CN117809083A - Cable joint fault detection method and system based on infrared or ultraviolet images - Google Patents
Cable joint fault detection method and system based on infrared or ultraviolet images Download PDFInfo
- Publication number
- CN117809083A CN117809083A CN202311713768.9A CN202311713768A CN117809083A CN 117809083 A CN117809083 A CN 117809083A CN 202311713768 A CN202311713768 A CN 202311713768A CN 117809083 A CN117809083 A CN 117809083A
- Authority
- CN
- China
- Prior art keywords
- infrared
- convolution
- cable joint
- fault
- ultraviolet
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 34
- 230000006870 function Effects 0.000 claims abstract description 33
- 238000003745 diagnosis Methods 0.000 claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 230000000694 effects Effects 0.000 claims abstract description 12
- 230000007246 mechanism Effects 0.000 claims description 22
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 8
- 238000010438 heat treatment Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 230000006872 improvement Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000000903 blocking effect Effects 0.000 claims description 3
- 238000005520 cutting process Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 230000009191 jumping Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 238000007599 discharging Methods 0.000 claims description 2
- 230000002708 enhancing effect Effects 0.000 claims description 2
- 238000009827 uniform distribution Methods 0.000 claims description 2
- 230000007547 defect Effects 0.000 description 6
- 238000009413 insulation Methods 0.000 description 6
- 238000012544 monitoring process Methods 0.000 description 5
- 238000007689 inspection Methods 0.000 description 4
- 239000000463 material Substances 0.000 description 3
- 238000002679 ablation Methods 0.000 description 2
- 230000003321 amplification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000003199 nucleic acid amplification method Methods 0.000 description 2
- 239000002689 soil Substances 0.000 description 2
- 206010000369 Accident Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000009529 body temperature measurement Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035882 stress Effects 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
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- 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
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a cable joint fault detection method and system based on infrared or ultraviolet images, wherein the method comprises the following steps: collecting infrared or ultraviolet image data of the cable connector through an infrared or ultraviolet imager; performing enhancement based on limiting contrast self-adaptive histogram equalization on the acquired data set, marking an original image by using image marking software LabelImg, and randomly dividing a training set and a testing set according to a ratio of 3:1; inputting the training set into an improved YOLOv7 model for fault diagnosis training to obtain a trained model; and performing effect test and fault diagnosis on the infrared or ultraviolet images in the test sample library by using the trained improved YOLOv7 model, and judging the severity of the fault by comparing the fault overlapping areas of the two types of images. The invention provides the functions of quickly and accurately positioning the fault point and effectively diagnosing and identifying the cable faults and fault disturbance.
Description
Technical Field
The invention relates to the technical field of computer image processing methods, in particular to a cable joint fault detection method and system based on infrared or ultraviolet images.
Background
Because the power cable is buried underground, the quality problem of the power cable, the capacity increase of the system and the damage caused by the action of external force are generated in the long-term operation process along with the wide use of the cable, the power cable is buried in soil, and the power cable is corroded by moisture in the soil for a long time to cause the links with weak insulation, such as cable accessories, to generate defects, and finally the insulation breakdown is caused to generate faults. Partial discharge occurs during the early stage due to insulation aging, and permanent faults are finally generated due to long-time discharge. The underground power cable is generally laid underground, and has great difficulty in positioning and diagnosing faults when faults occur, and a great amount of manpower, material resources and financial resources are consumed to find fault points. In addition, the cables have direct observability as they do not have overhead lines, making the fault detection and diagnosis task of the cables more difficult and challenging. Because the fault location and fault diagnosis of the power cable are important guarantees of safe and stable operation of the power system, fault points can be rapidly and accurately located, fault reasons can be rapidly analyzed through effective diagnosis and identification of cable faults and fault disturbance, insulation defects and hidden dangers can be discovered as soon as possible, further deterioration of faults is avoided, and loss of manpower, material resources and financial resources caused by the faults is remarkably reduced.
At present, the electric power maintenance of ordinary cables in China always adopts a manual regular inspection mode and performs preventive maintenance, however, the corrosion of moisture in a cable tunnel to a cable joint and the damage of mechanical stress to the cable joint are all performed slowly, and a worker can not expect when the cable joint fails despite regular inspection of the cable in the tunnel. The main reason for the failure of the cable joint is that insulation is aged and deteriorated, so that the temperature is increased and partial discharge is caused, and even fire accidents in the cable tunnel can be caused when the temperature is severe. Therefore, how to predict and pre-warn the cable status before a cable failure is a problem that needs to be solved by those skilled in the art.
At present, the method for detecting the state of the cable joint at home and abroad comprises the following steps: direct current component method, dielectric loss tangent detection method, partial discharge monitoring method, etc. In practical application, stray current exists between the cable shielding layer and the ground, so that the measurement of the direct current component is greatly interfered, and the monitoring result is influenced; in actual operation, the capacitive current on the cable metal sheath is far greater than the resistive current, so that the obtained cable insulation dielectric loss tangent value is very small, and in actual measurement, equipment is often difficult to obtain an accurate measured value; in practical application, the partial discharge signal monitoring is influenced by factors such as sensor performance, field noise, electromagnetic interference, signal attenuation and the like, and the acquisition difficulty is high.
In order to accurately measure and analyze the defect problem of the cable joint in daily work, domestic and foreign specialists research and manufacture a plurality of detection instruments (including infrared thermometers, ultraviolet imagers and the like) for defect detection of electrical equipment, and the method has great significance for inspecting the defects of the equipment and diagnosing faults in the inspection process of inspection personnel.
