CN111797890A - Method and system for detecting defects of power transmission line equipment - Google Patents
Method and system for detecting defects of power transmission line equipment Download PDFInfo
- Publication number
- CN111797890A CN111797890A CN202010419950.3A CN202010419950A CN111797890A CN 111797890 A CN111797890 A CN 111797890A CN 202010419950 A CN202010419950 A CN 202010419950A CN 111797890 A CN111797890 A CN 111797890A
- Authority
- CN
- China
- Prior art keywords
- equipment
- image
- transmission line
- power transmission
- defect
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C1/00—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people
- G07C1/20—Checking timed patrols, e.g. of watchman
-
- 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
- G01N2021/8854—Grading and classifying of flaws
-
- 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
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/888—Marking defects
-
- 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
Abstract
The invention discloses a method and a system for detecting defects of power transmission line equipment, and belongs to the technical field of power transmission line inspection. The method comprises the following steps: training by taking the equipment image as a training set to obtain a primary defect detection model; training by taking the generated defect identification result data as a training set to obtain a secondary defect detection model; and acquiring a patrol image of the target power transmission line as input data, inputting the patrol image into the primary defect detection model, acquiring an equipment image of the target power transmission line, inputting the equipment image of the target power transmission line as secondary input data into the secondary defect detection model, and acquiring an equipment defect identification result of the target power transmission line. The invention adopts two-stage detection and incremental detection model training modes, and can improve the accuracy of defect identification.
Description
Technical Field
The present invention relates to the field of transmission line inspection technologies, and more particularly, to a method and a system for detecting defects of transmission line equipment.
Background
In recent years, unmanned aerial vehicle is extensively used for electric power to patrol and examine, can follow the many angles and gather transmission line equipment image closely, more traditional manual work has promoted by a wide margin and has patrolled and examined efficiency. At present, most of inspection images in the industry depend on manual interpretation to find defects, and further improvement of inspection efficiency is limited.
The background texture of the power equipment is complex, so that the characteristic points of the power equipment are not easy to extract and distinguish, and the identification effect is influenced. The normal equipment on the transmission line accounts for the vast majority, and the defect identification false alarm rate is high. At present, the recognition effect of a mainstream deep learning network model cannot meet the industrial application requirements.
The traditional image recognition technology judges the defects of the equipment through indexes such as a matching template, a function and the like which are designed manually, and the recognition effect cannot be improved through incremental data. The recognition effect of the deep learning model can be improved along with the expansion of the number of the training sets, but the current manual labeling inspection image data is limited by the double cost and efficiency, and the data amount is difficult to give play to the theoretical detection precision of the existing deep learning network.
Disclosure of Invention
In view of the above problems, the present invention provides a method for detecting defects of power transmission line equipment, including:
carrying out hierarchical resampling processing on the power transmission line inspection image by using an initial primary equipment detection model, generating feature maps with different sizes in a scale space, establishing a feature pyramid, carrying out high-dimensional equipment feature extraction on primary power transmission equipment aiming at the feature pyramid, acquiring regional pixel coordinates and equipment classification data of the power transmission line inspection image according to the position of the power transmission line equipment positioned in the power transmission line inspection image by using the extracted high-dimensional equipment feature, generating an equipment image, training by using the equipment image as a training set, and acquiring a primary equipment detection model;
performing secondary high-dimensional feature defect extraction on the equipment image by using an initial secondary equipment detection model, identifying defect features, generating defect identification result data, and training by using the generated defect identification result data as a training set to obtain a secondary defect detection model;
and acquiring a patrol image of the target power transmission line as input data, inputting the patrol image into the primary equipment detection model, acquiring an equipment image of the target power transmission line, inputting the equipment image of the target power transmission line as secondary input data into the secondary defect detection model, and acquiring an equipment defect identification result of the target power transmission line.
Optionally, the high-dimensional device features include: geometric profiles, structures, colors or surface textures of towers, insulators, wires, hardware and dowel nuts.
Optionally, the high-dimensional defect feature includes: foreign matters of the pole tower, corrosion of the pole tower, self-explosion of the insulator, hardware cracks, broken strands and scattered strands of a lead, cracks of a pin nut and rusted geometric outlines, structures, colors or surface textures.
Optionally, the high-dimensional equipment feature extraction of the power transmission equipment is performed on the power transmission line inspection image, and the method includes the following steps:
carrying out feature extraction of layer-by-layer convolution on the power transmission line inspection feature image by using preset parameters;
classifying the extracted features, and classifying the probability of the occurrence of the features of each equipment or defect category to obtain feature maps of different layers;
and classifying and position regression are carried out on the feature maps of different layers, and according to the feature combination difference of the high-dimensional equipment, an image area of the electric transmission line equipment is extracted from the electric transmission line inspection image and corresponding equipment classification information is given.
Optionally, the method according to claim 1, further comprising: the extracted equipment image is free from background interference of the image, and only the equipment image area identified by the primary equipment model is subjected to convolution processing, so that the convolution operation range is narrowed, the calculation efficiency of the secondary defect detection model during operation is improved, and the false alarm rate is reduced.
