CN117911761A - YOLOV 5-based improved insulator defect detection method and system - Google Patents
YOLOV 5-based improved insulator defect detection method and system Download PDFInfo
- Publication number
- CN117911761A CN117911761A CN202410021297.3A CN202410021297A CN117911761A CN 117911761 A CN117911761 A CN 117911761A CN 202410021297 A CN202410021297 A CN 202410021297A CN 117911761 A CN117911761 A CN 117911761A
- Authority
- CN
- China
- Prior art keywords
- insulator
- image
- yolov
- feature
- defects
- 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
- 239000012212 insulator Substances 0.000 title claims abstract description 133
- 238000001514 detection method Methods 0.000 title claims abstract description 102
- 230000007547 defect Effects 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 claims abstract description 28
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 230000000007 visual effect Effects 0.000 claims abstract description 12
- 230000006872 improvement Effects 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 20
- 230000007246 mechanism Effects 0.000 claims description 9
- 230000010339 dilation Effects 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 239000003086 colorant Substances 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000009413 insulation Methods 0.000 abstract description 10
- 230000009466 transformation Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000001931 thermography Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- 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
-
- 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
- G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
-
- 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Computer Graphics (AREA)
- Computational Linguistics (AREA)
- Geometry (AREA)
- Signal Processing (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides a YOLOV-based improved insulator defect detection method and a YOLOV-based improved insulator defect detection system, wherein the method comprises the following steps: acquiring the data of the electric power system insulation sub-image, and preprocessing the data of the insulation sub-image; performing target detection on the preprocessed insulator image data by using an improved YOLOV model, and extracting the defect position and characteristic information of the insulator; classifying the detected defects of the insulator, and evaluating the type, severity and residual life of the defects of the insulator; marking the position of the insulator defect by means of a visual three-dimensional image, and displaying the severity and the residual life by means of text information description. Compared with the prior art, the technical scheme of the invention effectively improves the accuracy, the robustness and the efficiency of insulator defect detection by adopting advanced detection frames, improved data enhancement and other methods, and more accurately detects the defects, thereby improving the safety of the power equipment.
Description
Technical Field
The invention relates to the technical field of insulators, in particular to an improved insulator defect detection method and system based on YOLOV.
Background
Insulator defect detection is an important task in maintenance and operation of an electric power system, and aims to discover defects or damages on the surface of an insulator early and take appropriate measures for repair or replacement so as to ensure the reliability and safety of the insulator.
At present, common insulator defect detection includes the following methods:
First, visual inspection. The surface of the insulator is directly observed by personnel to find possible defects such as cracks, damages or dirt. The method is simple and easy to implement, but limited by human eye vision, lacks objectivity and consistency, and can limit the accuracy and efficiency of detection
And secondly, an infrared thermal imaging method. The thermal infrared imager is used for scanning the insulator, and possible defects are detected by measuring the heat distribution of the surface of the insulator. When there is local damage or corrosion on the surface of the insulator, different heat distribution is caused, so that abnormal areas are displayed in the infrared image. The method can be operated remotely and scanned rapidly, but has certain requirements on factors such as the heat conductivity of the insulator material, the ambient temperature and the like.
Thirdly, ultrasonic detection. By transmitting ultrasonic pulses to the insulator surface and then receiving the reflected signals, the internal structure and possible imperfections of the insulator are analyzed using the propagation velocity of the ultrasonic waves and the characteristics of the reflected signals. The method can detect the problems of cracks, bubbles, looseness and the like in the insulator, but needs more specialized equipment and operators.
And fourthly, a non-contact surge method. The defects existing on the surface of the insulator are excited by high-voltage electricity, and then the state of the insulator is analyzed by detecting the characteristics of surge reflected signals. Different types of defects may cause different forms of surge reflected waves so that the type and location of the defect may be identified. The method has better sensitivity to defects such as underbending, air bubbles, moisture and the like of the insulator.
In summary, the conventional insulator defect detection method generally needs to be performed on site, requires personnel to go up to stand or use special tools, is inconvenient to operate and has a certain safety risk, and the non-manual detection method needs to be improved in terms of detection precision and coverage range, so that tiny cracks or local damages may be ignored or misjudged. For insulators of complex structure, the accuracy of the detection may be limited, and more sensitive, high resolution detection methods have to be developed to cover a wider range of defect types and sizes.
Disclosure of Invention
In view of this, the invention proposes an insulator defect detection method and system based on YOLOV improvement, which effectively improves the accuracy, robustness and efficiency of insulator defect detection by adopting advanced detection frames, improved data enhancement and other methods, and more accurately detects defects, thereby improving the safety of power equipment.
The embodiment of the invention provides an insulator defect detection method based on YOLOV improvement, which comprises the following steps:
Acquiring the insulator image data of the power system, and preprocessing the insulator image data;
Performing target detection on the preprocessed insulator image data by using an improved YOLOV model, and extracting the defect position and characteristic information of the insulator;
Classifying the detected defects of the insulator, and evaluating the type, severity and residual life of the defects of the insulator;
marking the position of the insulator defect by means of a visual three-dimensional image, and displaying the severity and the residual life by means of text information description.
Illustratively, the insulator image data originates from an unmanned aerial vehicle aerial or artificial shooting, and the preprocessing the insulator image data includes:
carrying out data enhancement on the insulator image data, and randomly adding background, blurring and illumination with different angles on the basis of the insulator image data to prevent data from being over-fitted, wherein the data enhancement comprises brightness adjustment, contrast enhancement and histogram equalization;
and performing size standardization and noise removal on the data-enhanced insulator image data to obtain an insulator sample set.
Illustratively, the improved YOLOV model adopts a lightweight neural network architecture, and the target detection of the preprocessed insulator image data by using the improved YOLOV model includes:
Constructing a lightweight YOLOV model, inputting the insulator sample set to an input layer of the lightweight YOLOV model, and reducing the image dimension by adopting a slicing operation method;
Carrying out image feature extraction on the insulator sample set subjected to dimension reduction by adopting phantom convolution to obtain a first image feature and a second image feature;
And obtaining comprehensive characteristics by linearly combining the first image characteristics and the second image characteristics, and evaluating the target detection accuracy through the performance detection indexes.
Illustratively, the input layer of the lightweight YOLOV model includes a first channel, a second channel, and a third channel, where the first channel, the second channel, and the third channel respectively correspond to three primary colors of red, green, and blue, and the size of the input layer is 416×416 or 640×640.
Illustratively, the performing image feature extraction on the dimension-reduced insulator sample set by using phantom convolution to obtain a first image feature and a second image feature includes:
Dividing the insulator sample set into a first branch and a second branch;
performing low-rank standard convolution calculation on the first branch to obtain the first image feature;
And carrying out depth separable expansion convolution calculation on the second branch to obtain the second image feature, wherein the depth separable expansion convolution calculation combines the depth separable convolution and the expansion convolution into a single step.
Illustratively, said performing a depth separable dilation convolution calculation on said second branch to obtain said second image feature comprises:
before depth separable expansion convolution calculation, a quick region recommendation mechanism is introduced to identify a high-probability target region, the quick region recommendation mechanism is a candidate frame generator, and depth separable expansion convolution calculation is carried out on an identified result;
An attention module is added to each unit of the depth separable dilation convolution to adaptively adjust the channel weights of the feature map by employing a SENet mechanism.
Illustratively, the obtaining the integrated feature by linearly combining the first image feature and the second image feature, and the evaluating the target detection accuracy by the performance detection index includes:
the first image feature and the second image feature are linearly combined by adopting a mean value normalization method, a pixel point average value is calculated for each scale feature image, a weight is set as the ratio of the scale feature image pixel average value to the feature image pixel average value, and the calculated weight is used for linear combination;
and calculating the performance detection index by utilizing the comprehensive characteristics, judging the accuracy of target detection according to the size of the performance detection index, and multiplying the accuracy by a corresponding improvement factor to improve the importance of the target.
Illustratively, the integrated feature calculation is performed using the following formula:
Wherein F is the comprehensive feature, P i is the ith feature map, w i is the weight of the ith feature map P i, and n is the number of feature maps.
Illustratively, calculating the performance detection index by using the comprehensive features, determining the accuracy of target detection according to the size of the performance detection index, and multiplying the accuracy by the corresponding improvement factor to increase the importance of the target includes:
Calculating an intersection area of the prediction frame and the real frame and a union area of the prediction frame and the real frame, wherein the intersection area is an area of an overlapping part of the prediction frame and the real frame, and the union area is a sum of areas of all areas of the prediction frame and the real frame minus the intersection area;
calculating the intersection area divided by the union area to obtain the performance detection index;
judging whether the performance detection index is in a preset interval or not;
if yes, the performance detection index value is improved by multiplying the improvement factor, and if not, the original performance detection index value is kept unchanged.
Another embodiment of the present invention proposes an insulator defect detection system based on YOLOV improvement, comprising:
The acquisition unit is used for acquiring the insulator image data of the power system and preprocessing the insulator image data;
The target detection unit is used for carrying out target detection on the preprocessed insulator image data by utilizing the improved YOLOV model and extracting the defect position and characteristic information of the insulator;
the classification unit is used for classifying the detected insulator defects and evaluating the types, severity and residual life of the insulator defects;
The visual display unit is used for marking the positions of the defects of the insulators in a visual three-dimensional image mode and displaying the severity and the residual service life in a text information description mode.
The invention provides a YOLOV-based improved insulator defect detection method and a YOLOV-based improved insulator defect detection system, wherein the method comprises the following steps: acquiring the data of the electric power system insulation sub-image, and preprocessing the data of the insulation sub-image; performing target detection on the preprocessed insulator image data by using an improved YOLOV model, and extracting the defect position and characteristic information of the insulator; classifying the detected defects of the insulator, and evaluating the type, severity and residual life of the defects of the insulator; marking the position of the insulator defect by means of a visual three-dimensional image, and displaying the severity and the residual life by means of text information description. Compared with the prior art, the technical scheme of the invention effectively improves the accuracy, the robustness and the efficiency of insulator defect detection by adopting advanced detection frames, improved data enhancement and other methods, and more accurately detects the defects, thereby improving the safety of the power equipment.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are required for the embodiments will be briefly described, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope of the present invention. Like elements are numbered alike in the various figures.
FIG. 1 is a schematic flow chart of an insulator defect detection method based on YOLOV improvement according to an embodiment of the present invention;
Fig. 2 is a schematic flow chart of a method in step S102 according to an embodiment of the present invention;
fig. 3 is a schematic flow chart of a method in step S202 according to an embodiment of the present invention;
Fig. 4 is a flowchart of a method in step S203 according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments.
Example 1
Referring to fig. 1, the embodiment proposes an improved insulator defect detection method based on YOLOV, which includes:
Step S101, acquiring the data of the electric power system insulation sub-image, and preprocessing the data of the insulation sub-image;
Here, the insulator image data originates from unmanned aerial vehicle aerial photographing or artificial photographing. Preprocessing the insulator image data mainly comprises data enhancement of the insulator image data, wherein background, blurring and illumination with different angles are randomly added on the basis of the insulator image data to prevent data from being over-fitted, and the data enhancement comprises brightness adjustment, contrast enhancement, histogram equalization and the like. Meanwhile, size standardization and noise removal are required to be carried out on the data of the data-enhanced insulator image data to obtain the insulator sample set.
Step S102, performing target detection on the preprocessed insulator image data by using an improved YOLOV model, and extracting the defect position and characteristic information of an insulator;
specifically, the improved YOLOV model of the present embodiment employs a lightweight neural network architecture.
Step S103, classifying the detected insulator defects, and evaluating the type, severity and residual life of the insulator defects;
and step S104, marking the positions of the defects of the insulators in a visual three-dimensional image mode, and displaying the severity and the residual service life in a text information description mode.
Referring to fig. 2, step S102 includes:
step S201, a lightweight YOLOV model is built, an insulator sample set is input to an input layer of the lightweight YOLOV model, and the image dimension is reduced by adopting a slicing operation method;
Specifically, the input layer of the lightweight YOLOV model provided by the embodiment of the invention comprises a first channel, a second channel and a third channel, and the input layer can accept image data with different channel numbers. The input image typically comprises 3 channels corresponding to three colors RGB (red, green, blue), wherein the dimension of each channel may be 416×416×3 or 640×640×3. The dimensions of the image are reduced by selectively retaining part of the information of the image or reducing the size of the image, which may be by reducing the size of the image according to a fixed scale or extracting the channels of interest from the color image.
Step S202, performing image feature extraction on the insulator sample set subjected to dimension reduction by adopting phantom convolution to obtain a first image feature and a second image feature;
Specifically, the phantom convolution provided by the embodiment of the invention is realized by dividing the input feature map into two branches and applying different convolution operations on each branch, and can learn more features in different scales and directions by utilizing the phantom convolution.
Step S203, the first image feature and the second image feature are linearly combined to obtain a comprehensive feature, and the target detection accuracy is evaluated through the performance detection index.
Referring to fig. 3, step S202 includes:
step S301, dividing the insulator sample set into a first branch and a second branch;
Step S302, performing low-rank standard convolution calculation on a first branch to obtain a first image feature;
Step S303, performing a depth-separable expansion convolution calculation on the second branch to obtain a second image feature, wherein the depth-separable expansion convolution calculation combines the depth-separable convolution and the expansion convolution into a single step.
Specifically, before depth separable expansion convolution calculation, a quick region recommendation mechanism is introduced to identify a high-probability target region, the quick region recommendation mechanism is a candidate frame generator, and depth separable expansion convolution calculation is performed on an identified result. Here, the embodiment of the invention operates in a sliding window mode, slides windows with different sizes and proportions on the whole image, generates candidate frames by classifying or regressing the areas in the windows, effectively reduces the calculated amount and the searching range, and improves the speed and the accuracy of target detection.
At the same time, an attention module is added at each element of the depth separable dilation convolution, adaptively adjusting the channel weights of the feature map by employing a SENet mechanism. Here, the embodiment of the present invention first compresses each channel of the feature map into a scalar through global pooling, that is, calculates an average value of each channel to extract the relationship between channels. Then, a weight vector is generated through two fully connected layers according to the compressed feature map, and the weight vector is used for weighting the features of each channel in the feature map. The two full connection layers respectively perform nonlinear transformation and linear transformation to learn a method of adjusting channel weights. And finally, performing element multiplication on the obtained channel weight and the original feature map to generate a self-adaptive adjusted feature map. The attention module can adaptively adjust the relative importance of the channel according to the weight of the feature map, and improves the detection capability of the network for useful targets in complex scenes. In this way, the network can better adapt to different scenes and further improve the performance of the model.
Referring to fig. 4, step S203 includes:
Step S401, carrying out linear combination on the first image feature and the second image feature by adopting a mean normalization method, calculating a pixel point average value of the feature image of each scale, setting the weight as the ratio of the pixel average value of the feature image of the scale to the pixel average value of all feature images, and carrying out linear combination by utilizing the calculated weight;
Step S402, calculating the performance detection index for the comprehensive characteristics, judging the accuracy of target detection according to the size of the performance detection index, and multiplying the accuracy by a corresponding improvement factor to improve the importance of the target.
Specifically, the embodiment of the invention performs comprehensive feature calculation by the formula (1):
wherein F is the comprehensive feature, P i is the ith feature map, w i is the weight of the ith feature map P i, and n is the number of feature maps.
