CN109190712A - A kind of line walking image automatic classification system of taking photo by plane based on deep learning - Google Patents
A kind of line walking image automatic classification system of taking photo by plane based on deep learning Download PDFInfo
- Publication number
- CN109190712A CN109190712A CN201811103998.2A CN201811103998A CN109190712A CN 109190712 A CN109190712 A CN 109190712A CN 201811103998 A CN201811103998 A CN 201811103998A CN 109190712 A CN109190712 A CN 109190712A
- Authority
- CN
- China
- Prior art keywords
- image
- model
- classification
- line walking
- training
- 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
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
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of line walking image automatic classification system of taking photo by plane based on deep learning.This method comprises: establishing line walking image classification image library and its tag library；Deep learning model is established, the high performance disaggregated model such as including VGGNet, ResNet, DenseNet, NasNet；Train classification models, the input data of iteration all executes data enhancement operations at random every time in training process, including rotation, filled type are cut, gray processing；Trained four models are integrated, integrated model is obtained；Classifying rules is set, image to be detected is inputted into integrated model, takes classification results of first three three image categories of confidence level as image to be detected, copy stores to the server of three classifications and achieves ground；Continuous optimization disaggregated model periodically carries out performance upgrade to automatic classification system.
Description
Technical field
The invention belongs to ultra-high-tension power transmission line line walking technology, image recognition technology, machine learning techniques fields, and in particular to
A kind of line walking image automatic classification system of taking photo by plane based on deep learning.
Background technique
Transmission line of electricity carries the function of electrical energy transportation in electric system, and the fortune inspection maintenance for transmission line of electricity is various regions electricity
The focus of work of power department.In recent years, unmanned plane inspection is increasingly becoming transmission of electricity because its is economical, flexibility and safety
One of the main means of route fortune inspection maintenance.The image that unmanned plane obtains, both can allow ground handling operator straight at the scene
Connect the apparent line fault of confirmation, can also band backhaul inspection maintenance centre and be stored in server, be convenient for subsequent data diagnosis
Analysis.The variety of components of transmission line of electricity is various, and substantial amounts, and the diagnostic analysis cost for depending artificial progress data alone is high, needs
Want intelligent algorithm auxiliary diagnosis.And current intelligent trouble diagnosis algorithm has transmission line part and its locating scene
Higher specific aim and dependence, if occurring in picture to not go out in the characteristics of image isolog and training data of deagnostic package
Similar scene is now crossed, then is easily misidentified.The classification of unmanned plane image at present is mainly according to route and shaft tower number, all kinds of portions
Part mixes storage, greatly the diagnosis difficulty of intelligent algorithm.It therefore, should in line walking image store to server
Classification preservation is carried out according to transmission line part classification, by class diagnostic analysis, promotes the performance of intelligent diagnostics algorithm, while conveniently
Staff's maintenance management data.Unmanned plane line walking image data amount is huge, and the artificial data classification that carries out needs high cost,
How to be classified automatically according to component categories unmanned plane line walking image data, is the current technical issues that need to address.
Summary of the invention
The purpose of the present invention is to provide a kind of the line walking image automatic classification system of taking photo by plane based on deep learning, benefit
Line walking image generic is identified with the Image Classfication Technology based on deep learning, and line walking staff is helped to complete tentatively
Line walking image data is sorted out, and operation management is facilitated, and reduces data diagnosis and analyzes difficulty, establishes base for subsequent further diagnosis
Plinth mitigates person works' intensity and promotes the intelligent level of power-line patrolling.
To achieve the above object, the technical scheme is that a kind of line walking image of taking photo by plane based on deep learning is automatic
Categorizing system includes the following steps:
Step S1, line walking image classification image library and its tag library are established: the unmanned plane line walking picture including different classes of component,
All pictures are divided into eight classes: shaft tower, basis, insulator, earthing or grounding means, affiliated facility, grounded-line, small fitting, big fitting；Wherein
Affiliated facility includes preventive birds harm installation and shaft tower Sign Board, and small fitting includes the fasteners such as screw bolt and nut, and big fitting includes shockproof
Hammer, wire clamp, grading ring and conductor spacer；Every kind of image category all includes failure picture and non-faulting picture, and the two quantity is close, institute
There is picture not distinguished the mark of failure according to eight classifications；
Step S2, it establishes deep learning model: establishing depth convolutional neural networks VGGNet, ResNet, DenseNet, NasNet
High-performance disaggregated model；
Step S3, data enhance: expanding the image library of step S1 using data enhancing technology, increase the content multiplicity of image library
Property, specific practice is passed through an original image in image library including rotation, filled type cutting, the enhancing of gray processing data behaviour
Make, is transformed to a new image；Enhancing operation is only present in training process, and all operations are applied to original with certain probability
On figure, the then input data as this model repetitive exercise；
Step S4, training line walking image classification model: to step S1 establish line walking image classification image library and its tag library into
The division of row training set and test set, the multiple network model established using step S2 is by back-propagation algorithm in training set
Upper training utilizes the data enhancement method lift scheme performance of step S3 in training process, finally obtains the line walking of heterogeneous networks
Image classification model；
Step S5, model integrated: generating deep learning Multi net voting integrated model, and the classification confidence level of heterogeneous networks output is carried out
Weighted average, obtains final classification results；
Step S6, classifying rules is set: image to be detected being input in integrated model, first three three images of confidence level are taken
Classification results of the classification as image to be detected, copy are stored to the archive of three classifications；
Step S7, every several fortune overhaul periods, the number classified the Continuous optimization of disaggregated model: is taken out out of system database
According to after manual review, update line walking image classification image library and its tag library repeat step S2 to step S5, wherein will
The initialization model of step S4 trained disaggregated model before being changed to system, training after, by this trained four
A network model is integrated, the integrated model as system.
In an embodiment of the present invention, in step s 4, training initialization model used is in ImageNet data set
The upper resulting model of training.
Compared to the prior art, the invention has the following advantages: the innovation of the invention consists in that proposing based on defeated
The line walking image data of electric line component categories files mode classification, promotes data management efficiency, reduces line fault and intelligently examines
The identification difficulty of disconnected algorithm.In view of the high cost of manual sort, the present invention devises line walking image Algorithms for Automatic Classification, uses
Depth convolutional neural networks extract characteristics of image, compare traditional image-recognizing method, and this system is cumbersome without engineer
Image characteristics extraction device, but give the task of feature extraction to depth convolutional neural networks, obtain more comprehensively, can more describe to scheme
The depth characteristic information of picture.On this basis, multiple high-performance depth convolutional neural networks are integrated, single Network Recognition is reduced and brings
Risk, promoted line walking image classification accuracy and robustness.The database of line walking image is constantly updated, and transfer learning is utilized
Thought and model fine tuning means, regular Optimum Classification model, make model have inheritability, save training cost, constantly
Upgrade-system.The classification results of this system can be used for line walking data base administration, Analysis on Fault Diagnosis and fault location, be subsequent
Further application lay a good foundation.
Detailed description of the invention
Take photo by plane line walking image automatic classification system flow chart of the Fig. 1 based on deep learning.
Fig. 2 depth convolutional neural networks classification process figure.
Specific embodiment
With reference to the accompanying drawing, technical solution of the present invention is specifically described.
The present invention provides a kind of line walking image automatic classification system of taking photo by plane based on deep learning, includes the following steps:
Step S1, line walking image classification image library and its tag library are established: the unmanned plane line walking picture including different classes of component,
All pictures are divided into eight classes: shaft tower, basis, insulator, earthing or grounding means, affiliated facility, grounded-line, small fitting, big fitting；Wherein
Affiliated facility includes preventive birds harm installation and shaft tower Sign Board, and small fitting includes the fasteners such as screw bolt and nut, and big fitting includes shockproof
Hammer, wire clamp, grading ring and conductor spacer；Every kind of image category all includes failure picture and non-faulting picture, and the two quantity is close, institute
There is picture not distinguished the mark of failure according to eight classifications；
Step S2, it establishes deep learning model: establishing depth convolutional neural networks VGGNet, ResNet, DenseNet, NasNet
High-performance disaggregated model；
Step S3, data enhance: expanding the image library of step S1 using data enhancing technology, increase the content multiplicity of image library
Property, specific practice is passed through an original image in image library including rotation, filled type cutting, the enhancing of gray processing data behaviour
Make, is transformed to a new image；Enhancing operation is only present in training process, and all operations are applied to original with certain probability
On figure, the then input data as this model repetitive exercise；
Step S4, training line walking image classification model: to step S1 establish line walking image classification image library and its tag library into
The division of row training set and test set, the multiple network model established using step S2 is by back-propagation algorithm in training set
Upper training utilizes the data enhancement method lift scheme performance of step S3 in training process, finally obtains the line walking of heterogeneous networks
Image classification model；
Step S5, model integrated: generating deep learning Multi net voting integrated model, and the classification confidence level of heterogeneous networks output is carried out
Weighted average, obtains final classification results；
Step S6, classifying rules is set: image to be detected being input in integrated model, first three three images of confidence level are taken
Classification results of the classification as image to be detected, copy are stored to the archive of three classifications；
Step S7, every several fortune overhaul periods, the number classified the Continuous optimization of disaggregated model: is taken out out of system database
According to after manual review, update line walking image classification image library and its tag library repeat step S2 to step S5, wherein will
The initialization model of step S4 trained disaggregated model before being changed to system, training after, by this trained four
A network model is integrated, the integrated model as system.
In step s 4, training initialization model used is the resulting model of training on ImageNet data set.
The following are specific implementation processes of the invention.
In order to which the purpose of the present invention, technical solution and advantage is more clearly understood, below in conjunction with specific embodiment, and
Referring to attached drawing, the present invention is described in further detail, and system flow chart is as shown in Figure 1, being divided into off-line training step and dividing
Two relatively independent stages of class stage.In off-line training step, the corresponding label file of eight class line walking images passes through
It after data enhancing, is put into established four deep learning models one by one, respectively carries out backpropagation and update model parameter.?
After successive ignition training, the model file for participating in model integrated is preferentially determined, according to standard of each model on eight class line walking images
True rate ranking setting is weighted and averaged weight, generates line walking image classification integrated model, which completes once training
It is transplanted to different regions, uses in different UAV Intelligent detection terminal, be not necessarily to repetition training.At regular intervals, update is patrolled
Line image classification image library and its tag library, the means finely tuned using the thought and model of transfer learning, in classification before
Tuning on model, continually strengthens disaggregated model.In sorting phase, the line walking image that need to will only take photo by plane directly inputs trained classification
In model, the classification results of image can be obtained rapidly, result is finally recorded in the storing place for corresponding to classification on server.
1, line walking image classification image library and its tag library are established.Occurred according to the quantity of transmission line part and its failure
The collected unmanned plane line walking picture comprising different classes of component is divided into eight classes by the frequency and hazard rating: shaft tower, basis,
Insulator, earthing or grounding means, affiliated facility, grounded-line, small fitting, big fitting.Wherein affiliated facility includes preventive birds harm installation and shaft tower
Sign Board, small fitting include the fasteners such as screw bolt and nut, and big fitting includes stockbridge damper, wire clamp, grading ring and conductor spacer.One
Plurality of classes component copies more parts and simultaneously stamps respective labels picture according to the marking principles of figure one kind if it exists.Each classification
It all include failure picture and non-faulting picture, the two quantity is close, and failure picture number accounts for the ratio of current class total number of images
Not less than 40%.The mark of picture does not distinguish specific failure, is all labeled according to eight base parts.
2, deep learning model is established.The classification process of depth convolutional neural networks is as shown in Fig. 2, input picture passes through spy
Sign obtains profound characteristics of image figure after extracting, classifier using the information inference of characteristic pattern go out input picture belong to it is all kinds of
Confidence level.According to depth convolutional neural networks convolutional layer, pond layer, full articulamentum, special function unit various combination, network
It can difference on the recognition performance of different classes of image.The implementation case is using four kinds of high performance networks: VGGNet,
ResNet,DenseNet,NasNet.VGGNet includes 13 layers of 3*3 convolutional layer and 3 layers of full articulamentum, the last layer Softmax
Classification layer；ResNet includes 49 layers of convolutional layer, and other than first layer is 7*7 convolution, remaining is all 1*1 convolution sum 3*3 convolution, special
Different functional unit is residual unit, and the last layer is Softmax classification layer；DenseNet includes 33 layers of convolutional layer, in addition to first
Layer is outside 7*7 convolution, remaining is all 1*1 convolution sum 3*3 convolution, and special function unit is dense connection unit, and the last layer is
Softmax classification layer；NasNet uses Recursive Networks to generate the model description of neural network in the training process, and uses increasing
Strong learning training Recursive Networks, are automatically found suitable neural network structure.
3, data enhance.Enhance technology EDS extended data set using data, increases the content variety of image library, specific practice
It is an original image in database to be transformed to one newly by the data enhancement operations such as rotation, filled type cutting, gray processing
Image.All operations are applied in original image with 0.2 probability, i.e. a picture may be applied a variety of data enhancing behaviour simultaneously
Make.Enhancing is only present in training process, and all operations are applied in original image with predetermined probability, is then changed as this model
The input data of generation training, reduces EMS memory occupation with this.
4, training line walking image classification model.According to 8 to 2 ratio, to line walking image classification image library and its tag library
It is trained the division of collection and test set.Using VGGNet, ResNet, DenseNet, NasNet network model by reversely passing
Algorithm training on training set is broadcast, utilizes data enhancement method lift scheme performance in training process.With training iteration, model
Performance will gradually rise, every the training pattern file of preservation in ten minutes, its classifying quality is verified on test set, preferentially
Preservation model file finally obtains the line walking image classification model of heterogeneous networks.Wherein, training initialization model used be
The resulting model of training on ImageNet data set, parameter update mode is RMSProp, initial learning rate 0.003, momentum system
Number 0.9, batch size 1.
5, model integrated: the classification confidence level that VGGNet, ResNet, DenseNet, NasNet network export is added
Weight average, classification accuracy of each model of the basis of design of weight on test set, by taking insulator part classification as an example: ResNet
Accuracy rate highest in identification insulator classification, when the confidence level weighted average of four network output, ResNet weight is set
It is 0.4, the weight of excess-three network is 0.2, and the insulator confidence level after weighted average is integrated model in insulator class
Final confidence level output on not.
6, classifying rules is set.Image to be detected is input in integrated model, first three three images of confidence level are taken
Classification results of the classification as image to be detected, copy store to the server of three classifications and achieve ground.Server achieves ground
The routine data storing place in the maintenance centre Yun Jian, different from the line walking image classification image library established for training pattern and its
Tag library.In the case where memory space anxiety, confidence threshold value can be set, classification results are filtered, system will take and set
Reliability first three and numerical value are more than 0.5 classification results of three image categories as image to be detected, give up and do not meet confidence level
It is required that recognition result.
7, the Continuous optimization of disaggregated model.Every several fortune overhaul periods, from the server of line walking image automatic classification system
Middle to take out the data classified, data volume is consistent with the data volume that a preceding training pattern uses and data content does not repeat, with
This updates line walking image classification image library and its tag library, and the mode of update is to merge.Line walking image classification image library and its mark
The maximum size in label library is set as 10,000,000 pictures, when reaching the upper limit, abandons and this more amount of new data consistent early stage
Image data.In line walking image classification image library and its tag library in the updated carry out VGGNet, ResNet, DenseNet,
The training of NasNet even deep learning model, initialization model is primary trained disaggregated model before system, after training,
This trained four network model is integrated, the integrated model as system.After model tuning, after optimization
Integrated model subseries again is carried out to the image in system server, after tuning several times, the false recognition rate of system will not
Disconnected decline.Staff only need to check the image category correctness taken out from system server, really when system at regular intervals upgrades
Training data when protecting tuning model each time be it is reliable, without carrying out manual review to each genealogical classification result,
Human cost is saved with this.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made
When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.
Claims (2)
1. a kind of line walking image automatic classification system of taking photo by plane based on deep learning, which comprises the steps of:
Step S1, line walking image classification image library and its tag library are established: the unmanned plane line walking picture including different classes of component,
All pictures are divided into eight classes: shaft tower, basis, insulator, earthing or grounding means, affiliated facility, grounded-line, small fitting, big fitting；Wherein
Affiliated facility includes preventive birds harm installation and shaft tower Sign Board, and small fitting includes the fasteners such as screw bolt and nut, and big fitting includes shockproof
Hammer, wire clamp, grading ring and conductor spacer；Every kind of image category all includes failure picture and non-faulting picture, and the two quantity is close, institute
There is picture not distinguished the mark of failure according to eight classifications；
Step S2, it establishes deep learning model: establishing depth convolutional neural networks VGGNet, ResNet, DenseNet, NasNet
High-performance disaggregated model；
Step S3, data enhance: expanding the image library of step S1 using data enhancing technology, increase the content multiplicity of image library
Property, specific practice is passed through an original image in image library including rotation, filled type cutting, the enhancing of gray processing data behaviour
Make, is transformed to a new image；Enhancing operation is only present in training process, and all operations are applied to original with certain probability
On figure, the then input data as this model repetitive exercise；
Step S4, training line walking image classification model: to step S1 establish line walking image classification image library and its tag library into
The division of row training set and test set, the multiple network model established using step S2 is by back-propagation algorithm in training set
Upper training utilizes the data enhancement method lift scheme performance of step S3 in training process, finally obtains the line walking of heterogeneous networks
Image classification model；
Step S5, model integrated: generating deep learning Multi net voting integrated model, and the classification confidence level of heterogeneous networks output is carried out
Weighted average, obtains final classification results；
Step S6, classifying rules is set: image to be detected being input in integrated model, first three three images of confidence level are taken
Classification results of the classification as image to be detected, copy are stored to the archive of three classifications；
Step S7, every several fortune overhaul periods, the number classified the Continuous optimization of disaggregated model: is taken out out of system database
According to after manual review, update line walking image classification image library and its tag library repeat step S2 to step S5, wherein will
The initialization model of step S4 trained disaggregated model before being changed to system, training after, by this trained four
A network model is integrated, the integrated model as system.
2. a kind of line walking image automatic classification system of taking photo by plane based on deep learning according to claim 1, feature exist
In in step s 4, training initialization model used is the resulting model of training on ImageNet data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811103998.2A CN109190712A (en) | 2018-09-21 | 2018-09-21 | A kind of line walking image automatic classification system of taking photo by plane based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811103998.2A CN109190712A (en) | 2018-09-21 | 2018-09-21 | A kind of line walking image automatic classification system of taking photo by plane based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109190712A true CN109190712A (en) | 2019-01-11 |
Family
ID=64908999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811103998.2A Pending CN109190712A (en) | 2018-09-21 | 2018-09-21 | A kind of line walking image automatic classification system of taking photo by plane based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109190712A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109859190A (en) * | 2019-01-31 | 2019-06-07 | 北京工业大学 | A kind of target area detection method based on deep learning |
| CN109886922A (en) * | 2019-01-17 | 2019-06-14 | 丽水市中心医院 | Hepatocellular carcinoma automatic grading method based on SE-DenseNet deep learning frame and multi-modal Enhanced MR image |
| CN109903840A (en) * | 2019-02-28 | 2019-06-18 | 数坤（北京）网络科技有限公司 | A kind of model integration method and apparatus |
| CN109934222A (en) * | 2019-03-01 | 2019-06-25 | 长沙理工大学 | A kind of insulator chain self-destruction recognition methods based on transfer learning |
| CN110245644A (en) * | 2019-06-22 | 2019-09-17 | 福州大学 | A kind of unmanned plane image transmission tower lodging knowledge method for distinguishing based on deep learning |
| CN110276747A (en) * | 2019-06-06 | 2019-09-24 | 中南大学 | A kind of insulator breakdown detection and failure ranking method based on image analysis |
| CN110334775A (en) * | 2019-07-12 | 2019-10-15 | 广东工业大学 | A kind of recognition methods of unmanned plane line fault and device based on width study |
| CN110598638A (en) * | 2019-09-12 | 2019-12-20 | Oppo广东移动通信有限公司 | Model training method, face gender prediction method, device and storage medium |
| CN110619282A (en) * | 2019-08-26 | 2019-12-27 | 海南撰云空间信息技术有限公司 | Automatic extraction method for unmanned aerial vehicle orthoscopic image building |
| CN110826413A (en) * | 2019-10-10 | 2020-02-21 | 深圳供电局有限公司 | Transmission line damper fault detection method and system and storage medium |
| CN111199539A (en) * | 2019-12-25 | 2020-05-26 | 汕头大学 | Crack detection method based on integrated neural network |
| CN112000094A (en) * | 2020-07-20 | 2020-11-27 | 山东科技大学 | Single-and-double-eye combined high-voltage transmission line hardware fitting online identification and positioning system and method |
| CN113269139A (en) * | 2021-06-18 | 2021-08-17 | 中电科大数据研究院有限公司 | Self-learning large-scale police officer image classification model aiming at complex scene |
| CN113537044A (en) * | 2021-07-14 | 2021-10-22 | 哈尔滨理工大学 | Aircraft engine fault diagnosis method based on STFT and improved DenseNet |
| CN114047195A (en) * | 2021-11-11 | 2022-02-15 | 合肥工业大学智能制造技术研究院 | New energy battery cap defect detection method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426455A (en) * | 2015-11-12 | 2016-03-23 | 中国科学院重庆绿色智能技术研究院 | Method and device for carrying out classified management on clothes on the basis of picture processing |
| CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
| CN107153844A (en) * | 2017-05-12 | 2017-09-12 | 上海斐讯数据通信技术有限公司 | The accessory system being improved to flowers identifying system and the method being improved |
-
2018
- 2018-09-21 CN CN201811103998.2A patent/CN109190712A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426455A (en) * | 2015-11-12 | 2016-03-23 | 中国科学院重庆绿色智能技术研究院 | Method and device for carrying out classified management on clothes on the basis of picture processing |
| CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
| CN107153844A (en) * | 2017-05-12 | 2017-09-12 | 上海斐讯数据通信技术有限公司 | The accessory system being improved to flowers identifying system and the method being improved |
Non-Patent Citations (1)
| Title |
|---|
| 施彦等: "《神经网络设计方法与实例分析》", 31 December 2009 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886922A (en) * | 2019-01-17 | 2019-06-14 | 丽水市中心医院 | Hepatocellular carcinoma automatic grading method based on SE-DenseNet deep learning frame and multi-modal Enhanced MR image |
| CN109886922B (en) * | 2019-01-17 | 2023-08-18 | 丽水市中心医院 | Automatic grading method for hepatocellular carcinoma based on SE-DenseNet deep learning framework and enhanced MR image |
| CN109859190A (en) * | 2019-01-31 | 2019-06-07 | 北京工业大学 | A kind of target area detection method based on deep learning |
| CN109903840B (en) * | 2019-02-28 | 2021-05-11 | 数坤（北京）网络科技有限公司 | Model integration method and device |
| CN109903840A (en) * | 2019-02-28 | 2019-06-18 | 数坤（北京）网络科技有限公司 | A kind of model integration method and apparatus |
| CN109934222A (en) * | 2019-03-01 | 2019-06-25 | 长沙理工大学 | A kind of insulator chain self-destruction recognition methods based on transfer learning |
| CN110276747A (en) * | 2019-06-06 | 2019-09-24 | 中南大学 | A kind of insulator breakdown detection and failure ranking method based on image analysis |
| CN110245644A (en) * | 2019-06-22 | 2019-09-17 | 福州大学 | A kind of unmanned plane image transmission tower lodging knowledge method for distinguishing based on deep learning |
| CN110334775A (en) * | 2019-07-12 | 2019-10-15 | 广东工业大学 | A kind of recognition methods of unmanned plane line fault and device based on width study |
| CN110619282A (en) * | 2019-08-26 | 2019-12-27 | 海南撰云空间信息技术有限公司 | Automatic extraction method for unmanned aerial vehicle orthoscopic image building |
| CN110598638A (en) * | 2019-09-12 | 2019-12-20 | Oppo广东移动通信有限公司 | Model training method, face gender prediction method, device and storage medium |
| CN110826413A (en) * | 2019-10-10 | 2020-02-21 | 深圳供电局有限公司 | Transmission line damper fault detection method and system and storage medium |
| CN110826413B (en) * | 2019-10-10 | 2022-01-25 | 深圳供电局有限公司 | Transmission line damper fault detection method and system and storage medium |
| CN111199539A (en) * | 2019-12-25 | 2020-05-26 | 汕头大学 | Crack detection method based on integrated neural network |
| CN112000094A (en) * | 2020-07-20 | 2020-11-27 | 山东科技大学 | Single-and-double-eye combined high-voltage transmission line hardware fitting online identification and positioning system and method |
| CN113269139A (en) * | 2021-06-18 | 2021-08-17 | 中电科大数据研究院有限公司 | Self-learning large-scale police officer image classification model aiming at complex scene |
| CN113269139B (en) * | 2021-06-18 | 2023-09-26 | 中电科大数据研究院有限公司 | Self-learning large-scale police officer image classification model for complex scene |
| CN113537044A (en) * | 2021-07-14 | 2021-10-22 | 哈尔滨理工大学 | Aircraft engine fault diagnosis method based on STFT and improved DenseNet |
| CN114047195A (en) * | 2021-11-11 | 2022-02-15 | 合肥工业大学智能制造技术研究院 | New energy battery cap defect detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109190712A (en) | A kind of line walking image automatic classification system of taking photo by plane based on deep learning | |
| Pierdicca et al. | Deep convolutional neural network for automatic detection of damaged photovoltaic cells | |
| CN110070530B (en) | Transmission line icing detection method based on deep neural network | |
| CN111709361B (en) | Method for processing inspection data of unmanned aerial vehicle of power transmission line | |
| CN108108736A (en) | A kind of solar energy photovoltaic panel spot identification method | |
| Adou et al. | Insulator faults detection based on deep learning | |
| CN110309884A (en) | Electricity consumption data anomalous identification system based on ubiquitous electric power Internet of Things net system | |
| CN111126820A (en) | Electricity stealing prevention method and system | |
| CN113255661B (en) | Bird species image identification method related to bird-involved fault of power transmission line | |
| CN116191680B (en) | Monitoring management system applied to photovoltaic power generation | |
| Hao et al. | Detection of bird nests on power line patrol using single shot detector | |
| CN105303162A (en) | Target proposed algorithm-based insulator recognition algorithm for aerial images | |
| CN111680759B (en) | Power grid inspection insulator detection classification method | |
| CN115984158A (en) | Defect analysis method and device, electronic equipment and computer readable storage medium | |
| CN117200693A (en) | Photovoltaic module fault diagnosis method, device, equipment and storage medium | |
| CN112000708A (en) | Abnormal data processing method and system based on regulation and control adapted data fusion | |
| CN111209955A (en) | Airplane power supply system fault identification method based on deep neural network and random forest | |
| CN116029440A (en) | Ultra-short-term power prediction method and device for photovoltaic power station | |
| CN113822373B (en) | Image classification model training method based on integration and knowledge distillation | |
| CN114169623A (en) | Power equipment fault analysis method and device, electronic equipment and storage medium | |
| CN112669264A (en) | Artificial intelligence defect identification method and system for unmanned aerial vehicle routing inspection of distribution network line | |
| CN113920424A (en) | Method and device for extracting visual objects of power transformation inspection robot | |
| CN113657443A (en) | Online Internet of things equipment identification method based on SOINN network | |
| Rustia et al. | Tree-based deep convolutional neural network for hierarchical identification of low-resolution insect images | |
| CN117173448B (en) | Method and device for intelligently controlling and early warning progress of foundation engineering |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190111 |