CN114581419A - Transformer insulating sleeve defect detection method, related equipment and readable storage medium - Google Patents

Transformer insulating sleeve defect detection method, related equipment and readable storage medium Download PDF

Info

Publication number
CN114581419A
CN114581419A CN202210224935.2A CN202210224935A CN114581419A CN 114581419 A CN114581419 A CN 114581419A CN 202210224935 A CN202210224935 A CN 202210224935A CN 114581419 A CN114581419 A CN 114581419A
Authority
CN
China
Prior art keywords
defect detection
insulating sleeve
transformer
defect
transformer insulating
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.)
Pending
Application number
CN202210224935.2A
Other languages
Chinese (zh)
Inventor
王坚俊
孙林涛
张翾喆
邹晖
刘江明
彭晨光
周国伟
郭创新
李文燕
刘昌标
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Super High Voltage Branch Of State Grid Zhejiang Electric Power Co ltd
Original Assignee
Super High Voltage Branch Of State Grid Zhejiang Electric Power Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Super High Voltage Branch Of State Grid Zhejiang Electric Power Co ltd filed Critical Super High Voltage Branch Of State Grid Zhejiang Electric Power Co ltd
Priority to CN202210224935.2A priority Critical patent/CN114581419A/en
Publication of CN114581419A publication Critical patent/CN114581419A/en
Pending legal-status Critical Current

Links

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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • 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
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/20112—Image segmentation details
    • G06T2207/20132—Image cropping
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component
    • 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)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The application discloses a transformer insulating sleeve defect detection method, related equipment and a readable storage medium. After a transformer insulating sleeve image to be subjected to defect detection is obtained, the transformer insulating sleeve image is input into a transformer insulating sleeve defect detection model, the transformer insulating sleeve defect detection model extracts the features of components with different sizes in the transformer insulating sleeve image to obtain feature maps with multiple scales, and defect detection is performed on the feature maps with the scales to obtain a defect detection result. According to the scheme, the defect detection of the insulating sleeve of the transformer can be automatically realized, and the detection efficiency and the accuracy of the detection result are improved.

Description

Transformer insulating sleeve defect detection method, related equipment and readable storage medium
Technical Field
The application relates to the technical field of image processing, in particular to a transformer insulating sleeve defect detection method, related equipment and a readable storage medium.
Background
The industrial video monitoring or unmanned aerial vehicle inspection technology is widely applied to the field of transformer insulating sleeve defect detection at present. Utilize industry video, unmanned aerial vehicle's flexibility, can shoot a large amount of transformer bushing images, the backstage can be based on these defects of image detection transformer bushing.
At present, the image of the transformer insulating sleeve is mostly analyzed in a manual mode to detect the defects of the transformer insulating sleeve, but the mode has low efficiency, the detection result is greatly influenced by human factors, and the accuracy of the detection result cannot be ensured.
Therefore, how to provide a method for detecting defects of an insulating sleeve of a transformer becomes a technical problem to be solved urgently by those skilled in the art.
Disclosure of Invention
In view of the above problems, the present application provides a transformer bushing defect detection method, related devices and readable storage media. The specific scheme is as follows:
a transformer bushing defect detection method, the method comprising:
acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
inputting the transformer insulating sleeve inspection image into a transformer insulating sleeve defect detection model, performing feature extraction on the transformer insulating sleeve inspection image by the transformer insulating sleeve defect detection model to obtain feature maps of multiple scales, and performing defect detection on the feature maps of the scales to obtain a defect detection result.
Optionally, the transformer bushing defect detection model includes:
a multi-scale feature map extraction module and a defect detection module;
the transformer bushing defect detection model is right the transformer bushing patrols and examines the image and carry out feature extraction, obtains the characteristic map of a plurality of yards to carry out defect detection to the characteristic map of each yard, obtain the defect testing result, include:
the multi-scale feature map extraction module is used for extracting features of the transformer insulating sleeve inspection image to obtain feature maps of multiple scales;
and the defect detection module is used for detecting the defects of the characteristic diagrams of all scales to obtain a defect detection result.
Alternatively,
the multi-scale feature map extraction module comprises: the first 91 layers of the pretrained ResNet-101 network and the BiFPN network;
the multi-scale feature map extraction module is right the transformer bushing patrols and examines the image and carry out the feature extraction, obtains the feature map of a plurality of yards, includes:
the first 91 layers of the pre-trained ResNet-101 network and the BiFPN network extract the characteristics of the inspection image of the transformer insulating sleeve to obtain a characteristic diagram with five scales;
the defect detection module comprises: the back 10 layers of the RPN network and the pre-trained ResNet-101 network;
the defect detection module carries out defect detection on the feature map of each scale to obtain a defect detection result, and the defect detection result comprises the following steps:
and the RPN network and the rear 10 layers of the pre-trained ResNet-101 network carry out defect detection on the feature maps of the five scales to obtain a defect detection result.
Optionally, the training mode of the transformer bushing defect detection model includes:
determining a training set and a testing set, wherein the training set and the testing set respectively comprise a plurality of transformer insulating sleeve inspection images marked with defect types and defect positioning frames of all parts;
and training the transformer insulating sleeve defect detection model by taking each transformer insulating sleeve inspection image in the training set as a training sample, taking each part of defect types and defect positioning frames marked on each transformer insulating sleeve inspection image in the training set as sample labels, and taking the average precision mean value calculated based on the test set and reaching a set threshold value as a training finishing condition.
Optionally, the determining a training set and a test set includes:
acquiring a transformer insulating sleeve inspection image;
marking the defect types and the defect positioning frames of all the parts in the transformer insulating sleeve inspection image to obtain a transformer insulating sleeve inspection image data set;
performing data enhancement on a transformer insulating sleeve inspection image data set to obtain a sample data set;
determining the training set and the test set based on the sample data set.
Optionally, the data enhancement is performed on the transformer bushing inspection image data set to obtain a sample data set, and the method includes:
determining an original image to be subjected to data enhancement;
carrying out random cutting processing on the original image to be subjected to data enhancement to obtain a randomly cut image sample;
carrying out random brightness conversion processing on the randomly cut image to obtain an image with random brightness conversion;
and carrying out random hue and saturation transformation processing on the image after the random brightness transformation to obtain a sample data set.
Optionally, the calculating an average precision mean value based on the test set includes:
inputting each sample data in the test set into a transformer insulating sleeve defect detection model to obtain a detection result, wherein the detection result comprises a defect positioning frame corresponding to each defect type;
traversing each defect type, and calculating the average precision of the defect types;
and calculating the average value of the average precision of each defect type as the average precision average value.
A transformer bushing defect detection apparatus, the apparatus comprising:
the acquiring unit is used for acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
and the detection unit is used for inputting the transformer insulating sleeve inspection image into a transformer insulating sleeve defect detection model, the transformer insulating sleeve defect detection model is used for extracting the characteristics of the transformer insulating sleeve inspection image to obtain characteristic diagrams of multiple scales, and defect detection is carried out on the characteristic diagrams of all scales to obtain a defect detection result.
A transformer insulating sleeve defect detection device comprises a memory and a processor;
the memory is used for storing programs;
the processor is used for executing the program to realize the steps of the transformer insulating sleeve defect detection method.
A readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the transformer bushing defect detection method as described above.
By means of the technical scheme, the application discloses a transformer insulating sleeve defect detection method, related equipment and a readable storage medium. After a transformer insulating sleeve image to be subjected to defect detection is obtained, the transformer insulating sleeve image is input into a transformer insulating sleeve defect detection model, the transformer insulating sleeve defect detection model extracts the features of components with different sizes in the transformer insulating sleeve image to obtain feature maps with multiple scales, and defect detection is carried out on the feature maps with the scales to obtain a defect detection result. According to the scheme, the defect detection of the insulating sleeve of the transformer can be automatically realized, and the detection efficiency and the accuracy of the detection result are improved.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the application. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
fig. 1 is a schematic flowchart of a method for detecting a defect of an insulating sleeve of a transformer according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a transformer bushing defect detection model disclosed in an embodiment of the present application;
fig. 3 is a schematic structural diagram of a transformer bushing defect detection model disclosed in an embodiment of the present application;
fig. 4 is a schematic structural diagram of a transformer bushing defect detection apparatus disclosed in an embodiment of the present application;
fig. 5 is a block diagram of a hardware structure of a transformer bushing defect detection device disclosed in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only some embodiments of the present application, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Next, the method for detecting defects of the transformer bushing provided by the present application is described by the following embodiments.
Referring to fig. 1, fig. 1 is a schematic flowchart of a method for detecting a defect of an insulation sleeve of a transformer, which may include:
step S101: and acquiring an image of the transformer insulating sleeve to be subjected to defect detection.
Step S102: inputting the transformer insulating sleeve image into a transformer insulating sleeve defect detection model, extracting the features of parts with different sizes in the transformer insulating sleeve image by the transformer insulating sleeve defect detection model to obtain feature maps with multiple scales, and performing defect detection on the feature maps with all scales to obtain a defect detection result.
The embodiment discloses a transformer insulating sleeve defect detection method. After a transformer insulating sleeve image to be subjected to defect detection is obtained, the transformer insulating sleeve image is input into a transformer insulating sleeve defect detection model, the transformer insulating sleeve defect detection model extracts the features of components with different sizes in the transformer insulating sleeve image to obtain feature maps with multiple scales, and defect detection is carried out on the feature maps with the scales to obtain a defect detection result. According to the scheme, the defect detection of the insulating sleeve of the transformer can be automatically realized, and the detection efficiency and the accuracy of the detection result are improved.
In another embodiment of the application, the structure of a transformer bushing defect detection model is described. It should be noted that, the structure of the transformer bushing defect detection model can take various forms.
Referring to fig. 2, fig. 2 is a schematic structural diagram of a transformer bushing defect detection model disclosed in an embodiment of the present application. As shown in fig. 2, the model includes a multi-scale feature map extraction module, and a defect detection module.
The multi-scale feature map extraction module extracts features of components with different sizes in the transformer insulating sleeve inspection image to obtain feature maps with multiple scales; and the defect detection module is used for detecting the defects of the characteristic diagrams of all scales to obtain a defect detection result.
Referring to fig. 3, fig. 3 is a schematic structural diagram of a transformer bushing defect detection model disclosed in an embodiment of the present application. The model comprises a multi-scale feature map extraction module and a defect detection module, wherein the multi-scale feature map extraction module can comprise: the first 91 layers of the pre-trained ResNet-101 network (the depth residual network of layer 101) and BiFPN network (weighted bidirectional feature pyramid network); the defect detection module may include: the last 10 layers of the RPN network and the pre-trained ResNet-101 network. As an implementable manner, the ResNet-101 network can be pre-trained using the ImageNet dataset, and specifically, the ResNet-101 network can be used to train an image classification task on the ImageNet dataset. The specific implementation manner is the existing mature technology, and is not described herein again.
And the first 91 layers of the pre-trained ResNet-101 network and the BiFPN network extract the characteristics of the inspection image of the transformer insulating sleeve to obtain a characteristic diagram with five scales. And the RPN network and the rear 10 layers of the pre-trained ResNet-101 network carry out defect detection on the feature maps of the five scales to obtain a defect detection result.
It should be noted that the ResNet-101 network includes a convolution operation layer, four residual blocks, an average value pooling function, a full-connection network, and a softmax function, and the convolution operation layer and the four residual blocks are the first 91 layers of the ResNet-101 network. The four residual blocks contain 3, 4, 23 and 3 bottletech respectively. Two blocks are contained in each Bottleneck, Conv Block and Identity Block respectively.
The specific process of extracting the characteristics of the inspection image of the transformer insulating sleeve by the front 91 layer of the pre-trained ResNet-101 network and the BiFPN network to obtain the characteristic diagram with five scales can be as follows:
and the convolution operation layer is used for performing convolution operation on the transformer insulating sleeve inspection image by using a convolution kernel, and outputting a first initial characteristic diagram through Max Pooling after batch standardization and a nonlinear activation function are used. Processing the first initial characteristic diagram output by the convolution operation layer by four residual blocks to obtain a second initial characteristic diagram, a third initial characteristic diagram, a fourth initial characteristic diagram and a fifth initial characteristic diagram; and inputting the first initial feature map, the second initial feature map, the third initial feature map, the fourth initial feature map and the fifth initial feature map into a BiFPN network, and outputting the feature maps of the five scales by the BiFPN network.
For the sake of easy understanding, the first initial feature map, the second initial feature map, the third initial feature map, the fourth initial feature map and the fifth initial feature map are assumed to be
Figure BDA0003538838540000071
Then the intermediate feature map of the BiFPN output is
Figure BDA0003538838540000072
Wherein
Figure BDA0003538838540000073
For the ith intermediate feature map, Conv (. cndot.) represents a convolution operation, Resize (. cndot.) represents a Resize operation, and wi1And wi2The learnable feature map weight for the ith intermediate level,
Figure BDA0003538838540000078
for the ith input profile, e is taken to be 0.0001. Using the intermediate feature map, an output feature map can be obtained as
Figure BDA0003538838540000074
Figure BDA0003538838540000075
Figure BDA0003538838540000076
Wherein
Figure BDA0003538838540000077
Denotes the ith output feature map, w'i1、w′i2And w'i3Is the learnable feature map weight of the ith output layer.
The later 10 layers of the RPN network and the pre-trained ResNet-101 network perform defect detection on the feature maps of the five scales, and the specific process of obtaining the defect detection result can be as follows:
judging whether each anchor frame at each position in the feature maps with different scales contains the defect to be detected or not by using the RPN, regressing the coordinate offset of the anchor frame to obtain a region suggestion frame to obtain a region of interest (ROI), and removing repeated ROIs by using non-maximum suppression (NMS) to obtain the ROI to be detected;
and (4) using ROI Pooling to classify and position the defects of the ROI to be detected to obtain a defect detection result.
In another embodiment of the application, a training method of a transformer bushing defect detection model is introduced.
Step S201: and determining a training set and a testing set, wherein the training set and the testing set respectively comprise a plurality of transformer insulating sleeve inspection images marked with defect types and defect positioning frames of all parts.
Step S202: and training the transformer insulating sleeve defect detection model by taking each transformer insulating sleeve inspection image in the training set as a training sample, taking each part of defect types and defect positioning frames marked on each transformer insulating sleeve inspection image in the training set as sample labels, and taking the average accuracy average (mAP) calculated based on the test set as a set threshold value as a training end condition.
It should be noted that the loss function of the transformer bushing defect detection model is an RPN classification loss function LclsAnd a localization loss function LregThe total RPN loss function is:
Figure BDA0003538838540000081
wherein,nclsrepresenting the number of batches of training, NregThe number of anchor frames is shown, and lambda is a parameter for balancing the anchor frames and the lambda. p is a radical ofiPredicting the probability that the ith anchor frame contains the defects to be detected for the model; if the ith anchor frame contains a defect to be detected,
Figure BDA0003538838540000082
is 1, otherwise is 0. t is tiFor the correction parameter of the ith anchor frame,
Figure BDA0003538838540000083
is the actual correction parameter of the ith anchor frame.
In another embodiment of the present application, a specific implementation of determining a training set and a test set is described, which may include:
step S301: and acquiring a transformer insulating sleeve inspection image.
In this step, can utilize industry video, unmanned aerial vehicle, shoot and obtain transformer bushing and patrol and examine the image.
Step S302: and marking the defect types and the defect positioning frames of all the parts in the transformer insulating sleeve inspection image to obtain a transformer insulating sleeve inspection image data set.
Specifically, the transformer insulating sleeve inspection image marked manually in history can be acquired, the characteristics of the image of the defect part of the frame in the image are observed, and then the defect type and the defect positioning frame of each part in the transformer insulating sleeve inspection image collected newly are marked manually according to the rule for defect identification in transformer insulating sleeve inspection. The defect type label uses one-hot code and normal label code [00001]Branch occlusion Label coding [00010]Insulator defect label code [00100 ]]Pin defect label code [01000 ]]Tower lightning strike defect code [10000]. Use of defect positioning frame
Figure BDA0003538838540000084
Is shown in which
Figure BDA0003538838540000085
And
Figure BDA0003538838540000086
respectively representing the horizontal and vertical coordinates, w, of the center of the defect-locating framegtAnd hgtRespectively, the width and height of the defect localization box.
Step S303: and carrying out data enhancement on the transformer insulating sleeve inspection image data set to obtain a sample data set.
As an implementation mode, data enhancement can be performed on all data or part of data in the transformer insulating sleeve inspection image data set to obtain a sample data set. For example, 100 images can be randomly sampled in the transformer insulating sleeve inspection image data set for data enhancement. The specific implementation of the data enhancement will be described in detail by the following embodiments, which are not described in detail herein.
Step S304: determining the training set and the test set based on the sample data set.
As one possible implementation, the sample data set may be divided into a training set and a test set on an 8:2 scale.
In another embodiment of the present application, a specific implementation manner of performing data enhancement on the transformer bushing inspection image data set in step S303 to obtain a sample data set is described, where the implementation manner may include:
step S401: determining an original image to be subjected to data enhancement;
in the application, all data or part of data in the transformer insulating sleeve inspection image data set can be determined as an original image to be subjected to data enhancement.
Step S402: and carrying out random cutting processing on the original image to be subjected to data enhancement to obtain a randomly cut image sample.
Specifically, a random cutting algorithm is used for cutting out square areas from the original image to be subjected to data enhancement respectively, and a defect positioning frame corresponding to the original image is adjusted. The specific adjustment mode can be as follows: first, let the minimum value of the width and height of the original image be lminI.e. lmin＝min(wimg，himg) Wherein w isimgAnd himgRespectively the width and height of the original image. In a uniform distribution of l to (0.3 l)min，lmin) Sampling and randomly cutting a square area with the side length of l from an original image. And then calculating the coordinates of the defect positioning frame in the original image in the randomly cropped image. And finally, removing the incomplete defect positioning frame in the cutting area to finally obtain the randomly cut image.
Step S403: and carrying out random brightness conversion processing on the randomly cut image to obtain an image with random brightness conversion.
The pixel value of a certain pixel point in the randomly cut image in the RGB space is set as (r)0，g0，b0). First, the brightness variation delta is sampled according to the uniform distributionbI.e. by
Figure BDA0003538838540000091
Wherein U () represents Uniform Distribution (uniformity Distribution), bδand
Figure BDA0003538838540000092
respectively, a lower limit and an upper limit of the luminance change amount.
Then according to the formula
Figure BDA0003538838540000101
Determining a pixel value (r) of a random luminance transformed image1，g1，b1). Clip (·) represents a clipping function, and for 24-bit RGB images, the clipping function limits output pixel values to 0-255.
Step S404: and carrying out random hue and saturation transformation processing on the image after the random brightness transformation to obtain a sample data set.
Firstly, transforming each pixel point in the image after random brightness transformation to HSV space. Note pmax＝max(r1，g1，b1)，pmin＝min(r1，g1，b1) Then the image is randomly transformed in brightness from RGB space pixel point (r)1，g1，b1) Transforming to HSV space pixel (h)1，s1，v1) Is of the formula
Figure BDA0003538838540000102
Figure BDA0003538838540000103
v1＝pmax
Then, the hue and saturation changes δ are resampled in a uniform distributionhAnd deltasLightness v2Remain unchanged, i.e.
Figure BDA0003538838540000104
Figure BDA0003538838540000105
v2＝v1
Wherein hδAnd
Figure BDA0003538838540000106
the upper and lower limits of the hue change amount, sδand
Figure BDA0003538838540000107
upper and lower limits of saturation change.
Finally, the transformed image is converted from a pixel point (h) of the HSV space2，s2，v2) Conversion into RGB space (r)2，g2，b2) I.e. by
Figure BDA0003538838540000111
Figure BDA0003538838540000112
e2＝v2×(1-s2)
q2＝v2×(1-f2×s2)
t2＝v2×(1-(1-f2)×s2)
Figure BDA0003538838540000113
Wherein f is2，e2，q2，t2Is an intermediate variable.
In another embodiment of the present application, a specific implementation of calculating the average precision mean value based on the test set is described, which may include the following steps:
step S501: and inputting each sample data in the test set into a transformer insulating sleeve defect detection model to obtain a detection result, wherein the detection result comprises a defect positioning frame corresponding to each defect type.
Step S502: and traversing each defect type, and calculating the average precision of the defect types.
Specifically, for each defect type, sorting each defect positioning frame of the type from high to low according to the scores, traversing each defect positioning frame, marking the detection frame as a positive sample if the intersection union ratio (IoU) of the detection frame and a certain correctly labeled defect positioning frame is greater than 0.5, removing the correctly labeled defect positioning frame corresponding to the detection frame, and otherwise marking as a negative sample. The accuracy and recall are calculated and the plot is plotted on an accuracy-recall curve. The average accuracy, i.e., the area under the accuracy-recall curve, is calculated.
Step S503: and calculating the average value of the average precision of each defect type as the average precision average value.
The transformer bushing defect detection device disclosed in the embodiment of the present application is described below, and the transformer bushing defect detection device described below and the transformer bushing defect detection method described above may be referred to in a mutually corresponding manner.
Referring to fig. 4, fig. 4 is a schematic structural diagram of a transformer bushing defect detection apparatus disclosed in an embodiment of the present application. As shown in fig. 4, the apparatus for detecting a defect of an insulation sleeve of a transformer may include:
the acquiring unit 11 is used for acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
and the detection unit 12 is used for inputting the transformer insulating sleeve inspection image into a transformer insulating sleeve defect detection model, and the transformer insulating sleeve defect detection model is used for extracting the characteristics of the transformer insulating sleeve inspection image to obtain characteristic diagrams of multiple scales and detecting the defects of the characteristic diagrams of all scales to obtain a defect detection result.
Referring to fig. 5, fig. 5 is a block diagram of a hardware structure of a transformer bushing defect detection device according to an embodiment of the present application, and referring to fig. 5, the hardware structure of the transformer bushing defect detection device may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;
in the embodiment of the application, the number of the processor 1, the communication interface 2, the memory 3 and the communication bus 4 is at least one, and the processor 1, the communication interface 2 and the memory 3 complete mutual communication through the communication bus 4;
the processor 1 may be a central processing unit CPU, or an application Specific Integrated circuit asic, or one or more Integrated circuits configured to implement embodiments of the present invention, etc.;
the memory 3 may include a high-speed RAM memory, and may further include a non-volatile memory (non-volatile memory) or the like, such as at least one disk memory;
wherein the memory stores a program and the processor can call the program stored in the memory, the program for:
acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
inputting the transformer insulating sleeve inspection image into a transformer insulating sleeve defect detection model, performing feature extraction on the transformer insulating sleeve inspection image by the transformer insulating sleeve defect detection model to obtain feature maps of multiple scales, and performing defect detection on the feature maps of the scales to obtain a defect detection result.
Alternatively, the detailed function and the extended function of the program may be as described above.
Embodiments of the present application further provide a readable storage medium, which may store a program adapted to be executed by a processor, where the program is configured to:
acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
inputting the transformer insulating sleeve inspection image into a transformer insulating sleeve defect detection model, performing feature extraction on the transformer insulating sleeve inspection image by the transformer insulating sleeve defect detection model to obtain feature maps of multiple scales, and performing defect detection on the feature maps of the scales to obtain a defect detection result.
Alternatively, the detailed function and the extended function of the program may be as described above.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A transformer bushing defect detection method is characterized by comprising the following steps:
acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
inputting the transformer insulation sleeve inspection image into a transformer insulation sleeve defect detection model, performing feature extraction on the transformer insulation sleeve inspection image by the transformer insulation sleeve defect detection model to obtain feature maps of multiple scales, and performing defect detection on the feature maps of the scales to obtain a defect detection result.
2. The method of claim 1, wherein the transformer bushing defect detection model comprises:
a multi-scale feature map extraction module and a defect detection module;
the transformer bushing defect detection model is right the transformer bushing patrols and examines the image and carry out feature extraction, obtains the characteristic map of a plurality of yards to carry out defect detection to the characteristic map of each yard, obtain the defect testing result, include:
the multi-scale characteristic diagram extraction module is used for extracting the characteristics of the transformer insulating sleeve inspection image to obtain characteristic diagrams of multiple scales;
and the defect detection module is used for detecting the defects of the characteristic diagrams of all scales to obtain a defect detection result.
3. The method of claim 2,
the multi-scale feature map extraction module comprises: the first 91 layers of the pretrained ResNet-101 network and the BiFPN network;
the multi-scale feature map extraction module is right the transformer bushing patrols and examines the image and carry out the feature extraction, obtains the feature map of a plurality of yards, includes:
the first 91 layers of the pre-trained ResNet-101 network and the BiFPN network extract the characteristics of the inspection image of the transformer insulating sleeve to obtain a characteristic diagram with five scales;
the defect detection module comprises: the back 10 layers of the RPN network and the pre-trained ResNet-101 network;
the defect detection module carries out defect detection on the feature map of each scale to obtain a defect detection result, and the defect detection result comprises the following steps:
and the RPN network and the rear 10 layers of the pre-trained ResNet-101 network carry out defect detection on the feature maps of the five scales to obtain a defect detection result.
4. The method according to claim 2, wherein the training mode of the transformer bushing defect detection model comprises:
determining a training set and a testing set, wherein the training set and the testing set respectively comprise a plurality of transformer insulating sleeve inspection images marked with defect types and defect positioning frames of all parts;
and training the transformer insulating sleeve defect detection model by taking each transformer insulating sleeve inspection image in the training set as a training sample, taking each part of defect types and defect positioning frames marked on each transformer insulating sleeve inspection image in the training set as sample labels, and taking the average precision mean value calculated based on the test set and reaching a set threshold value as a training finishing condition.
5. The method of claim 4, wherein determining the training set and the test set comprises:
acquiring a transformer insulating sleeve inspection image;
marking the defect types and the defect positioning frames of all the parts in the transformer insulating sleeve inspection image to obtain a transformer insulating sleeve inspection image data set;
performing data enhancement on a transformer insulating sleeve inspection image data set to obtain a sample data set;
determining the training set and the test set based on the sample data set.
6. The method according to claim 5, wherein the data enhancement of the transformer bushing inspection tour image data set is performed to obtain a sample data set, and comprises:
determining an original image to be subjected to data enhancement;
carrying out random cutting processing on the original image to be subjected to data enhancement to obtain a randomly cut image sample;
carrying out random brightness conversion processing on the randomly cut image to obtain an image with random brightness conversion;
and carrying out random hue and saturation transformation processing on the image subjected to the random brightness transformation to obtain a sample data set.
7. The method of claim 4, wherein said computing an average precision mean based on said test set comprises:
inputting each sample data in the test set into a transformer insulating sleeve defect detection model to obtain a detection result, wherein the detection result comprises a defect positioning frame corresponding to each defect type;
traversing each defect type, and calculating the average precision of the defect types;
and calculating the average value of the average precision of each defect type as the average precision average value.
8. A transformer bushing defect detection device, characterized in that the device includes:
the acquiring unit is used for acquiring a transformer insulating sleeve inspection image to be subjected to defect detection;
and the detection unit is used for inputting the transformer insulating sleeve inspection image into a transformer insulating sleeve defect detection model, the transformer insulating sleeve defect detection model is used for extracting the characteristics of the transformer insulating sleeve inspection image to obtain characteristic diagrams of multiple scales, and defect detection is carried out on the characteristic diagrams of all scales to obtain a defect detection result.
9. The transformer insulating sleeve defect detection equipment is characterized by comprising a memory and a processor;
the memory is used for storing programs;
the processor is used for executing the program to realize the steps of the transformer bushing defect detection method according to any one of claims 1 to 7.
10. A readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the transformer bushing defect detection method according to any one of claims 1 to 7.
CN202210224935.2A 2022-03-09 2022-03-09 Transformer insulating sleeve defect detection method, related equipment and readable storage medium Pending CN114581419A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210224935.2A CN114581419A (en) 2022-03-09 2022-03-09 Transformer insulating sleeve defect detection method, related equipment and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210224935.2A CN114581419A (en) 2022-03-09 2022-03-09 Transformer insulating sleeve defect detection method, related equipment and readable storage medium

Publications (1)

Publication Number Publication Date
CN114581419A true CN114581419A (en) 2022-06-03

Family

ID=81773204

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210224935.2A Pending CN114581419A (en) 2022-03-09 2022-03-09 Transformer insulating sleeve defect detection method, related equipment and readable storage medium

Country Status (1)

Country Link
CN (1) CN114581419A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587975A (en) * 2022-09-28 2023-01-10 国网湖北省电力有限公司超高压公司 Oil-filled equipment casing dirt deposition defect diagnosis system, method and medium
CN115950887A (en) * 2022-12-23 2023-04-11 广东瑞根电子有限公司 Electronic transformer defect detection method, storage medium and device

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111583198A (en) * 2020-04-23 2020-08-25 浙江大学 Insulator picture defect detection method combining FasterR-CNN + ResNet101+ FPN
CN113205502A (en) * 2021-05-10 2021-08-03 内蒙古大学 Insulator defect detection method and system based on deep learning
CN113240642A (en) * 2021-05-13 2021-08-10 创新奇智(北京)科技有限公司 Image defect detection method and device, electronic equipment and storage medium
CN113592839A (en) * 2021-08-06 2021-11-02 广东电网有限责任公司 Distribution network line typical defect diagnosis method and system based on improved fast RCNN
CN114140446A (en) * 2021-09-29 2022-03-04 国网甘肃省电力公司检修公司 Insulator defect detection method based on semantic segmentation and target detection

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111583198A (en) * 2020-04-23 2020-08-25 浙江大学 Insulator picture defect detection method combining FasterR-CNN + ResNet101+ FPN
CN113205502A (en) * 2021-05-10 2021-08-03 内蒙古大学 Insulator defect detection method and system based on deep learning
CN113240642A (en) * 2021-05-13 2021-08-10 创新奇智(北京)科技有限公司 Image defect detection method and device, electronic equipment and storage medium
CN113592839A (en) * 2021-08-06 2021-11-02 广东电网有限责任公司 Distribution network line typical defect diagnosis method and system based on improved fast RCNN
CN114140446A (en) * 2021-09-29 2022-03-04 国网甘肃省电力公司检修公司 Insulator defect detection method based on semantic segmentation and target detection

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587975A (en) * 2022-09-28 2023-01-10 国网湖北省电力有限公司超高压公司 Oil-filled equipment casing dirt deposition defect diagnosis system, method and medium
CN115950887A (en) * 2022-12-23 2023-04-11 广东瑞根电子有限公司 Electronic transformer defect detection method, storage medium and device

Similar Documents

Publication Publication Date Title
US10977523B2 (en) Methods and apparatuses for identifying object category, and electronic devices
CN108009543B (en) License plate recognition method and device
CN108681692B (en) Method for identifying newly added buildings in remote sensing image based on deep learning
CN111310645B (en) Method, device, equipment and storage medium for warning overflow bin of goods accumulation
CN111680746B (en) Vehicle damage detection model training, vehicle damage detection method, device, equipment and medium
CN114581419A (en) Transformer insulating sleeve defect detection method, related equipment and readable storage medium
CN111507958A (en) Target detection method, training method of detection model and electronic equipment
CN108489996B (en) Insulator defect detection method and system and terminal equipment
CN113592839B (en) Distribution network line typical defect diagnosis method and system based on improved fast RCNN
CN111814905A (en) Target detection method, target detection device, computer equipment and storage medium
CN111985465A (en) Text recognition method, device, equipment and storage medium
CN112750113B (en) Glass bottle defect detection method and device based on deep learning and linear detection
CN116152261B (en) Visual inspection system for quality of printed product
CN115829995A (en) Cloth flaw detection method and system based on pixel-level multi-scale feature fusion
CN113469997A (en) Method, device, equipment and medium for detecting plane glass
CN111753794A (en) Fruit quality classification method and device, electronic equipment and readable storage medium
CN114511718A (en) Intelligent management method and system for materials for building construction
CN115239646A (en) Defect detection method and device for power transmission line, electronic equipment and storage medium
CN107993223A (en) Scratch detection method, apparatus and electronic equipment
CN115170512A (en) Defect classification and identification method and device, storage medium and electronic equipment
CN111738272A (en) Target feature extraction method and device and electronic equipment
CN110490170B (en) Face candidate frame extraction method
CN117351472A (en) Tobacco leaf information detection method and device and electronic equipment
CN117422677A (en) Method, device and system for detecting image defects of power line for airborne terminal
CN112001336A (en) Pedestrian boundary crossing alarm method, device, equipment and system

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