CN110133443A - Based on the transmission line part detection method of parallel vision, system, device - Google Patents
Based on the transmission line part detection method of parallel vision, system, device Download PDFInfo
- Publication number
- CN110133443A CN110133443A CN201910468525.0A CN201910468525A CN110133443A CN 110133443 A CN110133443 A CN 110133443A CN 201910468525 A CN201910468525 A CN 201910468525A CN 110133443 A CN110133443 A CN 110133443A
- Authority
- CN
- China
- Prior art keywords
- transmission line
- component
- model
- image
- line part
- 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
- 230000005540 biological transmission Effects 0.000 title claims abstract description 126
- 238000001514 detection method Methods 0.000 title claims abstract description 94
- 238000012360 testing method Methods 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 48
- 230000005611 electricity Effects 0.000 claims abstract description 46
- 238000000034 method Methods 0.000 claims abstract description 46
- 230000000007 visual effect Effects 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 230000007613 environmental effect Effects 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000007689 inspection Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000012986 modification Methods 0.000 claims description 3
- 230000004048 modification Effects 0.000 claims description 3
- 239000011800 void material Substances 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 4
- 238000007726 management method Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 7
- 238000013461 design Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000033228 biological regulation Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013145 classification model Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 210000004218 nerve net Anatomy 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 239000012212 insulator Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/085—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution lines, e.g. overhead
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to power system managements and detection field, and in particular to a kind of based on the transmission line part detection method of parallel vision, system, device, it is intended to solve the problems, such as that data acquisition is at high cost, model performance is limited and cannot adjust update in time.The method of the present invention includes: to obtain each component video image of transmission line of electricity;Pass through trained target detection model, obtaining widget bounding box size, position and component generic;Output block bounding box size, position and component generic.The scene that one aspect of the present invention is combined by establishing actual situation, extends the training and test sample of each component of transmission line of electricity, reduces the cost of data acquisition, improve the performance of model;On the other hand by the method for on-line study, constantly reach better detection effect using new data more new model.
Description
Technical field
The invention belongs to Operation of Electric Systems management and detection fields, and in particular to a kind of transmission of electricity based on parallel vision
Circuit parts detection method, system, device.
Background technique
Chinese Regional is vast, has very high requirement to electric system normal operation, guarantees that the normal operation of transmission line of electricity is closed
The life and work of hundreds of millions compatriots.Transmission line of electricity power transmission distance is long, and ultra-high-tension power transmission line is widely used, it is desirable that electric power overhaul
Personnel guarantee the normal use of transmission line of electricity at work, but due to extraneous crucial complicated and changeable, naturally destruction and transmission of electricity
Route various pieces itself may there is also problem of materials lead to transmission line malfunction.Therefore, transmission line of electricity is carried out timely
Maintenance it is extremely important, artificial power transmission lines overhauling used to be carried out, be on the one hand limited to human resources and staff peace
On the one hand full limitation is unable to satisfy timely even online inspection transmission line status again.
The core of computer vision is the image data set progress vision mode for relying on large-scale tape label at this stage
Study and assessment, but in being detected to complicated transmission line status, it is difficult to it is collected from actual scene sufficient high-quality
It measures real image data and artificial mark expense is high, it is difficult to which the detection met under complex situations wants data set with monitoring
It asks.The extensive picture number collected in transmission line of electricity as particular detection component is needed for the detection of transmission line of electricity specific objective
According to, but for complicated and diversified transmission line part, the data such as data collection, training pattern are carried out for single specific objective
Collection creation is still very difficult.The image data set that transmission line of electricity is obtained by unmanned plane or other photographic devices, by artificial
It is labeled, carries out image recognition and target detection using convolutional neural networks (CNN), YOLO algorithm is applied in transmission line of electricity
In insulator breakdown detection, using the picture construction data set of actual acquisition, good achievement is achieved, however this side operator
According to procurement cost height, model performance is limited [1].Using long-distance monitorng device, shape is run by video images detection transmission line of electricity
State improves optimization to the network model of existing target detection for breaking down or there are the feature of image of risk,
The pond ROI layer is increased to image characteristics extraction network and has modified loss function, is trained using great amount of samples, is obtained extensive
Effect is good, can detecte out the vision mode of transmission line status, however model cannot be with the perfect timely tune of data set
Whole update [2].Parallel vision is established on actual scene and artificial scene, is a kind of intelligent vision calculating side of actual situation interaction
Method.By the artificial scene that building color is true to nature, the environmental condition being likely to occur in actual scene is simulated, and automatically derive essence
True markup information combines the actual scene data set of large-scale artificial scene data set and appropriate scale, can train
More effective machine learning and vision computation model utilize artificial scene, are able to carry out various experiments with computing, thoroughly evaluating vision
Validity of the algorithm under complex environment, or the free parameter [3] of optimal setting model.
Following documents is technical background data related to the present invention:
[1] Yang Xiaoxu, Wen Zhaoyang deep learning are in the research and application [J] in electric transmission line isolator fault detection
State's new traffic, 2018 (10)
[2] broken image recognition technology [J] outside transmission line of electricity of Zhang Ji, Yu Juan, Wang Jinli, the et al. based on deep learning
Computer system application, 27 (8)
[3]Xing Y,Lv C,Chen L,et al.Advances in Vision-Based Lane Detection:
Algorithms,Integration,Assessment,and Perspectives on ACP-Based Parallel
Vision[J].IEEE/CAA Journal of Automatica Sinica,2018,5(3):645-661.
Summary of the invention
In order to solve the above problem in the prior art, i.e., data acquisition is at high cost, model performance is limited and cannot be timely
The problem of adjustment updates, the present invention provides a kind of transmission line part detection methods based on parallel vision, comprising:
Step S10 obtains each component video image of transmission line of electricity as data to be tested;
Step S20 passes through trained target detection mould to the video image of each component in the data to be tested
Type, obtaining widget bounding box size, position and component generic, status information;
Step S30, the part boundaries frame size, position and component generic, status information are testing result;
Wherein, the target detection model is constructed based on deep neural network, is instructed according to the following steps to each component
Practice:
Step B10 obtains training set, test set;The training set includes multiple training samples, and the training sample includes
Component virtual image and its corresponding label information;The test set includes multiple test samples, and the test sample includes portion
Part real scene image and its corresponding label information;
Step B20 randomly selects a training sample, a test sample from the training set, test set respectively, in parallel
Target detection model is inputted, the fitting degree of training sample and test sample testing result is calculated, and passes through deep neural network
Capacity adaptive characteristic optimization aim detection model model parameter;
Step B30 repeats step B20 until the training sample and the fitting degree of test sample testing result are big
In preset threshold value or reach preset frequency of training, obtains trained target detection model.
In some preferred embodiments, the virtual image subset, acquisition methods are as follows:
Step T10 constructs three-dimensional artificial environment, the model for the transmission line part that creation needs to detect;
It is imitative to carry out artificial scene for step T20, the model based on the three-dimensional artificial environment and each transmission line part
Very, and virtual photographic device is constructed, obtains the image and corresponding component locations at each visual angle of model of each transmission line part
And classification information, as component virtual image subset.
In some preferred embodiments, " the three-dimensional artificial environment and each transmission line part are based in step T20
Model, carry out artificial scene simulation ", method are as follows:
Based on three-dimensional artificial environment, carry out the environmental element emulation of real scene, reappear environmental factor, make virtual scene with
The similarity of real scene is greater than preset value.
In some preferred embodiments, in step T20 " virtual photographic device is constructed, each transmission line part is obtained
The image and corresponding component locations and classification information at each visual angle of model ", method are as follows:
Virtual photographic device is constructed, turns two-dimentional interface based on three-dimensional, photographic device is to each component in imitation real scene
Image Acquisition is carried out the acquisition of each image of component in virtual scene using virtual photographic device, obtains each transmission line part
The image and corresponding component locations and classification information at each visual angle of model.
In some preferred embodiments, the real scene image subset, acquisition methods are as follows:
Using the image of each component in the method for unmanned plane and/or manual inspection acquisition real scene, as real scene
Image subset.
In some preferred embodiments, the artificial markup information, comprising:
The corresponding target component bounding box size of image, position and component generic, status information.
It is additionally provided with after " obtaining trained target detection model " in some preferred embodiments, in step B30
The step of on-time model updates, method are as follows:
Step G10 carries out assessment test by testing the set pair analysis model, whether is judgment models performance at preset time point
Decline;
Step G20 terminates when step G10 judging result is "No";When judging result is "Yes", in the net of model
The level for increasing preset quantity after network calculates the fitting degree of the model modification image set of acquisition, and passes through depth nerve net
The performance of the capacity adaptive characteristic lift scheme of network.
Another aspect of the present invention proposes a kind of transmission line part detection system based on parallel vision, including defeated
Enter module, module of target detection, output module;
The input module is configured to obtain each component video image of transmission line of electricity as data to be tested and input;
The module of target detection is configured to the video image to each component in data to be tested, by training
Target detection model, obtaining widget bounding box size, position and component generic, status information;
The output module is configured to part boundaries frame size, position and the component generic that will acquire, state letter
Breath output.
The third aspect of the present invention proposes a kind of storage device, wherein be stored with a plurality of program, described program be suitable for by
Processor is loaded and is executed to realize the above-mentioned transmission line part detection method based on parallel vision.
The fourth aspect of the present invention proposes a kind of processing unit, including processor, storage device;The processor is fitted
In each program of execution;The storage device is suitable for storing a plurality of program;Described program be suitable for loaded by processor and executed with
Realize the above-mentioned transmission line part detection method based on parallel vision.
Beneficial effects of the present invention:
(1) the present invention is based on the transmission line part detection methods of parallel vision, are based on parallel theories of vision, on the one hand logical
Creation transmission line of electricity 3D simulating scenes are crossed, endlessly expert along training data are provided for transmission line faultlocating, compared to huge
Artificial mark cost, effectively reduce artificial mark cost, greatly improve database creation efficiency;On the other hand artificial transmission of electricity
Route scene can according to transmission line of electricity and critical component it is practical change be adjusted, guarantee detection model fitting effect and its
Generalization ability;The third aspect provides the unified platform that a kind of collection depth learning network is trained and assesses, and is effectively promoted and is guaranteed
Performance of the depth network model in transmission line faultlocating realizes the transmission line part detection of multiple target, while can not also
The disconnected update and extension for carrying out component detection range.
(2) the present invention is based on the transmission line part detection methods of parallel vision, are sat using each component of positioning three-dimensional scenic
The method for marking information obtains band mark image data set, and acquisition and creation to the data set of particular elements research have important reality
Meaning.
(3) the present invention is based on the transmission line part detection methods of parallel vision, are based on national standard specification creation component
Model, build transmission line of electricity scene and can be synchronized with relevant national standard Regulations carry out model library adjustment and scene more
Newly, so that artificial scene is closer in true.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is the techniqueflow schematic diagram of the transmission line part detection method the present invention is based on parallel vision;
Fig. 2 is a kind of examining for target for embodiment of the transmission line part detection method the present invention is based on parallel vision
The YOLO algorithm model schematic diagram of survey.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is only used for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to just
Part relevant to related invention is illustrated only in description, attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
The present invention provides a kind of transmission line part detection method based on parallel vision, imitative by creation transmission line of electricity 3D
True scene provides endlessly expert along training data for transmission line faultlocating, and is capable of providing a kind of collection deep learning net
The unified platform of network training and assessment effectively promotes and guarantees performance of the depth network model in transmission line faultlocating, real
The transmission line part detection of existing multiple target, while can also constantly carry out the update and extension of component detection range.
A kind of transmission line part detection method based on parallel vision of the invention, comprising:
Step S10 obtains each component video image of transmission line of electricity as data to be tested;
Step S20 passes through trained target detection mould to the video image of each component in the data to be tested
Type, obtaining widget bounding box size, position and component generic, status information;
Step S30, the part boundaries frame size, position and component generic, status information are testing result;
Wherein, the target detection model is constructed based on deep neural network, is instructed according to the following steps to each component
Practice:
Step B10 obtains training set, test set;The training set includes multiple training samples, and the training sample includes
Component virtual image and its corresponding label information;The test set includes multiple test samples, and the test sample includes portion
Part real scene image and its corresponding label information;
Step B20 randomly selects a training sample, a test sample from the training set, test set respectively, in parallel
Target detection model is inputted, the fitting degree of training sample and test sample testing result is calculated, and passes through deep neural network
Capacity adaptive characteristic optimization aim detection model model parameter;
Step B30 repeats step B20 until the training sample and the fitting degree of test sample testing result are big
In preset threshold value or reach preset frequency of training, obtains trained target detection model.
In order to be more clearly illustrated to the transmission line part detection method the present invention is based on parallel vision, tie below
It closes Fig. 1 and step each in embodiment of the present invention method is unfolded to be described in detail.
The transmission line part detection method based on parallel vision of an embodiment of the present invention, including step S10- step
S30, each step are described in detail as follows:
Step S10 obtains each component video image of transmission line of electricity as data to be tested.
Video image is exactly the sequence of continuous still image, is that a kind of pair of objective things are more vivid, vivo describes.
In one embodiment of the present of invention, using the image of the method for unmanned plane and/or manual inspection acquisition each component of transmission line of electricity.
Step S20 passes through trained target detection mould to the video image of each component in the data to be tested
Type, obtaining widget bounding box size, position and component generic, status information.
Image classification, detection and segmentation are three big tasks of computer vision field.Image classification model is to draw image
It is divided into single classification, generally corresponds to object most outstanding in image.But many pictures of real world generally comprise not only
One object, at this time if using image classification model be image distribute a single label be in fact it is very coarse, not
Accurately.In the case of such, it is necessary to use target detection model, target detection model can identify the multiple of a picture
Object, and different objects (providing bounding box) can be oriented.Target detection is widely used in many scenes, as nobody drives
Sail with security system etc..
Target detection has many methods, for example Haar feature+Adaboost algorithm of conventional method, Hog feature+SVM are calculated
Method (SVM, Support Vector Machine), DPM algorithm (DPM, Deformable Part Model) etc., for example be based on
Serial (RCNN, the Regions with Convolutional Neural Network) algorithm of the RCNN of deep neural network,
YOLO series (YOLO, You Only Look Once) etc..It is constructed in one embodiment of the present of invention based on deep neural network
YOLO network model, and training is optimized to it, target detection is carried out using the network after optimization training.
As shown in Fig. 2, for the present invention is based on a kind of use of embodiment of the transmission line part detection method of parallel vision
It is used to extract the convolutional network model of feature in the YOLO algorithm of target detection, YOLO network architecture reference GooLeNet model,
Conv.Layer represents convolutional layer, and lower number is convolution size, and Maxpool Layer represents maximum pond layer, under be most
Great Chiization parameter, each module bottom represent every layer of port number, and the number of horizontal edge and longitudinal edge represents every layer of length and width, internal
Module represents convolution, around number represent the length and width of convolution kernel, after braces × * represents and omit and repeat identical layer and carry out
Feature extraction.
The target detection model is constructed based on deep neural network, is trained according to the following steps to each component:
Step B10 obtains training set, test set;The training set includes multiple training samples, and the training sample includes
Component virtual image and its corresponding label information;The test set includes multiple test samples, and the test sample includes portion
Part real scene image and its corresponding label information.
The artificial markup information, comprising:
The corresponding target component bounding box size of image, position and component generic, status information.
The virtual image subset, acquisition methods are as follows:
Step T10 constructs three-dimensional artificial environment, the model for the transmission line part that creation needs to detect.
In one embodiment of the invention, three-dimensional artificial environment, the main mesh of Delta3D are constructed using Delta3D game engine
Mark is to provide the API function library of a set of simple possible, the fundamental for building any visual software is constituted, by power transmission line
Road real scene designing draft, is built according to scene sketches, threedimensional model is exported, for setting the base of transmission line of electricity background
This pattern scene has thus generated the artificial background of transmission line of electricity environment.
Transmission line of electricity single component detect object library according to the existing different overhead transmission line designing technique regulations of country into
Row fitting model builds library, and carries out the simulation of overhead transmission line and build, comprising: " 200kV~500kV compact is maked somebody a mere figurehead defeated
Electric line designing technique regulation " (DL/T 5217-2013), " repeating ice shelf sky Transmission Line Design technical regulation " (DL/T
5440-2009), " high voltage direct current overhead transmission line design discipline " (DL5479-2015), " mining influence area overhead transmission line
Road design specification " (DL/T5539-2018), " 110Kv~750kV overhead transmission line design specification " (GB 50545-2010),
" 1000kV overhead transmission line design specification " (GB 50665-2011) etc..
The transmission line of electricity critical component failure then transmission line of electricity according to detection required for corresponding national standard specification creation
The model of trouble unit, comprising: " power transmission and transformation primary equipment defect classification standard " (QGDW1906-2013 standard etc..
Each component of power delivery circuit and critical component fault message, the target and phase detected required for covering comprehensively as far as possible
Answer the various patterns of target object.
It is imitative to carry out artificial scene for step T20, the model based on the three-dimensional artificial environment and each transmission line part
Very, and virtual photographic device is constructed, obtains the image and corresponding component locations at each visual angle of model of each transmission line part
And classification information, as component virtual image subset.
Current many 3D scene of game production, which have reached, shows current three-dimensional scenic rendering very close to true effect
It can accomplish that, very close to real world, to break through current power transmission lines overhauling bottleneck, this method is managed based on parallel vision
By, a kind of method for proposing transmission line faultlocating, creation transmission line of electricity three-dimensional simulation scene acquisition data training transmission line of electricity inspection
Survey grid network realizes the detection and monitoring of timely and safe transmission line status.
In one embodiment of the invention, building for the artificial scene of transmission line of electricity is carried out using 3DMax software, by transmission line of electricity
Mode input is to artificial scene and creates, and when being input to artificial scene, inputs each component of transmission line of electricity in artifact background
Each position records its classification information and location information.
" it is imitative to carry out artificial scene for the model based on the three-dimensional artificial environment and each transmission line part in step T20
Very ", method are as follows:
Based on three-dimensional artificial environment, carry out the environmental element emulation of real scene, reappear environmental factor, make virtual scene with
The similarity of real scene is greater than preset value.
In one embodiment of the invention, using 3DMax software, true field is carried out to the artificial transmission line of electricity scene created
The elements simulation of scape reappears the environmental factors such as corresponding illumination, weather, and compares real scene and rendered, and creation is as far as possible
Close to the artificial scene of true environment.
In step T20 " virtual photographic device is constructed, the image at each visual angle of model of each transmission line part and right is obtained
The component locations and classification information answered ", method are as follows:
Virtual photographic device is constructed, turns two-dimentional interface based on three-dimensional, photographic device is to each component in imitation real scene
Image Acquisition is carried out the acquisition of each image of component in virtual scene using virtual photographic device, obtains each transmission line part
The image and corresponding component locations and classification information at each visual angle of model.
In one embodiment of the invention, turns the interface API of 2D using the correspondence 3D of OpenGL, can imitate in reality and carry out
The observation of transmission line faultlocating generates the image at all angles visual angle.It is equivalent to and is imaged in artificial scenario building virtual machine
Network, observation transmission line of electricity scene that can be comprehensive flexibly acquire in scene for simulating virtual condition observation transmission line of electricity
The case where, the video image data of extensive multiplicity is generated from each visual angle.3D, which turns 2D, then can directly be carried out using computer
Coordinate transform extracts target position and classification information.
The real scene image subset, acquisition methods are as follows:
Identical as the acquisition modes of each component detection process video image of above-mentioned power delivery circuit, details are not described herein.
Transmission line of electricity data belong to specific high-risk tasks in areas content, and image pattern is not as the target in daily life
Sample image is the same to be obtained easily from internet.Due to the limitation of data set, for transmission line part and critical component
Failure carry out state-detection task, simple data enhancement methods, be unable to satisfy occur in transmission line faultlocating it is varied
The problem of with complicated shape.
For this insufficient problem of training data, the present invention is based on the theories of parallel vision, establish the scene of actual situation combination,
For assisting generating the image data of extensive tape label vision mode training.
Step B20 randomly selects a training sample, a test sample from the training set, test set respectively, in parallel
Target detection model is inputted, the fitting degree of training sample and test sample testing result is calculated, and passes through deep neural network
Capacity adaptive characteristic optimization aim detection model model parameter.
Step B30 repeats step B20 until the training sample and the fitting degree of test sample testing result are big
In preset threshold value or reach preset frequency of training, obtains trained target detection model.
The step of on-time model updates is additionally provided with after " obtaining trained target detection model " in step B30,
Method are as follows:
Step G10 carries out assessment test by testing the set pair analysis model, whether is judgment models performance at preset time point
Decline.
Step G20 terminates when step G10 judging result is "No";When judging result is "Yes", in the net of model
The level for increasing preset quantity after network calculates the fitting degree of the model modification image set of acquisition, and passes through depth nerve net
The performance of the capacity adaptive characteristic lift scheme of network.
Step S30, the part boundaries frame size, position and component generic, status information are testing result.
Transmission line of electricity image large database concept can be created using the method for the present invention:
Firstly, controllably acquiring artificial scene analog sample image according to concrete application scene, artificial transmission line of electricity is created
Image pattern library.
Then, constantly expand unmanned plane or the collected true transmission line of electricity sample graph of manual inspection in practical application scene
Picture, as true picture sample database.
Third step, when there is the variation of component standards and increase in the variation of relevant specification of country, synchronized update 3D transmission of electricity
Route critical component model library and artificial scene, are periodically updated database.
Finally, artificial, true transmission line of electricity image pattern library data take dynamic change and expansion according to practical application scene
Fill scheme, creation dynamic, controllable artificial/authentic specimen database.
The transmission line part detection system based on parallel vision of second embodiment of the invention, including input module, mesh
Mark detection module, output module;
The input module is configured to obtain each component video image of transmission line of electricity as data to be tested and input;
The module of target detection is configured to the video image to each component in data to be tested, by training
Target detection model, obtaining widget bounding box size, position and component generic, status information;
The output module is configured to part boundaries frame size, position and the component generic that will acquire, state letter
Breath output.
Person of ordinary skill in the field can be understood that, for convenience and simplicity of description, foregoing description
The specific work process of system and related explanation, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
It should be noted that the transmission line part detection system provided by the above embodiment based on parallel vision, only with
The division of above-mentioned each functional module carries out for example, in practical applications, can according to need and by above-mentioned function distribution by
Different functional modules is completed, i.e., by the embodiment of the present invention module or step again decompose or combine, for example, above-mentioned
The module of embodiment can be merged into a module, can also be further split into multiple submodule, described above to complete
All or part of function.For module involved in the embodiment of the present invention, the title of step, it is only for distinguish each mould
Block or step, are not intended as inappropriate limitation of the present invention.
A kind of storage device of third embodiment of the invention, wherein being stored with a plurality of program, described program is suitable for by handling
Device is loaded and is executed to realize the above-mentioned transmission line part detection method based on parallel vision.
A kind of processing unit of fourth embodiment of the invention, including processor, storage device;Processor is adapted for carrying out each
Program;Storage device is suitable for storing a plurality of program;Described program is suitable for being loaded by processor and being executed to realize above-mentioned base
In the transmission line part detection method of parallel vision.
Person of ordinary skill in the field can be understood that, for convenience and simplicity of description, foregoing description
The specific work process and related explanation of storage device, processing unit, can refer to corresponding processes in the foregoing method embodiment,
Details are not described herein.
Those skilled in the art should be able to recognize that, mould described in conjunction with the examples disclosed in the embodiments of the present disclosure
Block, method and step, can be realized with electronic hardware, computer software, or a combination of the two, software module, method and step pair
The program answered can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electric erasable and can compile
Any other form of storage well known in journey ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field is situated between
In matter.In order to clearly demonstrate the interchangeability of electronic hardware and software, in the above description according to function generally
Describe each exemplary composition and step.These functions are executed actually with electronic hardware or software mode, depend on technology
The specific application and design constraint of scheme.Those skilled in the art can carry out using distinct methods each specific application
Realize described function, but such implementation should not be considered as beyond the scope of the present invention.
Term " includes " or any other like term are intended to cover non-exclusive inclusion, so that including a system
Process, method, article or equipment/device of column element not only includes those elements, but also including being not explicitly listed
Other elements, or further include the intrinsic element of these process, method, article or equipment/devices.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this field
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from this
Under the premise of the principle of invention, those skilled in the art can make equivalent change or replacement to the relevant technologies feature, these
Technical solution after change or replacement will fall within the scope of protection of the present invention.
Claims (10)
1. a kind of transmission line part detection method based on parallel vision, which is characterized in that the transmission line part detection side
Method includes:
Step S10 obtains each component video image of transmission line of electricity as data to be tested;
Step S20, to the video image of each component in the data to be tested, by trained target detection model,
Obtaining widget bounding box size, position and component generic, status information;
Step S30, the part boundaries frame size, position and component generic, status information are testing result;
Wherein, the target detection model is constructed based on deep neural network, is trained according to the following steps to each component:
Step B10 obtains training set, test set;The training set includes multiple training samples, and the training sample includes component
Virtual image and its corresponding label information;The test set includes multiple test samples, and the test sample includes that component is true
Real scene image and its corresponding label information;
Step B20 randomly selects a training sample, a test sample from the training set, test set respectively, parallel to input
Target detection model calculates the fitting degree of training sample and test sample testing result, and the appearance for passing through deep neural network
Measure the model parameter of adaptive characteristic optimization aim detection model;
Step B30 repeats step B20 until the fitting degree of the training sample and test sample testing result is greater than in advance
If threshold value or reach preset frequency of training, obtain trained target detection model.
2. the transmission line part detection method according to claim 1 based on parallel vision, which is characterized in that the void
Quasi- image subset, acquisition methods are as follows:
Step T10 constructs three-dimensional artificial environment, the model for the transmission line part that creation needs to detect;
Step T20, the model based on the three-dimensional artificial environment and each transmission line part carry out artificial scene simulation, and
Virtual photographic device is constructed, image and corresponding component locations and the classification at each visual angle of model of each transmission line part are obtained
Information, as component virtual image subset.
3. the transmission line part detection method according to claim 2 based on parallel vision, which is characterized in that step
" model based on the three-dimensional artificial environment and each transmission line part, carry out artificial scene simulation ", method in T20
Are as follows:
Based on three-dimensional artificial environment, the environmental element emulation of real scene is carried out, environmental factor is reappeared, makes virtual scene and true
The similarity of scene is greater than preset value.
4. the transmission line part detection method according to claim 2 based on parallel vision, which is characterized in that step
In T20 " virtual photographic device is constructed, the image and corresponding component locations at each visual angle of model of each transmission line part are obtained
And classification information ", method are as follows:
Virtual photographic device is constructed, turns two-dimentional interface based on three-dimensional, image of the photographic device to each component in imitation real scene
Acquisition is carried out the acquisition of each image of component in virtual scene using virtual photographic device, obtains the model of each transmission line part
The image at each visual angle and corresponding component locations and classification information.
5. the transmission line part detection method according to claim 1 based on parallel vision, which is characterized in that described true
Real field scape image subset, acquisition methods are as follows:
Using the image of each component in the method for unmanned plane and/or manual inspection acquisition real scene, as real scene image
Subset.
6. the transmission line part detection method according to claim 1 based on parallel vision, which is characterized in that the people
Work markup information, comprising:
The corresponding target component bounding box size of image, position and component generic, status information.
7. the transmission line part detection method according to claim 1 based on parallel vision, which is characterized in that step
The step of on-time model updates, method are additionally provided with after " obtaining trained target detection model " in B30 are as follows:
Step G10 carries out assessment test by testing the set pair analysis model, whether judgment models performance declines at preset time point;
Step G20 terminates when step G10 judging result is "No";When judging result be "Yes" when, model network it
The level for increasing preset quantity afterwards calculates the fitting degree of the model modification image set of acquisition, and passes through deep neural network
The performance of capacity adaptive characteristic lift scheme.
8. a kind of transmission line part detection system based on parallel vision, which is characterized in that including input module, target detection
Module, output module;
The input module is configured to obtain each component video image of transmission line of electricity as data to be tested and input;
The module of target detection is configured to the video image to each component in data to be tested, passes through trained mesh
Mark detection model, obtaining widget bounding box size, position and component generic, status information;
The output module, part boundaries frame size, position and component generic, the status information for being configured to will acquire are defeated
Out.
9. a kind of storage device, wherein being stored with a plurality of program, which is characterized in that described program is suitable for being loaded and being held by processor
Row is to realize the described in any item transmission line part detection methods based on parallel vision of claim 1-7.
10. a kind of processing unit, including
Processor is adapted for carrying out each program;And
Storage device is suitable for storing a plurality of program;
It is characterized in that, described program is suitable for being loaded by processor and being executed to realize:
The described in any item transmission line part detection methods based on parallel vision of claim 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910468525.0A CN110133443B (en) | 2019-05-31 | 2019-05-31 | Power transmission line component detection method, system and device based on parallel vision |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910468525.0A CN110133443B (en) | 2019-05-31 | 2019-05-31 | Power transmission line component detection method, system and device based on parallel vision |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110133443A true CN110133443A (en) | 2019-08-16 |
CN110133443B CN110133443B (en) | 2020-06-16 |
Family
ID=67583280
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910468525.0A Active CN110133443B (en) | 2019-05-31 | 2019-05-31 | Power transmission line component detection method, system and device based on parallel vision |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110133443B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110738164A (en) * | 2019-10-12 | 2020-01-31 | 北京猎户星空科技有限公司 | Part abnormity detection method, model training method and device |
CN111461209A (en) * | 2020-03-30 | 2020-07-28 | 深圳市凯立德科技股份有限公司 | Model training device and method |
CN112949849A (en) * | 2021-01-27 | 2021-06-11 | 武汉星巡智能科技有限公司 | Method and device for optimizing intelligent camera detection model by adopting edge calculation |
CN113569432A (en) * | 2021-09-22 | 2021-10-29 | 南通宏耀锅炉辅机有限公司 | Simulation detection method and system for liquid-air-tight element |
CN114264911A (en) * | 2022-03-03 | 2022-04-01 | 成都市克莱微波科技有限公司 | Power amplifier reliability test method and test system |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104978580A (en) * | 2015-06-15 | 2015-10-14 | 国网山东省电力公司电力科学研究院 | Insulator identification method for unmanned aerial vehicle polling electric transmission line |
CN106895824A (en) * | 2017-01-24 | 2017-06-27 | 国网四川省电力公司检修公司 | Unmanned plane localization method based on computer vision |
CN108010030A (en) * | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
US20190050667A1 (en) * | 2017-03-10 | 2019-02-14 | TuSimple | System and method for occluding contour detection |
CN109753929A (en) * | 2019-01-03 | 2019-05-14 | 华东交通大学 | A kind of united high-speed rail insulator inspection image-recognizing method of picture library |
CN109801284A (en) * | 2019-01-25 | 2019-05-24 | 华中科技大学 | A kind of high iron catenary insulator breakdown detection method based on deep learning |
CN109816724A (en) * | 2018-12-04 | 2019-05-28 | 中国科学院自动化研究所 | Three-dimensional feature extracting method and device based on machine vision |
-
2019
- 2019-05-31 CN CN201910468525.0A patent/CN110133443B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104978580A (en) * | 2015-06-15 | 2015-10-14 | 国网山东省电力公司电力科学研究院 | Insulator identification method for unmanned aerial vehicle polling electric transmission line |
CN106895824A (en) * | 2017-01-24 | 2017-06-27 | 国网四川省电力公司检修公司 | Unmanned plane localization method based on computer vision |
US20190050667A1 (en) * | 2017-03-10 | 2019-02-14 | TuSimple | System and method for occluding contour detection |
CN108010030A (en) * | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
CN109816724A (en) * | 2018-12-04 | 2019-05-28 | 中国科学院自动化研究所 | Three-dimensional feature extracting method and device based on machine vision |
CN109753929A (en) * | 2019-01-03 | 2019-05-14 | 华东交通大学 | A kind of united high-speed rail insulator inspection image-recognizing method of picture library |
CN109801284A (en) * | 2019-01-25 | 2019-05-24 | 华中科技大学 | A kind of high iron catenary insulator breakdown detection method based on deep learning |
Non-Patent Citations (4)
Title |
---|
WENWEN ZHANG,等: "A parallel vision approach to scene-specific pedestrian detection", 《NEUROCOMPUTING》 * |
王坤峰,等: "平行图像:图像生成的一个新型理论框架", 《模式识别与人工智能》 * |
王坤峰,等: "平行视觉:基于ACP的智能视觉计算方法", 《自动化学报》 * |
翟永杰,等: "平行视觉框架下深度卷积神经网络可视化分析", 《计算机工程与应用》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110738164A (en) * | 2019-10-12 | 2020-01-31 | 北京猎户星空科技有限公司 | Part abnormity detection method, model training method and device |
CN110738164B (en) * | 2019-10-12 | 2022-08-12 | 北京猎户星空科技有限公司 | Part abnormity detection method, model training method and device |
CN111461209A (en) * | 2020-03-30 | 2020-07-28 | 深圳市凯立德科技股份有限公司 | Model training device and method |
CN111461209B (en) * | 2020-03-30 | 2024-04-09 | 深圳市凯立德科技股份有限公司 | Model training device and method |
CN112949849A (en) * | 2021-01-27 | 2021-06-11 | 武汉星巡智能科技有限公司 | Method and device for optimizing intelligent camera detection model by adopting edge calculation |
CN112949849B (en) * | 2021-01-27 | 2024-03-26 | 武汉星巡智能科技有限公司 | Method and device for optimizing intelligent camera detection model by adopting edge calculation |
CN113569432A (en) * | 2021-09-22 | 2021-10-29 | 南通宏耀锅炉辅机有限公司 | Simulation detection method and system for liquid-air-tight element |
CN114264911A (en) * | 2022-03-03 | 2022-04-01 | 成都市克莱微波科技有限公司 | Power amplifier reliability test method and test system |
CN114264911B (en) * | 2022-03-03 | 2022-05-20 | 成都市克莱微波科技有限公司 | Power amplifier reliability test method and test system |
Also Published As
Publication number | Publication date |
---|---|
CN110133443B (en) | 2020-06-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110133443A (en) | Based on the transmission line part detection method of parallel vision, system, device | |
CN109643125A (en) | For training the 3D virtual world true to nature of automated driving system to create and simulation | |
CN109615611A (en) | A kind of insulator self-destruction defect inspection method based on inspection image | |
CN106920243A (en) | The ceramic material part method for sequence image segmentation of improved full convolutional neural networks | |
CN110378408A (en) | Power equipment image-recognizing method and device based on transfer learning and neural network | |
CN112819830A (en) | Individual tree crown segmentation method based on deep learning and airborne laser point cloud | |
CN110135437A (en) | For the damage identification method of vehicle, device, electronic equipment and computer storage medium | |
Lu et al. | P_SegNet and NP_SegNet: New neural network architectures for cloud recognition of remote sensing images | |
CN109712127A (en) | A kind of electric transmission line fault detection method for patrolling video flowing for machine | |
CN108280852A (en) | A kind of door and window point cloud shape detecting method and system based on laser point cloud data | |
Wu et al. | Detection of foreign objects intrusion into transmission lines using diverse generation model | |
CN109840905A (en) | Power equipment rusty stain detection method and system | |
Uggla et al. | Towards synthesized training data for semantic segmentation of mobile laser scanning point clouds: Generating level crossings from real and synthetic point cloud samples | |
CN115186473A (en) | Scene perception modeling and verifying method based on parallel intelligence | |
CN108717522A (en) | A kind of human body target tracking method based on deep learning and correlation filtering | |
CN109272560A (en) | A kind of power line vectorization method and device | |
Xia et al. | Neural detection of foreign objects for transmission lines in power systems | |
Li et al. | Speed regulation of overhead catenary system inspection robot for high-speed railway through reinforcement learning | |
Liu | Multidimensional nonlinear landscape planning based on parameter feature extraction and multimedia technology | |
CN112883785A (en) | Railway operation environment safety image detection system based on parallel vision theory | |
CN109670556B (en) | Global heat source heavy industry area identification method based on fire point and noctilucent data | |
Niu et al. | [Retracted] Analysis of the 3D Application Evaluation System of Landscape Based on Hybrid Cooperation of VR and AI | |
Alaya et al. | CNN-based tree model extraction | |
Bajcsy et al. | Image filtering-A context dependent process | |
Elliott et al. | An application of faster R-CNNs in object detection efforts applied to the critical infrastructure domain |
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 |