Optionally, when performing secondary high-dimensional feature defect extraction, performing convolution operation in the device image area, performing classification and position regression on feature maps of different layers, extracting a corresponding device defect area from the device image according to the high-dimensional defect feature combination difference, and providing corresponding defect classification information.
The invention also provides a system for detecting the defects of the transmission line equipment, which comprises the following steps:
the primary equipment identification model training module is used for carrying out layered resampling processing on the power transmission line inspection image by using an initial primary equipment detection model, generating feature maps with different sizes in a scale space, establishing a feature pyramid, carrying out high-dimensional equipment feature extraction on the power transmission line equipment for the first time aiming at the feature pyramid, obtaining regional pixel coordinates and equipment classification data of the power transmission line inspection image according to the power transmission line equipment position positioned in the power transmission line inspection image by using the extracted high-dimensional equipment feature, generating an equipment image, training by using the equipment image as a training set, and obtaining a primary equipment detection model;
the secondary defect identification model training module is used for extracting secondary high-dimensional feature defects from the equipment image by using the initial secondary equipment detection model, identifying defect features, generating defect identification result data, and training by taking the generated defect identification result data as a training set to obtain a secondary defect detection model;
and the defect identification module is used for acquiring the inspection image of the target power transmission line as input data, inputting the inspection image into the primary equipment detection model, acquiring the equipment image of the target power transmission line, inputting the equipment image of the target power transmission line into the secondary defect detection model by taking the equipment image of the target power transmission line as secondary input data, and acquiring an equipment defect identification result of the target power transmission line.
Optionally, the high-dimensional device features include: geometric profiles, structures, colors or surface textures of towers, insulators, wires, hardware and dowel nuts.
Optionally, the high-dimensional defect feature includes: foreign matters of the pole tower, corrosion of the pole tower, self-explosion of the insulator, hardware cracks, broken strands and scattered strands of a lead, cracks of a pin nut and rusted geometric outlines, structures, colors or surface textures.
Optionally, the high-dimensional equipment feature extraction of the power transmission equipment is performed on the power transmission line inspection image, and the method includes the following steps:
carrying out feature extraction of layer-by-layer convolution on the power transmission line inspection feature image by using preset parameters;
classifying the extracted features, and classifying the probability of the occurrence of the features of each equipment or defect category to obtain feature maps of different layers;
and classifying and position regression are carried out on the feature maps of different layers, and according to the feature combination difference of the high-dimensional equipment, an image area of the electric transmission line equipment is extracted from the electric transmission line inspection image and corresponding equipment classification information is given.
Optionally, the primary training module is further configured to: the extracted equipment image is free from background interference of the image, and only the equipment image area identified by the primary equipment model is subjected to convolution processing, so that the convolution operation range is narrowed, the calculation efficiency of the secondary defect detection model during operation is improved, and the false alarm rate is reduced.
Optionally, when performing secondary high-dimensional feature defect extraction, performing convolution operation in the device image area, performing classification and position regression on feature maps of different layers, extracting a corresponding device defect area from the device image according to the high-dimensional defect feature combination difference, and providing corresponding defect classification information.
The invention adopts a mode of twice detection, and can improve the accuracy of defect identification.
Drawings
FIG. 1 is a flow chart of a method for detecting defects in transmission line equipment according to the present invention;
FIG. 2 is a flowchart of an embodiment of a method for detecting defects in transmission line equipment according to the present invention;
fig. 3 is a structural diagram of a system for detecting defects of power transmission line equipment according to the present invention.
Detailed Description
The exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, however, the present invention may be embodied in many different forms and is not limited to the embodiments described herein, which are provided for complete and complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting of the invention. In the drawings, the same units/elements are denoted by the same reference numerals.
Unless otherwise defined, terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Further, it will be understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense.
The invention provides a method for detecting defects of transmission line equipment, which comprises the following steps of:
carrying out hierarchical resampling processing on the power transmission line inspection image by using an initial primary equipment detection model, generating feature maps with different sizes in a scale space, establishing a feature pyramid, carrying out high-dimensional equipment feature extraction on primary power transmission equipment aiming at the feature pyramid, acquiring regional pixel coordinates and equipment classification data of the power transmission line inspection image according to the position of the power transmission line equipment positioned in the power transmission line inspection image by using the extracted high-dimensional equipment feature, generating an equipment image, training by using the equipment image as a training set, and acquiring a primary equipment detection model;
performing secondary high-dimensional feature defect extraction on the equipment image by using an initial secondary equipment detection model, identifying defect features, generating defect identification result data, and training by using the generated defect identification result data as a training set to obtain a secondary defect detection model;
and acquiring a patrol image of the target power transmission line as input data, inputting the patrol image into the primary equipment defect detection model, acquiring an equipment image of the target power transmission line, inputting the equipment image of the target power transmission line into the secondary defect detection model by taking the equipment image of the target power transmission line as secondary input data, and acquiring an equipment defect identification result of the target power transmission line.
Wherein, the high dimensional device characteristics include: geometric profiles, structures, colors or surface textures of towers, insulators, wires, hardware and dowel nuts.
High dimensional defect features, including: foreign matters of the pole tower, corrosion of the pole tower, self-explosion of the insulator, hardware cracks, broken strands and scattered strands of a lead, cracks of a pin nut and rusted geometric outlines, structures, colors or surface textures.
High-dimensional equipment feature extraction of power transmission equipment is carried out to transmission line image of patrolling and examining, includes:
carrying out feature extraction of layer-by-layer convolution on the power transmission line inspection feature image by using preset parameters;
classifying the extracted features, and classifying the probability of the occurrence of the features of each equipment or defect category to obtain feature maps of different layers;
and classifying and position regression are carried out on the feature maps of different layers, and according to the feature combination difference of the high-dimensional equipment, an image area of the electric transmission line equipment is extracted from the electric transmission line inspection image and corresponding equipment classification information is given.
The extracted equipment image is free from background interference of the image, and only the equipment image area identified by the primary equipment model is subjected to convolution processing, so that the convolution operation range is narrowed, the calculation efficiency of the secondary defect detection model during operation is improved, and the false alarm rate is reduced.
And when secondary high-dimensional feature defect extraction is carried out, carrying out convolution operation in the equipment image area, carrying out classification and position regression on feature maps of different layers, extracting a corresponding equipment defect area from the equipment image according to the high-dimensional defect feature combination difference, and giving out corresponding defect classification information.
The invention is further illustrated by the following examples:
the flow of the embodiment, as shown in fig. 2, includes:
detecting a power inspection image by adopting a secondary deep learning model, wherein the input end of the power inspection image is an original power transmission line inspection image, the output end of the power inspection image is a power target detection result, and high-dimensional features are extracted through a depth convolution layer;
the accuracy of identifying large-size targets such as poles and towers, insulators and small-size targets such as hardware fittings, pin nuts and the like is ensured through a characteristic Pyramid network (FPN);
an improved SSD network is used as a primary equipment detection model for positioning the position of equipment in an image, and irrelevant image background interference is eliminated;
an improved fast RCNN model is adopted as a secondary equipment defect detection model, and a secondary model is called to identify equipment defects on the basis of primary model equipment positioning;
the initial primary equipment detection model and the secondary defect detection model need pre-training.
Constructing an incremental training environment, wherein the output of the detection model is subjected to manual examination and correct recognition results and is used for replacing the traditional manually marked training data and training the deep learning detection model;
wherein, adopt the secondary degree of depth learning model to detect electric power and patrol and examine the image, the input is the original power transmission line and patrols and examines the image, and the output is electric power target test result, includes:
aiming at the problems that the power line pole towers are usually in a normal state, the proportion of the image of the power line equipment with defects in the whole data set is too small, and the normal equipment and the irrelevant complex background cause interference on the identification effect, a deep convolution layer is adopted to extract high-dimensional features;
aiming at the problems that the proportion of an equipment image area in the whole inspection image is relatively small, the proportion of an irrelevant background area is large, and the blind feature extraction efficiency of convolution operation on the whole image is low during defect detection, the position of the equipment is positioned from the image through a primary detection model, and various types of equipment such as a tower, an insulator, a lead wire, a ground wire and hardware are identified through primary detection;
inputting the pixels of the device position area into a secondary defect detection model, respectively judging the device defects according to the typical characteristics of different devices, identifying multiple types of device defects through primary detection, detecting the position of the device in an image through an improved end-to-end model SSD, and generating a rectangular surrounding frame, thereby effectively reducing background interference and improving the accuracy of subsequent device defect identification;
the method comprises the steps of training a primary equipment detection model by marking typical equipment such as a tower and an insulator in a normal inspection image, and training a secondary defect equipment detection model by marking equipment defects such as tower foreign matters, tower corrosion, insulator spontaneous explosion, hardware cracks, broken strands and strand scattering of a lead and the like and typical equipment such as a normal tower and an insulator.
The method for identifying the large-size targets such as the towers and the insulators and the small-size targets such as the hardware fittings and the pin nuts by using the characteristic Pyramid network (FPN) comprises the following steps:
in order to ensure the effectiveness of high-dimensional feature extraction, a feature pyramid network is adopted to perform convolution on an equipment image region output from the upper section layer by layer, the sampling convolution kernel on each layer is 3 multiplied by 3, the resolution is 1/4 of the image of the upper layer, and the up-sampling is performed for 4 times, so that the target multilayer high-dimensional implicit feature information in the image is extracted;
extracting a target region for region screening, extracting and outputting a high-dimensional feature map, sorting in a descending order according to the probability of a target object appearing in the extraction region, outputting a high-probability region, performing target analysis according to the output of a region screening layer, and performing classification marking on pixel regions in each layer of the pyramid according to classification requirements;
the classification and position regression are simultaneously carried out on the feature maps of different layers of the pyramid network image, the lower-layer feature map is small in receptive field, the higher-layer feature map is large in receptive field, and the detection model can be more accurate in synchronous detection of a large target and a small target by using the feature maps with different sizes for target detection.
The method for eliminating background interference of irrelevant images by adopting an improved SSD network as a primary equipment detection model to position the equipment in the image comprises the following steps:
in order to improve the detection and positioning speed of equipment, the original SSD network is improved, a feature extraction basic network VGG-16 network with more weight parameters and larger calculation amount in the structure of the original SSD network is replaced by a lightweight MobileNet convolutional neural network with less weight parameters and high calculation speed, and a convolution filter method of the original MobileNet is deleted, so that a feature extraction model is further compressed and simplified;
in order to enable the convolutional neural network to extract more features and obtain feature maps with different scales, four convolutional layers are added after the thirteenth convolutional layer of the compressed Mobile Net network, wherein the fourteenth convolutional layer has 512 channels and a convolutional kernel size of 3 × 3, the convolutional layer Conv15 has 256 channels and a convolutional kernel size of 3 × 3, the convolutional layer Conv16 has 256 channels and a convolutional kernel size of 3 × 3, and the convolutional layer Conv17 has 128 channels and a convolutional kernel size of 3 × 3;
in order to reduce the calculation amount of the convolutional neural network model and reduce the number of channels of the convolutional layers, the convolutional layers with convolution kernels of 1 × 1 are added before each newly added convolutional layer, and the number of channels of each convolutional layer is transformed by using the 1 × 1 convolutional layer, so that the multiplication operation amount of the convolutional neural network is reduced by one order of magnitude.
After the improved MobileNet network extracts the feature map, the prediction frames are output to a classifier of an original SSD network, for each target prediction frame, the category and the confidence value of the target prediction frame are determined according to the maximum category confidence value, the prediction frames with lower thresholds and belonging to the background are filtered out and filtered, finally, descending order arrangement is carried out according to the confidence value, and only the equipment recognition frames larger than the thresholds are reserved and output as the detection result of a primary equipment recognition model.
Wherein, adopt through modified fast RCNN model as secondary defect detection model, on the basis of primary model equipment location, call secondary model discernment equipment defect, include:
inputting the image region identified by the primary detection model into an improved Faster-RCNN deep learning network for equipment defect detection;
the secondary defect detection model comprises a target region extraction layer, a region screening layer and a target analysis layer which are developed based on a universal fast RCNN deep learning network layer, wherein the target region extraction layer and the screening layer train network layer parameters according to a marking result region, a region to be detected is generated in an image feature pyramid network structure layer by layer, a feature image layer with the most appropriate size is selected to extract a feature block based on the size of the region to be detected, at the moment, the feature pyramid network is not a target detector alone, or is a target detector and a cooperative feature detector, and the result of feature detection is respectively transmitted to each feature image to complete small target detection;
aiming at the problem of characteristic degradation caused by a VGG-16 Network for characteristic extraction of an original fast-RCNN model, a ResNet101 Network which is more effective in fitting residual characteristic extraction is adopted, and after the characteristic extraction is completed, an intermediate result is input into a subsequent Region suggestion Network (RPN);
in the stage of identifying and detecting the defective target, aiming at the overfitting problem brought by the prior fast RCNN training area recommendation network only under the fixed value with the IOU of 0.5, a three-level cascade training mode is designed, detectors are respectively trained under three IOU thresholds, then the detectors are combined in a cascade mode to improve the detection effect, the IOU thresholds of all levels of cascade are respectively set to be 0.5,0.6 and 0.7, specifically, after one image passes through a feature extraction network, the area recommendation network is firstly trained under the IOU threshold of 0.5, the target category information and the boundary frame information detected in the first stage of the image are obtained, the outputs are used as the supervision signals of the area network of the second stage, a new area recommendation network sample is generated under the IOU threshold of 0.6, the next cascade area recommendation network is continuously trained, and the target category information and the boundary frame information detected in the second stage of the image are further obtained, and then inputting a training area suggestion network with an IOU threshold value of 0.7, thereby improving the precision of a defect identification target of the network, and inputting an intermediate result into a subsequent classification network.
According to the classification network in the original Faster R-CNN, target type attribution probability vectors and target detection frames in each area suggestion network are output, and the most probable type attribute and the most accurate target detection area are screened out for each target through a non-maximum suppression (NMS) algorithm, so that the multi-type device defect target is accurately identified, diagnosed and pixel-pole positioned.
The method comprises the steps of establishing an incremental training environment, manually checking a correct recognition result output by a detection model, replacing traditional manually marked training data, and training the deep learning detection model.
The equipment area identified by the primary model is used as a training set, and the primary equipment identification model is trained in an incremental mode; the defects identified by the secondary defect identification model are used as a training set, and the secondary defect identification model is trained, so that the workload of manual labeling in the process of manufacturing the deep learning training set can be reduced.
The defect information in the routing inspection big data is fully utilized in application, the problem that the recognition effect of the deep learning detection model depends on the training data volume is solved, and the recognition effect of the defect detection model of the embodiment on various equipment defects is further improved.
In the embodiment, the defects of typical transmission line equipment are manually marked, a deep learning model is trained as required to process the defects of the inspection image recognition equipment, and the equipment such as a tower, an insulator and the like is identified and positioned in the image for the first time in a mode of two-time detection, so that the retrieval range is reduced, and irrelevant background interference is avoided as much as possible; and defect detection is carried out on the identified target equipment, the equipment characteristics are extracted, whether the equipment defects exist or not is judged, and the accuracy of defect identification can be further improved.
An incremental training frame is built, a defect recognition result is used as a new training set to train a defect detection model, the characteristic that the detection effect of a deep learning model is increased along with the increase of the training data volume is exerted, the potential of a patrol big data mining deep learning network is fully utilized, and the accuracy of pole tower foreign matter, strand breakage of a ground wire, self-explosion of an insulator, hardware cracks and pin defect recognition in a patrol image is further improved.
The present invention further provides a system 200 for detecting defects of power transmission line equipment, as shown in fig. 3, including:
a primary equipment identification model training module 201, which performs hierarchical resampling processing on the power transmission line inspection image by using an initial primary equipment detection model, generates feature maps with different sizes in a scale space, establishes a feature pyramid, performs high-dimensional equipment feature extraction on the power transmission line equipment for the first time aiming at the feature pyramid, acquires region pixel coordinates and equipment classification data of the power transmission line inspection image according to the power transmission line equipment position located in the power transmission line inspection image by using the extracted high-dimensional equipment feature, generates an equipment image, trains by using the equipment image as a training set, and acquires a primary equipment detection model;
the secondary defect identification model training module 202 is used for performing secondary high-dimensional feature defect extraction on the equipment image by using the initial secondary equipment detection model, identifying defect features, generating defect identification result data, and training by using the generated defect identification result data as a training set to obtain a secondary defect detection model;
the defect identification module 203 acquires the inspection image of the target power transmission line as input data, inputs the inspection image into the primary equipment detection model, acquires the equipment image of the target power transmission line, and inputs the equipment image of the target power transmission line into the secondary defect detection model by taking the equipment image of the target power transmission line as secondary input data to acquire an equipment defect identification result of the target power transmission line.
High dimensional device features, comprising: geometric profiles, structures, colors or surface textures of towers, insulators, wires, hardware and dowel nuts.
High dimensional defect features, including: foreign matters of the pole tower, corrosion of the pole tower, self-explosion of the insulator, hardware cracks, broken strands and scattered strands of a lead, cracks of a pin nut and rusted geometric outlines, structures, colors or surface textures.
High-dimensional equipment feature extraction of power transmission equipment is carried out to transmission line image of patrolling and examining, includes:
carrying out feature extraction of layer-by-layer convolution on the power transmission line inspection feature image by using preset parameters;
classifying the extracted features, and classifying the probability of the occurrence of the features of each equipment or defect category to obtain feature maps of different layers;
and classifying and position regression are carried out on the feature maps of different layers, and according to the feature combination difference of the high-dimensional equipment, an image area of the electric transmission line equipment is extracted from the electric transmission line inspection image and corresponding equipment classification information is given.
The primary training module 201 is further configured to: the extracted equipment image is free from background interference of the image, and only the equipment image area identified by the primary equipment model is subjected to convolution processing, so that the convolution operation range is narrowed, the calculation efficiency of the secondary defect detection model during operation is improved, and the false alarm rate is reduced.
And when secondary high-dimensional feature defect extraction is carried out, carrying out convolution operation in the equipment image area, carrying out classification and position regression on feature maps of different layers, extracting a corresponding equipment defect area from the equipment image according to the high-dimensional defect feature combination difference, and giving out corresponding defect classification information.
The invention adopts a mode of twice detection, and can improve the accuracy of defect identification.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The scheme in the embodiment of the application can be implemented by adopting various computer languages, such as object-oriented programming language Java and transliterated scripting language JavaScript.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.
Claims (12)
1. A method for detecting a transmission line equipment defect, the method comprising:
carrying out hierarchical resampling processing on the power transmission line inspection image by using an initial primary equipment detection model, generating feature maps with different sizes in a scale space, establishing a feature pyramid, carrying out high-dimensional equipment feature extraction on primary power transmission equipment aiming at the feature pyramid, acquiring regional pixel coordinates and equipment classification data of the power transmission line inspection image according to the position of the power transmission line equipment positioned in the power transmission line inspection image by using the extracted high-dimensional equipment feature, generating an equipment image, training by using the equipment image as a training set, and acquiring a primary equipment detection model;
performing secondary high-dimensional feature defect extraction on the equipment image by using an initial secondary equipment detection model, identifying defect features, generating defect identification result data, and training by using the generated defect identification result data as a training set to obtain a secondary defect detection model;
and acquiring a patrol image of the target power transmission line as input data, inputting the patrol image into the primary equipment detection model, acquiring an equipment image of the target power transmission line, inputting the equipment image of the target power transmission line as secondary input data into the secondary defect detection model, and acquiring an equipment defect identification result of the target power transmission line.
2. The method of claim 1, the high-dimensional device features, comprising: geometric profiles, structures, colors or surface textures of towers, insulators, wires, hardware and dowel nuts.
3. The method of claim 1, the high dimensional defect feature, comprising: foreign matters of the pole tower, corrosion of the pole tower, self-explosion of the insulator, hardware cracks, broken strands and scattered strands of a lead, cracks of a pin nut and rusted geometric outlines, structures, colors or surface textures.
4. The method according to claim 1, wherein the extracting of the high-dimensional equipment features of the power transmission equipment aiming at the power transmission line inspection image comprises the following steps:
carrying out feature extraction of layer-by-layer convolution on the power transmission line inspection feature image by using preset parameters;
classifying the extracted features, and classifying the probability of the occurrence of the features of each equipment or defect category to obtain feature maps of different layers;
and classifying and position regression are carried out on the feature maps of different layers, and according to the feature combination difference of the high-dimensional equipment, an image area of the electric transmission line equipment is extracted from the electric transmission line inspection image and corresponding equipment classification information is given.
5. The method of claim 1, further comprising: the extracted equipment image is free from background interference of the image, and only the equipment image area identified by the primary equipment model is subjected to convolution processing, so that the convolution operation range is narrowed, the calculation efficiency of the secondary defect detection model during operation is improved, and the false alarm rate is reduced.
6. The method according to claim 1, wherein during the secondary high-dimensional feature defect extraction, a convolution operation is performed in an equipment image region, classification and position regression are performed on feature maps of different layers, and a corresponding equipment defect region is extracted from the equipment image and corresponding defect classification information is given according to a high-dimensional defect feature combination difference.
7. A system for detecting a defect in a piece of power transmission line equipment, the system comprising:
the primary equipment identification model training module is used for carrying out layered resampling processing on the power transmission line inspection image by using an initial primary equipment detection model, generating feature maps with different sizes in a scale space, establishing a feature pyramid, carrying out high-dimensional equipment feature extraction on the power transmission line equipment for the first time aiming at the feature pyramid, obtaining regional pixel coordinates and equipment classification data of the power transmission line inspection image according to the power transmission line equipment position positioned in the power transmission line inspection image by using the extracted high-dimensional equipment feature, generating an equipment image, training by using the equipment image as a training set, and obtaining a primary equipment detection model;
the secondary defect identification model training module is used for extracting secondary high-dimensional feature defects from the equipment image by using the initial secondary equipment detection model, identifying defect features, generating defect identification result data, and training by taking the generated defect identification result data as a training set to obtain a secondary defect detection model;
and the defect identification module is used for acquiring the inspection image of the target power transmission line as input data, inputting the inspection image into the primary defect detection model, acquiring the equipment image of the target power transmission line, inputting the equipment image of the target power transmission line as secondary input data into the secondary defect detection model and acquiring the equipment defect identification result of the target power transmission line.
8. The system of claim 7, the high-dimensional device features, comprising: geometric profiles, structures, colors or surface textures of towers, insulators, wires, hardware and dowel nuts.
9. The system of claim 7, the high dimensional defect feature, comprising: foreign matters of the pole tower, corrosion of the pole tower, self-explosion of the insulator, hardware cracks, broken strands and scattered strands of a lead, cracks of a pin nut and rusted geometric outlines, structures, colors or surface textures.
10. The system of claim 7, wherein the extracting of the high-dimensional device characteristics of the power transmission device for the power transmission line inspection image comprises:
carrying out feature extraction of layer-by-layer convolution on the power transmission line inspection feature image by using preset parameters;
classifying the extracted features, and classifying the probability of the occurrence of the features of each equipment or defect category to obtain feature maps of different layers;
and classifying and position regression are carried out on the feature maps of different layers, and according to the feature combination difference of the high-dimensional equipment, an image area of the electric transmission line equipment is extracted from the electric transmission line inspection image and corresponding equipment classification information is given.
11. The system of claim 7, the primary training module further to: the extracted equipment image is free from background interference of the image, and only the equipment image area identified by the primary equipment model is subjected to convolution processing, so that the convolution operation range is narrowed, the calculation efficiency of the secondary defect detection model during operation is improved, and the false alarm rate is reduced.
12. The system according to claim 7, wherein when performing secondary high-dimensional feature defect extraction, a convolution operation is performed in an equipment image region, classification and position regression are performed on feature maps of different layers, and a corresponding equipment defect region is extracted from the equipment image and corresponding defect classification information is given according to a high-dimensional defect feature combination difference.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010419950.3A CN111797890A (en) | 2020-05-18 | 2020-05-18 | Method and system for detecting defects of power transmission line equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010419950.3A CN111797890A (en) | 2020-05-18 | 2020-05-18 | Method and system for detecting defects of power transmission line equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111797890A true CN111797890A (en) | 2020-10-20 |
Family
ID=72806491
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010419950.3A Pending CN111797890A (en) | 2020-05-18 | 2020-05-18 | Method and system for detecting defects of power transmission line equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111797890A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112487916A (en) * | 2020-11-25 | 2021-03-12 | 上海电力大学 | Binocular three-dimensional identification system for electrical equipment |
| CN112926401A (en) * | 2021-01-29 | 2021-06-08 | 广州中科智巡科技有限公司 | Transmission line hardware corrosion detection method and system |
| CN112950584A (en) * | 2021-03-01 | 2021-06-11 | 哈尔滨工程大学 | Coating surface defect identification method based on deep learning |
| CN113112487A (en) * | 2021-04-20 | 2021-07-13 | 广东电网有限责任公司 | Method, device, equipment and storage medium for identifying thermal defects |
| CN113192222A (en) * | 2021-04-14 | 2021-07-30 | 山东理工大学 | Equipment-level alarm strategy method for power line visual alarm |
| CN113327255A (en) * | 2021-05-28 | 2021-08-31 | 宁波新胜中压电器有限公司 | Power transmission line inspection image processing method based on YOLOv3 detection, positioning and cutting and fine-tune |
| CN113361473A (en) * | 2021-06-30 | 2021-09-07 | 北京百度网讯科技有限公司 | Image processing method, model training method, device, apparatus, storage medium, and program |
| CN113469992A (en) * | 2021-07-16 | 2021-10-01 | 合肥中科融道智能科技有限公司 | Power equipment image defect detection method based on enhancement of different-level feature representation |
| CN113627378A (en) * | 2021-08-19 | 2021-11-09 | 国网上海市电力公司 | Line bolt defect identification method based on combination of Phash algorithm and deep learning |
| CN113780358A (en) * | 2021-08-16 | 2021-12-10 | 华北电力大学（保定） | Real-time hardware fitting detection method based on anchor-free network |
| CN114155246A (en) * | 2022-02-10 | 2022-03-08 | 国网江西省电力有限公司电力科学研究院 | Deformable convolution-based power transmission tower pin defect detection method |
| CN114397306A (en) * | 2022-03-25 | 2022-04-26 | 南方电网数字电网研究院有限公司 | Power grid grading ring hypercomplex category defect multi-stage model joint detection method |
| CN114486925A (en) * | 2021-12-25 | 2022-05-13 | 国网四川省电力公司电力科学研究院 | Panoramic information-based intelligent detection system and method for corrosion of inner side of power transmission line tower |
| CN114926723A (en) * | 2022-06-10 | 2022-08-19 | 国网江苏省电力有限公司电力科学研究院 | Method, terminal and storage medium for identifying and triggering alarm of interference objects around power transmission line |
| CN115937218A (en) * | 2023-03-13 | 2023-04-07 | 昆山市恒达精密机械工业有限公司 | Surface defect detection method and system for cold punching part |
| CN116385952A (en) * | 2023-06-01 | 2023-07-04 | 华雁智能科技(集团)股份有限公司 | Distribution network line small target defect detection method, device, equipment and storage medium |
| CN116674300A (en) * | 2023-05-24 | 2023-09-01 | 常州润来科技有限公司 | Automatic ink-jet marking system and method based on copper pipe flaw detection result |
| CN117058689A (en) * | 2023-10-09 | 2023-11-14 | 巴斯夫一体化基地(广东)有限公司 | Offline detection data processing method for chemical production |
| CN117078689A (en) * | 2023-10-17 | 2023-11-17 | 沈阳宏远电磁线股份有限公司 | Cable defect identification method and system based on machine vision |
| CN117132914A (en) * | 2023-10-27 | 2023-11-28 | 武汉大学 | Method and system for identifying large model of universal power equipment |
-
2020
- 2020-05-18 CN CN202010419950.3A patent/CN111797890A/en active Pending
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112487916A (en) * | 2020-11-25 | 2021-03-12 | 上海电力大学 | Binocular three-dimensional identification system for electrical equipment |
| CN112926401A (en) * | 2021-01-29 | 2021-06-08 | 广州中科智巡科技有限公司 | Transmission line hardware corrosion detection method and system |
| CN112950584A (en) * | 2021-03-01 | 2021-06-11 | 哈尔滨工程大学 | Coating surface defect identification method based on deep learning |
| CN113192222A (en) * | 2021-04-14 | 2021-07-30 | 山东理工大学 | Equipment-level alarm strategy method for power line visual alarm |
| CN113112487A (en) * | 2021-04-20 | 2021-07-13 | 广东电网有限责任公司 | Method, device, equipment and storage medium for identifying thermal defects |
| CN113327255A (en) * | 2021-05-28 | 2021-08-31 | 宁波新胜中压电器有限公司 | Power transmission line inspection image processing method based on YOLOv3 detection, positioning and cutting and fine-tune |
| CN113361473A (en) * | 2021-06-30 | 2021-09-07 | 北京百度网讯科技有限公司 | Image processing method, model training method, device, apparatus, storage medium, and program |
| CN113361473B (en) * | 2021-06-30 | 2023-12-08 | 北京百度网讯科技有限公司 | Image processing method, model training method, image processing device, model training apparatus, storage medium, and program |
| CN113469992A (en) * | 2021-07-16 | 2021-10-01 | 合肥中科融道智能科技有限公司 | Power equipment image defect detection method based on enhancement of different-level feature representation |
| CN113469992B (en) * | 2021-07-16 | 2022-02-18 | 合肥中科融道智能科技有限公司 | Power equipment image defect detection method based on enhancement of different-level feature representation |
| CN113780358A (en) * | 2021-08-16 | 2021-12-10 | 华北电力大学（保定） | Real-time hardware fitting detection method based on anchor-free network |
| CN113627378A (en) * | 2021-08-19 | 2021-11-09 | 国网上海市电力公司 | Line bolt defect identification method based on combination of Phash algorithm and deep learning |
| CN114486925A (en) * | 2021-12-25 | 2022-05-13 | 国网四川省电力公司电力科学研究院 | Panoramic information-based intelligent detection system and method for corrosion of inner side of power transmission line tower |
| CN114486925B (en) * | 2021-12-25 | 2023-10-31 | 国网四川省电力公司电力科学研究院 | Intelligent detection method for corrosion on inner side of power transmission line pole tower based on panoramic information |
| CN114155246A (en) * | 2022-02-10 | 2022-03-08 | 国网江西省电力有限公司电力科学研究院 | Deformable convolution-based power transmission tower pin defect detection method |
| CN114397306A (en) * | 2022-03-25 | 2022-04-26 | 南方电网数字电网研究院有限公司 | Power grid grading ring hypercomplex category defect multi-stage model joint detection method |
| CN114926723A (en) * | 2022-06-10 | 2022-08-19 | 国网江苏省电力有限公司电力科学研究院 | Method, terminal and storage medium for identifying and triggering alarm of interference objects around power transmission line |
| CN115937218A (en) * | 2023-03-13 | 2023-04-07 | 昆山市恒达精密机械工业有限公司 | Surface defect detection method and system for cold punching part |
| CN116674300A (en) * | 2023-05-24 | 2023-09-01 | 常州润来科技有限公司 | Automatic ink-jet marking system and method based on copper pipe flaw detection result |
| CN116674300B (en) * | 2023-05-24 | 2023-11-14 | 常州润来科技有限公司 | Automatic ink-jet marking system and method based on copper pipe flaw detection result |
| CN116385952B (en) * | 2023-06-01 | 2023-09-01 | 华雁智能科技(集团)股份有限公司 | Distribution network line small target defect detection method, device, equipment and storage medium |
| CN116385952A (en) * | 2023-06-01 | 2023-07-04 | 华雁智能科技(集团)股份有限公司 | Distribution network line small target defect detection method, device, equipment and storage medium |
| CN117058689A (en) * | 2023-10-09 | 2023-11-14 | 巴斯夫一体化基地(广东)有限公司 | Offline detection data processing method for chemical production |
| CN117058689B (en) * | 2023-10-09 | 2024-02-20 | 巴斯夫一体化基地(广东)有限公司 | Offline detection data processing method for chemical production |
| CN117078689A (en) * | 2023-10-17 | 2023-11-17 | 沈阳宏远电磁线股份有限公司 | Cable defect identification method and system based on machine vision |
| CN117078689B (en) * | 2023-10-17 | 2024-01-30 | 沈阳宏远电磁线股份有限公司 | Cable defect identification method and system based on machine vision |
| CN117132914A (en) * | 2023-10-27 | 2023-11-28 | 武汉大学 | Method and system for identifying large model of universal power equipment |
| CN117132914B (en) * | 2023-10-27 | 2024-01-30 | 武汉大学 | Method and system for identifying large model of universal power equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111797890A (en) | Method and system for detecting defects of power transmission line equipment | |
| CN108009591B (en) | Contact network key component identification method based on deep learning | |
| CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
| CN111507958B (en) | Target detection method, training method of detection model and electronic equipment | |
| CN110084165B (en) | Intelligent identification and early warning method for abnormal events in open scene of power field based on edge calculation | |
| CN103279765B (en) | Steel wire rope surface damage detection method based on images match | |
| Yang et al. | Deep learning‐based bolt loosening detection for wind turbine towers | |
| CN112884747A (en) | Automatic bridge crack detection system integrating cyclic residual convolution and context extractor network | |
| CN115187527B (en) | Separation and identification method for multi-source mixed ultrahigh frequency partial discharge spectrum | |
| CN108711148A (en) | A kind of wheel tyre defect intelligent detecting method based on deep learning | |
| CN114049565A (en) | Geological disaster identification method and device based on remote sensing image and DEM data | |
| CN111881970A (en) | Intelligent outer broken image identification method based on deep learning | |
| CN115471487A (en) | Insulator defect detection model construction and insulator defect detection method and device | |
| CN115171045A (en) | YOLO-based power grid operation field violation identification method and terminal | |
| CN110852318A (en) | Drainage pipeline defect accurate positioning method and system | |
| CN109102486B (en) | Surface defect detection method and device based on machine learning | |
| CN114596244A (en) | Infrared image identification method and system based on visual processing and multi-feature fusion | |
| CN115830302B (en) | Multi-scale feature extraction fusion power distribution network equipment positioning identification method | |
| CN116597411A (en) | Method and system for identifying traffic sign by unmanned vehicle in extreme weather | |
| CN115690659A (en) | Transformer substation safety detection method and device under shielding condition | |
| CN112990350B (en) | Target detection network training method and target detection network-based coal and gangue identification method | |
| CN115273009A (en) | Road crack detection method and system based on deep learning | |
| CN114897858A (en) | Rapid insulator defect detection method and system based on deep learning | |
| CN115937079A (en) | YOLO v 3-based rapid detection method for defects of power transmission line | |
| CN114119562A (en) | Brake disc outer surface defect detection method and system based on deep learning |
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 |