A Structured Extraction Method and System for Nameplate Information of Electrical Equipment

Abstract

The present invention discloses a method and system for structured extraction of nameplate information of power equipment, which relates to the technical field of image processing; obtaining an original image set of power equipment, and determining a target area of the nameplate for any original image; calculating the offset of the afterimage area relative to the main body area, determining the enhancement weight of the afterimage area according to the offset, and performing an optimization operation on the target area to obtain a target image; dividing each target image into regions to obtain a target image set, performing quality assessment on the regions in the target image set, retaining and splicing the regions that meet the preset conditions to obtain an enhanced image; extracting the nameplate text information in the enhanced image, and storing the nameplate text information in a structured form. Through precise positioning of the target area, calculation of afterimage offset and enhancement processing, regional quality assessment and intelligent splicing, the recognition accuracy of the nameplate text information is significantly improved, and the accuracy of power equipment information acquisition is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and particularly relates to a method and system for structurally extracting nameplate information of power equipment. Background Art

[0002] As a direct identifier of the identity and characteristics of power equipment, the nameplate of power equipment carries important information such as the basic information, technical parameters, operating conditions, and safety warnings of the equipment, and is an indispensable basis for guiding equipment selection, installation, operation and maintenance, fault analysis, and safety management. It not only helps operators quickly understand the equipment performance and ensure the efficient and stable operation of the equipment under suitable conditions, but also provides key references during equipment repair, replacement, or upgrade, promoting the improvement of the overall reliability, economy, and safety of the power system.

[0003] Workers use methods such as image processing technology, optical character recognition, and natural language processing to efficiently and accurately structurally extract the information on the nameplate of power equipment. This process converts the text, numbers, and symbols on the nameplate into editable and searchable structured data through automated scanning, recognition, and parsing, greatly facilitating the storage, management, and analysis of data, and improving the convenience and efficiency of the management and maintenance work of power equipment.

[0004] During the actual operation process, due to the vibration of the power equipment during operation, the shooting equipment cannot accurately focus, the captured images are prone to afterimages, and the nameplate information cannot be accurately recognized. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem that due to the vibration of the power equipment during operation, the shooting equipment cannot accurately focus, and the captured images are prone to afterimages, and to propose a method and system for structurally extracting nameplate information of power equipment.

[0006] In the first aspect of the implementation of the present invention, a method for structurally extracting nameplate information of power equipment is first proposed. The method includes:

[0007] Obtain the original image set of the power equipment, and determine the target area of the nameplate for any original image; the target area includes: the main area and the afterimage area;

[0008] Calculate the offset of the afterimage area relative to the main area, determine the enhancement weight of the afterimage area according to the offset, and perform an optimization operation on the target area to obtain a target image;

[0009] Perform region division on each target image to obtain a target image set, perform quality assessment on the regions in the target image set, and retain and splice the regions that meet the preset conditions to obtain an enhanced image;

[0010] Extract the nameplate text information from the enhanced image and store the nameplate text information in a structured form.

[0011] Optionally, calculate the offset of the afterimage area relative to the main area within the target area. The method includes:

[0012] Identify the positioning anchor points in the target area, use any positioning anchor point in the main area as the coordinate origin, and calculate the offset of this positioning anchor point in the afterimage area;

[0013] Determine the number of afterimage areas according to the coordinates of the positioning anchor points.

[0014] Optionally, determine the enhancement weight of the afterimage area according to the offset. The method includes:

[0015] Convert each afterimage area into a grayscale image, identify the pixel gradient area of the grayscale image. If there is a continuous area in the pixel gradient area, then determine this area as the edge area;

[0016] Calculate the pixel difference of the edge area. If the pixel difference meets the gradient threshold, then retain this afterimage area to obtain an edge area set;

[0017] Determine the image weight within the edge area set according to the offset, and perform edge enhancement according to the image weight to obtain a repaired image.

[0018] Optionally, perform an optimization operation on the target area to obtain a target image. The method includes:

[0019] Determine the edge enhancement coefficient of the main area according to the repaired image, and correct the edge enhancement range of the main area according to the edge enhancement coefficient;

[0020]

[0021] Wherein, F is the edge enhancement coefficient, d1 is the edge distance of the main area, d2 is the edge distance of the repaired image, and j is the number of repaired images.

[0022] Optionally, perform quality assessment on the areas in the target image set. The method includes:

[0023] Obtain information recognition data, and perform quality assessment on the areas in the target image set through a quality assessment formula; the information recognition data includes: meaningless information data, the number of invalid information extractions, and the total number of information extractions;

[0024] Quality assessment formula:

[0025]

[0026] Wherein, PSNRX Let \(R\) be the regional quality score, \(MAX\) be the maximum possible pixel value of the current image, \(MSE\) be the mean square error, \(W\) be the proportion of meaningless information, \(S\) be the number of ineffective information extraction times, and \(Z\) be the total number of information extraction times.

[0027] In the second aspect of the implementation of the present invention, a structured extraction system for nameplate information of power equipment is proposed, including: an image acquisition module, an image enhancement module, an image stitching module, and an information storage module.

[0028] The image acquisition module is used to acquire the original image set of the power equipment and determine the target area of the nameplate for any original image; the target area includes: the main area and the afterimage area.

[0029] The image enhancement module is used to calculate the offset of the afterimage area relative to the main area, determine the enhancement weight of the afterimage area according to the offset, and perform an optimization operation on the target area to obtain a target image.

[0030] The image stitching module is used to divide each target image into regions to obtain a target image set, evaluate the quality of the regions in the target image set, and retain and stitch the regions that meet the preset conditions to obtain an enhanced image.

[0031] The information storage module is used to extract the nameplate text information in the enhanced image and store the nameplate text information in a structured form.

[0032] Optionally, the image enhancement module includes: an offset calculation module and an afterimage quantity determination module.

[0033] The offset calculation module is used to identify the positioning anchor points in the target area, use any positioning anchor point in the main area as the coordinate origin, and calculate the offset of this positioning anchor point in the afterimage area.

[0034] The afterimage quantity determination module is used to determine the quantity of the afterimage area according to the coordinates of the positioning anchor points.

[0035] Optionally, the image enhancement module includes: an edge judgment module, an edge screening module, and an edge enhancement module.

[0036] The edge judgment module is used to convert each afterimage area into a grayscale image, identify the pixel gradient area of the grayscale image, and if there is a continuous area in the pixel gradient area, then determine this area as the edge area.

[0037] The edge screening module is used to calculate the pixel difference of the edge area, and if the pixel difference meets the gradient threshold, then retain this afterimage area to obtain a set of edge areas.

[0038] The edge enhancement module is used to determine the image weights within the edge region set according to the offset, and perform edge enhancement based on the image weights to obtain a repaired image.

[0039] Optionally, the image enhancement module includes: an edge coefficient enhancement module:

[0040] The edge coefficient enhancement module is used to determine the edge enhancement coefficient of the main body region according to the repaired image, and correct the edge enhancement range of the main body region according to the edge enhancement coefficient;

[0041]

[0042] Where F is the edge enhancement coefficient, d1 is the edge distance of the main body region, d2 is the edge distance of the repaired image, and j is the number of repaired images.

[0043] Optionally, the image stitching module includes: a quality assessment module:

[0044] The quality assessment module is used to obtain information recognition data, and perform quality assessment on the regions in the target image set through a quality assessment formula; the information recognition data includes: meaningless information data, the number of invalid information extractions, and the total number of information extractions;

[0045] Quality assessment formula:

[0046]

[0047] Where PSNR X is the regional quality score, MAX is the maximum possible pixel value of the current image, MSE is the mean square error, W is the proportion of meaningless information, S is the number of invalid information extractions, and Z is the total number of information extractions.

[0048] Advantages of the present invention:

[0049] The present invention proposes a method for structured extraction of nameplate information of power equipment. By obtaining the original image set of power equipment, the target region of the nameplate is determined for any original image; the offset of the afterimage region relative to the main body region is calculated, the enhancement weight of the afterimage region is determined according to the offset, and an optimization operation is performed on the target region to obtain a target image; the target image is divided into regions to obtain a target image set, the regions in the target image set are subjected to quality assessment, the regions meeting the preset conditions are retained and stitched to obtain an enhanced image; the nameplate text information in the enhanced image is extracted, and the nameplate text information is stored in a structured form. Through accurate positioning of the target region, calculation of afterimage offset and enhancement processing, regional quality assessment and intelligent stitching, the recognition accuracy of the nameplate text information is significantly improved, and the accuracy of power equipment information acquisition is improved. Description of the Drawings

[0050] The present invention will be further described below with reference to the accompanying drawings.

[0051] Figure 1 FIG. is a flowchart of a method for structured extraction of nameplate information of electrical equipment provided by an embodiment of the present invention;

[0052] Figure 2 FIG. is a schematic structural diagram of a system for structured extraction of nameplate information of electrical equipment provided by an embodiment of the present invention. Detailed implementation manners

[0053] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The term "and / or" in this article is only a description of the associated relationship of the associated objects, indicating that there may be three relationships. For example, A and B may represent: A exists alone, A and B exist simultaneously, and B exists alone. These three situations. In addition, the descriptions such as "first" and "second" in the present invention are only for the purpose of description, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments may be combined with each other, but it must be based on the fact that those skilled in the art can implement them. When the combination of the technical solutions appears to be contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist, nor is it within the protection scope required by the present invention.

[0054] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0055] An embodiment of the present invention provides a method for structured extraction of nameplate information of electrical equipment. Refer to Figure 1 , Figure 1 FIG. is a flowchart of a method for structured extraction of nameplate information of electrical equipment provided by an embodiment of the present invention. The method includes the following steps:

[0056] S101, obtain an original image set of the electrical equipment, and determine the target area of the nameplate for any original image.

[0057] S102, calculate the offset of the afterimage area relative to the main area, determine the enhancement weight of the afterimage area according to the offset, and perform an optimization operation on the target area to obtain a target image.

[0058] S103. Divide each target image into regions to obtain a set of target images. Perform quality assessment on the regions in the set of target images, retain the regions that meet the preset conditions, and splice them to obtain an enhanced image.

[0059] S104. Extract the nameplate text information from the enhanced image and store the nameplate text information in a structured form.

[0060] The target regions include: the main body region and the ghost region.

[0061] Based on a method for structured extraction of nameplate information of electrical equipment provided by an embodiment of the present invention, by accurately positioning the target region, calculating the ghost offset and performing enhancement processing, region quality assessment and intelligent splicing, the recognition accuracy of the nameplate text information is significantly improved, and the accuracy of obtaining electrical equipment information is improved.

[0062] In one implementation, for the main body region and the ghost region, the main body region is an image region in the image with obvious pixel values, but the edge of this image region will become blurred due to jitter; the pixel values of the ghost region are poor, but the edge is relatively clear; the nameplate is generally black on white or white on black, but may turn yellow due to long-term use. Converting it to a grayscale image can avoid this defect.

[0063] In one implementation, the positioning anchor point is the installation point or the fixed point of the nameplate of the electrical equipment. This point has obvious features on the image and is easy for the machine to recognize; for any original image, determine the target region of the nameplate, and the position of the nameplate can be determined by identifying the positioning anchor point.

[0064] In one implementation, the set of original images of the electrical equipment is generally taken by a camera; only horizontal displacement occurs in the main body region and the ghost region, and the horizontal displacement only includes: up and down displacement and left and right displacement.

[0065] In one implementation, perform an optimization operation on the target region to obtain a target image, and the optimization operation can be sharpening filtering, histogram equalization, contrast stretching, and frequency domain enhancement.