The existing fault diagnosis technology integrating the artificial intelligence algorithm has the relevant learning capacity and the real-time monitoring capacity, so that the cable fault is diagnosed online. The YOLO series algorithm is one of the deep learning single-stage algorithms, has the advantages of rapid, efficient and accurate detection performance, is widely focused and applied, and can adapt to target detection tasks under complex conditions of different scales, postures, shielding and the like.
Disclosure of Invention
In view of the foregoing, it is necessary to provide a cable joint fault detection method based on infrared or ultraviolet images, which is used for monitoring the fault condition of the cable joint in real time, and by improving the YOLOv7 algorithm to detect the cable joint fault of the infrared or ultraviolet images, the fault recognition precision is improved, and the defect that the infrared temperature measurement is easily affected by the environmental humidity is overcome.
The invention provides a cable joint fault detection method based on infrared or ultraviolet images, which comprises the following steps:
(1) Collecting infrared or ultraviolet image data of the cable connector as a data set through an infrared or ultraviolet imager;
(2) Performing enhancement based on limiting contrast self-adaptive histogram equalization on the acquired data set, marking an original image by using image marking software LabelImg, and randomly dividing a training set and a testing set according to a ratio of 3:1;
(3) Inputting the training set into an improved YOLOv7 model for fault diagnosis training to obtain a trained model;
(4) And performing effect test and fault diagnosis on the infrared or ultraviolet images in the test sample library by using the trained improved YOLOv7 model, and judging the severity of the fault by comparing the fault overlapping areas of the two types of images.
Further, enhancing the acquired dataset in step (2) based on limiting contrast adaptive histogram equalization comprises:
(2.1) blocking: dividing an input image into non-overlapping sub-blocks r of equal size k Where k=0, 1, …, l=1, L is the number of sub-blocks;
(2.2) calculating the sub-Block histogram h (r) k )；
(2.3) clipping the histogram of each sub-block with a clipping threshold: the cutting rule is
Wherein h' (r) k ) A clipping histogram for each sub-block; n (N) clip Is a defined actual shear threshold; n (N) avg An average value for pixels to be reassigned to each histogram;
(2.4) pixel point reassignment: for each sub-block, using the redundant pixel reassignment in the step (2.3) until all sheared pixel points are assigned;
(2.5) carrying out gray level histogram equalization treatment on each sub-region of the image after the steps so as to change the non-uniformly distributed histogram into uniform distribution;
and (2.6) reconstructing the gray value of the pixel point by adopting a bilinear interpolation method, taking the gray value of the obtained central pixel point of each sub-block as a reference point, and calculating the gray value of each point in the final output image.
Further, the improvement of the improved YOLOv7 model to the YOLOv7 model specifically comprises:
a lightweight network MobileOne is used as a backbone network of YOLOv 7;
adding a global attention mechanism GAM at the neck of the model to acquire richer cross-channel information, and improving the feature extraction capability of the model;
and a Focal-EIoU Loss function is introduced, so that the algorithm convergence rate is increased.
Further, the core module of the lightweight network MobileOne is designed based on MobileNet v1, and the structure is basically consistent with that of MobileNet v1, except that the depth separable convolution in MobileNet is replaced by a neural network structural block, the left part of the core module forms a complete structural block of MobileOne and is composed of an upper part and a lower part, wherein the upper part is based on the depth convolution, the lower part is based on the point convolution, and act represents an activation function; the depth convolution module consists of three branches, and the leftmost branch is a 1 multiplied by 1 convolution; the middle branch is a parameterized 3 x 3 convolution, i.e., k 3 x 3 convolutions; the right part is a jumping connection comprising a BN layer; the deep convolution is essentially a group convolution, the number of groups is the same as the number of channels, where both the 1 x 1 convolution and the 3 x 3 convolution are deep convolutions; the point convolution module consists of two branches, the left branch is a parameterized 1 x 1 convolution, and consists of k 1 x 1 convolutions, and the right branch is a jump connection containing a BN layer.
Furthermore, the global attention mechanism GAM consists of a channel attention sub-module and a space attention sub-module, and the mechanism optimally designs the sub-modules on the basis of a sequential channel-space attention mechanism in the CBAM;
wherein the input features are defined by F 1 Representing a series of intermediate operations in the GAM attention mechanism as intermediate state F 2 The output state is defined as F 3 The relationship among the three is as follows:
further, the original input features F of the channel attention sub-module 1 The dimension is C x W x H, the channel attention submodule firstly carries out three-dimensional channel replacement on the channel attention submodule, and the information is stored in a W x H x C form; with a two-layer MLP, the first layer performs an encoding operation to reduce the number of channels C to C/R and the second layer performs a decoding operation to obtain a result with the same number of channels as the input feature. Finally, the weighting coefficient M is obtained by carrying out Sigmoid activation function processing on the result c The dependency of the cross-dimensional channel space can be effectively enlarged.
Further, the input features F of the spatial attention sub-module 2 With double convolution, each convolution layer uses a 7*7 convolution kernel to arrive at spatial informationThe effect of fusion; the channel attention submodule reduces according to the importance of the features and obtains new scaled features; finally, the weighting factor M for this feature s And processing by adopting an activation function Sigmoid to obtain a more accurate weight value.
