A High- and Low-Voltage Line Strong Discharge Monitoring System and Method Based on Video Image Recognition

Abstract

This invention proposes a high- and low-voltage line strong discharge monitoring system and method based on video image recognition, relating to the field of strong discharge monitoring. It includes a camera for acquiring video streams of high- and low-voltage lines and sending them to an image analyzer; the image analyzer extracts keyframe images from the video streams, extracts static and dynamic spark features based on the keyframe images, and inputs a feature vector composed of these features into an extreme learning machine for spark recognition, generating a discharge warning signal, which is then sent to a remote server; the remote server receives and displays the discharge warning signal. This invention improves the accuracy of video spark image recognition and can quickly and accurately determine the discharge location, ensuring the safe and stable operation of the power grid and reducing economic losses caused by transmission line discharge faults, thus having significant social and economic implications.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage discharge monitoring technology, and particularly relates to a high-voltage and low-voltage line high-voltage discharge monitoring system and method based on video image recognition. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Currently, in the field of power technology, electrical equipment in high- and low-voltage transmission lines and transmission tower systems inevitably develops internal defects during manufacturing, on-site installation, and operation. These weak points, under the influence of strong electric fields, are prone to strong discharge phenomena. As the intensity of the discharge increases, insulation is further damaged, potentially leading to power equipment failure and power outages, seriously threatening the operational safety of the power grid. Therefore, discharge monitoring of high- and low-voltage transmission lines and power equipment is crucial.

[0004] Existing discharge detection methods generally employ ultra-high frequency (UHF), pulsed current, and ultrasonic methods. These methods only detect partial discharge signals, have a small measurement range, and cannot directly perceive the discharge phenomenon visually. Furthermore, these methods are all electrical detection methods, making the results susceptible to electromagnetic interference, difficult to wire, and unable to accurately locate the strong discharge position. Since high- and low-voltage transmission lines often traverse long paths when transmitting power, existing methods for detecting partial discharge signals are clearly insufficient to meet the requirements for monitoring strong discharges in high- and low-voltage transmission lines.

[0005] With the development of digital video processing and microprocessor technologies, video surveillance and digital image recognition methods and devices have attracted increasing attention from various industries. However, the inventors have discovered that most existing video surveillance systems only have monitoring functions and lack recognition functions; or they have the ability to recognize specific objects but cannot accurately and effectively identify strong discharge phenomena in power transmission lines. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a high- and low-voltage line strong discharge monitoring system and method based on video image recognition. By extracting the static and dynamic features of sparks from key frame images and inputting the static and dynamic features into a feature vector into an extreme learning machine for spark recognition, the accuracy of video spark image recognition is improved. At the same time, it can quickly and accurately determine the discharge location, ensure the safe and stable operation of the power grid, and reduce the economic losses caused by transmission line discharge faults, which has great social and economic significance.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0008] The first aspect of the present invention provides a high- and low-voltage line strong discharge monitoring system based on video image recognition.

[0009] A high- and low-voltage line strong discharge monitoring system based on video image recognition includes a camera, an image analyzer, and a remote server, wherein:

[0010] A camera is used to acquire video streams from high and low voltage lines and send these video streams to an image analyzer.

[0011] The image analyzer is used to extract key frame images from the video stream of high and low voltage lines. Based on the key frame images, it extracts static and dynamic features of sparks. The static and dynamic features of sparks are combined into a feature vector and input into the extreme learning machine for spark recognition, generating a discharge warning signal and sending the discharge warning signal to a remote server.

[0012] A remote server is used to receive and display discharge warning signals.

[0013] The second aspect of the present invention provides a method for monitoring strong discharge in high and low voltage lines based on video image recognition.

[0014] A method for monitoring strong discharges in high and low voltage lines based on video image recognition includes the following steps:

[0015] Acquire video streams from high and low voltage lines;

[0016] Based on the keyframe extraction algorithm, keyframes are extracted from the video streams of high and low voltage lines.

[0017] Spark static features are extracted from single-frame images in keyframes to obtain spark static features.

[0018] Spark dynamic features are extracted from the image sequence composed of keyframes to obtain spark dynamic features.

[0019] The static and dynamic features of the spark are combined to form a feature vector, which is then input into an extreme learning machine for spark recognition to generate a discharge warning signal.

[0020] The above one or more technical solutions have the following beneficial effects:

[0021] 1. This invention provides a high- and low-voltage line strong discharge monitoring system and method based on video image recognition. By extracting the static and dynamic features of sparks from key frame images and inputting the static and dynamic features into a feature vector into an extreme learning machine for spark recognition, the accuracy of video spark image recognition is improved. At the same time, it can quickly and accurately determine the discharge location, ensure the safe and stable operation of the power grid, and reduce the economic losses caused by transmission line discharge faults, which has great social and economic significance.

[0022] 2. Compared with existing technologies, the high and low voltage line strong discharge monitoring system based on video image recognition provided by this invention has a wider monitoring range. By connecting multiple intelligent high-definition network cameras, the detection area of ​​high and low voltage lines is expanded from the original single point to the area monitored by multiple cameras, which is very suitable for high and low voltage transmission lines with long-distance power transmission.

[0023] 3. The high and low voltage line strong discharge monitoring system based on video image recognition provided by this invention can be used in various occasions such as power distribution rooms, power distribution lines, and substations, and has strong versatility.

[0024] 4. This invention can more accurately identify the location of the discharge spark, and is applicable to the extraction of spark features in different scenarios and of different types, making it highly versatile. In addition, this invention uses an extreme learning machine for model training and prediction, which ensures high recognition accuracy even with a large amount of image data, and greatly reduces training and testing time.

[0025] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0027] Figure 1 This is a system structure diagram of the first embodiment.

[0028] Figure 2 This is a diagram of the Extreme Learning Machine network model structure.

[0029] Figure 3 This is a flowchart of the method in the second embodiment. Detailed Implementation

[0030] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0032] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0033] This invention employs a high-definition camera device conforming to State Grid enterprise standards, primarily for real-time monitoring of strong discharge conditions in high and low voltage lines and equipment. The high and low voltage line strong discharge monitoring system based on video image recognition provided by this invention connects the camera's monitoring data to an image analyzer via 4G / GSM / SMS or GPRS. The image analyzer transmits the analysis results to an on-site image analysis result collector, which then transmits them to a remote server via 5G communication. The image analyzer captures images of discharge sparks, determines the spark magnitude and accurately locates the fault, stores the spark images and video segments, generates discharge early warning signals, and finally provides timely warnings and alarms through various methods.

[0034] Example 1

[0035] This embodiment discloses a high- and low-voltage line strong discharge monitoring system based on video image recognition.

[0036] like Figure 1 As shown, the high- and low-voltage line strong discharge monitoring system based on video image recognition includes a camera, an image analyzer, and a remote server, wherein:

[0037] A camera is used to acquire video streams from high and low voltage lines and send these video streams to an image analyzer.

[0038] The image analyzer is used to extract key frame images from the video stream of high and low voltage lines. Based on the key frame images, it extracts static and dynamic features of sparks. The static and dynamic features of sparks are combined into a feature vector and input into the extreme learning machine for spark recognition, generating a discharge warning signal and sending the discharge warning signal to a remote server.

[0039] A remote server is used to receive and display discharge warning signals.

[0040] It also includes an image analysis result collector and a user display terminal. The image analysis result collector is used to communicate with the image analyzer and the remote server and forward the discharge warning signal sent by the image analyzer to the remote server. The user display terminal is used to communicate with the remote server and display the discharge warning signal.

[0041] In this invention, the camera is directly connected to the image analyzer via a wired Ethernet network. The image analyzer transmits the analysis results to an on-site image analysis result collector via 4G or a wired Ethernet network. The image analysis result collector then transmits the data to a remote server via 5G communication. In this embodiment, one image analysis result collector can simultaneously receive data from 35 image analyzers. The camera is powered by either automatic induction power supply technology or solar power combined with a battery.

[0042] The camera can be a general-purpose high-definition camera, which can be connected to the network. In order to improve the effectiveness of recognition, the camera is required to have a resolution of 1080P (1920*1080) or higher, an effective monitoring distance of 50 meters or less, automatic filtering of blurry images beyond 50 meters, and a light intensity of 0.5 lx or higher.

[0043] The image analyzer in this embodiment can filter out light emission interference and images of discharge operations such as welding at construction sites, and accurately capture strong discharge images in the video stream.

[0044] Furthermore, in this embodiment, the image analyzer uses a keyframe extraction algorithm to extract keyframe images from the high and low voltage line video stream. Specifically, it uses the color space total feature information of each frame in the video and the difference between adjacent frames to determine the keyframe.

[0045] The keyframe extraction algorithm is as follows:

[0046] (1) Input the high-voltage and low-voltage line video streams, extract frames from the high-voltage and low-voltage line video streams to obtain n frames of images, with the corresponding frame sequence being {f1, f2, f3, ... f n};

[0047] (2) Take the first frame image f1 as the keyframe, and let f c =f1;

[0048] (3) Let i = 2;

[0049] (4) Take the next frame image f from the video frame sequence i Calculate f i with f c The difference in the total amount of color space is calculated, and this difference is compared with a set threshold. If the difference in the total amount of color space is greater than the set threshold, then f is selected. i As a keyframe, and use fi To update f c Soon f i As the current keyframe;

[0050] (5) Let i = i + 1. If i ≤ n, then execute step (4); otherwise, the entire algorithm ends, and the keyframe extraction is completed.

[0051] The total color space mentioned in this embodiment is the total HSV color space, where H, S, and V represent hue, saturation, and brightness, respectively.

[0052] Furthermore, the image analyzer is also used to preprocess key frame images in high and low voltage line video streams using a bilateral filtering method.

[0053] Bilateral filtering can not only remove noise but also preserve image edge information. During extraction, bilateral filtering considers both the distance between pixels and their similarity. The specific formula is as follows:

[0054]

[0055]

[0056]

[0057] Where f(m,n) is the input keyframe image, f ′ (i,j) is the image after bilateral filtering.

[0058] w d (,n) and w r (,n) are the space domain kernel function and the range kernel function, σ d and σ r These are the spatial distance difference scale parameter and the pixel scale parameter. Ωp, is the set of pixels in the input keyframe image centered at (i,j) with a radius of 2p+1, where p is the filtering radius.

[0059] After extracting keyframes of the discharge spark and preprocessing the image, it can be determined whether there is a target area in the video that is suspected to be a spark. Then, the target spark can be accurately found by extracting the target image features, thus preventing unnecessary false alarms and underreporting.

[0060] Furthermore, the image analyzer performs static spark feature extraction on single-frame images in the keyframes to obtain static spark features, and performs dynamic spark feature extraction on the image sequence composed of keyframes to obtain dynamic spark features.

[0061] Specifically, static feature extraction is performed based on the brightness characteristics, texture characteristics, and circularity and matrix properties of the discharge spark. Circularity and matrix properties are used to describe the shape features of the spark image region, which are characterized by edge and region information of the spark image. The purpose of calculating circularity and matrix properties is to eliminate interference factors such as regularly shaped light sources and the sun. By setting the circularity, relatively regular interference areas in the image are removed, retaining only the complex-shaped flame region.

[0062] Dynamic feature extraction: Features are extracted based on spark area variation and stroboscopic characteristics. Spark area variation refers to the change in spark range over a period of time, obtained using the pixel difference between two adjacent keyframe images. Stroboscopic characteristics are obtained by extracting high-frequency components from the time series of spark region pixels, the target region area series, or by quantizing spark size changes. Irrelevant interference that does not produce these characteristics is then eliminated using the spark flicker features.

[0063] Extreme Learning Machine (ELM) is a machine learning method based on feedforward neural networks. The basic principle of ELM is to randomly generate the weights of the connections between the input layer and the hidden layer, as well as the thresholds of the hidden layer nodes, and set the number of neurons in the hidden layer. Then, a unique optimal solution can be obtained through simple matrix calculations.

[0064] like Figure 2 As shown, the Extreme Learning Machine (ELM) is a single-hidden-layer feedforward neural network with an nLm structure. The training set is input into the ELM. Number of hidden neurons L, output weight β, β = H + T, where H + It is the Moore-Penrose generalized inverse of H, and N is the number of training samples.

[0065] H + =(H T H+λI) -1 H T ,>N, where λ is a small positive number.

[0066] The cost function of the Extreme Learning Machine is:

[0067] J = (Hβ - T) T (Hβ-T)

[0068] Where β represents the parameters between the hidden layer and the output layer, H represents the output of the hidden layer, T represents the label of the sample, and the output layer has m neurons, corresponding to m output variables.

[0069]

[0070]

[0071] T is the label t of the training sample. j The matrix is ​​arranged such that each row represents a training sample label t. j H is the response h(x) of the hidden layer of the training samples. j The matrix is ​​arranged in rows, where each row represents the hidden layer response h(x) of a training sample. j ).

[0072] A feature vector is calculated by combining static and dynamic features of the spark. This feature vector includes brightness, texture, circularity and matrix properties, area change rate, and flicker characteristics. The resulting feature vector set is randomly divided into a training set and a test set. The training set is input into an Extreme Learning Machine (ELM) for training, and the test set is input into the trained ELM for testing and spark recognition. When the spark recognition result indicates the presence of a spark, the image analyzer generates a discharge warning signal.

[0073] The image analyzer is also used to determine the magnitude of the spark and accurately locate the spark position when a spark is present, store spark images and video segments, send the discharge warning signal to the image analysis result collector, which then forwards the discharge warning signal to the remote server. The remote server then displays the discharge warning signal and sends it to the user's display terminal.

[0074] Example 2

[0075] This embodiment discloses a method for monitoring strong discharge in high and low voltage lines based on video image recognition.

[0076] like Figure 3 As shown, the high- and low-voltage line strong discharge monitoring method based on video image recognition includes the following steps:

[0077] Acquire video streams from high and low voltage lines;

[0078] Based on the keyframe extraction algorithm, keyframes are extracted from the video streams of high and low voltage lines.

[0079] Spark static features are extracted from single-frame images in keyframes to obtain spark static features.

[0080] Spark dynamic features are extracted from the image sequence composed of keyframes to obtain spark dynamic features.

[0081] The static and dynamic features of the spark are combined to form a feature vector, which is then input into an extreme learning machine for spark recognition to generate a discharge warning signal.

[0082] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0083] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A high and low voltage line strong discharge monitoring system based on video image recognition, characterized in that, Includes cameras, image analyzers, and remote servers, among which: A camera is used to acquire video streams from high and low voltage lines and send these video streams to an image analyzer. The image analyzer is used to extract key frame images from the video stream of high and low voltage lines. Based on the key frame images, it extracts static and dynamic features of sparks. The static and dynamic features of sparks are combined into a feature vector and input into the extreme learning machine for spark recognition, generating a discharge warning signal and sending the discharge warning signal to a remote server. A remote server, used to receive and display discharge warning signals; The specific process by which the image analyzer extracts keyframe images from high and low voltage line video streams is as follows: (1) frame extraction is performed on the high and low voltage line video stream to obtain frame images; (2) Set the first frame image as the keyframe and calculate the difference in the total color space between the first frame image and the second frame image; (3) Compare the difference in the total amount of color space with the preset threshold. If the difference in the total amount of color space is greater than the preset threshold, then select the second frame image as the key frame. (4) repeat steps (1)-(3) until the frame images are all compared; The extraction of static features of the spark is based on the brightness characteristics, texture characteristics, and roundness and matrix degree of the discharge spark. The shape features of the spark image region are described by roundness and matrix degree, and the shape features are characterized by the edge information and region information of the spark image. The extraction of spark dynamic features is based on spark area change features and stroboscopic characteristics. Spark area change features refer to the changes in the spark range over a period of time, obtained by using the pixel difference between two adjacent keyframe images. Stroboscopic characteristics are obtained by extracting high-frequency components from the time series of spark region pixels and the area series of the target region, or by quantizing the changes in spark size.

2. The high-low voltage line high discharge monitoring system based on video image recognition of claim 1, wherein, The image analyzer is also used to preprocess key frame images in high and low voltage line video streams using a bilateral filtering method.

3. The high-low voltage line high discharge monitoring system based on video image recognition of claim 1, wherein, The image analyzer extracts static spark features from single frames in the keyframes to obtain static spark features, and extracts dynamic spark features from the image sequence composed of keyframes to obtain dynamic spark features.

4. The high-low voltage line high discharge monitoring system based on video image recognition of claim 1, wherein, The cost function of the Extreme Learning Machine is: in, This represents the parameters between the hidden layer and the output layer. This represents the output of the hidden layer. Labels representing samples.

5. The high- and low-voltage line strong discharge monitoring system based on video image recognition as described in claim 1, characterized in that, The image analyzer is also used to determine the magnitude of the spark and accurately locate the spark position when a spark is present, store spark images and video segments, and generate a discharge warning signal.

6. The high- and low-voltage line strong discharge monitoring system based on video image recognition as described in claim 1, characterized in that, It also includes an image analysis result collector and a user display terminal. The image analysis result collector is used to communicate with the image analyzer and the remote server and forward the discharge warning signal sent by the image analyzer to the remote server. The user display terminal is used to communicate with the remote server and display the discharge warning signal.

7. A method for monitoring strong discharge in high and low voltage lines based on video image recognition, characterized in that, The high- and low-voltage line strong discharge monitoring system based on video image recognition as described in any one of claims 1-6 includes the following steps: Acquire video streams from high and low voltage lines; Based on the keyframe extraction algorithm, keyframes are extracted from the video streams of high and low voltage lines. Spark static features are extracted from single-frame images in keyframes to obtain spark static features. Spark dynamic features are extracted from the image sequence composed of keyframes to obtain spark dynamic features. The static and dynamic features of the spark are combined to form a feature vector, which is then input into an extreme learning machine for spark recognition to generate a discharge warning signal.

Images