Equipment nameplate image recognition and calibration processing method

Abstract

The invention discloses an equipment nameplate image identification and calibration processing method, and relates to the technical field of image processing, and the method comprises the steps: obtaining a nameplate image set of target power equipment, screening the image set according to a preset mode to obtain a first image set, and further screening the first image set according to state data to obtain a second image set; and converting each image in the second image set to obtain a front-view image set, obtaining a final calibration image according to the front-view image set, and performing image identification on the final calibration image to determine the nameplate information of the power equipment, so that the obtained nameplate image of the target power equipment is corrected during identification, and the nameplate information of the target power equipment is obtained. The possibility of extraction errors or information loss of the final calibration image is small, and the influence on the accuracy and efficiency of equipment management and maintenance is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for recognizing and calibrating equipment nameplate images. Background Art

[0002] The nameplate of power equipment is a label that identifies the basic information of the equipment, usually including the equipment name, manufacturer information, technical parameters, production date, product number, certification mark and precautions. This information is crucial for the correct installation, operation and maintenance of the equipment; by taking images of the nameplate of power equipment and identifying and analyzing the images of the nameplate of power equipment taken, the nameplate information can be digitally recorded and archived, improving information management efficiency, ensuring data accuracy, supporting remote management and rapid troubleshooting, and facilitating efficient tracking, maintenance and overall management of equipment.

[0003] In the actual process of capturing the nameplate images of power equipment, various factors may affect the image quality, resulting in poor image quality and failure to reflect the equipment information in detail and accurately. In this case, directly identifying and analyzing the captured images may lead to data extraction errors or information missing, affecting the accuracy and efficiency of equipment management and maintenance. Summary of the invention

[0004] The purpose of the present invention is to solve the problem and propose a method for equipment nameplate image recognition and calibration processing.

[0005] The present invention proposes a method for equipment nameplate image recognition and calibration processing, the method comprising:

[0006] Acquire a nameplate image set of a target electric power device, and filter the image set according to a preset method to obtain a first image set;

[0007] For the first image set, extract state data of each image, obtain a state coefficient according to the state data, and determine the second image set according to the image corresponding to the state coefficient; the state data includes brightness uniformity, image clarity and feature matching completeness;

[0008] For each image in the second image set, determine its key point, and transform the image according to the key point to obtain a front view image set;

[0009] For each image in the front-view image set, a new coordinate system is established, and the position of each pixel is corrected according to the state of the actual object corresponding to the key point to obtain the position coordinates of each pixel. The final calibrated image is obtained based on the position coordinates of each pixel, and image recognition is performed on the final calibrated image to determine the nameplate information of the power equipment.

[0010] Optionally, filtering the image set according to a preset method to obtain the first image set includes:

[0011] Subjecting each image in the image set to a preset image feature processing condition to obtain a processed image set;

[0012] For the processed image set, obtain the value of each feature of each image before and after processing, and obtain the change rate of the feature according to the value before and after processing;

[0013] Comparing the change rate of each feature of the image with a preset normal feature change rate range, and if the change rates of all features of the image are within the preset normal feature change rate range, marking the image as a first usable image;

[0014] If the change rate of any feature of the image is not within the preset normal feature change rate range, marking the image as a first unusable image;

[0015] Take all first available images in the image set as the first image set.

[0016] Optionally, for the first image set, extracting status data of each image includes:

[0017] The state data includes brightness uniformity, and the method for obtaining the brightness uniformity is:

[0018] Each image in the first image set is converted into a grayscale image, and the brightness mean of the grayscale image is calculated using the formula:

[0019]

[0020] In the formula, μ is the mean brightness, M is the number of rows in the image, N is the number of columns in the image, and I ij is the brightness value of the pixel in the i-th row and j-th column in the grayscale image;

[0021] The brightness variance of the grayscale image is calculated based on the brightness value of each pixel and the brightness mean, and the inverse of the variance is recorded as the brightness uniformity. The calculation formula is:

[0022]

[0023] In the formula, E B is the variance of the changing image, brightness uniformity

[0024] Optionally, the status data includes image clarity, and the method for acquiring the image clarity is:

[0025] Applying the Laplacian operator to each image in the first image set to obtain a transformed image L;

[0026] Calculate the pixel mean of the transformed image L, and the calculation formula is:

[0027]

[0028] Where μL is the pixel mean of the image after Laplace transformation, A is the number of rows of the image after Laplace transformation, N is the number of columns of the image after Laplace transformation, and L is ab is the pixel value of the image at position (a, b) after Laplace transformation;

[0029] Calculate the variance of the image after Laplace transformation. The calculation formula can be:

[0030]

[0031] The variance of the image after Laplace transformation is taken as the image clarity Etk, Etk=Var(L).

[0032] Optionally, the status data includes feature matching completeness, and the method for acquiring the feature matching completeness is:

[0033] For each image in the first image set, a feature point detection algorithm is used to detect key feature points in the image as feature points of the image to be tested;

[0034] Using a matching algorithm, calculate the feature point matching pairs between the image to be tested and the reference image as the number of feature point matching;

[0035] The feature matching completeness is obtained by dividing the number of feature point matches of each image by the total number of feature points of the image to be tested. The calculation formula is: Ews=qa / nk, where Ews is the feature matching completeness, qa is the number of feature point matches, and nk is the total number of feature points of the image to be tested.

[0036] Optionally, obtaining a state coefficient according to the state data, and determining the second image set according to an image corresponding to the state coefficient includes:

[0037] The state data includes brightness uniformity, image clarity and feature matching completeness. The state data including brightness uniformity, image clarity and feature matching completeness are normalized, and the state coefficient is obtained according to the normalized state data. The calculation formula is:

[0038]

[0039] Where Pgt is the state coefficient, Efd, Etk and Ews represent brightness uniformity, image clarity and feature matching completeness respectively, b1, b2 and b3 are the proportional coefficients of brightness uniformity, image clarity and feature matching completeness, and b1, b2 and b3 are all greater than 0;

[0040] Compare the state coefficient with a preset state coefficient threshold, and if the state coefficient is not less than the preset state coefficient threshold, record the image corresponding to the state coefficient as the second available image;

[0041] If the state coefficient is less than a preset state coefficient threshold, the image corresponding to the state coefficient is recorded as a second unusable image;

[0042] All first available images in the first image set are taken as the second image set.

[0043] Optionally, the image is transformed according to the key points, and the obtained front view image set includes:

[0044] Marking the key points of each image in the second image set as marked key points;

[0045] Based on the marked key points, use perspective transformation to calculate the transformation matrix of the image from the current angle to the orthographic angle, and transform each image according to the calculated transformation matrix to obtain the orthographic angle image of the image;

[0046] The front-view angle image of each image in the second image set is taken as a new image set to obtain a front-view image set.

[0047] Optionally, performing position correction on each pixel point according to the state of the actual object corresponding to the key point to obtain the position coordinates of each pixel point, and obtaining the final calibration image according to the position coordinates of each pixel point includes:

[0048] For each image in the orthographic image set, the pixel coordinates corresponding to a key point are changed to the new coordinate circle center, and a new coordinate system is established based on the new coordinate circle center;

[0049] Mapping the pixel points corresponding to the origin in the new coordinate system to the actual nameplate of the electric power equipment, and establishing a coordinate system for the actual nameplate of the electric power equipment according to a number of key points, recorded as the actual coordinate system, the origin of the actual coordinate system is the key point corresponding to the origin of the new coordinate system;

[0050] Get the coordinates of the four corner points of the image in the actual coordinate system, the coordinates of the four corner points in the new coordinate system, and calculate the scaling ratio based on the coordinates in the new coordinate system and the coordinates in the actual coordinate system;

[0051] Obtain the coordinates of the electric equipment nameplate position corresponding to each pixel point of each image in the new coordinate system in the actual coordinate system, and determine the coordinates of each pixel point in the new coordinate system according to the coordinates in the actual coordinate system and the scaling ratio to obtain a corrected image of the image;

[0052] Obtain the corrected image of each image in the orthographic image set, and overlap the images with the same new coordinate system, calculate the average value of the coordinates of the same pixel points as the final position coordinates of the pixel points, calculate the final position coordinates of each pixel point, and obtain the final corrected image.

[0053] Beneficial effects of the present invention:

[0054] The present invention proposes a method for equipment nameplate image recognition and calibration processing, which obtains a nameplate image set of a target power equipment, and filters the image set according to a preset method to obtain a first image set, and further filters the first image set according to status data to obtain a second image set, and transforms each image in the second image set to obtain a front view image set, and obtains a final calibration image based on the front view image set, and performs image recognition on the final calibration image to determine the nameplate information of the power equipment. In this way, the nameplate image of the target power equipment obtained is corrected during recognition, so that the possibility of extraction error or information missing in the final calibration image is small, reducing the impact on the accuracy and efficiency of equipment management and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention will be further described below in conjunction with the accompanying drawings.

[0056] Figure 1 A flowchart of a method for equipment nameplate image recognition and calibration provided by the present invention. DETAILED DESCRIPTION

[0057] 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.

[0058] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0059] The embodiment of the present invention provides a method for calibrating and processing equipment nameplate image recognition. Figure 1 , Figure 1 A flowchart of a method for device nameplate image recognition and calibration processing provided by an embodiment of the present invention. The method comprises the following steps:

[0060] Acquire a nameplate image set of a target electric power device, and filter the image set according to a preset method to obtain a first image set;

[0061] For the first image set, extract state data of each image, obtain a state coefficient according to the state data, and determine the second image set according to the image corresponding to the state coefficient;

[0062] For each image in the second image set, determine its key point, and transform the image according to the key point to obtain a front view image set;

[0063] For each image in the front-view image set, a new coordinate system is established, and the position of each pixel is corrected according to the state of the actual object corresponding to the key point to obtain the position coordinates of each pixel. The final calibrated image is obtained based on the position coordinates of each pixel, and image recognition is performed on the final calibrated image to determine the nameplate information of the power equipment.

[0064] Based on a method for equipment nameplate image recognition and calibration processing provided in an embodiment of the present invention, the nameplate image of the target power equipment obtained is corrected during recognition in the above manner, so that the possibility of extraction errors or information missing in the final calibration image is smaller, thereby reducing the impact on the accuracy and efficiency of equipment management and maintenance.

[0065] In one embodiment, filtering the image set according to a preset method to obtain the first image set includes:

[0066] Subjecting each image in the image set to a preset image feature processing condition to obtain a processed image set;

[0067] For the processed image set, obtain the value of each feature of each image before and after processing, and obtain the change rate of the feature according to the value before and after processing;

[0068] Comparing the change rate of each feature of the image with a preset normal feature change rate range, and if the change rates of all features of the image are within the preset normal feature change rate range, marking the image as a first usable image;

[0069] If the change rate of any feature of the image is not within the preset normal feature change rate range, marking the image as a first unusable image;

[0070] Take all first available images in the image set as the first image set.

[0071] It should be noted that the preset image feature processing conditions are set by professionals based on actual circumstances, such as edge detection, contrast enhancement, noise removal, etc., and may also include other methods, which are not limited here; in addition, the preset normal feature change rate range is also set by professionals based on actual circumstances, and is not specifically limited.

[0072] For example: Assume that there are two images A and B. Under the preset image feature processing conditions of the two images A and B, the preset normal feature change rate range is:

[0073] Number of edge pixels: the rate of change is between 0.2 and 0.5;

[0074] Contrast value: the rate of change is between 0.5 and 0.8;

[0075] Noise level: The rate of change is between -0.3 and -0.1;

[0076] Filter the initial available images:

[0077] The processing of image A,

[0078] Edge detection: Original image: Number of edge pixels = 800;

[0079] Processed image: number of edge pixels = 960;

[0080] Change rate: (960-800) / 800=0.2;

[0081] Contrast enhancement: Original image: contrast value = 0.4;

[0082] Processed image: contrast value = 0.7;

[0083] Rate of change: (0.7-0.4) / 0.4=0.75;

[0084] Noise removal: Original image: Noise level = 0.6;

[0085] Processed image: noise level = 0.45;

[0086] Rate of change: (0.45-0.6) / 0.6=-0.25;

[0087] The change rate results of image A: the change rate of the number of edge pixels: 0.2 (within the normal range);

[0088] Contrast value change rate: 0.75 (within the normal range);

[0089] Noise level change rate: -0.25 (within the normal range);

[0090] Therefore, image A meets the normal variation range of all processing conditions and is marked as a preliminary usable image.

[0091] Image B processing, edge detection, original image: number of edge pixels = 900;

[0092] Processed image: number of edge pixels = 1400;

[0093] Rate of change: (1400-900) / 900=0.56;

[0094] Contrast enhancement, original image: contrast value = 0.5; processed image: contrast value = 1.0; change rate: (1.0-0.5) / 0.5 = 1.0

[0095] Noise removal, original image: noise level = 0.7; processed image: noise level = 0.6 change rate: (0.6-0.7) / 0.7 = -0.14;

[0096] The change rate results of image B; the change rate of the number of edge pixels: 0.56 (outside the normal range);

[0097] Contrast value change rate: 1.0 (outside the normal range);

[0098] Noise level change rate: -0.14 (within the normal range);

[0099] Since the change rate of the edge pixel number and the contrast value of image B are both beyond the normal range, it is marked as an unusable image.

[0100] In one implementation, if the rate of change of the basic eigenvalues ​​of an image under preset image feature processing conditions is reasonable and within the preset normal feature change rate range, this means that the image performs well under these processing conditions, has consistency and reliability, and therefore, such an image can be considered preliminarily usable and is more likely to identify useful information in subsequent correction processing. This method evaluates the various features of the image through a variety of preset processing conditions, and determines whether the image is within a reasonable range by the rate of change of the eigenvalues. This multi-dimensional screening method can not only filter out images that perform abnormally under certain specific processing conditions, but also comprehensively evaluate the overall quality of the image under different processing conditions, thereby screening out a set of images with higher quality and suitable for further processing. In this way, the interference of invalid images can be effectively reduced, and the accuracy and efficiency of subsequent image processing and analysis can be improved.

[0101] In one embodiment, for the first set of images, extracting status data of each image includes:

[0102] The status data includes brightness uniformity, and the method for obtaining brightness uniformity is:

[0103] Each image in the first image set is converted into a grayscale image, and the brightness mean of the grayscale image is calculated using the formula:

[0104]

[0105] In the formula, μ is the mean brightness, M is the number of rows in the image, N is the number of columns in the image, and I ij is the brightness value of the pixel in the i-th row and j-th column in the grayscale image;

[0106] The brightness variance of the grayscale image is calculated based on the brightness value of each pixel and the brightness mean, and the inverse of the variance is recorded as the brightness uniformity. The calculation formula is:

[0107]

[0108] In the formula, E B is the variance of the changing image, brightness uniformity

[0109] It should be noted that the larger the brightness uniformity Efd is, the more uniform the brightness of the corresponding image of the power equipment nameplate is. This is especially important for the captured image of the power equipment nameplate, because uniform brightness distribution helps the stability and accuracy of subsequent image processing and information extraction. In practical applications, images with higher brightness uniformity can ensure that the text and symbols on the nameplate are clearly visible, and will not cause information loss or difficulty in identification due to local darkness or brightness. It can more accurately identify and extract key features on the power equipment nameplate, which further improves the efficiency and accuracy of image processing and ensures that the final corrected image has high-quality available information.

[0110] In one implementation, through the above method, the calculated brightness uniformity is more accurate, and can more accurately reflect the brightness uniformity of the photographed power equipment nameplate image, and further screen out images that can identify effective information.

[0111] In one embodiment, the status data includes image clarity, and the method for obtaining the image clarity is:

[0112] Applying the Laplacian operator to each image in the first image set to obtain a transformed image L;

[0113] Calculate the pixel mean of the transformed image L, and the calculation formula is:

[0114]

[0115] Where μL is the pixel mean of the image after Laplace transformation, A is the number of rows of the image after Laplace transformation, N is the number of columns of the image after Laplace transformation, and L is ab is the pixel value of the image at position (a, b) after Laplace transformation;

[0116] Calculate the variance of the image after Laplace transformation. The calculation formula can be:

[0117]

[0118] The variance of the image after Laplace transformation is taken as the image clarity Etk, Etk=Var(L).

[0119] It should be noted that the greater the image clarity, the clearer the corresponding image of the nameplate of the power equipment. This is especially important for the image of the nameplate of the power equipment, because a clear image can more accurately capture and identify the text, symbols and other important information on the nameplate. In the management and maintenance of power equipment, the information on the nameplate is the key data source for equipment identification, parameter query and maintenance records. If the image is not clear, it may cause the text to be unclear and the symbols to be difficult to identify, thereby increasing the difficulty of information entry and identification, and may even cause misreading and erroneous records; high-definition images help improve the effect of subsequent image processing algorithms, such as character recognition and feature extraction. These algorithms are usually more stable and accurate on clear images, and can more reliably extract key information on the nameplate, ensuring efficient management and maintenance of power equipment. At the same time, clear images also have higher usability and reference value in archiving and long-term preservation, which is convenient for future reference and verification.

[0120] In one implementation, through the above method, the calculated image clarity is more accurate, and can more accurately reflect the clarity of the image of the photographed electric power equipment nameplate image, and further screen out images that can identify effective information.

[0121] In one embodiment, the status data includes feature matching completeness, and the method for obtaining the feature matching completeness is:

[0122] For each image in the first image set, a feature point detection algorithm is used to detect key feature points in the image as feature points of the image to be tested;

[0123] Using a matching algorithm, calculate the feature point matching pairs between the image to be tested and the reference image as the number of feature point matching;

[0124] The feature matching completeness is obtained by dividing the number of feature point matches of each image by the total number of feature points of the image to be tested. The calculation formula is: Ews=qa / nk, where Ews is the feature matching completeness, qa is the number of feature point matches, and nk is the total number of feature points of the image to be tested.

[0125] It should be noted that the reference image is set by professionals according to the actual situation and is not specifically limited; the feature point detection algorithm may be SIFT or SURF, and the matching algorithm may be KNN or BFMatcher, which may be selected and applied according to the actual situation and are not limited or elaborated.

[0126] It should be noted that the greater the feature matching completeness, the more key features the corresponding photographed power equipment nameplate contains. This is especially important for the photographed image of the power equipment nameplate, because the nameplate usually contains key information such as the equipment model, serial number, manufacturer information, technical parameters, etc. This information is an important basis for equipment identification, maintenance and management. If the feature matching degree in the image is high, it means that the captured image contains more key features of the nameplate, which helps to improve the accuracy and reliability of subsequent processing such as character recognition (OCR) and feature extraction. Complete feature matching can also ensure the accuracy of the image during comparison and archiving, thereby ensuring the integrity and traceability of the information. Images with high feature matching can more effectively support the management of power equipment, reduce errors and omissions, and improve work efficiency.

[0127] In one implementation, through the above method, the feature matching completeness of the calculated image is more accurate, and can more accurately reflect the degree of inclusion of key features of the image of the captured power equipment nameplate image, and further screen out images that can identify valid information.

[0128] In one embodiment, obtaining a state coefficient according to the state data, and determining the second image set according to the image corresponding to the state coefficient comprises:

[0129] The state data includes brightness uniformity, image clarity and feature matching completeness. The state data including brightness uniformity, image clarity and feature matching completeness are normalized, and the state coefficient is obtained according to the normalized state data. The calculation formula is:

[0130]

[0131] Where Pgt is the state coefficient, Efd, Etk and Ews represent brightness uniformity, image clarity and feature matching completeness respectively, b1, b2 and b3 are the proportional coefficients of brightness uniformity, image clarity and feature matching completeness, and b1, b2 and b3 are all greater than 0;

[0132] Compare the state coefficient with a preset state coefficient threshold, and if the state coefficient is not less than the preset state coefficient threshold, record the image corresponding to the state coefficient as the second available image;

[0133] If the state coefficient is less than a preset state coefficient threshold, the image corresponding to the state coefficient is recorded as a second unusable image;

[0134] All first available images in the first image set are taken as the second image set.

[0135] In one implementation, the preset state coefficient threshold is set by professionals according to actual conditions, and is not specifically limited or elaborated.

[0136] In one implementation, the second image set is screened from the first image set in the above manner, ensuring the quality and consistency of the image. This screening method based on brightness uniformity, image clarity, and feature matching completeness can effectively filter out images that do not meet the standards, thereby improving the accuracy and efficiency of subsequent processing and analysis. The screened images are not only clearer and more uniform in visual effect, but also contain more key features, ensuring the integrity and reliability of the nameplate information. This is of great significance for the management, maintenance, and data archiving of power equipment, and can significantly reduce misreading and erroneous records caused by image quality problems, and improve the work efficiency and data accuracy of the entire system.

[0137] In one embodiment, transforming the image according to the key points to obtain a front view image set includes:

[0138] Marking the key points of each image in the second image set as marked key points;

[0139] Based on the marked key points, use perspective transformation to calculate the transformation matrix of the image from the current angle to the orthographic angle, and transform each image according to the calculated transformation matrix to obtain the orthographic angle image of the image;

[0140] The front-view angle image of each image in the second image set is taken as a new image set to obtain a front-view image set.

[0141] It should be noted that the image can be corrected to a normal viewing angle image as follows:

[0142] 1. Mark key points: Determine the key points on the nameplate image of the power equipment. These key points can be the four corners of the nameplate or other obvious marking points. The selection of key points must ensure that they can be accurately located in all images.

[0143] 2. Calculate the transformation matrix: Use image processing algorithms (such as perspective transformation) to calculate the transformation matrix that converts the image from the current angle to the normal angle. This matrix needs to be calculated based on the marked key points.

[0144] 3. Apply transformation: Use the calculated transformation matrix to transform each image so that the nameplate appears at a normal viewing angle in the image.

[0145] For example, assuming a nameplate image, by marking its four corner points and calculating the perspective transformation matrix, the following are the specific steps:

[0146] Mark the key points, assuming the coordinates of the original image corners: upper left corner (100,150), upper right corner (400,120), lower right corner (420,300), lower left corner (110,320);

[0147] Calculate the transformation matrix: The target image size is (500,400).

[0148] Target point coordinates: upper left corner (0,0), upper right corner (500,0), lower right corner (500,400), lower left corner (0,400);

[0149] Apply transformation: Use the perspective transformation algorithm to calculate the transformation matrix, apply the transformation matrix to the original image, and obtain an image at a normal viewing angle.

[0150] The front-view angle image of each image in the second image set is taken as a new image set to obtain a front-view image set.

[0151] In one implementation, each image in the second image set is converted into an image at a normal viewing angle through the above method. This significantly improves the consistency, readability and recognition accuracy of the image, simplifies the subsequent processing process, and optimizes data storage and retrieval. This process ensures that the information on the nameplate of the power equipment is clear and complete, and provides a reliable foundation for subsequent image processing and information extraction, thereby greatly improving the overall work efficiency and data processing accuracy.

[0152] In one embodiment, the position of each pixel point is corrected according to the state of the actual object corresponding to the key point to obtain the position coordinates of each pixel point, and the final calibration image is obtained according to the position coordinates of each pixel point, including:

[0153] For each image in the orthographic image set, the pixel coordinates corresponding to a key point are changed to the new coordinate circle center, and a new coordinate system is established based on the new coordinate circle center;

[0154] Mapping the pixel points corresponding to the origin in the new coordinate system to the actual nameplate of the electric power equipment, and establishing a coordinate system for the actual nameplate of the electric power equipment according to a number of key points, recorded as the actual coordinate system, the origin of the actual coordinate system is the key point corresponding to the origin of the new coordinate system;

[0155] Get the coordinates of the four corner points of the image in the actual coordinate system, the coordinates of the four corner points in the new coordinate system, and calculate the scaling ratio based on the coordinates in the new coordinate system and the coordinates in the actual coordinate system;

[0156] Obtain the coordinates of the electric equipment nameplate position corresponding to each pixel point of each image in the new coordinate system in the actual coordinate system, and determine the coordinates of each pixel point in the new coordinate system according to the coordinates in the actual coordinate system and the scaling ratio to obtain a corrected image of the image;

[0157] Obtain the corrected image of each image in the orthographic image set, and overlap the images with the same new coordinate system, calculate the average value of the coordinates of the same pixel points as the final position coordinates of the pixel points, calculate the final position coordinates of each pixel point, and obtain the final corrected image.

[0158] In one implementation, a key point is generally set as the four corner points of the image. These four corner points are located at the four corner points of the image (such as the upper left corner, the upper right corner, the lower left corner, and the lower right corner). These points are set as the key points of the new coordinate system, and these corner points are usually the most obvious and easiest to identify. The positions of these points in the image are fixed and are not easily disturbed by other objects or noise, making them stable reference points. When the corner points are used as key points, the scaling and other geometric transformations of the image can be more easily calculated. Through the coordinates of the corner points, the width and height of the image, as well as other transformation parameters such as rotation angle, scaling ratio, etc., can be directly calculated, so as to perform accurate corrections.

[0159] In one implementation, the final corrected image obtained by the above method has high precision and consistency. The geometric correction of the image is achieved by accurate coordinate mapping and scaling calculation of key points. The position of each pixel is calculated by the average value of multiple images, which reduces noise and errors, thereby providing a more accurate and consistent image view. This corrected image can effectively improve the recognition and analysis accuracy of power equipment nameplates and improve the reliability and efficiency of image processing.

[0160] The above is 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 to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for equipment nameplate image recognition and calibration, characterized in that: The following steps are involved: Acquire a nameplate image set of a target electric power device, and filter the image set according to a preset method to obtain a first image set; For the first image set, extract state data of each image, obtain a state coefficient according to the state data, and determine the second image set according to the image corresponding to the state coefficient; the state data includes brightness uniformity, image clarity and feature matching completeness; For each image in the second image set, determine its key point, and transform the image according to the key point to obtain a front view image set; For each image in the front-view image set, a new coordinate system is established, and the position of each pixel is corrected according to the state of the actual object corresponding to the key point to obtain the position coordinates of each pixel. The final calibrated image is obtained based on the position coordinates of each pixel, and image recognition is performed on the final calibrated image to determine the nameplate information of the power equipment.

2. The method for calibrating and processing equipment nameplate images according to claim 1, characterized in that: The first image set is obtained by screening the image set according to a preset method, including: Subjecting each image in the image set to a preset image feature processing condition to obtain a processed image set; For the processed image set, obtain the value of each feature of each image before and after processing, and obtain the change rate of the feature according to the value before and after processing; Comparing the change rate of each feature of the image with a preset normal feature change rate range, and if the change rates of all features of the image are within the preset normal feature change rate range, marking the image as a first usable image; If the change rate of any feature of the image is not within the preset normal feature change rate range, marking the image as a first unusable image; Take all first available images in the image set as the first image set.

3. The method for equipment nameplate image recognition and calibration according to claim 1, characterized in that: For the first set of images, extracting status data of each image includes: The state data includes brightness uniformity, and the method for obtaining the brightness uniformity is: Each image in the first image set is converted into a grayscale image, and the brightness mean of the grayscale image is calculated as follows: In the formula, μ is the mean brightness, M is the number of rows in the image, N is the number of columns in the image, and I ij is the brightness value of the pixel in the i-th row and j-th column in the grayscale image; The brightness variance of the grayscale image is calculated based on the brightness value of each pixel and the brightness mean, and the inverse of the variance is recorded as the brightness uniformity. The calculation formula is: In the formula, E B is the variance of the changing image, brightness uniformity 4. The method for calibrating and processing equipment nameplate images according to claim 3, characterized in that: The state data includes image clarity, and the method for obtaining the image clarity is: Applying the Laplacian operator to each image in the first image set to obtain a transformed image L; Calculate the pixel mean of the transformed image L, and the calculation formula is: Where μL is the pixel mean of the image after Laplace transformation, A is the number of rows of the image after Laplace transformation, N is the number of columns of the image after Laplace transformation, and L is ab is the pixel value of the image at position (a, b) after Laplace transformation; Calculate the variance of the image after Laplace transformation. The calculation formula can be: The variance of the image after Laplace transformation is taken as the image clarity Etk, Etk=Var(L).

