derivative convolution operation on the original image, the derivative of the original image on the x-axis can be expressed as:

[0056] The derivative on the y-axis can be expressed as:

[0057] Analogous to the Sobel operator, the modulus of the derivative and the maximum angle of the direction change rate can be expressed as follows:

[0058] According to the definition of the Canny edge detection operator, if a point in each gradient direction is the maximum value point in the neighborhood, it must be the edge of the image. By calculating the extreme value and direction of each derivative point, it can be determined whether the point is an edge point.

[0059] After completing the edge detection of the image, it is necessary to outline the edge of the image, and it is hoped that the outlined curve can be fitted into several regular fitting circles, and the coordinates of the center of the fitting circle can be obtained.

[0060] According to three random known points in the image after edge detection, any two perpendicular bisectors corresponding to the three points are obtained, and a fitting circle is obtained according to the intersection of the two perpendicular bisectors, and circles with diameters close to the known image are screened out, while circles with diameters not close to the known image are discarded.

[0061] After the fitting circle is generated, the coordinates of the center of the circle and the radius of the circle are determined, and the preprocessed image is cropped with the coordinates of the circle as the center and a number of pixels as the side length to obtain the feature image.

[0062] S3 trains the material matching model based on the support vector machine according to the feature image and the preset distribution network material drawings to generate the material matching degree.

[0063] Among them, the specific method of obtaining the material matching model is as follows: Extract features from feature images and preset distribution network material drawings and perform normalization processing; The features in the feature image are considered to be matched if they are consistent with those in the preset material drawing, otherwise they are considered to be mismatched; A data set is constructed according to the feature image and the preset distribution network material drawing, the data set includes positive samples and negative samples, the positive samples represent instances where the feature image and the material drawing match, and the negative samples represent instances where the feature image and the material drawing do not match; Use kernel functions to train material matching models based on data sets and adjust model parameters in real time during the training process to obtain the optimal matching algorithm; Applying the optimal matching algorithm to the matching of the initial image; The material matching model outputs the predicted matching result, namely the material matching degree.

[0064] The method for obtaining the optimal matching algorithm is as follows: Map the training samples in the training set to the high-order feature space through the kernel function; In the high-order feature space, a non-negative relaxation coefficient is introduced to determine the linear separability of the training samples and obtain the decision function; Based on the decision function, the quadratic programming operation is performed on the feature space, and the dual problem of quadratic programming is solved by introducing Lagrange multipliers, thereby obtaining the optimal matching algorithm. The quadratic programming operation is specifically as follows: in a high-dimensional feature space, the SVM searches for a plane that maximizes the distance to the nearest training sample.

[0065] Then for an n-dimensional nonlinear classification, if a training sample is given

[0066] Then mapping the training samples to the high-order feature space using the kernel function can be expressed as

[0067] In high-order feature space, if the linear separability of training samples is unknown in advance, it is necessary to introduce non-negative relaxation coefficients , then the decision function can be expressed as:

[0068] in >0, i=1,2,...,n, q represents the normal vector, and a represents the error.

[0069] To maximize the classification interval, it is equivalent to a quadratic programming operation, as shown below

[0070] Among them, b is defined as a penalty coefficient greater than 0, which represents the degree of constraint on misidentified samples. The Lagrange multiplier is introduced and equivalently transformed into a dual form, that is, in Represents the Lagrange multiplier Solving the above formula, the optimal matching algorithm is as follows:

[0071] Where sgn() represents the sign function, and the nature of the sample can be distinguished by the sign of z(x).

[0072] A model for training samples was built through support vector machine. First, the parameters of the support vector machine were configured, and then the sample information was classified and imported into the support vector machine in batches, so that the machine could fully learn the sample information and save the learned information in an xml file.

[0073] Using support vector machines (SVM) to train material matching models and generate material matching degrees has the following advantages: High-precision classification: SVM separates data points of different categories by constructing the optimal hyperplane, which achieves high-precision classification in the feature space and makes very accurate judgments on material matching.

[0074] Handling nonlinear data: By using kernel functions, SVM is able to handle nonlinear relationships, allowing it to work effectively even in complex material matching situations.

[0075] Strong generalization ability: SVM has good generalization ability. Even when there are differences between training data and actual data, the model can still provide reliable matching predictions and reduce the risk of overfitting.

[0076] Automated process: By automating the calculation of material matching, the SVM model reduces the need for manual inspection, improves the efficiency and consistency of the matching process, and saves time and costs.

[0077] Anomaly detection: SVM can identify and mark abnormal matching materials, thereby helping to discover potential problems, ensuring that the material usage of the distribution network project meets expectations and ensuring project quality.

[0078] Strong adaptability: Support vector machines can adapt to different material characteristics and matching criteria by adjusting kernel functions and parameter settings, providing flexible solutions.

[0079] S4, judging whether the material matching of the completed distribution network project is abnormal based on the comparison between the material matching degree and the preset material matching threshold.

[0080] A preset material matching threshold is set based on historical data and experience, and the material matching degree is compared with the preset material matching threshold. If the material matching degree is greater than the preset material matching threshold, the material matching degree is considered normal; if the material matching degree is less than the preset material matching threshold, the material matching degree is considered abnormal.

[0081] S5, when material matching is abnormal, the completed distribution network project will be reworked, and the upcoming distribution network project will be warned and adjusted.

[0082] The specific steps of rework are as follows: Check and confirm anomalies: Conduct a detailed inspection of the completed project to confirm the specific circumstances of the material matching anomalies.

[0083] Develop rework plan: Develop a detailed rework plan based on the abnormal situation, including the materials that need to be replaced, the estimated time and cost.

[0084] Implement rework: Carry out rework as planned to ensure material matching issues are resolved.

[0085] Verify improvements: Verify after completing the rework to ensure that the fixes are effective and meet the expected standards.

[0086] Early warning and adjustment of upcoming projects: Analyze the causes of exceptions: Analyze the root causes of material matching exceptions, such as supply chain issues, drawing errors, or inaccurate data.

[0087] Update distribution network material drawings: Update and optimize material drawings and matching standards based on anomaly causes to reduce future matching issues.

[0088] Implement early warning mechanism: Establish an early warning system in the upcoming project to monitor material matching in real time and promptly identify and deal with potential problems.

[0089] Adjust procurement and supply chain: Optimize procurement and supply chain management based on the analysis results to ensure that the quality and matching of materials meet the requirements.

[0090] Training and improvement: Provide training to relevant personnel to enhance their understanding of material matching standards and avoid similar problems.

[0091] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0092] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0093] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0094] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0095] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0096] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A material matching method for a distribution network project, characterized in that: The following steps are involved: Real-time collection of material images of the completed parts of the distribution network project construction process as initial images; Preprocess the initial image and extract the target to obtain the feature image; According to the feature image and the preset distribution network material drawings, the material matching model is trained based on the support vector machine to generate the material matching degree; By comparing the material matching degree with the preset material matching threshold, it is determined whether the material matching of the completed distribution network project is abnormal; When material matching is abnormal, the completed distribution network project will be reworked, and the upcoming distribution network project will be warned and adjusted.

2. A material matching method for distribution network engineering according to claim 1, characterized in that: The preprocessing includes image grayscale, image noise reduction and edge detection.

3. A material matching method for distribution network engineering according to claim 2, characterized in that: The specific steps of edge detection are as follows: Preprocess the initial image and convert it into a grayscale image; The grayscale image is smoothed and filtered using an edge filter to remove noise to obtain a filtered image; Using threshold processing, the highlight edges in the filtered image are extracted and the gradient amplitude and direction of the highlight edges are calculated based on the Canny edge detection operator; Suppress non-edge pixels along the gradient direction and retain the local maximum value of the gradient amplitude of the highlight edge; The edge points are classified based on a preset threshold value, and the extracted edges are processed according to the detection requirements, including edge thinning, contour generation, merging discontinuous edges and separating backgrounds.

4. A material matching method for distribution network engineering according to claim 1, characterized in that: The method for acquiring the characteristic image is as follows: Outline the edge of the image after edge detection, and fit the outlined curve into several regular fitting circles; After the fitting circle is generated, the coordinates of the center of the circle and the radius of the circle are determined, and the preprocessed image is cropped with the coordinates of the circle as the center and a number of pixels as the side length to obtain the feature image.

5. A material matching method for distribution network engineering according to claim 4, characterized in that: The method for obtaining the fitting circle is as follows: According to three random known points in the image after edge detection, any two perpendicular bisectors corresponding to the three points are obtained, and a fitting circle is obtained according to the intersection of the two perpendicular bisectors, and circles with diameters close to the known image are screened out, while circles with diameters not close to the known image are discarded.

6. A material matching method for distribution network engineering according to claim 1, characterized in that: The specific method for obtaining the material matching model is as follows: Extract features from feature images and preset distribution network material drawings and perform normalization processing; The features in the feature image are considered to be matched if they are consistent with those in the preset material drawing, otherwise they are considered to be mismatched; A data set is constructed according to the feature image and the preset distribution network material drawing, the data set includes positive samples and negative samples, the positive samples represent instances where the feature image and the material drawing match, and the negative samples represent instances where the feature image and the material drawing do not match; Use kernel functions to train material matching models based on data sets and adjust model parameters in real time during the training process to obtain the optimal matching algorithm; Applying the optimal matching algorithm to the matching of the initial image; The material matching model outputs the predicted matching result, namely the material matching degree.

7. A material matching method for distribution network engineering according to claim 6, characterized in that: The specific method for obtaining the optimal matching algorithm is as follows: Map the training samples in the training set to the high-order feature space through the kernel function; In the high-order feature space, a non-negative relaxation coefficient is introduced to determine the linear separability of the training samples and obtain the decision function; Based on the decision function, the quadratic programming operation is performed on the feature space, and the dual problem of quadratic programming is solved by introducing Lagrange multipliers, thereby obtaining the optimal matching algorithm. The quadratic programming operation is specifically as follows: in a high-dimensional feature space, the SVM searches for a plane that maximizes the distance to the nearest training sample.

8. A material matching method for distribution network engineering according to claim 2, characterized in that: The specific calculation formula for the image grayscale is as follows: Among them, i and j represent the horizontal and vertical coordinate points in the image respectively, R is the red channel information, G is the green channel information, and B is the blue channel information.

9. A material matching method for distribution network engineering according to claim 7, characterized in that: The specific calculation formula of the decision function is as follows: in >0, i=1,2,...,n, q represents the normal vector, a represents the error, represents the i-th image sample in the training data, Represents the category label of the i-th image sample.

10. A material matching method for distribution network engineering according to claim 6, characterized in that: The specific calculation formula of the optimal matching algorithm is as follows: Among them, sgn() is the signal function, a represents the error, represents the Lagrange multiplier, x represents the input feature image, and k represents the kernel function.

Images