Further, the introduction of the Focal-EIoU Loss function specifically includes:
the improvement of the EIoU Loss over the CIoU Loss used by original YOLOv7 is to split the Loss function into overlapping area Loss, center point distance Loss and aspect ratio Loss 3 parts, and modify a and v in CIoU Loss, C w And C h For the width and height of the smallest box containing the real box, predicted box, EIoU Loss utilizesAnddirectly calculating the true value of the width and the height of the bounding box, and solving the problem of interference optimization caused by using the aspect ratio by CIoU Loss and the problem of divergence in the training process; considering the problem that training samples are unbalanced in regression of the prediction frames, the number of high-quality anchor frames with small regression errors in an input image is far less than the number of low quality anchor frames with large errors, and samples with poor quality can generate excessive gradient to influence optimization of parameters; by integrating the EIoU Loss function and the Focal Loss function, a final Focal-EIoU Loss expression is obtained, as shown in the formula (5):
L Focal-EIoU ＝IoU γ L EIoU (5)
in which L IoU 、L dis 、L asp The overlap area penalty, center point distance penalty, and aspect ratio penalty are shown, respectively.
Further, the fault diagnosis in the step (4) includes:
diagnosing whether a cable joint to be detected has a fault or not by calculating the overlapping degree of the identified cable joint area, the heating area and the facula area; the overlapping degree refers to the overlapping rate of a cable joint target window, a heating area target window and a facula area target window generated by the YOLOv7 model, namely the overlapping of the two detection frame selection areas is the overlapping degree of the fault diagnosis accuracy rate, and for the fault diagnosis, when the overlapping degree is lower than a certain value X, the possibility of abnormal heating and abnormal discharging of the power equipment is considered to be small, and at the moment, secondary manual diagnosis is usually needed; when the overlapping degree is 0, the cable connector can be basically judged that no abnormal area exists, and secondary diagnosis is not needed at the moment; when the overlapping degree is larger than a certain value, the abnormal area of the cable joint can be judged.
The cable joint fault detection system based on the infrared or ultraviolet image comprises a processor and a memory, wherein a computer program is stored in the memory, and the cable joint fault detection method based on the infrared or ultraviolet image is realized when the computer program is executed by the processor.
Compared with the prior art, the invention has the beneficial effects that:
because the imaging effect of the infrared or ultraviolet imager is influenced by factors such as the emissivity of the shooting object material, the similarity of the background and the target, the detection distance and the like, compared with a visible light image, the infrared or ultraviolet imager has lower quality and is mainly characterized by lower contrast, poorer detail resolution and lower signal-to-noise ratio, therefore, the infrared or ultraviolet image is required to be subjected to image enhancement based on the adaptive histogram equalization of limited contrast, the definition of the object outline in the image is effectively improved, and the follow-up data set labeling and object identification accuracy are facilitated; the method has the advantages that the number of channels of the YoleOv 7 main network is set to be large, the model complexity is high, in order to reduce the model complexity, the algorithm is more suitable for completing the image detection task of the cable joint, and the lightweight network MobileOne is used as the YoleOv 7 main network to accelerate the recognition speed and meet the requirement of real-time detection; in order to reduce network information reduction and enlarge global dimension interaction characteristics, a GAM attention mechanism is added to the neck of the model to acquire richer cross-channel information, so that the characteristic extraction capability of the model is improved; in order to increase the algorithm convergence rate and improve the original algorithm performance and detection accuracy, a Focal-EIoU Loss function is adopted to replace the CIoU Loss function. The GAM attention mechanism and the Focal-EIoU Loss function further increase the accuracy of target identification, the MobileOne backbone network increases the detection speed, and the three are mutually matched, so that the comprehensive performance is superior to that of the original algorithm, and a lightweight YOLOv7 network is provided.
Drawings
FIG. 1 is a flow chart of a cable joint fault detection method based on infrared or ultraviolet images;
fig. 2 is a schematic diagram of a MobileOne neural network structure block provided by the invention;
FIG. 3 is a schematic diagram of the attention mechanism of the GAM provided by the present invention;
FIG. 4 is a schematic illustration of the channel attention provided by the present invention;
FIG. 5 is a schematic view of spatial attention provided by the present invention;
fig. 6 is a flow chart of fault diagnosis provided by the present invention.
Detailed Description
The invention is described below in connection with specific embodiments.
The embodiment of the invention provides a cable joint fault detection method based on infrared or ultraviolet images, and in combination with fig. 1, fig. 1 is a flow chart of the cable joint fault detection method based on infrared or ultraviolet images, which comprises steps (1) to (4).
(1) Collecting infrared or ultraviolet image data of the cable connector through an infrared or ultraviolet imager;
(2) Performing enhancement based on limiting contrast self-adaptive histogram equalization on the acquired data set, marking an original image by using image marking software LabelImg, and randomly dividing a training set and a testing set according to a ratio of 3:1;
the step of image enhancement based on limiting contrast adaptive histogram equalization includes:
and (3) blocking: dividing an input imageDivided into non-overlapping sub-blocks r of equal size k (k=0, 1, …, l=1, L is the number of sub-blocks);
computing a sub-block histogram h (r) k )；
Clipping the histogram of each sub-block with a clipping threshold: the cutting rule is
Wherein h' (r) k ) A clipping histogram for each sub-block; n (N) clip Is a defined actual shear threshold; n (N) avg An average value for pixels to be reassigned to each histogram;
pixel point reassignment: for each sub-block, using the redundant pixels in the step (3) to redistribute until all sheared pixel points are distributed;
equalizing the gray level histogram of each sub-area of the image after the steps, wherein even if the unevenly distributed histogram becomes evenly distributed;
reconstructing the gray value of the pixel point by adopting a bilinear interpolation method, taking the gray value of the obtained central pixel point of each sub-block as a reference point, and calculating the gray value of each point in the final output image.
In comparison with general histogram equalization, adaptive histogram equalization calculates a plurality of histograms, each corresponding to a portion of an image, and then uses them to redistribute image brightness to improve image quality, but it tends to amplify the contrast in the near-constant region of the image due to the high concentration of histograms in the near-constant region of the image, resulting in noise being amplified in the near-constant region, while limiting contrast adaptive histogram equalization limits contrast amplification, thereby reducing noise amplification problems. Therefore, the image is enhanced by using the self-adaptive histogram equalization of limited contrast, the definition of the outline of the object in the image can be improved, and the data set labeling is convenient.
(3) Inputting the training set into an improved YOLOv7 model for fault diagnosis training to obtain a trained model;
(4) And performing effect test and fault diagnosis on the infrared or ultraviolet images in the test sample library by using the trained improved YOLOv7 model, and judging the severity of the fault by comparing the fault overlapping areas of the two types of images.
The improvement points of the YOLOv7 model in the steps (3) and (4) comprise:
a lightweight network MobileOne is used as a backbone network of YOLOv 7;
adding a GAM attention mechanism to the neck of the model to acquire richer cross-channel information, and improving the feature extraction capability of the model;
and a Focal-EIoU Loss function is introduced, so that the algorithm convergence rate is increased.
As seen in connection with fig. 2, the lightweight network MobileOne includes:
the core module of the MobileOne is designed based on the MobileNet V1, the structure is basically consistent with the MobileNet V1, the difference is that the depth separable convolution in the MobileNet is replaced by a neural network structural block, the left part of the core module forms a complete structural block of the MobileOne and consists of an upper part and a lower part, wherein the upper part is based on the depth convolution, the lower part is based on the point convolution, and act represents an activation function; the depth convolution module consists of three branches, and the leftmost branch is a 1 multiplied by 1 convolution; the middle branch is a parameterized 3 x 3 convolution, i.e., k 3 x 3 convolutions; the right part is a jumping connection comprising a BN layer; the deep convolution is essentially a group convolution, the number of groups is the same as the number of channels, where both the 1 x 1 convolution and the 3 x 3 convolution are deep convolutions; the point convolution module consists of two branches, the left branch is a parameterized 1 x 1 convolution, and consists of k 1 x 1 convolutions, and the right branch is a jump connection containing a BN layer.
As seen in connection with fig. 3, the global attention mechanism GAM includes:
global attention mechanisms (Global Attention Mechanism, GAM) may function to reduce network information curtailment and to magnify global dimensional interaction features. The global attention mechanism is composed of a channel attention sub-module and a spatial attention sub-module. The mechanism optimally designs the submodules of the mechanism on the basis of a sequential channel-space attention mechanism in the CBAM;
wherein the input features are defined by F 1 Representing a series of intermediate operations in the GAM attention mechanism as intermediate state F 2 The output state is defined as F 3 The relationship among the three is as follows:
as seen in connection with fig. 4, the channel attention submodule includes:
original input feature F 1 The dimension is C x W x H, the channel attention submodule firstly carries out three-dimensional channel replacement on the channel attention submodule, and the information is stored in a W x H x C form; with a two-layer MLP, the first layer performs an encoding operation to reduce the number of channels C to C/R and the second layer performs a decoding operation to obtain a result with the same number of channels as the input feature. Finally, the weighting coefficient M is obtained by carrying out Sigmoid activation function processing on the result c The dependency of the cross-dimensional channel space can be effectively enlarged.
As seen in connection with fig. 5, the spatial attention sub-module includes:
input feature F 2 Double convolution is adopted, and each convolution layer uses a 7*7 convolution kernel to achieve the effect of spatial information fusion. The channel attention submodule reduces according to the importance of the features and obtains new scaled features. Finally, the weighting factor M for this feature s And processing by adopting an activation function Sigmoid to obtain a more accurate weight value.
The Focal-EIoU Loss function includes:
original YOLOv7 uses CIoU as a coordinate loss function, which takes into account the 3 geometric factors of overlap area, center point distance, and aspect ratio. However, in the process of object recognition and detection, the bounding box is one of the important factors determining its accuracy, and the Loss function of the original YOLOv7 algorithm is not considered, so the Focal-EIoU Loss function is introduced instead of the original Loss function.
The improvement of EIoU Loss over CIoU Loss is to split the Loss function into overlapping area Loss, center point distance Loss and aspect ratio Loss 3 parts, and modify a and v in CIoU Loss, C w And C h For the width and height of the smallest frame that can contain the real frame, the predicted frame, EIoU Loss utilizationAnd->The method directly calculates the true value of the width and the height of the bounding box, so that the problem of interference optimization caused by the aspect ratio of the CIoU Loss is solved, and the problem of divergence in the training process is also solved. In addition, considering the problem that training samples are unbalanced in regression of a prediction frame, namely the number of high-quality anchor frames with small regression errors in an input image is far less than the number of low-quality anchor frames with large errors, samples with poor quality can generate excessive gradient to influence optimization of parameters, and the problem of Focal Loss for solving unbalance of positive and negative samples is also solved, and larger gradient optimization is arranged at the place with large deviation so as to facilitate detection of difficult samples and reduce influence of samples with poor quality on algorithm performance. By integrating the EIoU Loss function and the Focal Loss function, a final Focal-EIoU Loss expression is obtained, as shown in the formula (5):
L Focal-EIoU ＝IoU γ L EIoU (5)
in which L IoU 、L dis 、L asp The overlap area penalty, center point distance penalty, and aspect ratio penalty are shown, respectively.
The Focal-EIoU Loss function not only contains beneficial characteristics of the CIoU, but also focuses on a high-quality bounding box, improves model detection precision, and can accelerate model convergence.
As seen in connection with fig. 6, the fault diagnosis in step (4) includes:
firstly, extracting coordinates of a target detection frame, and storing coordinate information of each row of a cable connector, a heating area and a facula area into a matrix in three types; and extracting coordinates of the three types of matrixes to obtain overlapping degree, and carrying out automatic classification of the coordinate file corresponding to the fault type and the normal condition of each type of cable joint to finally obtain a fault diagnosis effect.
The specific case of the data set in this embodiment is as follows:
the method comprises the steps of collecting 2240 infrared or ultraviolet images of a cable joint of a certain 220kV tunnel, carrying out denoising treatment on the collected images in batches, carrying out enhancement on an acquired data set based on limited contrast self-adaptive histogram equalization, marking an original image by using image marking software LabelImg, and randomly dividing a training set and a testing set according to a ratio of 3:1.
Training parameters and evaluation indexes:
for the training parameters, the initial learning rate was set to 0.005, the momentum term was 0.9, and the weight decay regularization term was 0.0005,Batch size set to 8.
In order to verify the actual recognition effect of the improved YOLOv 7-based cable splice target detection framework, the accuracy, recall and average accuracy index were passed on the test set. AP (averageprecision) index. The calculation formula is as follows:
wherein P is the accuracy of the network alignment sample prediction, R is the retrieval effect of the network alignment sample, TP is the number of cable joints accurately judged, FP is the number of cable joints misjudged as cable joints by non-cable joints, and FN is the number of cable joints misjudged as non-cable joints.
The weight ratio of the improved structure in the model performance is explored by adopting an ablation experimental rule. The results of the above experiments are shown in Table 1.
Table 1 ablation experiment results for improving YOLOv7 algorithm
From the above table, the accuracy AP of the improved YOLOv7 algorithm is 94.66%, which is improved by 8.05% compared with the original model of YOLOv 7. And, the addition of each module has a positive effect. Therefore, the invention can achieve ideal detection effect.
Claims (10)
1. The cable joint fault detection method based on the infrared or ultraviolet image is characterized by comprising the following steps of:
(1) Collecting infrared or ultraviolet image data of the cable connector as a data set through an infrared or ultraviolet imager;
(2) Performing enhancement based on limiting contrast self-adaptive histogram equalization on the acquired data set, marking an original image by using image marking software LabelImg, and randomly dividing a training set and a testing set according to a ratio of 3:1;
(3) Inputting the training set into an improved YOLOv7 model for fault diagnosis training to obtain a trained model;
(4) And performing effect test and fault diagnosis on the infrared or ultraviolet images in the test sample library by using the trained improved YOLOv7 model, and judging the severity of the fault by comparing the fault overlapping areas of the two types of images.
2. The method of claim 1, wherein the enhancing of the collected data set in step (2) based on limiting contrast adaptive histogram equalization comprises:
(2.1) blocking: dividing an input image into non-overlapping sub-blocks r of equal size k Where k=0, 1, …, l=1, L is the number of sub-blocks;
(2.2) calculating the sub-Block histogram h (r) k )；
(2.3) clipping the histogram of each sub-block with a clipping threshold: the cutting rule is
Wherein h' (r) k ) A clipping histogram for each sub-block; n (N) clip Is a defined actual shear threshold; n (N) avg An average value for pixels to be reassigned to each histogram;
(2.4) pixel point reassignment: for each sub-block, using the redundant pixel reassignment in the step (2.3) until all sheared pixel points are assigned;
(2.5) carrying out gray level histogram equalization treatment on each sub-region of the image after the steps so as to change the non-uniformly distributed histogram into uniform distribution;
and (2.6) reconstructing the gray value of the pixel point by adopting a bilinear interpolation method, taking the gray value of the obtained central pixel point of each sub-block as a reference point, and calculating the gray value of each point in the final output image.
3. The cable joint fault detection method based on infrared or ultraviolet images according to claim 1, wherein the improvement of the improved YOLOv7 model to the YOLOv7 model specifically comprises:
a lightweight network MobileOne is used as a backbone network of YOLOv 7;
adding a global attention mechanism GAM at the neck of the model to acquire richer cross-channel information, and improving the feature extraction capability of the model;
and a Focal-EIoU Loss function is introduced, so that the algorithm convergence rate is increased.
4. The infrared or ultraviolet image-based cable joint fault detection method according to claim 3, wherein the core module of the lightweight network MobileOne is designed based on MobileNet v1, and the structure is basically consistent with that of MobileNet v1, except that depth separable convolution in MobileNet is replaced by a neural network structural block, the left part of the neural network structural block forms a complete structural block of MobileOne and is formed by an upper part and a lower part, wherein the upper part is based on depth convolution, the lower part is based on point convolution, and act; the depth convolution module consists of three branches, and the leftmost branch is a 1 multiplied by 1 convolution; the middle branch is a parameterized 3 x 3 convolution, i.e., k 3 x 3 convolutions; the right part is a jumping connection comprising a BN layer; the deep convolution is essentially a group convolution, the number of groups is the same as the number of channels, where both the 1 x 1 convolution and the 3 x 3 convolution are deep convolutions; the point convolution module consists of two branches, the left branch is a parameterized 1 x 1 convolution, and consists of k 1 x 1 convolutions, and the right branch is a jump connection containing a BN layer.
5. The method for detecting cable joint failure based on infrared or ultraviolet images according to claim 3, wherein the global attention mechanism GAM is composed of a channel attention sub-module and a spatial attention sub-module, and the mechanism is optimally designed based on a sequential channel-spatial attention mechanism in CBAM;
wherein the input features are defined by F 1 Representing a series of intermediate operations in the GAM attention mechanism as intermediate state F 2 The output state is defined as F 3 The relationship among the three is as follows:
6. root of Chinese characterThe method for detecting a cable joint failure based on an infrared or ultraviolet image according to claim 5, wherein the original input feature F of the channel attention sub-module 1 The dimension is C x W x H, the channel attention submodule firstly carries out three-dimensional channel replacement on the channel attention submodule, and the information is stored in a W x H x C form; with a two-layer MLP, the first layer performs an encoding operation to reduce the number of channels C to C/R and the second layer performs a decoding operation to obtain a result with the same number of channels as the input feature. Finally, the weighting coefficient M is obtained by carrying out Sigmoid activation function processing on the result c The dependency of the cross-dimensional channel space can be effectively enlarged.
7. The method for detecting a cable joint failure based on infrared or ultraviolet images according to claim 5, wherein the input features F of the spatial attention sub-module 2 Double convolution is adopted, and each convolution layer uses a 7*7 convolution kernel to achieve the effect of spatial information fusion; the channel attention submodule reduces according to the importance of the features and obtains new scaled features; finally, the weighting factor M for this feature s And processing by adopting an activation function Sigmoid to obtain a more accurate weight value.
8. A method for detecting a cable joint failure based on infrared or ultraviolet images according to claim 3, wherein the introduction of the Focal-EIoU Loss function specifically comprises:
the improvement of EIoU Loss over CIoU Loss used by original YOLOv7 is to split the Loss function into overlapping area Loss, center point distance Loss, and aspect ratio Loss 3 parts, and modify a and v in CIoULoss, C w And C h For the width and height of the smallest box containing the real box, predicted box, EIoU Loss utilizesAnddirectly calculating the true value of the width and the height of the bounding box, and solving the problem of interference optimization caused by using the aspect ratio by CIoU Loss and the problem of divergence in the training process; considering the problem that training samples are unbalanced in regression of the prediction frames, the number of high-quality anchor frames with small regression errors in an input image is far less than the number of low quality anchor frames with large errors, and samples with poor quality can generate excessive gradient to influence optimization of parameters; by integrating the EIoU Loss function and the Focal Loss function, a final Focal-EIoU Loss expression is obtained, as shown in the formula (5):
L Focal-EIoU ＝IoU γ L EIoU (5)
in which L IoU 、L dis 、L asp The overlap area penalty, center point distance penalty, and aspect ratio penalty are shown, respectively.
9. The method for detecting a cable joint failure based on an infrared or ultraviolet image according to claim 6, wherein the failure diagnosis in step (4) comprises:
diagnosing whether a cable joint to be detected has a fault or not by calculating the overlapping degree of the identified cable joint area, the heating area and the facula area; the overlapping degree refers to the overlapping rate of a cable joint target window, a heating area target window and a facula area target window generated by the YOLOv7 model, namely the overlapping of the two detection frame selection areas is the overlapping degree of the fault diagnosis accuracy rate, and for the fault diagnosis, when the overlapping degree is lower than a certain value X, the possibility of abnormal heating and abnormal discharging of the power equipment is considered to be small, and at the moment, secondary manual diagnosis is usually needed; when the overlapping degree is 0, the cable connector can be basically judged that no abnormal area exists, and secondary diagnosis is not needed at the moment; when the overlapping degree is larger than a certain value, the abnormal area of the cable joint can be judged.
10. A cable joint fault detection system based on infrared or ultraviolet images, characterized by comprising a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the cable joint fault detection method based on infrared or ultraviolet images according to any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311713768.9A CN117809083B (en) | 2023-12-14 | 2023-12-14 | Cable joint fault detection method and system based on infrared or ultraviolet images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311713768.9A CN117809083B (en) | 2023-12-14 | 2023-12-14 | Cable joint fault detection method and system based on infrared or ultraviolet images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117809083A true CN117809083A (en) | 2024-04-02 |
| CN117809083B CN117809083B (en) | 2024-08-30 |
Family
ID=90426354
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311713768.9A Active CN117809083B (en) | 2023-12-14 | 2023-12-14 | Cable joint fault detection method and system based on infrared or ultraviolet images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117809083B (en) |
Citations (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105425123A (en) * | 2015-11-20 | 2016-03-23 | 国网福建省电力有限公司泉州供电公司 | Method and system for collaboratively detecting power equipment failure through ultraviolet imaging and infrared imaging |
| CN109785289A (en) * | 2018-12-18 | 2019-05-21 | 中国科学院深圳先进技术研究院 | A kind of transmission line of electricity defect inspection method, system and electronic equipment |
| CN110795991A (en) * | 2019-09-11 | 2020-02-14 | 西安科技大学 | Mining locomotive pedestrian detection method based on multi-information fusion |
| CN111948501A (en) * | 2020-08-05 | 2020-11-17 | 广州市赛皓达智能科技有限公司 | Automatic inspection equipment for power grid operation |
| CN112184601A (en) * | 2020-09-09 | 2021-01-05 | 中国计量大学 | Method for enhancing vein image under near infrared light source by utilizing improved CLAHE algorithm |
| CN112233073A (en) * | 2020-09-30 | 2021-01-15 | 国网山西省电力公司大同供电公司 | Real-time detection method for infrared thermal imaging abnormity of power transformation equipment |
| US20210020360A1 (en) * | 2019-07-15 | 2021-01-21 | Wuhan University | Internal thermal fault diagnosis method of oil-immersed transformer based on deep convolutional neural network and image segmentation |
| CN112380952A (en) * | 2020-11-10 | 2021-02-19 | 广西大学 | Power equipment infrared image real-time detection and identification method based on artificial intelligence |
| CN112381784A (en) * | 2020-11-12 | 2021-02-19 | 国网浙江省电力有限公司信息通信分公司 | Equipment detecting system based on multispectral image |
| CN112562255A (en) * | 2020-12-03 | 2021-03-26 | 国家电网有限公司 | Intelligent image detection method for cable channel smoke and fire condition in low-light-level environment |
| US20210174149A1 (en) * | 2018-11-20 | 2021-06-10 | Xidian University | Feature fusion and dense connection-based method for infrared plane object detection |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
| CN113159334A (en) * | 2021-02-24 | 2021-07-23 | 广西大学 | Electrical equipment infrared image real-time detection and diagnosis method based on light-weight deep learning |
| US11194997B1 (en) * | 2020-08-04 | 2021-12-07 | Nanjing Huatu Information Technology Co., Ltd. | Method and system for thermal infrared facial recognition |
| CN114264915A (en) * | 2021-12-13 | 2022-04-01 | 国网甘肃省电力公司临夏供电公司 | Power distribution network cable joint operation condition assessment early warning device and method |
| CN115546565A (en) * | 2022-11-09 | 2022-12-30 | 上海朗驰佰特智能技术有限公司 | YOLOCBF-based power plant key area pipeline oil leakage detection method |
| US20230039196A1 (en) * | 2021-08-09 | 2023-02-09 | The United States Of America, As Represented By The Secretary Of The Navy | Small unmanned aerial systems detection and classification using multi-modal deep neural networks |
| CN115909093A (en) * | 2022-10-21 | 2023-04-04 | 前郭富汇风能有限公司 | Power equipment fault detection method based on unmanned aerial vehicle inspection and infrared image semantic segmentation |
| US20230186439A1 (en) * | 2021-06-28 | 2023-06-15 | Zhejiang Gongshang University | Lane detection method integratedly using image enhancement and deep convolutional neural network |
| CN116434119A (en) * | 2023-04-21 | 2023-07-14 | 河北工程大学 | Method and system for detecting target in mine roadway |
| CN116758393A (en) * | 2023-05-24 | 2023-09-15 | 淮阴工学院 | Flame detection method based on MPGD-YOLO network |
| CN116824341A (en) * | 2023-07-03 | 2023-09-29 | 海南电网有限责任公司电力科学研究院 | YOLOv 7-based improved insulator abnormal temperature rise detection method |
| CN116883873A (en) * | 2023-07-12 | 2023-10-13 | 哈尔滨工业大学 | Infrared small target detection method for air-ground application |
| CN117078695A (en) * | 2023-08-18 | 2023-11-17 | 内蒙古工业大学 | Carotid plaque ultrasonic image identification segmentation method based on deep learning |
-
2023
- 2023-12-14 CN CN202311713768.9A patent/CN117809083B/en active Active
Patent Citations (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105425123A (en) * | 2015-11-20 | 2016-03-23 | 国网福建省电力有限公司泉州供电公司 | Method and system for collaboratively detecting power equipment failure through ultraviolet imaging and infrared imaging |
| US20210174149A1 (en) * | 2018-11-20 | 2021-06-10 | Xidian University | Feature fusion and dense connection-based method for infrared plane object detection |
| CN109785289A (en) * | 2018-12-18 | 2019-05-21 | 中国科学院深圳先进技术研究院 | A kind of transmission line of electricity defect inspection method, system and electronic equipment |
| US20210020360A1 (en) * | 2019-07-15 | 2021-01-21 | Wuhan University | Internal thermal fault diagnosis method of oil-immersed transformer based on deep convolutional neural network and image segmentation |
| CN110795991A (en) * | 2019-09-11 | 2020-02-14 | 西安科技大学 | Mining locomotive pedestrian detection method based on multi-information fusion |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
| US11194997B1 (en) * | 2020-08-04 | 2021-12-07 | Nanjing Huatu Information Technology Co., Ltd. | Method and system for thermal infrared facial recognition |
| CN111948501A (en) * | 2020-08-05 | 2020-11-17 | 广州市赛皓达智能科技有限公司 | Automatic inspection equipment for power grid operation |
| CN112184601A (en) * | 2020-09-09 | 2021-01-05 | 中国计量大学 | Method for enhancing vein image under near infrared light source by utilizing improved CLAHE algorithm |
| CN112233073A (en) * | 2020-09-30 | 2021-01-15 | 国网山西省电力公司大同供电公司 | Real-time detection method for infrared thermal imaging abnormity of power transformation equipment |
| CN112380952A (en) * | 2020-11-10 | 2021-02-19 | 广西大学 | Power equipment infrared image real-time detection and identification method based on artificial intelligence |
| CN112381784A (en) * | 2020-11-12 | 2021-02-19 | 国网浙江省电力有限公司信息通信分公司 | Equipment detecting system based on multispectral image |
| CN112562255A (en) * | 2020-12-03 | 2021-03-26 | 国家电网有限公司 | Intelligent image detection method for cable channel smoke and fire condition in low-light-level environment |
| CN113159334A (en) * | 2021-02-24 | 2021-07-23 | 广西大学 | Electrical equipment infrared image real-time detection and diagnosis method based on light-weight deep learning |
| US20230186439A1 (en) * | 2021-06-28 | 2023-06-15 | Zhejiang Gongshang University | Lane detection method integratedly using image enhancement and deep convolutional neural network |
| US20230039196A1 (en) * | 2021-08-09 | 2023-02-09 | The United States Of America, As Represented By The Secretary Of The Navy | Small unmanned aerial systems detection and classification using multi-modal deep neural networks |
| CN114264915A (en) * | 2021-12-13 | 2022-04-01 | 国网甘肃省电力公司临夏供电公司 | Power distribution network cable joint operation condition assessment early warning device and method |
| CN115909093A (en) * | 2022-10-21 | 2023-04-04 | 前郭富汇风能有限公司 | Power equipment fault detection method based on unmanned aerial vehicle inspection and infrared image semantic segmentation |
| CN115546565A (en) * | 2022-11-09 | 2022-12-30 | 上海朗驰佰特智能技术有限公司 | YOLOCBF-based power plant key area pipeline oil leakage detection method |
| CN116434119A (en) * | 2023-04-21 | 2023-07-14 | 河北工程大学 | Method and system for detecting target in mine roadway |
| CN116758393A (en) * | 2023-05-24 | 2023-09-15 | 淮阴工学院 | Flame detection method based on MPGD-YOLO network |
| CN116824341A (en) * | 2023-07-03 | 2023-09-29 | 海南电网有限责任公司电力科学研究院 | YOLOv 7-based improved insulator abnormal temperature rise detection method |
| CN116883873A (en) * | 2023-07-12 | 2023-10-13 | 哈尔滨工业大学 | Infrared small target detection method for air-ground application |
| CN117078695A (en) * | 2023-08-18 | 2023-11-17 | 内蒙古工业大学 | Carotid plaque ultrasonic image identification segmentation method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 李庆忠;李宜兵;牛炯;: "基于改进YOLO和迁移学习的水下鱼类目标实时检测", 模式识别与人工智能, no. 03, 15 March 2019 (2019-03-15) * |
| 武建华;梁利辉;纪欣欣;刘云鹏;裴少通;: "基于YOLOv3算法的绝缘子红外图像故障检测方法", 广东电力, no. 09, 30 September 2020 (2020-09-30) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117809083B (en) | 2024-08-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111797890A (en) | Method and system for detecting defects of power transmission line equipment | |
| CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
| CN113838054B (en) | Mechanical part surface damage detection method based on artificial intelligence | |
| CN104964886A (en) | Welded member fatigue stress and strain real-time non-contact type monitoring method | |
| CN113344475A (en) | Transformer bushing defect identification method and system based on sequence modal decomposition | |
| CN114897855A (en) | Method for judging defect type based on X-ray picture gray value distribution | |
| CN111753877B (en) | Product quality detection method based on deep neural network migration learning | |
| CN117541534A (en) | Power transmission line inspection method based on unmanned plane and CNN-BiLSTM model | |
| CN114170478A (en) | Defect detection and positioning method and system based on cross-image local feature alignment | |
| CN118334007A (en) | Crack detection and early warning method and system for hydraulic concrete structure | |
| CN115937079A (en) | YOLO v 3-based rapid detection method for defects of power transmission line | |
| Chen et al. | Automatic crack segmentation and feature extraction in electroluminescence images of solar modules | |
| CN118014949A (en) | X-ray image quality evaluation system, method and training method | |
| CN117809083B (en) | Cable joint fault detection method and system based on infrared or ultraviolet images | |
| CN117484031A (en) | Photovoltaic module welding processing equipment | |
| Ramlal et al. | Toward automated utility pole condition monitoring: A deep learning approach | |
| CN116484184A (en) | Method and device for enhancing partial discharge defect sample of power equipment | |
| CN110610136A (en) | Transformer substation equipment identification module and identification method based on deep learning | |
| CN116338545A (en) | Method, system, equipment and medium for identifying metering error state of current transformer | |
| Xia et al. | A multi-target detection based framework for defect analysis of electrical equipment | |
| CN114529543A (en) | Installation detection method and device for peripheral screw gasket of aero-engine | |
| Zhang et al. | Angle steel tower bolt defect detection based on YOLO-V3 | |
| Momot et al. | Analysis of the influence of the threshold level of binarization on the efficiency of segmentation of images of surface defects in steel by the U-NET network | |
| CN117351240B (en) | Positive sample sampling method, system, storage medium and electronic equipment | |
| CN118314488B (en) | Extra-high voltage transformer station space-earth multi-scale re-decision target detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |