Power grid terminal detection method and system based on target detection and image comparison

Through object detection and image comparison technology, faults of power grid terminals are identified, the problem of low efficiency of traditional manual inspection is solved, fast and accurate fault detection is achieved, and the maintenance efficiency of power grid terminals is improved.

CN119027696APending Publication Date: 2024-11-26ANHUI ZENITH ELECTRICITY & ELECTRONICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411126468.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Traditional power grid terminal fault detection relies on manual inspection, which is inefficient and difficult to detect and report faults in a timely manner.

Method used

Using a method based on object detection and image comparison, by acquiring real-time and historical meter box images, using trained object detection and image comparison networks, identifying meta-targets and calculating feature similarity, and determining whether a fault exists or not.

Benefits of technology

It greatly improves detection efficiency, reduces the workload of on-site inspection personnel, can quickly and intuitively obtain detection results, and reduces the time for fault detection and processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027696A_ABST
    Figure CN119027696A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of power grid maintenance, in particular to a power grid terminal detection method and system based on target detection and image comparison. According to the method, a target detection and image comparison mode is adopted, the element targets are recognized based on the real-time image and the historical image of the electric meter box, and the feature similarity of the two element targets corresponding to the real-time image and the historical image is analyzed to judge whether a fault exists or not, so that a detection result can be quickly and intuitively obtained; compared with a traditional manual troubleshooting mode, the method greatly reduces the workload of field troubleshooting personnel, improves the detection efficiency, and has important practical significance and economic value. According to the invention, the problems of low efficiency and incapability of ensuring the effect of the existing manual fault detection mode of the power grid terminal are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power grid maintenance, and more specifically to: 1. a power grid terminal detection method based on target detection and image comparison; 2. a power grid terminal detection system based on target detection and image comparison. Background Art

[0002] The power grid terminal refers to the meter box where the electricity meter is installed in a certain area.

[0003] With the implementation of the new generation of power grid standards, the number of meter boxes installed has continued to increase, and the difficulty of management and maintenance has increased linearly. The traditional maintenance method of meter boxes mainly relies on manual inspections, which requires trained electricians to open the meter boxes one by one and conduct full-area inspections. This method has high technical requirements for electricians and is inefficient, making it difficult to detect and report faults in a timely manner. Summary of the invention

[0004] Based on this, it is necessary to provide a power grid terminal detection method and system based on target detection and image comparison to address the problems of low efficiency and unguaranteed effect of the existing manual fault detection method for power grid terminals.

[0005] The present invention is implemented by the following technical solutions:

[0006] In a first aspect, the present invention discloses a power grid terminal detection method based on target detection and image comparison, comprising the following steps:

[0007] Step 1: obtain a real-time image of the target electric meter box, and use the trained target detection network to process the real-time image to identify all meta-targets in the real-time image.

[0008] Among them, all meta-targets in the real-time image include: X1, X2, ..., X N ;X i =(x i ,y i ,w i ,h i ), i∈[1,N]; x i ,y i 、w i 、h i Respectively represent the X-axis center coordinate, Y-axis center coordinate, width, and length of the i-th meta-target in the real-time image;

[0009] Among them, the types of meta-targets include: meter box, meter, wiring location, and screw.

[0010] Step 2: Obtain the historical image of the target electric meter box and all meta-targets therein, and perform meta-target matching with the real-time image to find the meta-target in the historical image corresponding to the meta-target in the real-time image;

[0011] The two meta-targets corresponding to the real-time image and the historical image are combined into a meta-target pair.

[0012] Among them, the historical image is the image of the target electric meter box when there is no fault in the past;

[0013] All meta-objects in the historical image include: Y1, Y2, ..., Y M ; Y j =(x' j ,y' j ,w' j ,h' j ), j∈[1,M]; x' j ,y' j 、w' j 、h' j They respectively represent the X-axis center coordinate, Y-axis center coordinate, width, and length of the j-th meta-target in the historical image.

[0014] Step 3: Use the trained image contrast network to process all meta-target pairs and calculate their feature similarity;

[0015] If the feature similarity of a pair of meta-targets is less than a preset similarity threshold K, the meta-targets in the pair of meta-targets are marked as faulty in the real-time image; otherwise, the pair of meta-targets is not faulty.

[0016] This power grid terminal detection method based on target detection and image comparison implements the method or process according to the embodiment of the present disclosure.

[0017] In a second aspect, the present invention discloses a power grid terminal detection system based on target detection and image comparison, comprising: a cloud server and a handheld device. The handheld device is connected to the cloud server via wireless communication.

[0018] The power grid terminal detection system based on target detection and image comparison uses the power grid terminal detection method based on target detection and image comparison disclosed in the first aspect.

[0019] The power grid terminal detection system based on target detection and image comparison implements the method or process according to the embodiment of the present disclosure.

[0020] In a third aspect, the present invention discloses a computer program product, including a computer program. When the computer program is executed by a processor, the steps of the power grid terminal detection method based on target detection and image comparison disclosed in the second aspect are implemented.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] The present invention adopts the method of target detection and image comparison, identifies meta-targets based on the real-time image and historical image of the meter box, and analyzes the feature similarity of the two meta-targets corresponding to the real-time image and the historical image to determine whether there is a fault, and can quickly and intuitively obtain the detection results; compared with the traditional manual troubleshooting method, it greatly reduces the workload of on-site troubleshooting personnel and improves the detection efficiency, which has important practical significance and economic value. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0024] Figure 1 A flow chart of a power grid terminal detection method based on target detection and image comparison provided in Example 1 of the present invention;

[0025] Figure 2 A data flow diagram of a power grid terminal detection method based on target detection and image comparison provided in Embodiment 1 of the present invention;

[0026] Figure 3 for Figure 2 Schematic diagram of meta-target recognition results for real-time images. DETAILED DESCRIPTION

[0027] 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.

[0028] It should be noted that when a component is referred to as being "mounted on" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be a central component at the same time. When a component is considered to be "fixed to" another component, it may be directly fixed on the other component or there may be a central component at the same time.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0030] Example 1

[0031] See also Figure 1 , Figure 2 , shows the process of the power grid terminal detection method based on target detection and image comparison, which includes the following steps:

[0032] Step 1: obtain a real-time image of the target electric meter box, and use the trained target detection network to process the real-time image to identify all meta-targets in the real-time image.

[0033] Generally speaking, a handheld device with a camera (such as a power grid operation and maintenance mobile phone, or a smart phone with an operation and maintenance APP installed) can be used to take real-time photos of the target meter box to obtain a real-time image of the target meter box. This method has lower requirements for power grid operation and maintenance personnel, but it is necessary to ensure the angle and clarity of the image shooting.

[0034] The target detection network aims to find out which meta-targets (classifications) are in the target image and their corresponding position data (coordinates, size, etc.), which can be: R-CNN network, SPP-Net network, SSD network, YOLO network, or other mature networks. In this embodiment 1, it is recommended to use Yolov5 network as the target detection network, and use historical data set 1 (i.e., the meter box image set with correct classification information) to train it to obtain a trained target detection network.

[0035] In addition, due to the large size of the Yolov5 network, it is not suitable for being built into handheld devices at this stage. Therefore, it is recommended to store the trained target detection network in a cloud server, which has sufficient performance to support the processing load of the target detection network. In this way, the handheld device and the cloud server are connected through wireless communication to establish two-way communication. Then, the real-time image of the target meter box is obtained through the handheld terminal and uploaded to the cloud server through wireless communication; the cloud server uses the trained target detection network to process the real-time image to identify all meta-targets in the real-time image.

[0036] Specifically, all meta-targets in the real-time image include: X1, X2, …, X N .

[0037] Among them, X i=(x i ,y i ,w i ,h i ), i∈[1,N]; x i ,y i 、w i 、h i They respectively represent the X-axis center coordinate, Y-axis center coordinate, width, and length of the i-th meta-target in the real-time image.

[0038] See also Figure 3 ,The types of meta-targets include: meter box, meter, wiring location, screws, which are the places where faults may occur, so identifying them facilitates subsequent fault location.

[0039] If X i It is an electric meter box, then x i ,y i They are the X-axis center coordinates and Y-axis center coordinates of the meter box respectively; w i 、h i are the width and length of the meter box, respectively. This is equivalent to dividing a rectangular area in the real-time image (i.e. Figure 3 The outermost rectangle) represents the location of the meter box; other types of meta-targets are similar. Figure 3 There is also a corresponding rectangular area in . In this way, the real-time image is processed by the target detection network to identify all meta-targets in the real-time image.

[0040] Step 2: Obtain the historical image of the target electric meter box and all meta-targets therein, and perform meta-target matching with the real-time image to find the meta-target in the historical image corresponding to the meta-target in the real-time image;

[0041] The two meta-targets corresponding to the real-time image and the historical image are combined into a meta-target pair.

[0042] It should be noted that historical images actually provide a control group for fault location - historical images are images of the target meter box when there was no fault in the past (it is recommended to use historical images that are close in time to the real-time images to minimize errors).

[0043] The type of the meta-target remains unchanged. Since the object is still the target electric meter box, all meta-targets in the historical image include: Y1, Y2, ..., Y M ; Y j =(x' j ,y' j ,w' j ,h' j ), j∈[1,M]; x' j ,y' j、w' j 、h' j They respectively represent the X-axis coordinate, Y-axis coordinate, width, and length of the j-th meta-target in the historical image.

[0044] Historical images can be stored in mobile terminals or cloud servers. Of course, if historical images are stored in mobile terminals, the mobile terminals upload historical images to the cloud server through wireless communication. The cloud server pre-processes historical images using a trained target detection network to identify all meta-targets in the historical images; the cloud server performs meta-target matching based on all meta-targets in real-time images and all meta-targets in historical images - the cloud server has sufficient performance to support the processing load of meta-target matching.

[0045] Specifically, the method for performing meta-target matching includes:

[0046] For any meta-object X in the real-time image i , calculate its difference with any meta-target Y in the historical image j is the probability P(X i =Y j );

[0047] in, θ j ∈R M×1 Represents the softmax function parameter matrix; Represents θ j The transpose of

[0048] Let X i Keep unchanged, take P(X i =Y j ) when Y is maximum j As X i The corresponding meta-object in the historical images.

[0049] In this way, traverse X1, X2, ..., X N and Y1, Y2, ..., Y M , that is, we get N pairs of meta-target pairs.

[0050] Step 3: Use the trained image contrast network to process all meta-target pairs and calculate their feature similarity;

[0051] If the feature similarity of a pair of meta-targets is less than a preset similarity threshold K, the meta-targets in the pair of meta-targets are marked as faulty in the real-time image; otherwise, the pair of meta-targets is not faulty.

[0052] In this way, the operation and maintenance personnel can intuitively know the location of the fault in the target meter box, so there is no need to conduct an inspection of the entire area, and they only need to repair the location of the fault, which greatly improves efficiency.

[0053] The value of K is determined based on experience - the larger its value, the higher the sensitivity of fault detection, so K is generally set to 0.9 or above.

[0054] Specifically, the image contrast network aims to calculate the feature similarity of two input images - it generally uses a twin network. The image contrast network has two different inputs, through two feature extraction networks with the same architecture - the two feature extraction networks output an encoding to calculate the difference between the two inputs; the goal of the image contrast network is to use a similarity score to classify whether the two inputs are the same or different - binary cross entropy, contrast function or triple loss can be used to calculate the similarity score.

[0055] In this embodiment 1, the two feature extraction networks can use the feature extraction part of the Yolov5 network or other mature target detection networks, and use the historical data set 2 (i.e., the meter box image set with similarity information) to train them to obtain a trained image comparison network.

[0056] In the present embodiment 1, the feature similarity adopts structural similarity (SSIM), which comprehensively measures the similarity of two images through the structure of the images and is an image quality evaluation standard that conforms to human intuition.

[0057] That is to say, for any meta-target pair [X, Y], the calculation formula for its feature similarity is:

[0058]

[0059] Where, L SSIM (X,Y) represents the feature similarity of X and Y;

[0060] X and Y represent the meta-target of real-time images and the meta-target of historical images, respectively;

[0061] μ X represents the mean of X, μ Y represents the mean of Y, σ XY Represents the covariance of X and Y; c1 and c2 represent default constants; represents the variance of X; represents the variance of Y.

[0062] Similar to Yolov5, the image contrast network is also large in size and is not suitable for being built into handheld devices at this stage. Therefore, it is recommended to store the trained image contrast network in a cloud server, which has sufficient performance to support the processing load of the image contrast network. In this way, the cloud server uses the trained image contrast network to process all meta-target pairs and calculate their feature similarity; the cloud server selects meta-target pairs whose feature similarity is less than the preset similarity threshold K, and marks the meta-target in the meta-target pair as faulty in the real-time image; the cloud server sends the marking results to the handheld terminal for operation and maintenance personnel to view.

[0063] This embodiment 1 simultaneously provides a power grid terminal detection system based on target detection and image comparison, which includes: a cloud server and a handheld device. The handheld device is connected to the cloud server via wireless communication to establish two-way communication.

[0064] This power grid terminal detection system based on target detection and image comparison uses the above-mentioned power grid terminal detection method based on target detection and image comparison, which can help operation and maintenance personnel to detect the target meter box to determine the location of the fault, making it convenient for operation and maintenance personnel to carry out targeted maintenance.

[0065] Example 2

[0066] This embodiment 2 discloses a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of the power grid terminal detection method based on target detection and image comparison disclosed in embodiment 1 are implemented.

[0067] This embodiment 2 also discloses a readable storage medium, in which computer program instructions are stored. When the computer program instructions are read and executed by a processor, the steps of the power grid terminal detection method based on target detection and image comparison disclosed in embodiment 1 are executed.

[0068] This embodiment 2 also discloses a computer program product, including a computer program. When the computer program is executed by a processor, the steps of the power grid terminal detection method based on target detection and image comparison disclosed in embodiment 1 are implemented.

[0069] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A power grid terminal detection method based on target detection and image comparison, characterized in that: The following steps are involved: Step 1: obtain a real-time image of the target electric meter box, and use the trained target detection network to process the real-time image to identify all meta-targets in the real-time image; Among them, all meta-targets in the real-time image include: X1, X2, ..., X N ;X i =(x i ,y i ,w i ,h i ), i∈[1,N]; x i ,y i 、w i 、h i Respectively represent the X-axis center coordinate, Y-axis center coordinate, width, and length of the i-th meta-target in the real-time image; Among them, the types of meta-targets include: meter box, meter, connection point, screw; Step 2: Obtain the historical image of the target electric meter box and all meta-targets therein, and perform meta-target matching with the real-time image to find the meta-target in the historical image corresponding to the meta-target in the real-time image; The two meta-targets corresponding to the real-time image and the historical image are combined into a meta-target pair; Among them, the historical image is the image of the target electric meter box when there is no fault in the past; All meta-objects in the historical image include: Y1, Y2, ..., Y M ; Y j =(x' j ,y' j ,w' j ,h' j ), j∈[1,M]; x' j ,y' j 、w' j 、h' j Respectively represent the X-axis center coordinate, Y-axis center coordinate, width, and length of the j-th meta-target in the historical image; Step 3: Use the trained image contrast network to process all meta-target pairs and calculate their feature similarity; If the feature similarity of a pair of meta-targets is less than a preset similarity threshold K, the meta-targets in the pair of meta-targets are marked as faulty in the real-time image; otherwise, the pair of meta-targets is not faulty.

2. The power grid terminal detection method based on target detection and image comparison according to claim 1 is characterized in that: Target detection networks include but are not limited to: R-CNN network, SPP-Net network, SSD network, and YOLO network.

3. The power grid terminal detection method based on target detection and image comparison according to claim 1 is characterized in that: The target detection network is the Yolov5 network.

4. The power grid terminal detection method based on target detection and image comparison according to claim 1 is characterized in that: In step 2, the method for performing meta-target matching includes: For any meta-object X in the real-time image i , calculate its difference with any meta-target Y in the historical image j is the probability P(X i =Y j ); in, Represents the softmax function parameter matrix; Represents θ j The transpose of Let X i Keep unchanged, take P(X i =Y j ) when Y is maximum j As X i The corresponding meta-object in the historical images.

5. The power grid terminal detection method based on target detection and image comparison according to claim 4 is characterized in that: In step three, the image comparison network uses a twin network.

6. The power grid terminal detection method based on target detection and image comparison according to claim 1 or 5, characterized in that: In step 3, for any meta-target pair [X,Y], The calculation formula of its feature similarity is: Where, L SSIM (X,Y) represents the feature similarity of X and Y; X and Y represent the meta-target of real-time images and the meta-target of historical images, respectively; μ X represents the mean of X, μ Y represents the mean of Y, σ XY Represents the covariance of X and Y; c1 and c2 represent default constants; represents the variance of X; represents the variance of Y.

7. The power grid terminal detection method based on target detection and image comparison according to claim 6 is characterized in that: K=0.9。 8. The power grid terminal detection method based on target detection and image comparison according to claim 1 is characterized in that: In step 1, the trained object detection network is stored in the cloud server; The real-time image of the target meter box is obtained through the handheld terminal and uploaded to the cloud server through wireless communication; The cloud server uses the trained target detection network to process the real-time image to identify all meta-targets in the real-time image; In step 2, the historical images are stored in the mobile terminal or cloud server; If the historical images are stored in the mobile terminal, the mobile terminal will upload the historical images to the cloud server via wireless communication. The cloud server pre-processes the historical images using a trained object detection network to identify all meta-objects in the historical images; The cloud server performs meta-target matching based on all meta-targets in the real-time image and all meta-targets in the historical image; In step 3, the trained image comparison network is stored in the cloud server; The cloud server uses the trained image comparison network to process all meta-target pairs and calculate their feature similarity; The cloud server selects a meta-target pair whose feature similarity is less than a preset similarity threshold K, and marks the meta-target in the meta-target pair as faulty in the real-time image; The cloud server sends the marking results to the handheld terminal.

9. A power grid terminal detection system based on target detection and image comparison, characterized in that: include: Cloud Server, A handheld device connected to a cloud server via wireless communication; The power grid terminal detection system based on target detection and image comparison uses the power grid terminal detection method based on target detection and image comparison as described in claim 8.

10. A computer program product, characterized in that The method comprises a computer program; when the computer program is executed by a processor, the steps of the power grid terminal detection method based on target detection and image comparison as described in any one of claims 1 to 8 are implemented.