CN111507975B - Method for detecting abnormity of outdoor insulator of traction substation - Google Patents
Method for detecting abnormity of outdoor insulator of traction substation Download PDFInfo
- Publication number
- CN111507975B CN111507975B CN202010324733.6A CN202010324733A CN111507975B CN 111507975 B CN111507975 B CN 111507975B CN 202010324733 A CN202010324733 A CN 202010324733A CN 111507975 B CN111507975 B CN 111507975B
- Authority
- CN
- China
- Prior art keywords
- insulator
- image
- network
- images
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 239000012212 insulator Substances 0.000 title claims abstract description 123
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000002159 abnormal effect Effects 0.000 claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 239000013598 vector Substances 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 6
- 230000005856 abnormality Effects 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 2
- 238000005520 cutting process Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 claims description 2
- 238000001914 filtration Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 claims description 2
- 238000011176 pooling Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000003860 storage Methods 0.000 claims description 2
- 238000003909 pattern recognition Methods 0.000 abstract description 2
- 238000007689 inspection Methods 0.000 description 22
- 229910052573 porcelain Inorganic materials 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 5
- 230000007547 defect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000011179 visual inspection Methods 0.000 description 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical group [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000001680 brushing effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method for detecting the abnormity of an outdoor insulator of a traction substation. Relates to the technical field of computer vision, pattern recognition and intelligent systems. Respectively constructing a data set of an insulator positioning network and an insulator image generation network; an insulator positioning network is constructed, and the network obtains the positioning capacity of the insulators in the image through training; an insulator image generation network is constructed, and the insulator image generation network is trained to obtain the reconstruction capability of insulator images; inputting the traction substation image into a network model; positioning the insulator through an insulator positioning network, and extracting an insulator image; and (4) carrying out abnormity detection on the insulator, and giving an abnormity score to each picture by the insulator image generation network. And setting an abnormal judgment threshold, judging as an abnormal sample if the abnormal score exceeds the set threshold, and judging as a normal sample if the abnormal score is lower than the threshold. And finally, extracting the characteristics of the judged abnormal image and the generated image thereof, and comparing the difference between the two images to locate an abnormal area.
Description
Technical Field
The invention relates to the technical field of computer vision, pattern recognition and intelligent systems.
Background
The insulator is an important device in the power transmission line, and has important significance for maintaining the stability of the power transmission line and ensuring the normal operation of a power transmission network. If the insulator fails, serious transmission faults and economic losses can be caused. Therefore, in the process of power transmission line inspection, it is an important task to check whether the insulator works normally. According to routine inspection and inspection items and requirements of the power transformation station, the routine inspection and inspection content items and requirements of the routine inspection and inspection: 1) the porcelain skirt surface pollution degree has no discharge phenomenon. 2) The porcelain skirt and the flange have no crack or damage. 3) The high-voltage porcelain knob insulating support is free of stress (lead), the support is free of inclination, and the base is fastened by bolts. 4) The flange and iron parts of the equipment have no cracks or fissures. The unit of the patrol time, the number of times and the content under various duty modes is clearly specified. The routine inspection tour includes normal tour, full tour and light-off tour. And (4) normal inspection: 1) there is post insulator equipment of transformer substation on duty, once at least every day: night patrol is performed at least once a week: 2) and carrying out secondary inspection tour every week on the post porcelain insulator equipment in the unattended substation. The overall patrol content mainly refers to overall external inspection of the equipment. The light-off inspection should be performed once a week, and the inspection equipment has no corona, discharge, and overheating of the connector. In order to improve the pollution resistance level of the post porcelain insulator, according to the running condition of equipment, the post porcelain insulator which can run in a region with serious pollution can adopt the technical measures of brushing and staring V paint on the surface and the like. When a defect is found in the inspection, the device abnormality and the defect record are recorded in detail and reported to the upper level. Therefore, the traditional power inspection method mainly depends on manual inspection, and the mode is time-consuming, low in efficiency and long in inspection period. The method for detecting the insulator abnormity, which can reduce human resources, is convenient and efficient and solves the problem of insufficient data, has great significance. The presence of unmanned aerial vehicles and high definition cameras makes it possible to shoot insulator images instead of the original visual inspection. Compared with the traditional mode, the mode of patrolling and examining through the intelligent equipment has the advantages of high safety, convenience in implementation, high efficiency and the like, the equipment is used for shooting the image of the insulator, analysis and processing are carried out, the utilization efficiency of data is improved, and a basis is provided for the realization of the whole intelligent system. With the development of deep learning, it also becomes an important detection means. However, in the training process of the deep neural network model, the insufficient number of abnormal insulator images is a great problem.
Disclosure of Invention
The invention aims to provide a method for detecting the abnormity of an outdoor insulator of a traction substation, which can effectively solve the technical problem of judging whether the working state of the outdoor insulator of the traction substation is normal or not.
The purpose of the invention is realized by the following technical scheme: a method for detecting the abnormity of an outdoor insulator of a traction substation comprises the following steps:
step one, constructing an insulator positioning network data set
The method comprises the steps of adopting images shot by a tower-mounted monitoring camera in a traction substation to form an image library, and carrying out median filtering processing on each image in the image library to suppress noise, wherein the method specifically comprises the following steps: traversing the image by using a 3 x 3 pixel sliding window algorithm, sequencing 9 pixel points in a sliding window according to the pixel values, and taking a median value as a pixel value of a central point of the sliding window; marking an insulator region in the image by using a surrounding frame capable of surrounding the complete insulator in a manual marking mode, wherein all the images and marking results thereof jointly form an insulator positioning network data set;
step two, insulator image generation network data set construction
Cutting the images in the image library, extracting independent insulator images, and manually manufacturing corresponding crack, magnet dropping and flashover abnormal insulator images of each normal insulator image by considering that the number of the abnormal insulator images is small in practice so that each normal insulator image corresponds to one abnormal insulator image; all the two images jointly form an insulator image generation network data set;
step three, construction and training of insulator positioning network
The insulator positioning network adopts a fast R-CNN network model, and comprises a feature extraction network, a candidate area generation network and a classification and bounding box regression network; adopting the pre-trained front 16 layers of VGG-16 on a large-scale classified recognition data set ImageNet as a feature extraction network to generate a feature map; generating areas with different areas and aspect ratios at corresponding positions of an original image, calculating the probability P that each area contains an insulator by using a softmax function, and taking the probability P as a candidate area generation network; inputting the candidate region into a classification and bounding box regression network to classify the target, simultaneously regressing the target bounding box, and outputting the position of the insulator bounding box and the probability P of the insulator bounding box being classified as an insulator through the full connecting layer0The method specifically comprises the following steps: after the feature map enters a candidate region generation network, traversing the feature map by using a 3 x 3 pixel sliding window algorithm, generating a plurality of regions at each sliding window position based on 9 default bounding boxes with fixed proportion, wherein each region may or may not contain an insulator, performing secondary classification on all the generated regions by using a softmax function, and selecting the first 300 regions with the highest probability P as candidate regions; setting the aspect ratio of the default bounding box of the candidate area to be 0.5, 1 and 2, and the size to be 8 times, 16 times and 32 times; setting the coordinates of the basic bounding box as [0,0,15 ]]Wherein (0,0) is the coordinate of the upper left corner of the boundary box, and (15,15) is the coordinate of the lower right corner of the boundary box; keeping the area and the central point of the boundary frame unchanged, obtaining three new boundary frames according to three length-width ratios, multiplying three sizes by the length and the width of the three new boundary frames, and keeping the central point unchanged to obtain 9 default boundary frames; acting the obtained candidate region on the feature map, extracting region features through RoI pooling, and performing classification and boundary frame prediction on the features through a classification and boundary frame regression network; training an insulator positioning network by using an insulator positioning network data set to obtain the capacity of positioning the insulator in the image;
step four, construction and training of insulator image generation network
The insulator image generation network is built based on a GAN network structure and comprises three parts: a generator, a discriminator and a reconstruction encoder; the generator is built based on an automatic encoder structure and comprises an encoder and a decoder; the encoder maps a 3-channel picture X into a 32-dimensional vector Z; adding a U-Net cross-layer connection mode at one end of a decoder to obtain a better image generation effect, and finally reconstructing a 32-dimensional vector Z obtained by coding into a 3-channel picture X'; the discriminator adopts the same network structure as the encoder in the generator and is used for distinguishing the original image from the reconstructed image generated by the generator; the reconstruction encoder recompresses the reconstructed picture X 'generated by the generator into a 32-dimensional vector Z', which has the same structure as the encoder in the generator, and is used for comparing the difference between the original picture and the reconstructed picture in a higher-level abstract space, specifically: the encoder consists of 4 × 4 convolutional layers, outputs a vector of size 1 × 1 × 32, with width and height of 1 and dimension of 32; the decoder performs deconvolution operation on the vector, performs upsampling once every time of deconvolution operation, then splices the upsampling with the characteristics of the same size of the corresponding layer in the encoder, and outputs a reconstructed image after 4 multiplied by 4 deconvolution layers; the reconstructed image is encoded again by a reconstruction encoder to obtain vectors with the same size of 1 multiplied by 32; the network loss consists of three parts, respectively: 1) the reconstruction loss of the generator is compared with the original image and the reconstruction image, and the L1 loss is adopted; 2) the loss of the coding network, namely the loss of the generator and the reconstruction coder, is compared with the coding characteristics of the original image and the reconstruction image, and the loss of L2 is adopted; 3) judging network loss, and adopting cross entropy loss of two categories; training an insulator image generation network by using an insulator image generation network training set, wherein the training method adopts an Adam training method to obtain the reconstruction capability of insulator images;
step five, image input
Under the condition of real-time processing, extracting an original video image which is acquired by a camera and stored in a storage area as an input image to be subjected to anomaly detection; under the condition of offline processing, decomposing the video file which is acquired into an image sequence consisting of a plurality of frames, extracting the frame images one by one as input images according to a time sequence, and stopping the whole process if the input images are empty;
sixthly, positioning the insulator
Normalizing the size of an input image into 224 multiplied by 224 pixels of the size required by the input end of the insulator positioning network, and then obtaining the position and the probability P of an insulator surrounding frame after the forward processing of the network0(ii) a If the probability P0If the number of the insulator images is greater than the preset threshold value of 0.9, the positioning is considered to be successful, and all insulator images which are successfully positioned are extracted;
seventhly, detecting the abnormality of the insulator
Inputting the insulator image extracted in the step six into an insulator image generation network for anomaly detection, taking the coding network loss in the network loss as an anomaly score of the input image, and simultaneously setting an anomaly judgment threshold with a threshold value of 0.2, if the anomaly score of the input image is smaller than the anomaly judgment threshold, judging that the input image is normal, otherwise, judging that the input image is abnormal; performing LBP feature extraction on the image judged to be abnormal and the image generated by the insulator image generation network to obtain respective LBP feature maps, optionally performing pixel-by-pixel search on one image, and comparing the image with a pixel at a corresponding position of the other image; if a pixel satisfies: the pixel difference value of the pixel point in the two images is more than or equal to 30, and the pixel difference values of 8 pixel points in a 3 multiplied by 3 pixel window taking the pixel point as the center are more than or equal to 30, and then the pixel point is judged to have abnormity; and then, carrying out binarization processing on the image judged to be abnormal, and enabling the gray value of the pixel point judged to be abnormal to be 255 and the gray values of other pixel points to be 0, so that an abnormal area can be obtained, and skipping to the fifth step.
The invention has the advantages and positive effects that: the method comprises the steps of firstly constructing an insulator positioning network, and extracting an insulator in a traction substation image shot by a camera to obtain an insulator image. And then constructing an insulator image generation network which is built based on a GAN network structure, wherein a generator part is built based on an automatic encoder structure, a U-Net cross-layer connection mode is added at one end of a decoder, a discriminator part adopts the same network structure as an encoder in the generator, and an encoder is added behind a picture generated by the generator to serve as a reconstruction encoder. In the training process, the normal insulator image and the corresponding abnormal insulator image are used for training, so that the network has the capability of reconstructing the input insulator image into the normal insulator image. In the process of anomaly detection, any insulator image is input into a network, each image is given an anomaly score, if the score exceeds a set anomaly judgment threshold value, the image is judged to be an abnormal sample, and if the score is lower than the anomaly judgment threshold value, the image is judged to be a normal sample. And finally, performing feature extraction on the judged abnormal image and the generated image thereof, and positioning an abnormal area by comparing the difference of the two images. The method can realize higher abnormality detection accuracy and meet the intelligent trend of power routing inspection.
Drawings
FIG. 1 is a diagram of a network structure for generating an insulator image according to the present invention
FIG. 2 is a technical flow chart of the present invention
Detailed Description
The method can be used for a traction substation containing the insulator, the traditional power inspection method mainly depends on manual inspection, and the method is time-consuming, low in efficiency and long in inspection period. The advent of helicopters, unmanned aerial vehicles, high-definition cameras made it possible to shoot insulator images instead of the original visual inspection. Compared with the traditional mode, the mode of patrolling and examining through the intelligent equipment has the advantages of high safety, convenience in implementation, high efficiency and the like. The abnormity detection of the outdoor insulator of the traction substation can be realized by adopting the detection method provided by the invention.
Specifically, an insulator positioning network is constructed at first, and the insulators in the traction substation images shot by the camera are subjected to target extraction to obtain insulator images. And then constructing an insulator image generation network which is built based on a GAN network structure, wherein a generator part is built based on an automatic encoder structure, a cross-layer connection mode is added at one end of a decoder, a discriminator part adopts the same network structure as an encoder in the generator, and an encoder is added behind the image generated by the generator to serve as a reconstruction encoder. In the training process, the obtained normal insulator image and the corresponding abnormal insulator image are used for training, the data distribution of the trained network model is suitable for reconstructing an input image into the corresponding normal insulator image, when the abnormal insulator image is input into the model, the reconstructed image loses abnormal features, and further the vector encoded by the reconstructed encoder loses abnormal features, so that the vector is different from the encoding vector of the original image. In the testing process, the insulator images are simultaneously input into the network, each image is given an abnormal score, if the score exceeds a set threshold value, the abnormal sample is judged, and if the score is lower than the threshold value, the normal sample is judged. And finally, performing feature extraction on the judged abnormal image and the generated image thereof to locate an abnormal area. The method can realize higher abnormality detection accuracy and meet the intelligent trend of power routing inspection.
The method can be realized by programming in any computer programming language (such as Python language), and the detection system software based on the method can realize real-time anomaly detection application in any PC or embedded system.
Claims (1)
1. A method for detecting the abnormity of an outdoor insulator of a traction substation comprises the following steps:
step one, constructing an insulator positioning network data set
The method comprises the steps of adopting images shot by a tower-mounted monitoring camera in a traction substation to form an image library, and carrying out median filtering processing on each image in the image library to suppress noise, wherein the method specifically comprises the following steps: traversing the image by using a 3 x 3 pixel sliding window algorithm, sequencing 9 pixel points in a sliding window according to the pixel values, and taking a median value as a pixel value of a central point of the sliding window; marking an insulator region in the image by using a surrounding frame capable of surrounding the complete insulator in a manual marking mode, wherein all the images and marking results thereof jointly form an insulator positioning network data set;
step two, insulator image generation network data set construction
Cutting the images in the image library, extracting independent insulator images, and manually manufacturing corresponding crack, magnet dropping and flashover abnormal insulator images of each normal insulator image by considering that the number of the abnormal insulator images is small in practice so that each normal insulator image corresponds to one abnormal insulator image; all the two images jointly form an insulator image generation network data set;
step three, construction and training of insulator positioning network
The insulator positioning network adopts a fast R-CNN network model, and comprises a feature extraction network, a candidate area generation network and a classification and bounding box regression network; adopting the pre-trained front 16 layers of VGG-16 on a large-scale classified recognition data set ImageNet as a feature extraction network to generate a feature map; generating areas with different areas and aspect ratios at corresponding positions of an original image, and calculating the probability P of containing insulators in each area by using a softmax function to serve as a candidate area generation network; inputting the candidate region into a classification and bounding box regression network to classify the target, simultaneously regressing the target bounding box, and outputting the position of the insulator bounding box and the probability P of the insulator bounding box being classified as an insulator through the full connecting layer0The method specifically comprises the following steps: after the feature map enters a candidate region generation network, traversing the feature map by using a 3 x 3 pixel sliding window algorithm, generating a plurality of regions at each sliding window position based on 9 default bounding boxes with fixed proportion, and selecting the first 300 regions with the highest probability P as candidate regions; setting the aspect ratio of the default bounding box of the candidate area to be 0.5, 1 and 2, and the size to be 8 times, 16 times and 32 times; setting the coordinates of the basic bounding box as [0,0,15 ]]Wherein (0,0) is the coordinate of the upper left corner of the boundary box, and (15,15) is the coordinate of the lower right corner of the boundary box; keeping the area and the central point of the boundary frame unchanged, obtaining three new boundary frames according to three length-width ratios, multiplying three sizes by the length and the width of the three new boundary frames, and keeping the central point unchanged to obtain 9 default boundary frames; acting the obtained candidate region on the feature map, extracting region features through RoI pooling, and performing classification and boundary frame prediction on the features through a classification and boundary frame regression network; training an insulator positioning network using an insulator positioning network dataset to obtain a pair imageThe middle insulator positioning capability;
step four, construction and training of insulator image generation network
The insulator image generation network is built based on a GAN network structure and comprises three parts: a generator, a discriminator and a reconstruction encoder; the generator is built based on an automatic encoder structure and comprises an encoder and a decoder; the encoder maps a 3-channel picture X into a 32-dimensional vector Z; adding a U-Net cross-layer connection mode at one end of a decoder to obtain a better image generation effect, and finally reconstructing a 32-dimensional vector Z obtained by coding into a 3-channel picture X'; the discriminator adopts the same network structure as the encoder in the generator and is used for distinguishing the original image from the reconstructed image generated by the generator; the reconstruction encoder re-maps the reconstructed picture X 'generated by the generator into a 32-dimensional vector Z', which has the same structure as the encoder in the generator, and is used for comparing the difference between the original picture and the reconstructed picture in a high-level abstract space, specifically: the encoder consists of 4 × 4 convolutional layers, outputs a vector of size 1 × 1 × 32, with width and height of 1 and dimension of 32; the decoder performs deconvolution operation on the vector, performs upsampling once every time of deconvolution operation, then splices the upsampling with the characteristics of the same size of the corresponding layer in the encoder, and outputs a reconstructed image after 4 multiplied by 4 deconvolution layers; the reconstructed image is encoded again by a reconstruction encoder to obtain vectors with the same size of 1 multiplied by 32; the network loss consists of three parts, respectively: 1) the reconstruction loss of the generator is compared with the original image and the reconstruction image, and the L1 loss is adopted; 2) the loss of the coding network, namely the loss of the generator and the reconstruction coder, is compared with the coding characteristics of the original image and the reconstruction image, and the loss of L2 is adopted; 3) judging network loss, and adopting cross entropy loss of two categories; training an insulator image generation network by using an insulator image generation network training set, wherein the training method adopts an Adam training method to obtain the reconstruction capability of insulator images;
step five, image input
Under the condition of real-time processing, extracting an original video image which is acquired by a camera and stored in a storage area as an input image to be subjected to anomaly detection; under the condition of off-line processing, decomposing an acquired video file into an image sequence consisting of a plurality of frames, extracting the frame images one by one as input images according to a time sequence, and stopping the whole process if the input images are empty;
sixthly, positioning the insulator
Normalizing the size of an input image into 224 multiplied by 224 pixels of the size required by the input end of the insulator positioning network, and then obtaining the position and the probability P of an insulator surrounding frame after the forward processing of the network0(ii) a If the probability P0If the number of the insulator images is greater than the preset threshold value of 0.9, the positioning is considered to be successful, and all insulator images which are successfully positioned are extracted;
seventhly, detecting the abnormality of the insulator
Inputting the insulator image extracted in the step six into an insulator image generation network for anomaly detection, taking the coding network loss in the network loss as an anomaly score of the input image, and simultaneously setting an anomaly judgment threshold with a threshold value of 0.2, if the anomaly score of the input image is smaller than the anomaly judgment threshold, judging that the input image is normal, otherwise, judging that the input image is abnormal; performing LBP feature extraction on the image judged to be abnormal and the image generated by the insulator image generation network to obtain respective LBP feature maps, optionally performing pixel-by-pixel search on one image, and comparing the image with a pixel at a corresponding position of the other image; if a pixel satisfies: the pixel difference value of the pixel point in the two images is more than or equal to 30, and the pixel difference values of 8 pixel points in a 3 multiplied by 3 pixel window taking the pixel point as the center are more than or equal to 30, and then the pixel point is judged to have abnormity; and then, carrying out binarization processing on the image judged to be abnormal, and enabling the gray value of the pixel point judged to be abnormal to be 255 and the gray values of other pixel points to be 0, so that an abnormal area can be obtained, and skipping to the fifth step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010324733.6A CN111507975B (en) | 2020-04-23 | 2020-04-23 | Method for detecting abnormity of outdoor insulator of traction substation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010324733.6A CN111507975B (en) | 2020-04-23 | 2020-04-23 | Method for detecting abnormity of outdoor insulator of traction substation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111507975A CN111507975A (en) | 2020-08-07 |
| CN111507975B true CN111507975B (en) | 2022-02-08 |
Family
ID=71877851
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010324733.6A Expired - Fee Related CN111507975B (en) | 2020-04-23 | 2020-04-23 | Method for detecting abnormity of outdoor insulator of traction substation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111507975B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114140731A (en) * | 2021-12-08 | 2022-03-04 | 西南交通大学 | Traction substation abnormity detection method |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113379729B (en) * | 2021-07-02 | 2023-07-25 | 四川启睿克科技有限公司 | Image tiny abnormality detection method, device and computer readable storage medium |
| CN117315354B (en) * | 2023-09-27 | 2024-04-02 | 南京航空航天大学 | Insulator anomaly detection method based on multi-discriminant composite coding GAN network |
| CN117315368B (en) * | 2023-10-23 | 2024-04-23 | 龙坤(无锡)智慧科技有限公司 | Intelligent operation and maintenance inspection method for large-scale data center |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9865057B2 (en) * | 2013-09-30 | 2018-01-09 | Brainlab Ag | Generating supplementary slice images based on atlas data |
| JP6486069B2 (en) * | 2014-10-31 | 2019-03-20 | 株式会社東芝 | Image processing apparatus, inspection apparatus, inspection method, and image processing program |
| CN107274451A (en) * | 2017-05-17 | 2017-10-20 | 北京工业大学 | Isolator detecting method and device based on shared convolutional neural networks |
| CN107358636B (en) * | 2017-06-16 | 2020-04-28 | 华南理工大学 | Loose defect image generation method based on texture synthesis |
| US10275872B2 (en) * | 2017-08-24 | 2019-04-30 | Applied Materials Israel Ltd. | Method of detecting repeating defects and system thereof |
| KR102175286B1 (en) * | 2018-10-11 | 2020-11-06 | 라온피플 주식회사 | Apparatus and method for inspecting for defects |
| CN110853013A (en) * | 2019-11-11 | 2020-02-28 | 四川大学青岛研究院 | Method for detecting spraying quality of outer wall of white household appliance |
-
2020
- 2020-04-23 CN CN202010324733.6A patent/CN111507975B/en not_active Expired - Fee Related
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114140731A (en) * | 2021-12-08 | 2022-03-04 | 西南交通大学 | Traction substation abnormity detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111507975A (en) | 2020-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111507975B (en) | Method for detecting abnormity of outdoor insulator of traction substation | |
| Hnewa et al. | Object detection under rainy conditions for autonomous vehicles: A review of state-of-the-art and emerging techniques | |
| CN109753929B (en) | High-speed rail insulator inspection image recognition method | |
| CN111784685B (en) | Power transmission line defect image identification method based on cloud edge cooperative detection | |
| CN111402248B (en) | Power transmission line wire defect detection method based on machine vision | |
| CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
| CN106407928B (en) | Transformer composite insulator casing monitoring method and system based on raindrop identification | |
| Sohn et al. | Automatic powerline scene classification and reconstruction using airborne lidar data | |
| CN110569841B (en) | Contact gateway key component target detection method based on convolutional neural network | |
| CN112734692A (en) | Transformer equipment defect identification method and device | |
| CN109815800A (en) | Object detection method and system based on regression algorithm | |
| CN114743119A (en) | High-speed rail contact net dropper nut defect detection method based on unmanned aerial vehicle | |
| CN110751630A (en) | Power transmission line foreign matter detection method and device based on deep learning and medium | |
| CN113515655A (en) | Fault identification method and device based on image classification | |
| CN111079645A (en) | Insulator self-explosion identification method based on AlexNet network | |
| CN115965578A (en) | Binocular stereo matching detection method and device based on channel attention mechanism | |
| CN108921147B (en) | Black smoke vehicle identification method based on dynamic texture and transform domain space-time characteristics | |
| Manninen et al. | Multi-stage deep learning networks for automated assessment of electricity transmission infrastructure using fly-by images | |
| CN116485802B (en) | Insulator flashover defect detection method, device, equipment and storage medium | |
| CN111160087A (en) | Method and system for automatically identifying porcelain insulator cap and disc surface in infrared image and storage medium | |
| CN114359167A (en) | Insulator defect detection method based on lightweight YOLOv4 in complex scene | |
| CN113160202A (en) | Crack detection method and system | |
| CN111310899B (en) | Power defect identification method based on symbiotic relation and small sample learning | |
| Lu et al. | An anomaly detection method for outdoors insulator in high-speed railway traction substation | |
| CN111402223B (en) | Transformer substation defect problem detection method using transformer substation video image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220208 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |