CN117437482A - Cable main insulation part defect identification method and system - Google Patents
Cable main insulation part defect identification method and system Download PDFInfo
- Publication number
- CN117437482A CN117437482A CN202311520856.7A CN202311520856A CN117437482A CN 117437482 A CN117437482 A CN 117437482A CN 202311520856 A CN202311520856 A CN 202311520856A CN 117437482 A CN117437482 A CN 117437482A
- Authority
- CN
- China
- Prior art keywords
- image
- training
- neural network
- texture features
- preset
- 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
Links
- 230000007547 defect Effects 0.000 title claims abstract description 133
- 238000009413 insulation Methods 0.000 title claims abstract description 101
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000012549 training Methods 0.000 claims abstract description 161
- 238000013528 artificial neural network Methods 0.000 claims abstract description 84
- 238000007781 pre-processing Methods 0.000 claims abstract description 54
- 238000005070 sampling Methods 0.000 claims abstract description 30
- 210000001503 joint Anatomy 0.000 claims abstract description 3
- 238000012545 processing Methods 0.000 claims description 52
- 238000012216 screening Methods 0.000 claims description 24
- 238000010586 diagram Methods 0.000 claims description 19
- 230000004075 alteration Effects 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 12
- 238000003708 edge detection Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 abstract description 2
- 230000008859 change Effects 0.000 description 6
- 238000010276 construction Methods 0.000 description 6
- 230000002950 deficient Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013211 curve analysis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000012212 insulator Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for identifying defects of a main insulation part of a cable, which relate to the technical field of power cable detection, wherein when a training image set is received, the training image set is subjected to image preprocessing to generate corresponding training image texture features, the training image texture features are input into a preset initial neural network identification model for training, a target neural network identification model is generated, when a sampling image of a middle joint of the cable to be identified is received, the sampling image is subjected to image preprocessing, and a corresponding defect identification result is output through the target neural network identification model; the defect identification method solves the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well identified, the defects are required to be confirmed manually, the defects cannot be identified or the defects are identified incorrectly easily, and the operation reliability of the distribution network cable is reduced.
Description
Technical Field
The invention relates to the technical field of power cable detection, in particular to a method and a system for identifying defects of a main insulation part of a cable.
Background
Currently, most of the cable line faults are caused by improper operation in the construction process. On the one hand, the cable is damaged by being pulled or excessively bent due to the overlarge mechanical traction in the process of laying the cable. On the other hand, improper operation in installing the intermediate joint results in construction defects at the joint. Among the two factors, the construction defect of the cable intermediate joint is most hidden, the duty ratio is the largest in cable accidents caused by the existence of the construction defect of the joint, the existence of the construction defect directly causes the reduction of the insulation performance of the joint, the service life is shortened, and breakdown easily occurs after the joint is put into operation.
The defects of the main insulation part of the cable mainly comprise main insulation stain defects and main insulation scratch defects according to statistics, the identification efficiency of the two defects is low at present, the types of the defects cannot be well identified, the defects need to be confirmed manually, the defects are easy to be easily identified or the situation of identification errors is easy to be caused, and the operation reliability of the distribution network cable is reduced.
Disclosure of Invention
The invention provides a method and a system for identifying defects of a main insulation part of a cable, which solve the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well identified, the defects are required to be confirmed manually, the defects cannot be identified or the defects are identified incorrectly easily, and the operation reliability of a distribution network cable is reduced.
The invention provides a defect identification method for a main insulation part of a cable, which comprises the following steps:
when a training image set is received, carrying out image preprocessing on the training image set to generate corresponding training image texture features;
inputting the texture features of the training image into a preset initial neural network recognition model for training, and generating a target neural network recognition model;
when a sampling image of the cable middle joint to be identified is received, the sampling image is subjected to image preprocessing, and a corresponding defect identification result is output through the target neural network identification model.
Optionally, when receiving the training image set, performing image preprocessing on the training image set to generate corresponding training image texture features, including:
when a training image set is received, a corresponding RGB color space component square distribution diagram is established for the training image set, wherein the training image comprises a cable intermediate joint image and an associated main insulation part image;
extracting corresponding RGB component values from the RGB color space component square distribution map associated with the main insulation part image;
determining a target area where the main insulation part is located based on the RGB component value, and taking an image of the target area as a preliminary image, wherein the preliminary image is an image only containing the main insulation part;
and performing image processing on the preliminary image based on a preset image processing strategy to generate corresponding training image texture features.
Optionally, the preset image processing policy includes a first image processing policy and a second image processing policy, and the step of performing image processing on the preliminary image based on the preset image processing policy to generate a corresponding training image texture feature includes:
Performing image processing on the preliminary image by adopting the first image processing strategy to generate corresponding first image texture features, wherein the first image texture features comprise first average brightness, first average contrast, symmetry of a first gray histogram, first gray randomness and first chromatic aberration;
performing image processing on the preliminary image by adopting the second image processing strategy to generate corresponding second image texture features, wherein the second image texture features comprise second average brightness, second average contrast, symmetry of a second gray histogram, second gray randomness and second chromatic aberration;
and taking the first image texture feature and the second image texture feature as training image texture features.
Optionally, the step of performing image processing on the preliminary image by using the first image processing policy to generate a corresponding first image texture feature includes:
cutting the preliminary image by adopting a maximum inter-class variance method to generate a first cutting image;
obtaining a vertical direction projection image of the first cutting image in the three components of the RGB color space in the horizontal direction according to a projection method;
Extracting an image between two inflection points in the vertical direction projection image to serve as a first preprocessing image;
and extracting a first image texture feature associated with the first preprocessing image.
Optionally, the step of performing image processing on the preliminary image by using the second image processing policy to generate a corresponding second image texture feature includes:
cutting the preliminary image by adopting a maximum inter-class variance method to generate a second cut image;
performing edge detection on the second cut image by adopting a Prewitt operator to generate a binary image;
multiplying the binary image and the second cutting image to generate a second preprocessing image;
and extracting a second image texture feature associated with the second preprocessing image.
Optionally, the step of inputting the training image texture features into a preset initial neural network recognition model to train and generate a target neural network recognition model includes:
inputting the texture features of the training image into a preset initial neural network recognition model for training, and outputting a corresponding training result;
calculating an error value between the training result and a preset standard training result;
And outputting the target neural network identification model when the error value is smaller than a preset standard error value.
Optionally, the method further comprises a model optimization process, wherein the model optimization process specifically comprises the following steps:
grouping the training image texture features according to a preset grouping condition to generate a first grouping training image texture feature and a second grouping training image texture feature;
acquiring a first weight ratio value associated with the texture features of the first grouping training image;
acquiring a second weight ratio value associated with the texture features of the second packet training image;
adopting a preset first adjustment value to adjust the first weight ratio, and calculating a first error rate associated with all the first weight ratio values through traversal;
and screening a first target error rate meeting a preset screening condition from the plurality of first error rates, and feedback optimizing the target neural network identification model according to the first weight ratio related to the first target error rate, wherein the preset screening condition is that the first error rate is smaller than or equal to a preset error rate tolerance value and the first error rate is the minimum value.
Optionally, the method further comprises:
when any one of the first error rates does not meet the preset screening conditions, a preset second adjustment value is adopted to adjust the first weight ratio, and the second error rates related to all the first weight ratio values are calculated through traversal;
And screening the second error rates meeting preset screening conditions from the plurality of second error rates, and feeding back and optimizing the target neural network identification model according to the first weight ratio related to the second error rates.
A second aspect of the present invention provides a system for identifying defects in a main insulation portion of a cable, including:
the image preprocessing module is used for carrying out image preprocessing on the training image set when the training image set is received, and generating corresponding training image texture features;
the target neural network recognition model module is used for inputting a preset initial neural network recognition model to train by adopting the texture characteristics of the training image to generate a target neural network recognition model;
and the defect identification result output module is used for carrying out image preprocessing on the sampling image when receiving the sampling image of the cable intermediate joint to be identified, and outputting a corresponding defect identification result through the target neural network identification model.
Optionally, the image preprocessing module includes:
the training image receiving sub-module is used for establishing a corresponding RGB color space component square distribution diagram for a training image set when the training image set is received, wherein the training image comprises a cable intermediate joint image and an associated main insulation part image;
An RGB component value sub-module, configured to extract a corresponding RGB component value from the RGB color space component square distribution map associated with the main insulation part image;
a preliminary image sub-module, configured to determine a target area where the main insulation part is located based on the RGB component values, and take an image of the target area as a preliminary image, where the preliminary image is an image that only includes the main insulation part;
and the training image texture feature sub-module is used for carrying out image processing on the preliminary image based on a preset image processing strategy to generate corresponding training image texture features.
From the above technical scheme, the invention has the following advantages:
when a training image set is received, image preprocessing is carried out on the training image set to generate corresponding training image texture features, the training image texture features are input into a preset initial neural network recognition model to carry out training to generate a target neural network recognition model, when a sampling image of a cable intermediate joint to be recognized is received, image preprocessing is carried out on the sampling image, and a corresponding defect recognition result is output through the target neural network recognition model; the defect identification method solves the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well distinguished, the defects are required to be confirmed manually, the defects cannot be identified or the errors are identified easily, and the operation reliability of the distribution network cable is reduced; the invention uses the projection method to perform primary positioning of the main insulation, extracts the main insulation area, identifies the type of the defect through the neural network identification algorithm, and has accurate identification result.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained from these drawings without inventive faculty for a person skilled in the art.
Fig. 1 is a flowchart of a method for identifying defects of a main insulation portion of a cable according to an embodiment of the present invention;
fig. 2 is a flowchart of a method for identifying defects of a main insulation part of a cable according to a second embodiment of the present invention;
FIG. 3 is a graph of R color space component square distribution of a main insulating portion;
FIG. 4 is a graph of the G color space component square distribution of the main insulating portion;
FIG. 5 is a B color space component square distribution diagram of a main insulating portion;
FIG. 6 is a plot of the R color space component square of the isolated main insulation portion of FIG. 3;
FIG. 7 is a graph of the G color space component square distribution of the isolated main insulating portion of FIG. 4;
FIG. 8 is a B color space component square distribution plot of the isolated main insulation portion of FIG. 5;
FIG. 9 is a first schematic view of a stain defect;
FIG. 10 is a second schematic view of a stain defect;
FIG. 11 is a third schematic illustration of a stain defect;
FIG. 12 is a projection view of a RGB color space perpendicular comprising a first schematic of smudge defects;
FIG. 13 is a projection view in the vertical direction of the RGB color space containing a second schematic of smudge defects;
FIG. 14 is a projection view in the vertical direction of the RGB color space containing a third schematic with smudge defects;
FIG. 15 is a first schematic view of a scratch defect;
FIG. 16 is a second schematic diagram including a scratch defect;
FIG. 17 is a third schematic diagram including a scratch defect;
FIG. 18 is a perspective view in the RGB component radial direction of a first schematic including scratch defects;
FIG. 19 is a perspective view in the RGB component radial direction of a second schematic containing scratch defects;
FIG. 20 is a perspective view in the radial direction of RGB components of a third schematic containing scratch defects;
fig. 21 is a block diagram of a defect identifying system for a main insulation part of a cable according to a third embodiment of the present invention.
Detailed Description
The embodiment of the invention provides a method and a system for identifying defects of a main insulation part of a cable, which are used for solving the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well identified, the defects are required to be confirmed manually, the defects cannot be identified or the defects are identified incorrectly easily, and the operation reliability of a distribution network cable is reduced.
In order to make the objects, features and advantages of the present invention more comprehensible, the technical solutions in the embodiments of the present invention are described in detail below with reference to the accompanying drawings, and it is apparent that the embodiments described below are only some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a method for identifying defects of a main insulation portion of a cable according to an embodiment of the invention.
The invention provides a defect identification method for a main insulation part of a cable, which comprises the following steps:
and 101, when a training image set is received, performing image preprocessing on the training image set to generate corresponding training image texture features.
Training images refer to a set of images consisting of cable intermediate joint images.
Training image texture features refer to extracting texture features of images of a training image set subjected to image preprocessing, and are used for training a model by taking the texture features as feature quantities input by the model.
In the embodiment of the invention, when a training image set formed by cable intermediate joint images is received, image preprocessing is carried out on the training image set, and training image texture features for training a model by taking the feature quantity input by the model are generated.
And 102, inputting a preset initial neural network recognition model by using texture features of the training image to train, and generating a target neural network recognition model.
And the target neural network recognition model is used for carrying out defect recognition on the sampling image of the cable intermediate connector to be recognized, so as to output a recognition model of a defect recognition result.
In the embodiment of the invention, training image texture features are used as input feature quantities to input a preset initial neural network recognition model for training, and a target neural network recognition model is generated.
And 103, when a sampling image of the cable intermediate joint to be identified is received, performing image preprocessing on the sampling image, and outputting a corresponding defect identification result through the target neural network identification model.
In the embodiment of the invention, when a sampling image of a cable intermediate joint to be identified is received, image preprocessing is carried out on the sampling image, sampling image texture features related to the sampling image are generated, and the sampling image texture features are input into a target neural network identification model for identification, so that a corresponding defect identification result is output.
When a training image set is received, image preprocessing is carried out on the training image set to generate corresponding training image texture features, the training image texture features are input into a preset initial neural network recognition model to carry out training to generate a target neural network recognition model, when a sampling image of a cable intermediate joint to be recognized is received, image preprocessing is carried out on the sampling image, and a corresponding defect recognition result is output through the target neural network recognition model; the defect identification method solves the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well distinguished, the defects are required to be confirmed manually, the defects cannot be identified or the errors are identified easily, and the operation reliability of the distribution network cable is reduced; the invention uses the projection method to perform primary positioning of the main insulation, extracts the main insulation area, identifies the type of the defect through the neural network identification algorithm, and has accurate identification result.
Referring to fig. 2, fig. 2 is a flowchart illustrating a method for identifying defects of a main insulation portion of a cable according to a second embodiment of the present invention.
The invention provides a defect identification method for a main insulation part of a cable, which comprises the following steps:
step 201, when a training image set is received, a corresponding RGB color space component square distribution diagram is established for the training image set, wherein the training image comprises a cable intermediate joint image and an associated main insulation part image.
In the embodiment of the invention, corresponding RGB color space component straight distribution diagrams are respectively established for the cable intermediate joint image and the main insulation part image of the cable intermediate joint image, wherein, fig. 3-5 are the corresponding RGB color space component straight distribution diagrams of the main insulation part image.
Step 202, extracting corresponding RGB component values from the RGB color space component square distribution diagram associated with the main insulation part image.
In the embodiment of the invention, the RGB component values of the main insulation part images separated according to fig. 3-5 are extracted from the RGB color space component square distribution diagram corresponding to the main insulation part images as shown in fig. 6-8.
It is worth mentioning that the cable intermediate connector image can be separated from the main insulation part image by RGB component values.
Step 203, determining a target area where the main insulation part body is located based on the RGB component values, and taking an image of the target area as a preliminary image, wherein the preliminary image is an image only containing the main insulation part.
In the embodiment of the invention, the RGB component values are R epsilon [100,161], G epsilon [90,160], B epsilon [100, 65]. Each gradation component can effectively determine the region of the main insulation as the main color feature of the main insulation portion.
Step 201 to step 203 are specifically to obtain a cable intermediate joint image and a main insulation part image associated with the cable intermediate joint image, respectively establishing an associated RGB color space component square distribution diagram for the cable intermediate joint image and the main insulation part image, where the RGB color space component square distribution diagram of the main insulation part image is shown in fig. 3 to 5, firstly separating a main insulation part body from the RGB color space component square distribution diagram of the main insulation part image, that is, extracting RGB component values associated with the main insulation part, then extracting associated RGB component values from the RGB color space component square distribution diagram of the cable intermediate joint image, and judging a region of the main insulation part in the cable intermediate joint image by the separated RGB component values associated with the main insulation part image and the RGB component values associated with the cable intermediate joint image; and determining a target area where the main insulation part body is located based on the RGB component values, and taking an image of the target area as a preliminary image, wherein the preliminary image is an image only containing the main insulation part.
And 204, performing image processing on the preliminary image based on a preset image processing strategy to generate corresponding training image texture features.
Further, the preset image processing policies include a first image processing policy and a second image processing policy, and step 204 may include the sub-steps of:
s11, performing image processing on the preliminary image by adopting a first image processing strategy to generate corresponding first image texture features, wherein the first image texture features comprise first average brightness, first average contrast, symmetry of a first gray histogram, first gray randomness and first chromatic aberration.
Further, S11 may comprise the sub-steps of:
s111, cutting the preliminary image by adopting a maximum inter-class variance method, and generating a first cutting image.
S112, obtaining a vertical projection image of the first cutting image in the three components of the RGB color space in the horizontal direction according to a projection method.
S113, extracting an image between two inflection points in the vertical projection image as a first preprocessing image.
S114, extracting first image texture features associated with the first preprocessing image.
Since the preliminary image extracted in step 203 is an image in which the main insulating portion is mixed with the background, the preliminary image needs to be preprocessed, and the background is removed, so as to obtain an accurate clear image of the main insulating portion.
It should be noted that the first cut image is in a matrix form composed of a plurality of pixels, and projection is performed, that is, the gray matrix of the first cut image is calculated, so as to obtain n columns of straight projection images.
In the embodiment of the invention, the mixed image of the main insulation and the background is segmented by adopting a maximum inter-class variance method, the image is segmented according to a gray threshold e by adopting the maximum inter-class variance method, a first segmented image is generated, a vertical projection image of the first segmented image in three components of a horizontal RGB color space is obtained according to a projection method, inflection points of the vertical projection image are searched, partial images between the two inflection points are independently extracted, a first preprocessed image is obtained, first image texture features associated with the first preprocessed image are extracted, the first image texture features comprise first average brightness, first average contrast, symmetry of a first gray histogram, first gray randomness and first color difference, and the image texture features of the image with the main insulation part with the stain defects and the image texture features of the image without the stain defects can be better distinguished and used as a training set of a model.
The first image texture feature includes a first average brightness, a first average contrast, symmetry of a first gray histogram, first gray randomness, and a first color difference, where the first color difference is an important feature quantity for determining whether a main insulation stain defect exists; referring to fig. 9-11, fig. 9-11 show images of stains on the surface of the main insulating portion, wherein fig. 9-11 show images of only the main insulating portion after the background is cut, the white portion is the main insulating portion, the black portion in the white portion package is the portion with stains, fig. 12-14 show vertical projection images, and the change of the images is usually continuous due to the fact that only the main insulating portion is in the images, and the stains on the surface of the main insulating portion and the color difference between the main insulating portion are larger in the power cable construction site, so that whether the stains exist on the surface of the images can be effectively shown by a projection method.
In FIG. 12, the curve has obvious inflection points, which are larger than the average value of the whole gray scale, so that the surface of the main insulation part can be judged to have non-main insulation type color objects, and the gray scale is smaller; the curve of FIG. 13 is divided into two sections, the whole change of the curve is not large, but the middle has obvious difference, and as only a main insulating part exists in the graph, the surface conditions are similar, and the large change proves that the color mutation exists in the graph; the curve in fig. 14 varies greatly and has a "V" shaped sharp corner. Compared with the original image, the region with mutation is the stain defect region through curve analysis, the projection image can better reflect the region with stains on the main insulator, and the defect region can be independently extracted through projection positioning and color gray scale difference.
The defective area image was extracted, the average gray level of the area was obtained, the gray level of the main insulating non-defective area was obtained, the ratio of the mode defining the difference between the main insulating average gray level and the average gray level of the defective area to the main insulating average gray level was defined as the color difference, and the color difference and the gray level were obtained as shown in table 1.
TABLE 1 Gray-scale of defective area and Main insulation Gray-scale
The chromatic aberration is defined as normal between [ -0.1,0.1], and when the chromatic aberration is beyond the range, the chromatic aberration is larger, and the defect is a stain defect.
S12, performing image processing on the preliminary image by adopting a second image processing strategy to generate a corresponding second image texture feature, wherein the second image texture feature comprises second average brightness, second average contrast, symmetry of a second gray histogram, second gray randomness and second chromatic aberration.
Further, S12 may comprise the sub-steps of:
s121, cutting the preliminary image by using a maximum inter-class variance method to generate a second cut image.
S122, performing edge detection on the second cut image by adopting a Prewitt operator to generate a binary image.
S123, multiplying the binary image and the second cutting image to generate a second preprocessing image.
S124, extracting second image texture features associated with the second preprocessed image.
In the embodiment of the invention, the mixed image of the main insulation and the background is segmented by adopting a maximum inter-class variance method, the image is segmented according to a gray threshold e by adopting the maximum inter-class variance method, a second cutting image which is free of the background and only contains the main insulation part is generated, the edge detection is carried out on the second cutting image by adopting a Prewitt operator, a binary image is generated, the edge detection is used for better extracting the edges of the main insulation part and the area to be scratched, a binary image is generated, multiplication operation is carried out on the binary image and the second cutting image, a second preprocessing image is generated, and a second image texture feature associated with the second preprocessing image is extracted, wherein the second image texture feature comprises second average brightness, second average contrast, symmetry of a second gray histogram, second gray randomness and second color difference, and the image texture feature of the image with the scratch defect and the image without the scratch defect of the main insulation part can be better distinguished, and the image texture feature without the scratch defect image is taken as a training set of a model.
It should be noted that, as shown in fig. 15 to 17, the image in which the scratch exists on the surface of the main insulation is an internal change of the main insulation, the scratch is almost similar to the color of the main insulation body, and the scratch usually causes erroneous judgment by the projection method, but the scratch usually has large differences in average brightness, average contrast, symmetry of the gray histogram and gray randomness, so that the scratch cannot be accurately judged by gray color difference, and the judgment needs to be performed by texture characteristics.
The axial gray level change along the cable in the image is represented by radial projection, the position of the cable in the whole image can be judged by the axial projection, the gray level at the dividing point can be suddenly changed, and the main body part of the cable intermediate joint is positioned and separated based on the gray level change, as shown in fig. 18-20.
Similarly, the segmentation threshold e is calculated based on the maximum inter-class variance method, and the binary image can be better calculated by the automatically calculated threshold segmentation image than the artificial set value, and the region with the main insulation part can be obtained by multiplying the binary image obtained by segmentation with the original image.
S13, taking the first image texture feature and the second image texture feature as training image texture features.
In the embodiment of the invention, the first image texture feature and the second image texture feature are used as training image texture features.
And 205, inputting a preset initial neural network recognition model by using texture features of the training image to train, and generating a target neural network recognition model.
Further, step 205 may comprise the sub-steps of:
s21, training is carried out by inputting training image texture features into a preset initial neural network recognition model, and corresponding training results are output.
S22, calculating an error value between the training result and a preset standard training result.
S23, outputting a target neural network recognition model when the error value is smaller than a preset standard error value.
In the embodiment of the invention, five characteristic quantities of average brightness, average contrast, symmetry of gray level histogram, gray level randomness and chromatic aberration are taken as the neuron number of an input layer, and are output as two types of main insulation stain defects and main insulation scratch defects, and the two types of main insulation stain defects are output through a formulaSelecting the number of hidden layer nodes, wherein m and n are the number of input and output nodes respectively, a is an integer between 1 and 10, training a preset initial neural network identification model, and ending when the error value of the result of the neural network training is smaller than a set value Training to obtain the target neural network recognition model.
TABLE 2 training image texture feature data
By comparing the correlations of the feature amounts, it is found that the standard deviation σ is not within the range of [0,1], and it is generally necessary to ensure that the result is between [0,1] in the output, and therefore it is necessary to use normalization operation for the parameters that are not within the prescribed range. There are mainly two types of stains and scratches for the main insulation, namely, the two types of defects are expected to be output, and as the excitation function of the neuron has an S-shaped characteristic, the result is between (0, 1), and 0.001 represents 0 and 0.999 represents 1, so that the code of the stain of the main insulation is expected to be output (0.001,0.999), and the code of the scratch of the main insulation is expected to be output (0.999,0.001).
The training process of the neural network is as follows:
(1) The initial weight is given a random number, the maximum iteration is set for 1000 times, the minimum error is 0.001, the training step length is 0.01, and the number of input samples is 20, as shown in table 2;
(2) Inputting the sample parameters into a network, and inputting the corresponding two types of defect coding values into the network as output;
(3) And learning and training according to the designed network structure, and ending training when the error value is smaller than the stipulation value to obtain a trained network.
TABLE 3 test data
TABLE 4 test results
After the network training is finished, 10 images with defects of main insulation are randomly extracted, wherein 5 images are respectively extracted from main insulation stains and main insulation scratches, corresponding characteristic quantities of defect areas are respectively extracted, the images are input into a network, the reliability of the network is tested, and the test results are shown in table 3.
As shown in Table 4, the results obtained in group 3 and group 5 are different from the expected results, and erroneous judgment occurs, so that the recognition effect is better in the overall view, the recognition rate reaches 80%, and the basic requirement of the early-stage experiment can be met. However, the problem that the two sets of defect experimental data are not recognized still exists in the experiment, and the symmetry characteristic quantity of the average brightness and gray level histogram in the texture characteristic is greatly changed by comparing the two sets of defect experimental data with the similar defect data.
By analysis, the reasons for possible erroneous judgment are:
(1) The extraction of the defect area is not accurate enough, and certain error exists when the characteristic quantity is extracted;
(2) The number of samples selected during training of the neural network may be insufficient, resulting in insufficient accuracy of the data calculated by the trained network;
(3) The feature quantity may be insufficiently selected, and the defect feature quantity may be reflected, so that the defect feature quantity cannot be accurately identified.
Therefore, for possible reasons, a sample library needs to be further enlarged in the later stage, the extraction method of the feature quantity is optimized, the data source is more accurate, the defect features can be deeply analyzed, more feature quantities are extracted for calculation, and the recognition rate is improved.
Further, the method also comprises a model optimization process, wherein the model optimization process specifically comprises the following steps:
a11, grouping the texture features of the training images according to preset grouping conditions to generate a first grouping of texture features of the training images and a second grouping of texture features of the training images;
a12, acquiring a first weight ratio value associated with texture features of the first grouping training image;
a13, obtaining a second weight ratio value associated with texture features of the second grouping training image;
a14, adjusting the first weight ratio by adopting a preset first adjustment value, and calculating first error rates related to all the first weight ratio values through traversal;
and A15, screening out a first target error rate meeting a preset screening condition from the plurality of first error rates, and feeding back an optimized target neural network identification model according to a first weight ratio related to the first target error rate, wherein the preset screening condition is that the first error rate is smaller than or equal to a preset error rate tolerance value and the first error rate is the minimum value.
In the embodiment of the invention, the error rate of the neural network recognition result is judged, if the error rate is larger than the set error rate tolerance value, feedback optimization is carried out on the target neural network recognition model, the overall weight ratio of average brightness, average contrast, symmetry of a gray histogram and gray randomness is set as a first weight ratio alpha, and the weight ratio of chromatic aberration is set as a second weight ratio beta, alpha+beta=1; and adjusting the alpha value by a preset first adjustment value of 0.05, traversing the first error rates of all the alpha values, and taking the target neural network recognition model corresponding to a certain alpha value as an optimized target neural network recognition model if the first error rate of the target neural network recognition model corresponding to the alpha value is minimum and the first error rate is smaller than or equal to a set error rate tolerance value.
Further, the method further comprises the following steps:
a16, when any first error rate does not meet the preset screening condition, adopting a preset second adjustment value to adjust the first weight ratio, and calculating second error rates related to all the first weight ratio by traversing;
and A17, screening out second error rates meeting preset screening conditions from the plurality of second error rates, and feeding back and optimizing the target neural network identification model according to the first weight ratio related to the second error rates.
In the embodiment of the invention, if the error rate of the neural network identification algorithm corresponding to a certain alpha value is minimum and the error rate is larger than a set error rate tolerance value, performing secondary optimization;
the method comprises the following steps: setting the weights corresponding to the average brightness, the average contrast, the symmetry of the gray level histogram and the gray level randomness as alpha 1, alpha 2, alpha 3 and alpha 4, wherein alpha 1+alpha 2+alpha 3+alpha 4 = alpha;
obtaining all the combinations of alpha 1+ alpha 2+ alpha 3+ alpha 4 = alpha with a preset second adjustment value 0.01, and calculating a second error rate of the target neural network recognition model corresponding to the combinations;
if the second error rate of the target neural network identification model corresponding to a certain combination is minimum and the second error rate is smaller than or equal to the set error rate tolerance value, the target neural network identification model corresponding to the combination is used as the optimized target neural network identification model.
And 206, when a sampling image of the cable intermediate joint to be identified is received, performing image preprocessing on the sampling image, and outputting a corresponding defect identification result through the target neural network identification model.
In the embodiment of the present invention, when a sampled image of a cable intermediate connector to be identified is received, image preprocessing is performed on the sampled image, where the image preprocessing is consistent with the processes of steps 201 to 204, and is not repeated herein, a sampled image texture feature associated with the sampled image is generated, and the sampled image texture feature is input into a target neural network identification model for identification, so as to output a corresponding defect identification result.
When a training image set is received, image preprocessing is carried out on the training image set to generate corresponding training image texture features, the training image texture features are input into a preset initial neural network recognition model to carry out training to generate a target neural network recognition model, when a sampling image of a cable intermediate joint to be recognized is received, image preprocessing is carried out on the sampling image, and a corresponding defect recognition result is output through the target neural network recognition model; the defect identification method solves the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well distinguished, the defects are required to be confirmed manually, the defects cannot be identified or the errors are identified easily, and the operation reliability of the distribution network cable is reduced; the invention uses the projection method to perform primary positioning of the main insulation, extracts the main insulation area, identifies the type of the defect through the neural network identification algorithm, and has accurate identification result.
Referring to fig. 21, fig. 21 is a block diagram illustrating a defect identifying system for a main insulation portion of a cable according to a third embodiment of the present invention.
The invention provides a defect identification system for a main insulation part of a cable, which comprises the following components:
the image preprocessing module 301 is configured to perform image preprocessing on the training image set when receiving the training image set, and generate corresponding training image texture features;
the target neural network recognition model module 302 is configured to input a preset initial neural network recognition model to train by using texture features of a training image, so as to generate a target neural network recognition model;
and the defect identification result output module 303 is configured to, when receiving a sampling image of the cable intermediate connector to be identified, perform image preprocessing on the sampling image, and output a corresponding defect identification result through the target neural network identification model.
Further, the image preprocessing module 301 includes:
the training image receiving sub-module is used for establishing a corresponding RGB color space component square distribution diagram for the training image set when the training image set is received, wherein the training image comprises a cable intermediate joint image and an associated main insulation part image;
an RGB component value sub-module for extracting corresponding RGB component values from the RGB color space component square distribution map associated with the main insulation part image;
The primary image sub-module is used for determining a target area where the main insulation part is located based on the RGB component value, and taking an image of the target area as a primary image, wherein the primary image is an image only containing the main insulation part;
and the training image texture feature sub-module is used for carrying out image processing on the preliminary image based on a preset image processing strategy to generate corresponding training image texture features.
Further, the preset image processing strategy includes a first image processing strategy and a second image processing strategy, and the training image texture feature submodule includes:
the first image texture feature unit is used for performing image processing on the preliminary image by adopting a first image processing strategy to generate corresponding first image texture features, wherein the first image texture features comprise first average brightness, first average contrast, symmetry of a first gray histogram, first gray randomness and first chromatic aberration;
the second image texture feature unit is used for performing image processing on the preliminary image by adopting a second image processing strategy to generate corresponding second image texture features, wherein the second image texture features comprise second average brightness, second average contrast, symmetry of a second gray histogram, second gray randomness and second chromatic aberration;
And the integration unit is used for taking the first image texture feature and the second image texture feature as training image texture features.
Further, the first image texture feature unit includes:
the first cutting image subunit is used for cutting the preliminary image by adopting a maximum inter-class variance method to generate a first cutting image;
a vertical direction projection image subunit, configured to obtain a vertical direction projection image of the first cut image in the three components of the horizontal direction RGB color space according to a projection method;
a first preprocessing image subunit, configured to extract an image between two inflection points in the vertical projection image as a first preprocessing image;
and the first extraction subunit is used for extracting the first image texture features associated with the first preprocessing image.
Further, the second image texture feature unit includes:
the second cutting image subunit is used for cutting the preliminary image by adopting a maximum inter-class variance method to generate a second cutting image;
the binary image subunit is used for carrying out edge detection on the second cut image by adopting a Prewitt operator to generate a binary image;
the second preprocessing image subunit is used for performing multiplication operation on the binary image and the second cutting image to generate a second preprocessing image;
And the second extraction subunit is used for extracting the second image texture features associated with the second preprocessed image.
Further, the target neural network recognition model module 302 includes:
the training result submodule is used for inputting a preset initial neural network identification model to train by adopting texture features of training images and outputting a corresponding training result;
the error value submodule is used for calculating an error value between the training result and a preset standard training result;
and the model output sub-module is used for outputting a target neural network identification model when the error value is smaller than a preset standard error value.
Further, the method also comprises a model optimization process, wherein the model optimization process specifically comprises the following steps:
the grouping sub-module is used for grouping the texture features of the training images according to preset grouping conditions to generate a first grouping training image texture feature and a second grouping training image texture feature;
the first weight ratio submodule is used for acquiring a first weight ratio related to the texture characteristics of the first grouping training image;
the second weight ratio submodule is used for acquiring a second weight ratio related to the texture characteristics of the second packet training image;
the first error rate sub-module is used for adjusting the first weight ratio by adopting a preset first adjustment value, and calculating the first error rate associated with all the first weight ratio by traversing;
And the first feedback optimization sub-module is used for screening out a first target error rate meeting a preset screening condition from the plurality of first error rates, and feeding back and optimizing a target neural network identification model according to a first weight ratio related to the first target error rate, wherein the preset screening condition is that the first error rate is smaller than or equal to a preset error rate tolerance value and the first error rate is the minimum value.
Further, the method further comprises the following steps:
the second error rate sub-module is used for adjusting the first weight ratio by adopting a preset second adjustment value when any first error rate does not meet the preset screening condition, and calculating the second error rate associated with all the first weight ratio by traversal;
and the second feedback optimization sub-module is used for screening out second error rates meeting preset screening conditions from the plurality of second error rates, and feeding back and optimizing the target neural network identification model according to the first weight ratio related to the second error rates.
When a training image set is received, image preprocessing is carried out on the training image set to generate corresponding training image texture features, the training image texture features are input into a preset initial neural network recognition model to carry out training to generate a target neural network recognition model, when a sampling image of a cable intermediate joint to be recognized is received, image preprocessing is carried out on the sampling image, and a corresponding defect recognition result is output through the target neural network recognition model; the defect identification method solves the technical problems that the existing defects are low in identification efficiency, the types of the defects cannot be well distinguished, the defects are required to be confirmed manually, the defects cannot be identified or the errors are identified easily, and the operation reliability of the distribution network cable is reduced; the invention uses the projection method to perform primary positioning of the main insulation, extracts the main insulation area, identifies the type of the defect through the neural network identification algorithm, and has accurate identification result.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for identifying defects in a main insulation portion of a cable, comprising:
when a training image set is received, carrying out image preprocessing on the training image set to generate corresponding training image texture features;
inputting the texture features of the training image into a preset initial neural network recognition model for training, and generating a target neural network recognition model;
when a sampling image of the cable middle joint to be identified is received, the sampling image is subjected to image preprocessing, and a corresponding defect identification result is output through the target neural network identification model.
2. The method for identifying defects of a main insulation portion of a cable according to claim 1, wherein the step of performing image preprocessing on a training image set to generate corresponding training image texture features when the training image set is received comprises:
when a training image set is received, a corresponding RGB color space component square distribution diagram is established for the training image set, wherein the training image comprises a cable intermediate joint image and an associated main insulation part image;
extracting corresponding RGB component values from the RGB color space component square distribution map associated with the main insulation part image;
Determining a target area where the main insulation part is located based on the RGB component value, and taking an image of the target area as a preliminary image, wherein the preliminary image is an image only containing the main insulation part;
and performing image processing on the preliminary image based on a preset image processing strategy to generate corresponding training image texture features.
3. The method for identifying defects of a main insulation portion of a cable according to claim 2, wherein the preset image processing strategies include a first image processing strategy and a second image processing strategy, and the step of performing image processing on the preliminary image based on the preset image processing strategies to generate corresponding training image texture features includes:
performing image processing on the preliminary image by adopting the first image processing strategy to generate corresponding first image texture features, wherein the first image texture features comprise first average brightness, first average contrast, symmetry of a first gray histogram, first gray randomness and first chromatic aberration;
performing image processing on the preliminary image by adopting the second image processing strategy to generate corresponding second image texture features, wherein the second image texture features comprise second average brightness, second average contrast, symmetry of a second gray histogram, second gray randomness and second chromatic aberration;
And taking the first image texture feature and the second image texture feature as training image texture features.
4. A method of identifying defects in a main insulating portion of a cable according to claim 3, wherein the step of image processing the preliminary image using the first image processing strategy to generate corresponding first image texture features comprises:
cutting the preliminary image by adopting a maximum inter-class variance method to generate a first cutting image;
obtaining a vertical direction projection image of the first cutting image in the three components of the RGB color space in the horizontal direction according to a projection method;
extracting an image between two inflection points in the vertical direction projection image to serve as a first preprocessing image;
and extracting a first image texture feature associated with the first preprocessing image.
5. A method of identifying defects in a main insulating portion of a cable according to claim 3, wherein the step of image processing the preliminary image using the second image processing strategy to generate corresponding second image texture features comprises:
cutting the preliminary image by adopting a maximum inter-class variance method to generate a second cut image;
Performing edge detection on the second cut image by adopting a Prewitt operator to generate a binary image;
multiplying the binary image and the second cutting image to generate a second preprocessing image;
and extracting a second image texture feature associated with the second preprocessing image.
6. The method for identifying defects of a cable main insulation part according to claim 3, wherein the step of training by inputting the training image texture features into a preset initial neural network identification model to generate a target neural network identification model comprises the following steps:
inputting the texture features of the training image into a preset initial neural network recognition model for training, and outputting a corresponding training result;
calculating an error value between the training result and a preset standard training result;
and outputting the target neural network identification model when the error value is smaller than a preset standard error value.
7. The method for identifying defects in a main insulating portion of a cable according to claim 6, further comprising a model optimization process, said model optimization process comprising in particular:
grouping the training image texture features according to a preset grouping condition to generate a first grouping training image texture feature and a second grouping training image texture feature;
Acquiring a first weight ratio value associated with the texture features of the first grouping training image;
acquiring a second weight ratio value associated with the texture features of the second packet training image;
adopting a preset first adjustment value to adjust the first weight ratio, and calculating a first error rate associated with all the first weight ratio values through traversal;
and screening a first target error rate meeting a preset screening condition from the plurality of first error rates, and feedback optimizing the target neural network identification model according to the first weight ratio related to the first target error rate, wherein the preset screening condition is that the first error rate is smaller than or equal to a preset error rate tolerance value and the first error rate is the minimum value.
8. The method for identifying defects in a main insulating portion of a cable according to claim 7, further comprising:
when any one of the first error rates does not meet the preset screening conditions, a preset second adjustment value is adopted to adjust the first weight ratio, and the second error rates related to all the first weight ratio values are calculated through traversal;
and screening the second error rates meeting preset screening conditions from the plurality of second error rates, and feeding back and optimizing the target neural network identification model according to the first weight ratio related to the second error rates.
9. A system for identifying defects in a primary insulation of a cable, comprising:
the image preprocessing module is used for carrying out image preprocessing on the training image set when the training image set is received, and generating corresponding training image texture features;
the target neural network recognition model module is used for inputting a preset initial neural network recognition model to train by adopting the texture characteristics of the training image to generate a target neural network recognition model;
and the defect identification result output module is used for carrying out image preprocessing on the sampling image when receiving the sampling image of the cable intermediate joint to be identified, and outputting a corresponding defect identification result through the target neural network identification model.
10. The system for identifying defects in a main insulation portion of a cable according to claim 9, wherein the image preprocessing module comprises:
the training image receiving sub-module is used for establishing a corresponding RGB color space component square distribution diagram for a training image set when the training image set is received, wherein the training image comprises a cable intermediate joint image and an associated main insulation part image;
an RGB component value sub-module, configured to extract a corresponding RGB component value from the RGB color space component square distribution map associated with the main insulation part image;
A preliminary image sub-module, configured to determine a target area where the main insulation part is located based on the RGB component values, and take an image of the target area as a preliminary image, where the preliminary image is an image that only includes the main insulation part;
and the training image texture feature sub-module is used for carrying out image processing on the preliminary image based on a preset image processing strategy to generate corresponding training image texture features.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311520856.7A CN117437482A (en) | 2023-11-15 | 2023-11-15 | Cable main insulation part defect identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311520856.7A CN117437482A (en) | 2023-11-15 | 2023-11-15 | Cable main insulation part defect identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117437482A true CN117437482A (en) | 2024-01-23 |
Family
ID=89551345
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311520856.7A Pending CN117437482A (en) | 2023-11-15 | 2023-11-15 | Cable main insulation part defect identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117437482A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117934158A (en) * | 2024-03-13 | 2024-04-26 | 湖南三湘银行股份有限公司 | Credit data automatic examination method based on RPA |
-
2023
- 2023-11-15 CN CN202311520856.7A patent/CN117437482A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117934158A (en) * | 2024-03-13 | 2024-04-26 | 湖南三湘银行股份有限公司 | Credit data automatic examination method based on RPA |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113689428B (en) | Mechanical part stress corrosion detection method and system based on image processing | |
| CN116309537B (en) | Defect detection method for oil stain on surface of tab die | |
| CN110008680B (en) | Verification code generation system and method based on countermeasure sample | |
| EP3343440A1 (en) | Identifying and excluding blurred areas of images of stained tissue to improve cancer scoring | |
| CN117437482A (en) | Cable main insulation part defect identification method and system | |
| US20190220981A1 (en) | Systems and methods for adaptive histopathology image unmixing | |
| CN115294113A (en) | Wood veneer quality detection method | |
| CN115511889A (en) | Method for detecting welding defects on surface of solar cell panel bracket | |
| CN114882265B (en) | Method for detecting quality of finished oil | |
| US20230222643A1 (en) | Semantic deep learning and rule optimization for surface corrosion detection and evaluation | |
| US20190272627A1 (en) | Automatically generating image datasets for use in image recognition and detection | |
| CN114782329A (en) | Bearing defect damage degree evaluation method and system based on image processing | |
| CN116703911B (en) | LED lamp production quality detecting system | |
| CN110913210A (en) | Automatic main viewing image quality evaluation method based on standard data acquisition | |
| CN116612098B (en) | Insulator RTV spraying quality evaluation method and device based on image processing | |
| CN113109240A (en) | Method and system for determining imperfect grains of grains implemented by computer | |
| CN116452581B (en) | Intelligent voltage source state detection system and method based on machine vision | |
| CN117576121A (en) | Automatic segmentation method, system, equipment and medium for microscope scanning area | |
| CN117173130A (en) | Pipeline detection method, system, terminal equipment and storage medium | |
| US20050069200A1 (en) | Systems and methods for illuminant model estimation | |
| CN111310899A (en) | Electric power defect identification method based on symbiotic relationship and small sample learning | |
| CN116778357A (en) | Power line unmanned aerial vehicle inspection method and system utilizing visible light defect identification | |
| CN112508946B (en) | Cable tunnel anomaly detection method based on antagonistic neural network | |
| CN116051497A (en) | Intelligent analysis method for power transmission and transformation images of power grid based on data processing | |
| CN115310505A (en) | Automatic identification method and system for secondary circuit wiring terminal of mutual inductor |
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 |