Cable fingerprint-based identification method and system

Abstract

A cable fingerprint-based identification method and system. Rough matching is performed by means of rotated features to reduce computational load. Small-range matching is performed by means of the curve fitting of positional center points to obtain the rotation angle of each wire. The rotation angle of each wire at both ends of a cable is measured to construct a cable fingerprint database. Environmental features of the cable can reflect damage conditions in different environments, and a wire angle variation can represent the wire connection status at both ends of the cable. Meanwhile, on the basis of different angles, damage statuses can also be predicted. Cross-sectional features are cross-sectional features prior to the locating and classification of a wire, and can be used for determining the position and category of the wire. The constructed cable fingerprint database can realize more accurate cable fault detection.

Description

A method and system for cable fingerprint identification

[0001] This application claims priority to Chinese Patent Application No. 202410677751.0, filed on May 29, 2024, entitled "A Method and System for Recognizing Cable Fingerprints", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This invention relates to the field of computer technology, and more specifically, to a cable fingerprint recognition method and system. Background Technology

[0003] Currently, with the advent of intelligent and information-based technologies, we have entered the data era. Taking the power system as an example, cables contain a wealth of data about themselves. Just as human fingerprints may appear similar on the surface, each person's fingerprint is unique at birth. Cable fingerprinting refers to the process of obtaining characteristic data formed during the manufacturing, transportation, and installation of cables through specific testing and analysis, and combining this data to form a unique "fingerprint" used to identify and locate the cable's condition, location, and defects. Cable fingerprinting technology can help achieve full lifecycle management of cables, improve cable safety, reliability, and service life, and reduce the incidence of power accidents. By collecting the implicit information of cables and establishing a feature library that can describe them, it is called a power fingerprint. However, this data is often wasted, as it needs to be acquired every time a cable needs to be judged, tested, or predicted, greatly increasing our computational workload and prolonging the calculation time. Therefore, it is crucial to mine the specific implicit characteristics of cables and construct cable fingerprints that can meet the needs of production and daily life services. Furthermore, accurate judgment of cable installation location and matching based on the cable fingerprint is also necessary. Summary of the Invention

[0004] The purpose of this invention is to provide a cable fingerprint recognition method and system to solve the above-mentioned problems existing in the prior art.

[0005] In a first aspect, embodiments of the present invention provide a cable fingerprint recognition method, comprising:

[0006] The system acquires multiple geographic environment images of the cable installation location at multiple time points and two cross-sectional images of both ends of the cable; the geographic environment images represent the environmental conditions of different geographical locations at multiple time periods within a year; the cross-sectional images represent images of the exposed cross-sections of both ends of the cable taken by a movable camera device on the same horizontal plane; and the cable installation location represents the geographical location of the cable installation.

[0007] Based on multiple geographic environment images at the aforementioned multiple time points and the cable installation location, the cable loss in different environments is predicted, and the cable environmental loss value is obtained.

[0008] Based on the cross-sectional image, the shape and position of the wire in the cross-section are detected to obtain cross-sectional features; two cross-sectional images correspond to two cross-sectional features.

[0009] Using a feature rotation algorithm, based on the two cross-sectional features and two cross-sectional images, the changing position of the conductor in the cable is detected, and the changing angle of the conductor is obtained.

[0010] The cable environmental characteristics, conductor angle variation, and two cross-sectional characteristics are used to construct key-value pairs with the corresponding cables to form a cable fingerprint; one cable corresponds to one cable environmental characteristic and one conductor angle variation corresponds to two cross-sectional characteristics; multiple cables correspond to multiple cable fingerprints;

[0011] Cable fingerprinting identifies the time and location of cable failures.

[0012] Optionally, a cable fingerprint library can be constructed based on multiple cable fingerprints.

[0013] Optionally, the step of detecting the changing position of the conductor in the cable and obtaining the changing angle of the conductor based on the two cross-sectional features and two cross-sectional images using a feature rotation algorithm includes:

[0014] The cross-sectional features are divided into a first cross-sectional feature and a second cross-sectional feature; the cross-sectional image corresponding to the first cross-sectional feature is taken as the first cross-sectional image; the cross-sectional image corresponding to the second cross-sectional feature is taken as the second cross-sectional image.

[0015] The first cross-sectional feature is input into the classification network to detect the distribution position of the conductor, and the first conductor position detection feature is obtained.

[0016] The second cross-sectional feature is input into the classification network to detect the distribution position of the conductor, thereby obtaining the second conductor position detection feature and the center point of the second conductor position.

[0017] By using a feature rotation algorithm, multiple rotation features are obtained based on the position detection features of the first conductor.

[0018] By using a matching neural network, the position and shape of the conductor are matched based on the second conductor position detection feature and the rotation feature to obtain a conductor matching value; multiple rotation features correspond to multiple conductor matching values.

[0019] Based on the first cross-sectional feature and multiple wire matching values, the first rotating cross-sectional feature is obtained;

[0020] The change angle of the conductor is obtained by matching the first cross-sectional image, the second cross-sectional image, the first rotating cross-sectional feature, and the center point of the second conductor position.

[0021] Optionally, the step of obtaining multiple rotation features based on the first conductor position detection features using a feature rotation algorithm includes:

[0022] The sum of the height and width of the first conductor position detection feature minus 1 is used as the column number, the sum of the width and height of the first conductor position detection feature minus 1 is used as the column number, and the zero matrix corresponding to the page number of the length of the first conductor position detection feature is used as the increment matrix.

[0023] Based on the center point of the matrix corresponding to the center point of the first conductor position detection feature, the first conductor position detection feature is copied into the matrix to obtain the conductor augmentation matrix.

[0024] Multiple rotation angles are obtained; the rotation angles are 45 degrees, 90 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees; the rotation angle represents the angle by which the first conductor position detection feature is rotated.

[0025] Using the center point of the conductor addition matrix as a reference, the values ​​in the conductor addition matrix are rotated according to the rotation angle to obtain rotation features; multiple rotation angles correspond to multiple rotation features.

[0026] Optionally, the step of matching the first cross-sectional image, the second cross-sectional image, the first rotated cross-sectional feature, and the center point of the second conductor position to obtain the conductor change angle includes:

[0027] The first cross-sectional feature is input into a classification network for classification to obtain the center point of the first conductor position.

[0028] The center point of the first conductor is marked in the first cross-sectional image to obtain the first conductor image; the center point of the second conductor is marked in the second cross-sectional image to obtain the second conductor image.

[0029] Based on the first traverse image, value lines marking the changes in the positions of multiple center points in the first traverse image are used to obtain a first curve image; a second curve image is obtained correspondingly from the second traverse image.

[0030] Based on the first curve image and the second curve image, curve matching is performed through the input curve matching network to determine the angle difference and obtain the change angle of the conductor.

[0031] Optionally, the step of marking the value lines indicating the changes in the positions of multiple center points in the first traverse image to obtain a first curve image based on the first traverse image includes:

[0032] The first conductor image is divided into multiple segmentation sets by dividing it at different angles based on the center point;

[0033] The values ​​in the segmentation set are fitted to obtain a fitting curve; multiple segmentation sets correspond to multiple fitting curves.

[0034] Multiple fitted curves are marked on the first guideline image to obtain the first curve image.

[0035] Optionally, the step of matching the position and shape of the conductor using a matching neural network based on the second conductor position detection feature and the rotation feature to obtain a conductor matching value includes:

[0036] The matching neural network is a discriminative network; the matching neural network has two outputs, one output representing the probability of a match and the other output representing the probability of no match;

[0037] If the second conductor position detection feature and the rotation feature are input into the matching neural network, the matching feature is extracted to obtain the first matching probability and the first non-matching probability;

[0038] The first matching probability divided by the first unmatched probability is used as the wire matching value.

[0039] Optionally, obtaining the first rotating cross-sectional feature based on the first cross-sectional feature and multiple wire matching values ​​includes:

[0040] The wire matching value that is greater than the other wire matching values ​​among the plurality of wire matching values ​​shall be taken as the first wire matching value;

[0041] The rotation angle corresponding to the rotation feature corresponding to the first wire matching value is taken as the first rotation angle;

[0042] The first cross-sectional feature is rotated according to the first rotation angle to obtain the first rotated cross-sectional feature.

[0043] Optionally, the prediction of cable loss under different environments based on multiple geographic environment images at multiple time points and the cable installation location, to obtain cable environmental characteristics, includes:

[0044] The cable installation location is segmented in the geographic environment image to obtain a cable installation image;

[0045] The cable installation images are input into an environmental monitoring network to predict cable loss under different environments, thus obtaining environmental convolutional features; multiple geographical environment images at multiple time points correspond to multiple environmental convolutional features.

[0046] The multiple environmental convolutional features are input into a temporal convolutional network for convolution to obtain the cable environmental features.

[0047] Secondly, embodiments of the present invention provide a cable fingerprint recognition system, comprising:

[0048] Acquisition Module: Acquires multiple geographic environment images of the cable installation location at multiple time points and two cross-sectional images of both ends of the cable; the geographic environment images represent the environmental conditions of different geographical locations at multiple time periods within a year; the cross-sectional images represent images of the exposed cross-sections of both ends of the cable taken by a movable camera device on the same horizontal plane; the cable installation location represents the geographical location of the cable installation.

[0049] Environmental prediction module: Based on multiple geographic environment images at multiple time points and the cable installation location, predict the cable loss in different environments and obtain the cable environmental loss value;

[0050] Cross-section detection module: Based on the cross-section image, detect the shape and position of the wire in the cross-section to obtain cross-section features; two cross-section images correspond to two cross-section features;

[0051] Feature rotation module: Based on the two cross-sectional features and two cross-sectional images, the feature rotation algorithm is used to detect the changing position of the conductor in the cable and obtain the changing angle of the conductor.

[0052] Key-value pair module: Constructs key-value pairs with the cable environmental features, conductor angle variation, and two cross-sectional features to form a cable fingerprint; one cable corresponds to one cable environmental feature and one conductor angle variation corresponds to two cross-sectional features; multiple cables correspond to multiple cable fingerprints;

[0053] Identification module: Identifies the time and location of cable failures based on cable fingerprints.

[0054] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0055] This invention also provides a cable fingerprinting method and system. The method includes: obtaining the cable installation location, multiple geographical environment images at multiple time points, and two cross-sectional images of both ends of the cable. The geographical environment images represent the environmental conditions of different geographical locations over multiple time periods within a year. The cross-sectional images represent images of the exposed cross-sections of both ends of the cable, taken by a movable camera device on the same horizontal plane. The cable installation location represents the geographical location of the cable installation. Based on the multiple geographical environment images at multiple time points and the cable installation location, the cable loss in different environments is predicted to obtain a cable environmental loss value. Based on the cross-sectional images, the shape and position of the conductors in the cross-section are detected to obtain cross-sectional features. Two cross-sectional images correspond to two cross-sectional features. Using a feature rotation algorithm, based on the two cross-sectional features and the two cross-sectional images, the changing position of the conductors in the cable is detected to obtain the changing angle of the conductors. The cable environmental features, the changing angle of the conductors, and the two cross-sectional features are used to construct key-value pairs with the corresponding cables as cable fingerprints. One cable corresponds to one cable environmental feature and one changing angle of the conductor corresponds to two cross-sectional features. Multiple cables correspond to multiple cable fingerprints. The cable fingerprints are used to identify the fault time and location of the cable.

[0056] This invention obtains the predicted damage status of cables under different environments, determines the position of conductors in the cable, and obtains cross-sectional features based on the shape and color of the conductors. These cross-sectional features are rotated, and approximate matching is performed using the rotated features to reduce computational load. Curve fitting at the position center point is used for small-range matching to obtain the conductor rotation angle. The rotation angles of the conductors at both ends of the cable are detected to construct a cable fingerprint database. The cable environmental features reflect the damage status in different environments, and the changing conductor angles represent the state of the connecting conductors at both ends of the cable. Furthermore, the damage status can be predicted for different angles. The cross-sectional features are the cross-sectional features before the conductor position is classified, and can be used to determine the position and category of the conductor. The constructed cable fingerprint database can meet the requirements of reducing installation matching and replacing cables based on environmental damage, reducing computational load, and accelerating calculation speed. This achieves the technical effect of more accurately detecting cable faults and accurately identifying the location of faults at both ends of an already installed cable. Attached Figure Description

[0057] Figure 1 is a flowchart of a cable fingerprint recognition method provided by an embodiment of the present invention. Detailed Implementation

[0058] The present invention will now be described in detail with reference to the accompanying drawings.

[0059] Example 1

[0060] As shown in Figure 1, this embodiment of the invention provides a cable fingerprint recognition method, the method comprising:

[0061] S101: Obtain multiple geographic environment images of the cable installation location at multiple time points, and two cross-sectional images of both ends of the cable. The geographic environment images represent the environmental conditions at multiple time points within a year at different geographic locations. The cross-sectional images represent images of the exposed cross-sections of both ends of the cable, taken by a movable camera device on the same horizontal plane. The cable installation location represents the geographic location of the cable installation.

[0062] In this embodiment, the environment includes temperature and humidity. It may also include the acidity and alkalinity of the soil in the geographical area.

[0063] In this embodiment, the time interval of a period is 6 hours. The temperature corresponding to a time point is the average temperature over the 6 hours between the previous time point and the current time point, and the humidity corresponding to a time point is the average humidity over the 6 hours between the previous time point and the current time point.

[0064] The geographic environment image is obtained by capturing the time point corresponding to each geographic location, and the cable installation location can be found in the geographic environment image.

[0065] S102: Based on multiple geographic environment images at the multiple time points and the cable installation location, predict the cable loss in different environments and obtain the cable environmental loss value.

[0066] S103: Based on the cross-sectional image, detect the shape and position of the wire in the cross-section to obtain cross-sectional features. Two cross-sectional features are obtained corresponding to two cross-sectional images.

[0067] S104: Using a feature rotation algorithm, based on the two cross-sectional features and two cross-sectional images, detect the changing position of the conductor in the cable and obtain the changing angle of the conductor.

[0068] S105: The cable environmental characteristics, conductor change angle, and two cross-sectional characteristics are paired with the corresponding cable to form a cable fingerprint. One cable corresponds to one cable environmental characteristic and one conductor change angle corresponds to two cross-sectional characteristics. Multiple cables correspond to multiple cable fingerprints.

[0069] The cable environmental characteristics reflect damage levels in different environments, the conductor angle variation indicates the state of the connecting conductors at both ends of the cable, and the damage state can be predicted for different angles. The cross-sectional characteristics are those before the conductor's location is classified, and can be used to determine the conductor's location and category.

[0070] S106: Identify the time and location of cable failures based on cable fingerprints.

[0071] Among these methods, the timing of cable faults can be predicted based on cable environmental characteristics. The cross-sectional image can be rotated based on the changing angle of the conductor, and conductors at similar positions can be marked in two cross-sectional images to obtain matching conductor numbers. The energized end of a conductor can be located and matched with its counterpart. The discharge status of the conductors at both ends of a connected cable can be obtained to determine if each conductor is faulty. If a conductor is faulty, the location of the fault can be found based on the matching conductor number, indicating, for example, which conductor is faulty.

[0072] In this embodiment of the invention, the constructed fingerprint database contains standard fingerprints corresponding to cable fingerprints and cable fault locations. Specifically, for S106: the cable fingerprint is matched with the standard fingerprint. Specifically, the cosine of the vector angle between the matched cable fingerprint and the standard fingerprint can be calculated. If the cosine value is less than 0.5, the cable fingerprint is determined to match the standard fingerprint. Then, the fault location corresponding to the standard fingerprint is taken as the fault location of the cable fingerprint, and the detection time point (camera time point) is taken as the cable fault time.

[0073] Optionally, the method further includes: constructing a cable fingerprint library based on multiple cable fingerprints.

[0074] Optionally, the step of detecting the changing position of the conductor in the cable and obtaining the changing angle of the conductor based on the two cross-sectional features and two cross-sectional images using a feature rotation algorithm includes:

[0075] The cross-sectional features are divided into a first cross-sectional feature and a second cross-sectional feature. The cross-sectional image corresponding to the first cross-sectional feature is taken as the first cross-sectional image. The cross-sectional image corresponding to the second cross-sectional feature is taken as the second cross-sectional image.

[0076] The first cross-sectional feature is input into the classification network to detect the distribution position of the conductor, thereby obtaining the first conductor position detection feature.

[0077] The second cross-sectional feature is input into the classification network to detect the distribution position of the conductor, thereby obtaining the second conductor position detection feature and the center point of the second conductor position.

[0078] Specifically, the second cross-sectional feature is input into a classification network to detect the center point of all conductors in the cross-sectional image. The feature output of the convolution layer preceding the last convolution layer used to obtain the center point is used as the second conductor position detection feature.

[0079] In this embodiment, the classification network is a convolutional neural network (CNN) used to detect the location of each wire.

[0080] By using a feature rotation algorithm, multiple rotation features are obtained based on the position detection features of the first conductor.

[0081] The second conductor position detection feature and the rotation feature are input into a matching neural network to match the conductor position and shape, obtaining a conductor matching value. Multiple rotation features correspond to multiple conductor matching values.

[0082] Based on the multiple conductor matching values, a first rotating cross-sectional image and a second cross-sectional image are obtained.

[0083] The change angle of the conductor is obtained by matching the first cross-sectional image, the second cross-sectional image, the first rotating cross-sectional feature, and the center point of the second conductor position.

[0084] Optionally, based on the position detection features of the first conductor, multiple rotation features are obtained using a feature rotation algorithm, including:

[0085] The sum of the height and width of the first conductor position detection feature minus 1 is used as the column number, the sum of the width and height of the first conductor position detection feature minus 1 is used as the column number, and the zero matrix corresponding to the page number of the length of the first conductor position detection feature is used as the increment matrix.

[0086] The first conductor position detection feature is represented by a three-dimensional matrix, where the number of columns of the three-dimensional matrix corresponds to the height after feature extraction from the cross-sectional image, and the number of rows of the three-dimensional matrix corresponds to the width after feature extraction from the cross-sectional image.

[0087] By using the above method, after adding the boundary, the value in the first wire position detection feature can be placed in a position during the rotation process.

[0088] The first conductor position detection feature is copied into the augmentation matrix according to the center point of the matrix corresponding to the center point of the first conductor position detection feature, thus obtaining the conductor augmentation matrix.

[0089] Specifically, the center point of the first conductor position detection feature and the center point of the augmenting matrix are found. Then, the value of each value in the first conductor position detection feature on the augmenting matrix is ​​found, such as the feature vector corresponding to the upper left corner of the first conductor position detection feature being placed in the matrix at the position of the sum of the difference between the horizontal coordinate of the center point of the augmenting matrix and half the width of the first conductor position detection feature, and the vertical coordinate of the center point of the augmenting matrix and half the height of the first conductor position detection feature.

[0090] If one of the pages in the first conductor position detection feature is: If the center point is in the second row and second column, then the matrix increases to 5 rows and 5 columns. Therefore, one page of the matrix for increasing the traverse is... The center point is located in the third row and third column.

[0091] Multiple rotation angles are obtained. The rotation angles are 45 degrees, 90 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees. The rotation angle represents the angle by which the first conductor position detection feature is rotated.

[0092] Using the center point of the conductor augmentation matrix as a reference, the values ​​in the conductor augmentation matrix are rotated according to the rotation angle to obtain rotation features. Multiple rotation angles correspond to multiple rotation features.

[0093] For example, if the rotation angle is 45 degrees, then one of the corresponding rotation features is... Indicates the addition of a matrix to the wires In the middle, the 1 in the second row and second column is rotated 45 degrees to the third row and second column according to the position of the third row and third column.

[0094] The rotation feature indicates that the corresponding image has been rotated.

[0095] Using the method described above, multiple positions of the rotated image can be obtained with only one convolution. This reduces the computational cost by 1 and enables multi-angle matching.

[0096] Optionally, the step of matching the first cross-sectional image, the second cross-sectional image, the first rotated cross-sectional feature, and the center point of the second conductor position to obtain the conductor change angle includes:

[0097] The first cross-sectional feature is input into a classification network for classification to obtain the center point of the first conductor position.

[0098] In this embodiment, the classification network is a convolutional neural network (CNN) used to detect the location of each wire.

[0099] The center point of the first conductor is marked in the first cross-sectional image to obtain the first conductor image. The center point of the second conductor is marked in the second cross-sectional image to obtain the second conductor image.

[0100] Based on the first traverse image, value lines marking the changes in the positions of multiple center points in the first traverse image are used to obtain a first curve image. A second curve image is then obtained correspondingly from the second traverse image.

[0101] Based on the first curve image and the second curve image, curve matching is performed through the input curve matching network to determine the angle difference and obtain the change angle of the conductor.

[0102] Optionally, the step of marking the value lines indicating the changes in the positions of multiple center points in the first traverse image to obtain a first curve image based on the first traverse image includes:

[0103] The first conductor image is divided into multiple segmentation sets by dividing it at different angles based on the center point.

[0104] In this embodiment, the division is done in 5-degree increments.

[0105] The values ​​in the segmentation set are fitted to obtain a fitted curve. Multiple segmentation sets correspond to multiple fitted curves.

[0106] The least squares method was used for fitting.

[0107] Multiple fitted curves are marked on the first guideline image to obtain the first curve image.

[0108] Optionally, the step of matching the position and shape of the conductor using a matching neural network based on the second conductor position detection feature and the rotation feature to obtain a conductor matching value includes:

[0109] The matching neural network is a discriminative network. The matching neural network has two outputs: one representing the probability of a match, and the other representing the probability of no match.

[0110] The discriminant network is a discriminator network.

[0111] If the second conductor position detection feature and the rotation feature are input into the matching neural network, the matching feature is extracted to obtain the first matching probability and the first non-matching probability.

[0112] The first matching probability divided by the first unmatched probability is used as the wire matching value.

[0113] Optionally, obtaining the first rotating cross-sectional feature based on the first cross-sectional feature and multiple wire matching values ​​includes:

[0114] The conductor matching value that is greater than the other conductor matching values ​​among the plurality of conductor matching values ​​is taken as the first conductor matching value.

[0115] The rotation angle corresponding to the rotation feature of the first wire matching value is taken as the first rotation angle.

[0116] The first cross-sectional feature is rotated according to the first rotation angle to obtain the first rotated cross-sectional feature.

[0117] Optionally, the prediction of cable loss under different environments based on multiple geographic environment images at multiple time points and the cable installation location, to obtain cable environmental characteristics, includes:

[0118] The cable installation location is segmented in the geographic environment image to obtain a cable installation image.

[0119] The cable installation images are input into an environmental monitoring network to predict cable loss under different environments, resulting in environmental convolutional features. Multiple geographical environment images from multiple time points correspond to multiple environmental convolutional features.

[0120] In this embodiment, the environmental monitoring network is a recurrent neural network (RNN).

[0121] The multiple environmental convolutional features are input into a temporal convolutional network for convolution to obtain the cable environmental features.

[0122] In this embodiment, the temporal convolutional network is a temporal convolutional network (TCN).

[0123] Optionally, the environmental monitoring network can be trained using a training set of multiple regions containing different environments from geographic environmental images.

[0124] During training, values ​​for different times in different environments are used for training.

[0125] Example 2

[0126] Based on the above-described cable fingerprinting method, this invention also provides a cable fingerprinting system, the system comprising:

[0127] The acquisition module is used to obtain multiple geographic environment images of the cable installation location at multiple time points, and two cross-sectional images of both ends of the cable. The geographic environment images represent the environmental conditions at different geographical locations over multiple time periods within a year. The cross-sectional images represent images of the exposed cross-sections of both ends of the cable, taken by a movable camera device on the same horizontal plane. The cable installation location represents the geographical location of the cable installation.

[0128] The environmental prediction module is used to predict cable loss in different environments based on multiple geographic environment images at multiple time points and the cable installation location, and obtain cable environmental loss values.

[0129] The cross-section detection module is used to detect the shape and position of the wires in the cross-section based on the cross-section image, thereby obtaining cross-section features. Two cross-section images correspond to two cross-section features.

[0130] The feature rotation module is used to detect the changing position of the conductor in the cable and obtain the changing angle of the conductor based on the two cross-sectional features and two cross-sectional images using a feature rotation algorithm.

[0131] The key-value pair module is used to construct key-value pairs with the cable's environmental characteristics, conductor angle variation, and two cross-sectional features to create a cable fingerprint. One cable corresponds to one cable environmental feature, and one conductor angle variation corresponds to two cross-sectional features. Multiple cables can be associated with multiple cable fingerprints.

[0132] The identification module is used to identify the time and location of cable failures based on cable fingerprints.

[0133] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0134] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0135] Similarly, it should be understood that, in order to simplify this disclosure and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this method of disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into this detailed description, wherein each claim itself is a separate embodiment of the invention.

[0136] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0137] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0138] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the apparatus according to embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0139] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. A cable fingerprint recognition method, characterized in that, include: Obtain multiple geographic environment images of the cable installation location at multiple time points, and two cross-sectional images of both ends of the cable; The geographic environment image represents the environmental conditions of different geographic locations at multiple time periods within a year; the cross-sectional image represents the cross-sectional image of both ends of the cable exposed, captured by a movable camera device on the same horizontal plane; the cable installation location represents the geographic location of the cable installation. Based on multiple geographic environment images at the aforementioned multiple time points and the cable installation location, the cable loss in different environments is predicted, and the cable environmental loss value is obtained. Based on the cross-sectional image, the shape and position of the wires in the cross-section are detected to obtain the cross-sectional features; Two cross-sectional images correspond to two cross-sectional features; Using a feature rotation algorithm, based on the two cross-sectional features and two cross-sectional images, the changing position of the conductor in the cable is detected, and the changing angle of the conductor is obtained. The cable environmental characteristics, conductor angle variation, and two cross-sectional characteristics are used as key-value pairs with the corresponding cable to form a cable fingerprint. One cable corresponds to one cable environmental characteristic, and a change in conductor angle corresponds to two cross-sectional characteristics; Multiple cable fingerprints are obtained for each corresponding cable; Cable fingerprinting identifies the time and location of cable failures.

2. The cable fingerprint recognition method according to claim 1, characterized in that, The method further includes: A cable fingerprint library is constructed based on multiple cable fingerprints.

3. The cable fingerprint recognition method according to claim 1, characterized in that, The method of detecting the changing position of the conductor in the cable and obtaining the changing angle of the conductor through the feature rotation algorithm based on the two cross-sectional features and two cross-sectional images includes: The cross-sectional features are divided into a first cross-sectional feature and a second cross-sectional feature; the cross-sectional image corresponding to the first cross-sectional feature is taken as the first cross-sectional image; the cross-sectional image corresponding to the second cross-sectional feature is taken as the second cross-sectional image. The first cross-sectional feature is input into the classification network to detect the distribution position of the conductor, and the first conductor position detection feature is obtained. The second cross-sectional feature is input into the classification network to detect the distribution position of the conductor, thereby obtaining the second conductor position detection feature and the center point of the second conductor position. By using a feature rotation algorithm, multiple rotation features are obtained based on the position detection features of the first conductor. By using a matching neural network, the position and shape of the conductor are matched based on the second conductor position detection feature and the rotation feature to obtain a conductor matching value; multiple rotation features correspond to multiple conductor matching values. Based on the first cross-sectional feature and multiple wire matching values, the first rotating cross-sectional feature is obtained; The change angle of the conductor is obtained by matching the first cross-sectional image, the second cross-sectional image, the first rotating cross-sectional feature, and the center point of the second conductor position.

4. The cable fingerprint recognition method according to claim 3, characterized in that, The feature rotation algorithm, based on the position detection features of the first conductor, yields multiple rotation features, including: The sum of the height and width of the first conductor position detection feature minus 1 is used as the column number, the sum of the width and height of the first conductor position detection feature minus 1 is used as the column number, and the zero matrix corresponding to the page number of the length of the first conductor position detection feature is used as the increment matrix. Based on the center point of the matrix corresponding to the center point of the first conductor position detection feature, the first conductor position detection feature is copied into the matrix to obtain the conductor augmentation matrix. Multiple rotation angles are obtained; the rotation angles are 45 degrees, 90 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees; the rotation angle represents the angle by which the first conductor position detection feature is rotated. Using the center point of the conductor addition matrix as a reference, the values ​​in the conductor addition matrix are rotated according to the rotation angle to obtain rotation features; multiple rotation angles correspond to multiple rotation features.

5. The cable fingerprint recognition method according to claim 3, characterized in that, The step of matching the first cross-sectional image, the second cross-sectional image, the first rotated cross-sectional feature, and the center point of the second conductor position to obtain the conductor change angle includes: The first cross-sectional feature is input into a classification network for classification to obtain the center point of the first conductor position. The center point of the first conductor is marked in the first cross-sectional image to obtain the first conductor image; the center point of the second conductor is marked in the second cross-sectional image to obtain the second conductor image. Based on the first traverse image, value lines marking the changes in the positions of multiple center points in the first traverse image are used to obtain a first curve image; a second curve image is obtained correspondingly from the second traverse image. Based on the first curve image and the second curve image, curve matching is performed through the input curve matching network to determine the angle difference and obtain the change angle of the conductor.

6. The cable fingerprint recognition method according to claim 5, characterized in that, The step of marking the value lines indicating the changes in the positions of multiple center points in the first traverse image based on the first traverse image to obtain a first curve image includes: The first conductor image is divided into multiple segmentation sets by dividing it at different angles based on the center point; The values ​​in the segmentation set are fitted to obtain a fitting curve; multiple segmentation sets correspond to multiple fitting curves. Multiple fitted curves are marked on the first guideline image to obtain the first curve image.

7. The cable fingerprint recognition method according to claim 3, characterized in that, The step of matching the position and shape of the conductor using a matching neural network based on the second conductor position detection feature and the rotation feature to obtain a conductor matching value includes: The matching neural network is a discriminative network; the matching neural network has two outputs, one output representing the probability of a match and the other output representing the probability of no match; If the second conductor position detection feature and the rotation feature are input into the matching neural network, the matching feature is extracted to obtain the first matching probability and the first non-matching probability; The first matching probability divided by the first unmatched probability is used as the wire matching value.

8. The cable fingerprint recognition method according to claim 3, characterized in that, The process of obtaining the first rotating cross-sectional feature based on the first cross-sectional feature and multiple wire matching values ​​includes: The wire matching value that is greater than the other wire matching values ​​among the plurality of wire matching values ​​shall be taken as the first wire matching value; The rotation angle corresponding to the rotation feature corresponding to the first wire matching value is taken as the first rotation angle; The first cross-sectional feature is rotated according to the first rotation angle to obtain the first rotated cross-sectional feature.

9. The cable fingerprint recognition method according to claim 1, characterized in that, Based on multiple geographic environment images at multiple time points and the cable installation location, the cable loss is predicted under different environments to obtain cable environmental characteristics, including: The cable installation location is segmented in the geographic environment image to obtain a cable installation image; The cable installation images are input into an environmental monitoring network to predict cable loss under different environments, thus obtaining environmental convolutional features; multiple geographical environment images at multiple time points correspond to multiple environmental convolutional features. The multiple environmental convolutional features are input into a temporal convolutional network for convolution to obtain the cable environmental features.

10. A cable fingerprint recognition system, characterized in that, include: Acquisition module: Obtains multiple geographic environment images of the cable installation location at multiple time points and two cross-sectional images of both ends of the cable; The geographic environment image represents the environmental conditions of different geographic locations at multiple time periods within a year; the cross-sectional image represents the cross-sectional image of both ends of the cable exposed, captured by a movable camera device on the same horizontal plane; the cable installation location represents the geographic location of the cable installation. Environmental prediction module: Based on multiple geographic environment images at multiple time points and the cable installation location, predict the cable loss in different environments and obtain the cable environmental loss value; Cross-section detection module: Based on the cross-section image, detects the shape and position of the wires in the cross-section to obtain cross-sectional features; Two cross-sectional images correspond to two cross-sectional features; Feature rotation module: Based on the two cross-sectional features and two cross-sectional images, the feature rotation algorithm is used to detect the changing position of the conductor in the cable and obtain the changing angle of the conductor. Key-value pair module: Constructs key-value pairs with the cable environmental characteristics, conductor angle variation, and two cross-sectional characteristics to form a cable fingerprint; One cable corresponds to one cable environmental characteristic, and a change in conductor angle corresponds to two cross-sectional characteristics; Multiple cable fingerprints are obtained for each corresponding cable; Identification module: Identifies the time and location of cable failures based on cable fingerprints.

Images