5. The method for calibrating and processing equipment nameplate images according to claim 3, characterized in that: The state data includes feature matching completeness, and the method for obtaining the feature matching completeness is: For each image in the first image set, a feature point detection algorithm is used to detect key feature points in the image as feature points of the image to be tested; Using a matching algorithm, calculate the feature point matching pairs between the image to be tested and the reference image as the number of feature point matching; The feature matching completeness is obtained by dividing the number of feature point matches of each image by the total number of feature points of the image to be tested. The calculation formula is: Ews=qa / nk, where Ews is the feature matching completeness, qa is the number of feature point matches, and nk is the total number of feature points of the image to be tested.

6. The method for calibrating and processing equipment nameplate images according to claim 1, characterized in that: Obtaining a state coefficient according to the state data, and determining a second image set according to an image corresponding to the state coefficient includes: The state data includes brightness uniformity, image clarity and feature matching completeness. The state data including brightness uniformity, image clarity and feature matching completeness are normalized, and the state coefficient is obtained according to the normalized state data. The calculation formula is: Where Pgt is the state coefficient, Efd, Etk and Ews represent brightness uniformity, image clarity and feature matching completeness respectively, b1, b2 and b3 are the proportional coefficients of brightness uniformity, image clarity and feature matching completeness, and b1, b2 and b3 are all greater than 0; Compare the state coefficient with a preset state coefficient threshold, and if the state coefficient is not less than the preset state coefficient threshold, record the image corresponding to the state coefficient as the second available image; If the state coefficient is less than a preset state coefficient threshold, the image corresponding to the state coefficient is recorded as a second unusable image; All first available images in the first image set are taken as the second image set.

7. The method for equipment nameplate image recognition and calibration according to claim 1, characterized in that: The image is transformed according to the key points, and the orthographic image set includes: Marking the key points of each image in the second image set as marked key points; Based on the marked key points, use perspective transformation to calculate the transformation matrix of the image from the current angle to the orthographic angle, and transform each image according to the calculated transformation matrix to obtain the orthographic angle image of the image; The front-view angle image of each image in the second image set is taken as a new image set to obtain a front-view image set.

8. The method for calibrating and processing equipment nameplate images according to claim 1, characterized in that: According to the state of the actual object corresponding to the key point, each pixel point is positionally corrected to obtain the position coordinates of each pixel point. The final calibration image is obtained according to the position coordinates of each pixel point, including: For each image in the orthographic image set, the pixel coordinates corresponding to a key point are changed to the new coordinate circle center, and a new coordinate system is established based on the new coordinate circle center; Mapping the pixel points corresponding to the origin in the new coordinate system to the actual nameplate of the electric power equipment, and establishing a coordinate system for the actual nameplate of the electric power equipment according to a number of key points, recorded as the actual coordinate system, the origin of the actual coordinate system is the key point corresponding to the origin of the new coordinate system; Get the coordinates of the four corner points of the image in the actual coordinate system, the coordinates of the four corner points in the new coordinate system, and calculate the scaling ratio based on the coordinates in the new coordinate system and the coordinates in the actual coordinate system; Obtain the coordinates of the electric equipment nameplate position corresponding to each pixel point of each image in the new coordinate system in the actual coordinate system, and determine the coordinates of each pixel point in the new coordinate system according to the coordinates in the actual coordinate system and the scaling ratio to obtain a corrected image of the image; Obtain the corrected image of each image in the orthographic image set, and overlap the images with the same new coordinate system, calculate the average value of the coordinates of the same pixel points as the final position coordinates of the pixel points, calculate the final position coordinates of each pixel point, and obtain the final corrected image.

Images