[0066] In one implementation, these scattered but crucial information on the nameplate is converted into a structured data format that can be understood by a computer through technical means, such as OCR recognition, natural language processing, etc. This process not only improves the efficiency of data processing but also makes these data more convenient to be used in multiple aspects such as equipment management, fault troubleshooting, data analysis, etc. Through structured extraction, power enterprises can establish detailed equipment files and achieve the full-life cycle management of equipment. At the same time, based on these structured data, enterprises can also conduct data analysis, discover potential problems in equipment operation, optimize equipment maintenance strategies, and improve the reliability and economy of equipment operation. The structured extraction of power equipment nameplate information is an important part of the modernization and intelligentization of power equipment management and is of great significance for improving the level of power equipment management and ensuring the safe and stable operation of the power grid.

[0067] In one embodiment, step S102 includes:

[0068] Identify the positioning anchor points in the target area, use any positioning anchor point in the main area as the coordinate origin, and calculate the offset of this positioning anchor point in the afterimage area.

[0069] Determine the number of afterimage areas according to the coordinates of the positioning anchor points.

[0070] In one implementation, use any positioning anchor point in a main area as the coordinate origin, calculate the displacement of this positioning anchor point on the horizontal plane, that is, the position of this positioning anchor point on the afterimage area. The offset is obtained by calculating the distance between two positioning anchor points, that is, the distance between the positioning anchor point in the afterimage area and the coordinate origin in the main area. These two positioning anchor points must correspond one by one.

[0071] In one implementation, determine the number of afterimage areas according to the coordinates of the positioning anchor points. Since each positioning anchor point in the main area can be used as the coordinate origin, an afterimage area can be located through the coordinates of the positioning anchor points. When there are different coordinates, the position and number of afterimage areas can be determined according to the number of coordinates. For example, coordinates (-1, 2), (3, 4) represent two afterimage areas.

[0072] In one implementation, if the afterimage area coincides too much with the main area, resulting in the inability to identify the positioning anchor points of the afterimage area, then this afterimage area can be discarded. If there is only partial coincidence, calculate the recognizable positioning anchor points and calculate the corresponding positioning anchor points in the main area to obtain the offset.

[0073] In one embodiment, step S102 includes:

[0074] Convert each residual image area into a grayscale image, identify the pixel gradient area of the grayscale image. If there is a continuous area in the pixel gradient area, then determine this area as the edge area.

[0075] Calculate the pixel difference of the edge area. If the pixel difference meets the gradient threshold, then retain this residual image area to obtain the edge area set.

[0076] Determine the image weight within the edge area set according to the offset, and perform edge enhancement according to the image weight to obtain the restored image.

[0077] In one implementation, for the pixel gradient area in the grayscale image, the pixel gradient area refers to the area in the grayscale image where the pixel grayscale values change significantly, that is, the area with a large grayscale gradient. For example: in the grayscale image, the edge is the place where the pixel grayscale value changes suddenly, that is, the place with a large grayscale gradient. Therefore, the pixel gradient area often corresponds to feature areas such as the edges and textures of the image; it is beneficial for the machine to identify and judge the edge area within the grayscale image.

[0078] In one implementation, the condition for judging that there is a continuous area in the pixel gradient area and the continuous area is the edge area. If it is not continuous, it may be noise, resulting in incorrect machine recognition; the continuous area must meet the preset length and be an uninterrupted linear interval, and have a pixel gradient.

[0079] In one implementation, determine the image weight within the edge area set according to the offset. The larger the offset, the smaller the weight. The edges farther away are relatively blurred and do not have good enhancement conditions. On the contrary, the smaller the offset, the larger the weight.

[0080] In one embodiment, step S102 further includes:

[0081] Determine the edge enhancement coefficient of the main area according to the restored image, and correct the edge enhancement range of the main area according to the edge enhancement coefficient.

[0082]

[0083] Wherein, F is the edge enhancement coefficient, d1 is the edge distance of the main area, d2 is the edge distance of the restored image, and j is the number of restored images.

[0084] In one implementation, because the camera captures multiple original images, the number of restored images is also large. The number of restored images is determined by the number of original images; d1 is the edge distance of the main area, that is, the distance of the area with blurred edges in the main area. For example: the pixel values in this area have a flexible change, and generally it is an image with a gradual change in human vision. The length of this area is the edge distance of the main area.

[0085] In one implementation, the sharpening parameter can be adjusted through the edge enhancement coefficient. For example, the sharpening radius is corrected. The larger the sharpening radius, the wider the area affected by sharpening, and the more intuitive the clarity effect of the edge. However, an overly large radius may cause a halo effect on the edge, making the image details blurred, so correction is needed.

[0086] In one embodiment, step S103 includes:

[0087] Obtain information recognition data, and perform quality assessment on the regions in the target image set through a quality assessment formula; the information recognition data includes: meaningless information data, the number of ineffective information extractions, and the total number of information extractions.

[0088] Quality assessment formula:

[0089]

[0090] where PSNR X is the regional quality score, MAX is the maximum possible pixel value of the current image, MSE is the mean square error, W is the proportion of meaningless information, S is the number of ineffective information extractions, and Z is the total number of information extractions.

[0091] In one implementation, PSNR is a commonly used image quality assessment metric, namely peak signal-to-noise ratio, which is used to measure the quality of image restoration; PSNR X is the modified peak signal-to-noise ratio, S is the number of ineffective information extractions, that is, the number of times the machine cannot recognize information; W is the proportion of meaningless information, that is, the recognized information is garbled and does not have any meaning or valid meaning.

[0092] In one implementation, MSE is the mean square error, and the formula is where M and N are the width and height of the image respectively, and I qp is the pixel value of the original image at position (q, p), and K qp is the pixel value of the image after enhancement at position (q, p).

[0093] In one implementation, the original image set contains multiple original images, and the optimized target images also have the same number. For example: if there are five original images, then there are also five target images. Each target image is cut and divided into a target image set in equal proportion, and position labels are added to each image during division, which is beneficial for subsequent splicing operations; when splicing, if there are multiple optimal regions in each region, the region with the highest score is used. If the score of any region does not meet the preset conditions, the region with the highest score in that region is retained.

[0094] Based on the same inventive concept, the embodiments of the present invention also provide a structured extraction system for power equipment nameplate information. SeeFigure 2 , Figure 2 It is a schematic structural diagram of a system for structured extraction of nameplate information of power equipment provided by an embodiment of the present invention, including: an image acquisition module, an image enhancement module, an image splicing module, and an information storage module:

[0095] The image acquisition module is used to acquire the original image set of the power equipment and determine the target area of the nameplate for any original image; the target area includes: the main area and the afterimage area.

[0096] The image enhancement module is used to calculate the offset of the afterimage area relative to the main area, determine the enhancement weight of the afterimage area according to the offset, and perform an optimization operation on the target area to obtain the target image.

[0097] The image splicing module is used to perform area division on each target image to obtain a set of target images, evaluate the quality of the areas in the set of target images, and retain and splice the areas that meet the preset conditions to obtain the enhanced image.

[0098] The information storage module is used to extract the nameplate text information in the enhanced image and store the nameplate text information in a structured form.

[0099] Based on a system for structured extraction of nameplate information of power equipment provided by an embodiment of the present invention, by accurately positioning the target area, calculating and enhancing the afterimage offset, evaluating the area quality, and performing intelligent splicing, the recognition accuracy of the nameplate text information is significantly improved, and the accuracy of obtaining power equipment information is enhanced.

[0100] In one embodiment, the image enhancement module includes: an offset calculation module and an afterimage quantity determination module:

[0101] The offset calculation module is used to identify the positioning anchor points in the target area, use any positioning anchor point in the main area as the coordinate origin, and calculate the offset of this positioning anchor point in the afterimage area.

[0102] The afterimage quantity determination module is used to determine the quantity of the afterimage areas according to the coordinates of the positioning anchor points.

[0103] In one embodiment, the image enhancement module includes: an edge judgment module, an edge screening module, and an edge enhancement module:

[0104] The edge judgment module is used to convert each afterimage area into a grayscale image, identify the pixel gradient area of the grayscale image, and if there is a continuous area in the pixel gradient area, then determine this area as the edge area.

[0105] The edge screening module is used to calculate the pixel difference of the edge area, and if the pixel difference meets the gradient threshold, then retain this afterimage area to obtain a set of edge areas.

[0106] An edge enhancement module for determining image weights within an edge region set based on an offset and performing edge enhancement based on the image weights to obtain a restored image.

[0107] In one embodiment, the image enhancement module includes: an edge coefficient enhancement module:

[0108] The edge coefficient enhancement module is configured to determine an edge enhancement coefficient of a main region based on the restored image and correct the edge enhancement range of the main region according to the edge enhancement coefficient.

[0109]

[0110] Where F is the edge enhancement coefficient, d1 is the edge distance of the main region, d2 is the edge distance of the restored image, and j is the number of restored images.

[0111] In one embodiment, the image stitching module includes: a quality assessment module:

[0112] The quality assessment module is configured to obtain information recognition data and perform quality assessment on regions in a target image set through a quality assessment formula; the information recognition data includes: meaningless information data, the number of invalid information extractions, and the total number of information extractions.

[0113] Quality assessment formula:

[0114]

[0115] Where PSNR X is the region quality score, MAX is the maximum possible pixel value of the current image, MSE is the mean square error, W is the proportion of meaningless information, S is the number of invalid information extractions, and Z is the total number of information extractions.

[0116] The above has described a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the patent coverage scope of the present invention.

Claims

1. A method for structurally extracting nameplate information of electrical equipment, characterized in that, The method includes: Obtaining an original image set of a power device, and determining a target area of a nameplate for any original image; The target area includes: a main body area and a residual image area; The original image set of the power device is captured by a camera; only horizontal displacement occurs in the main body area and the residual image area, and the horizontal displacement only includes: up-and-down displacement and left-and-right displacement; Calculating the offset of the residual image area relative to the main body area, determining the enhancement weight of the residual image area according to the offset, and performing an optimization operation on the target area to obtain a target image; the optimization operations are sharpening filtering, histogram equalization, contrast stretching, and frequency domain enhancement; Performing region division on each target image to obtain a target image set, performing quality assessment on the regions in the target image set, and retaining and splicing the regions that meet the preset conditions to obtain an enhanced image; The original image set contains multiple original images, and the optimized target images also have the same number. Each target image is cut and divided in equal proportion to obtain a target image set, and a position label is added to each image during division; when splicing, if there are multiple optimal regions in each region, the region with the highest score is used, and if the score of any region does not meet the preset conditions, the region with the highest score of the region is retained; Extracting the nameplate text information in the enhanced image, and storing the nameplate text information in a structured form; Determining the enhancement weight of the residual image area according to the offset, including: Converting each residual image area into a grayscale image, identifying the pixel gradient area of the grayscale image, and if there is a continuous area in the pixel gradient area, determining the area as an edge area; Calculating the pixel difference of the edge area, and if the pixel difference meets the gradient threshold, retaining the residual image area to obtain an edge area set; Determining the image weight in the edge area set according to the offset, and performing edge enhancement according to the image weight to obtain a restored image; Calculating the offset of the residual image area relative to the main body area in the target area, including: Identifying the positioning anchor points in the target area, taking any positioning anchor point in the main body area as the coordinate origin, and calculating the offset of the positioning anchor point in the residual image area; Determining the number of residual image areas according to the coordinates of the positioning anchor points; Taking any positioning anchor point in a main body area as the coordinate origin, calculating the displacement of the positioning anchor point on the horizontal plane, and obtaining the offset by calculating the distance between two positioning anchor points, and the two positioning anchor points must correspond one by one; Determining the number of residual image areas according to the coordinates of the positioning anchor points; when there are different coordinates, determining the position and number of the residual image areas according to the number of coordinates; If the residual image area coincides too much with the main body area, resulting in the positioning anchor point of the residual image area not being recognizable, the residual image area is discarded. If only a part coincides, the recognizable positioning anchor points are calculated, and the corresponding positioning anchor points in the main body area are calculated to obtain the offset.

2. The structured extraction method of the nameplate information of a power device according to claim 1, wherein, Performing an optimization operation on the target area to obtain a target image, the method includes: Determining the edge enhancement coefficient of the main body area according to the restored image, and correcting the edge enhancement range of the main body area according to the edge enhancement coefficient; ； where F is the edge enhancement coefficient, is the edge distance of the main region, is the edge distance of the repaired image, and j is the number of repaired images.

3. A method for structured extraction of nameplate information of electrical equipment according to claim 1, characterized in that, Perform quality assessment on the regions in the target image set, the method comprising: Obtain information recognition data, and perform quality assessment on the regions in the target image set through a quality assessment formula; the information recognition data includes: meaningless information data, the number of ineffective information extractions, and the total number of information extractions; Quality assessment formula: ； Among them, is the regional quality score, MAX is the maximum pixel value of the current image, MSE is the mean square error, W is the proportion of meaningless information, S is the number of ineffective information extractions, and Z is the total number of information extractions.

4. A structured extraction system for power equipment nameplate information, which is used to implement the structured extraction method for power equipment nameplate information according to any one of claims 1 to 3, characterized in that, The system includes: an image acquisition module, an image enhancement module, an image stitching module, and an information storage module: The image acquisition module is configured to acquire an original image set of a power device, and determine a target region of a nameplate for any one of the original images; the target region includes: a main region and a ghost region; The image enhancement module is configured to calculate an offset of the ghost region relative to the main region, determine an enhancement weight of the ghost region according to the offset, and perform an optimization operation on the target region to obtain a target image; the optimization operation is sharpening filtering, histogram equalization, contrast stretching, and frequency domain enhancement; The image stitching module is configured to perform region division on each target image to obtain a target image set, perform quality assessment on the regions in the target image set, and retain and stitch the regions that meet the preset conditions to obtain an enhanced image; The information storage module is configured to extract the nameplate text information in the enhanced image and store the nameplate text information in a structured form; The image enhancement module includes: an edge judgment module, an edge screening module, and an edge enhancement module: The edge judgment module is configured to convert each ghost region into a grayscale image, identify the pixel gradient region of the grayscale image, and if there is a continuous region in the pixel gradient region, determine the region as an edge region; The edge screening module is configured to calculate the pixel difference of the edge region, and if the pixel difference meets the gradient threshold, retain the ghost region to obtain a set of edge regions; The edge enhancement module is configured to determine the image weight in the set of edge regions according to the offset, and perform edge enhancement according to the image weight to obtain a restored image.

5. The structured extraction system for nameplate information of electrical equipment according to claim 4, characterized in that, The image enhancement module includes: an offset calculation module and a ghost number determination module: The offset calculation module is configured to identify the positioning anchor points in the target region, use any positioning anchor point of the main region as the coordinate origin, and calculate the offset of the positioning anchor point in the ghost region; The ghost number determination module is configured to determine the number of ghost regions according to the coordinates of the positioning anchor points.

6. The structured extraction system for power equipment nameplate information according to claim 4, characterized in that, The image enhancement module includes: an edge coefficient enhancement module: The edge coefficient enhancement module is configured to determine the edge enhancement coefficient of the main region according to the restored image, and correct the edge enhancement range of the main region according to the edge enhancement coefficient; ； Among them, F is the edge enhancement coefficient, is the edge distance of the main region, is the edge distance of the repaired image, and j is the number of repaired images.

7. The structured extraction system for nameplate information of a power device according to claim 4, characterized in that, The image stitching module includes: a quality assessment module: The quality assessment module is configured to obtain information recognition data, and perform quality assessment on the regions in the target image set through a quality assessment formula; the information recognition data includes: meaningless information data, the number of ineffective information extractions, and the total number of information extractions; Quality assessment formula: ； Among them, is the regional quality score, MAX is the maximum pixel value of the current image, MSE is the mean square error, W is the proportion of meaningless information, S is the number of ineffective information extraction times, and Z is the total number of information extraction times.

Images