It should be noted that, the performance detection index provided by the embodiment of the present invention is obtained by dividing the intersection area of the prediction frame and the real frame by the union area, where the intersection area is the area of the overlapping portion of the prediction frame and the real frame, and the union area is the sum of the areas of all the regions of the prediction frame and the real frame minus the intersection area. Generally, the value range of the performance detection index is between 0 and 1, 0 indicates no intersection, 1 indicates complete overlapping, the higher the value is, the closer the predicted result is to the true value, the better the performance of the algorithm is, when the value of the performance detection index exceeds a certain threshold, the value of the threshold is 0.5, 0.6 or 0.7, and the value of the embodiment of the invention is 0.5. Whether the performance detection index value needs to be improved by the improvement factor is judged by judging whether the performance detection index value is in a preset interval or not.
Here, since the area of the small target is relatively small, the calculation result of the performance detection index tends to be low. The embodiment of the invention provides an improvement factor for optimizing a loss function in a target detection task, balances the detection accuracy of a small target and a large target, and improves the performance of the detector on various target sizes. For each target frame, calculating the performance detection index between the prediction frame and the real frame, dividing the performance detection index into two threshold ranges such as 0.0,0.5 and 0.5,1.0 according to the size of the performance detection index, multiplying the performance detection index value of the small target falling in the range of 0.0,0.5 by an improvement factor alpha to increase the performance detection index value of the small target, so that the performance detection index of the small target is more approximate to 1 and higher importance is given to the small target, and similarly, for the large target with the performance detection index falling in the range of 0.5,1.0, the original performance detection index value is kept unchanged and no modification is carried out.
The invention provides a YOLOV-based improved insulator defect detection method and a YOLOV-based improved insulator defect detection system, wherein the method comprises the following steps: acquiring the data of the electric power system insulation sub-image, and preprocessing the data of the insulation sub-image; performing target detection on the preprocessed insulator image data by using an improved YOLOV model, and extracting the defect position and characteristic information of the insulator; classifying the detected defects of the insulator, and evaluating the type, severity and residual life of the defects of the insulator; marking the position of the insulator defect by means of a visual three-dimensional image, and displaying the severity and the residual life by means of text information description. Compared with the prior art, the technical scheme of the invention effectively improves the accuracy, the robustness and the efficiency of insulator defect detection by adopting advanced detection frames, improved data enhancement and other methods, and more accurately detects the defects, thereby improving the safety of the power equipment.
Example 2
The insulator defect detection system based on YOLOV improvement provided by the embodiment of the invention comprises:
the acquisition unit is used for acquiring the electric power system insulation sub-image data and preprocessing the insulation sub-image data;
The target detection unit is used for carrying out target detection on the preprocessed insulator image data by utilizing the improved YOLOV model and extracting the defect position and characteristic information of the insulator;
the classification unit is used for classifying the detected insulator defects and evaluating the types, severity and residual life of the insulator defects;
The visual display unit is used for marking the positions of the defects of the insulators in a visual three-dimensional image mode and displaying the severity and the residual service life in a text information description mode.
It is understood that the above-described YOLOV-based modified insulator defect detection system corresponds to the YOLOV-based modified insulator defect detection method of embodiment 1. Any of the alternatives in embodiment 1 are also applicable to this embodiment and will not be described in detail here.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention.
Claims (10)
1. An improved insulator defect detection method based on YOLOV is characterized by comprising the following steps:
Acquiring the insulator image data of the power system, and preprocessing the insulator image data;
Performing target detection on the preprocessed insulator image data by using an improved YOLOV model, and extracting the defect position and characteristic information of the insulator;
Classifying the detected defects of the insulator, and evaluating the type, severity and residual life of the defects of the insulator;
marking the position of the insulator defect by means of a visual three-dimensional image, and displaying the severity and the residual life by means of text information description.
2. The method for detecting defects of insulators based on YOLOV improvement as set forth in claim 1, wherein the insulator image data is derived from unmanned aerial vehicle or artificial photographing, and the preprocessing the insulator image data includes:
carrying out data enhancement on the insulator image data, and randomly adding background, blurring and illumination with different angles on the basis of the insulator image data to prevent data from being over-fitted, wherein the data enhancement comprises brightness adjustment, contrast enhancement and histogram equalization;
and performing size standardization and noise removal on the data-enhanced insulator image data to obtain an insulator sample set.
3. The method for detecting the defects of the insulators based on YOLOV improvement as set forth in claim 2, wherein the improved YOLOV model adopts a lightweight neural network architecture, the using the improved YOLOV model to perform target detection on the preprocessed insulator image data, and extracting the insulator defect positions and the feature information includes:
Constructing a lightweight YOLOV model, inputting the insulator sample set to an input layer of the lightweight YOLOV model, and reducing the image dimension by adopting a slicing operation method;
Carrying out image feature extraction on the insulator sample set subjected to dimension reduction by adopting phantom convolution to obtain a first image feature and a second image feature;
And obtaining comprehensive characteristics by linearly combining the first image characteristics and the second image characteristics, and evaluating the target detection accuracy through the performance detection indexes.
4. The method of claim 3, wherein the input layer of the lightweight YOLOV model comprises a first channel, a second channel, and a third channel, the first channel, the second channel, and the third channel respectively correspond to three primary colors of red, green, and blue, and the size of the input layer is 416×416 or 640×640.
5. The method for detecting an insulator defect based on YOLOV improvement as set forth in claim 3, wherein the performing image feature extraction on the reduced-dimension insulator sample set by using phantom convolution to obtain a first image feature and a second image feature includes:
Dividing the insulator sample set into a first branch and a second branch;
performing low-rank standard convolution calculation on the first branch to obtain the first image feature;
And carrying out depth separable expansion convolution calculation on the second branch to obtain the second image feature, wherein the depth separable expansion convolution calculation combines the depth separable convolution and the expansion convolution into a single step.
6. The YOLOV5 improved insulator defect detection method as claimed in claim 5, wherein said performing depth-separable dilation convolution on said second branch to obtain said second image feature comprises:
before depth separable expansion convolution calculation, a quick region recommendation mechanism is introduced to identify a high-probability target region, the quick region recommendation mechanism is a candidate frame generator, and depth separable expansion convolution calculation is carried out on an identified result;
An attention module is added to each unit of the depth separable dilation convolution to adaptively adjust the channel weights of the feature map by employing a SENet mechanism.
7. The method for detecting defects of insulators based on YOLOV improvement as defined in claim 6, wherein the obtaining the integrated feature by linearly combining the first image feature and the second image feature and evaluating the accuracy of target detection by the performance detection index includes:
the first image feature and the second image feature are linearly combined by adopting a mean value normalization method, a pixel point average value is calculated for each scale feature image, a weight is set as the ratio of the scale feature image pixel average value to the feature image pixel average value, and the calculated weight is used for linear combination;
and calculating the performance detection index by utilizing the comprehensive characteristics, judging the accuracy of target detection according to the size of the performance detection index, and multiplying the accuracy by a corresponding improvement factor to improve the importance of the target.
8. The method for detecting defects in insulators based on YOLOV improvement as defined in claim 7, wherein the integrated feature calculation is performed using the following formula:
Wherein F is the comprehensive feature, P i is the ith feature map, w i is the weight of the ith feature map P i, and n is the number of feature maps.
9. The method for detecting defects of insulators based on YOLOV improvement as set forth in claim 7, wherein said calculating the performance detection index using the integrated feature, determining the accuracy of target detection according to the size of the performance detection index, and multiplying the accuracy by the corresponding improvement factor to increase the importance of the target comprises:
Calculating an intersection area of the prediction frame and the real frame and a union area of the prediction frame and the real frame, wherein the intersection area is an area of an overlapping part of the prediction frame and the real frame, and the union area is a sum of areas of all areas of the prediction frame and the real frame minus the intersection area;
calculating the intersection area divided by the union area to obtain the performance detection index;
judging whether the performance detection index is in a preset interval or not;
if yes, the performance detection index value is improved by multiplying the improvement factor, and if not, the original performance detection index value is kept unchanged.
10. An improved insulator defect detection system based on YOLOV, comprising:
The acquisition unit is used for acquiring the insulator image data of the power system and preprocessing the insulator image data;
The target detection unit is used for carrying out target detection on the preprocessed insulator image data by utilizing the improved YOLOV model and extracting the defect position and characteristic information of the insulator;
the classification unit is used for classifying the detected insulator defects and evaluating the types, severity and residual life of the insulator defects;
The visual display unit is used for marking the positions of the defects of the insulators in a visual three-dimensional image mode and displaying the severity and the residual service life in a text information description mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410021297.3A CN117911761A (en) | 2024-01-05 | 2024-01-05 | YOLOV 5-based improved insulator defect detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410021297.3A CN117911761A (en) | 2024-01-05 | 2024-01-05 | YOLOV 5-based improved insulator defect detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117911761A true CN117911761A (en) | 2024-04-19 |
Family
ID=90686388
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410021297.3A Pending CN117911761A (en) | 2024-01-05 | 2024-01-05 | YOLOV 5-based improved insulator defect detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117911761A (en) |
-
2024
- 2024-01-05 CN CN202410021297.3A patent/CN117911761A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tsanakas et al. | Fault diagnosis of photovoltaic modules through image processing and Canny edge detection on field thermographic measurements | |
| CN111047568B (en) | Method and system for detecting and identifying steam leakage defect | |
| CN105865723B (en) | Leakage inspection non-uniform correction method and gas leak detection apparatus | |
| US20100283847A1 (en) | Visual inspection apparatus and visual inspection method | |
| JP6548157B2 (en) | Degradation diagnostic apparatus and degradation diagnostic method | |
| CN105678700A (en) | Image interpolation method and system based on prediction gradient | |
| CN109345586A (en) | Electrical equipment discharge characteristic extracting method based on ultraviolet imagery technology | |
| CN103093458A (en) | Detecting method and detecting device for key frame | |
| CN111896549A (en) | Building crack monitoring system and method based on machine learning | |
| CN112541478A (en) | Insulator string stain detection method and system based on binocular camera | |
| CN105023270A (en) | Proactive 3D stereoscopic panorama visual sensor for monitoring underground infrastructure structure | |
| CN117576014A (en) | Ceramic substrate quality detection method, system, electronic equipment and storage medium | |
| CN115546716A (en) | Binocular vision-based method for positioning fire source around power transmission line | |
| CN115393747A (en) | Photovoltaic fault detection method based on deep learning | |
| CN103200349B (en) | Scanned image color cast automatic detection method | |
| CN114022435A (en) | Method for establishing pipeline leakage identification system based on machine vision | |
| CN116580363B (en) | Photovoltaic panel hot spot identification method, storage medium and electronic equipment | |
| CN117911761A (en) | YOLOV 5-based improved insulator defect detection method and system | |
| CN117308881A (en) | Detection device and detection method for inclination angle of tunnel inspection and repair cover plate | |
| CN113689399B (en) | Remote sensing image processing method and system for power grid identification | |
| O'Byrne et al. | Effects of turbidity and lighting on the performance of an image processing based damage detection technique | |
| CN110930447A (en) | Android-based unattended snow depth measurement method | |
| CN110427911A (en) | A kind of Approach for road detection, device, equipment and storage medium | |
| CN118154600B (en) | String falling detection method and device of photovoltaic power generation system, electronic equipment and medium | |
| CN116843691B (en) | Photovoltaic panel hot spot detection method, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |