CN109614969A - Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method - Google Patents
Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method Download PDFInfo
- Publication number
- CN109614969A CN109614969A CN201811400394.4A CN201811400394A CN109614969A CN 109614969 A CN109614969 A CN 109614969A CN 201811400394 A CN201811400394 A CN 201811400394A CN 109614969 A CN109614969 A CN 109614969A
- Authority
- CN
- China
- Prior art keywords
- training
- data
- distribution line
- deep learning
- condition
- 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
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000013135 deep learning Methods 0.000 title claims abstract description 22
- 230000008439 repair process Effects 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 54
- 238000005516 engineering process Methods 0.000 claims abstract description 12
- 230000000694 effects Effects 0.000 claims abstract description 11
- 238000004364 calculation method Methods 0.000 claims abstract description 6
- 230000003252 repetitive effect Effects 0.000 claims abstract description 4
- 238000001514 detection method Methods 0.000 claims description 17
- 239000000463 material Substances 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 10
- 238000010200 validation analysis Methods 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 3
- 238000013434 data augmentation Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000007423 decrease Effects 0.000 claims description 2
- 230000005611 electricity Effects 0.000 claims description 2
- 239000003550 marker Substances 0.000 claims description 2
- 238000009877 rendering Methods 0.000 claims description 2
- 230000001360 synchronised effect Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 8
- 238000000605 extraction Methods 0.000 abstract description 4
- 238000004321 preservation Methods 0.000 abstract description 3
- 238000010276 construction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000011897 real-time detection Methods 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Neurology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the extensive distribution wire the condition of a disaster emergency first-aid repair image-recognizing methods based on deep learning, comprising steps of including a large amount of electric pole photo on collection distribution line as training data and verify data；Data are labeled using the marking software of profession；The data set that mark is finished inputs repetitive exercise in specific model system；When meeting to data set precision of prediction, then training finishes preservation parameter；Extensive distribution wire the condition of a disaster emergency first-aid repair image-recognizing method based on deep learning of the invention is different from traditional computer processing technology, realize treatment effect end to end, it improves work efficiency, while feature extraction and calculation complicated inside model, effectively increases the robustness of model.
Description
Technical field
The present invention relates to the computer vision of deep learning and field of image processings, and in particular to based on the big of deep learning
Scale distribution line the condition of a disaster repairs image-recognizing method.
Background technique
Electric pole on power circuit is classified and counted, is the key component of power grid maintenance and construction, directly
Affect the efficiency of power grid maintenance and construction.Various electric pole on power circuit is because of more difficult differentiation, and the material category of use is not
Easily, and it is large number of, it needs professional to take a significant amount of time and examines on the spot, power circuit passes through nobody, geographical ring more
The region of border complexity brings great difficulty to examine on the spot.Replace manpower using the method for Computer Image Processing, to place
Carrying out prospecting in the power circuit of complicated landform can be to avoid these problems.Current electric pole detection system is mostly based on laser
Scanning system has research institute to propose a kind of electric pole automatic identification and localization method based on Vehicle-borne Laser Scanning data, energy
Enough automatic identification electric poles, but this method is affected by geographic factor and environment, can not adapt to environmental aspect complicated and changeable.
Deep learning is an important breakthrough in machine learning field in recent years, it makes computer in voice, image, language
Reason and good sense solution etc. achieves major progress, is widely applied in various fields.Deep learning is applied to match by this project
Electric line detects, and the distribution lines bar in detection image identifies its working condition.Using unmanned plane picture, to picture
In distribution line electric pole identified, classified, significantly reduce manpower burden.Detection model is constructed, experiments prove that
The validity of model.
Summary of the invention
In view of this, the present invention provides based on the extensive of deep learning to solve above-mentioned the problems of the prior art
Distribution line the condition of a disaster repairs image-recognizing method, promotes processing data speed using yolo algorithm, while model accuracy rate is high, energy
The position of electric pole in captured in real-time photo is accurately marked out, this method is had any different in traditional computer processing technology, realized
Treatment effect end to end, improves work efficiency, while feature extraction and calculation complicated inside model, effectively increases mould
The robustness of type.
To achieve the above object, technical scheme is as follows.
Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method, comprising the following steps:
Step 1 carries out tilt and laser scanning progress 3D modeling to distribution line, collects electric pole photo as the knowledge
The training material of other method；
Step 2, in professional annotation tool labelImg, the electric pole station location marker in picture is gone out using the rectangle tool
Come, and manual sort is carried out to electric pole, corresponding location information and class code are generated in annotation tool labelImg；
Step 3, will generate corresponding location information and class code input yolo detection model in be trained, in training process
In, using L2 regularization make neuron weight decay, achieve the effect that prevent and treat over-fitting, while calculate yolo detection model for
The training error of training data and validation error for verify data, rendering error curve；
Step 4, analytical error curve, in training process early period, training error can keep synchronous decline with validation error, rear
In phase training process, training error can tend to reach a limit, and validation error can slowly rise, so when training error and testing
When card error respectively reaches setting ratio, deconditioning system preserves the model parameter that training obtains；
Step 5, using the model parameter preserved as archetype, and using test data as the defeated of the recognition methods
Enter, input is pre-processed, form multiple test objects, which carries out convolutional calculation to multiple test objects respectively
Calculate with pondization, finally predicted, then take the average value of a variety of results, the location information of output distribution line electricity line bar and
Classification information.
Further, specifically included in the step 1: using fire balloon and dirigible to distribution line carry out tilt with
And laser scanning carries out 3D modeling, collects 10,000, resolution ratio is the electric pole photo of 3840*2160 as the recognition methods
Training material.
Further, the form of the electric pole in the step 1 includes normal, bar, disconnected bar and inclination；The step 5
In pretreatment include picture with 10 degree to 20 degree of range continuous overturning clockwise and the pictorial information for cutting out middle section.
Further, include: to the mark of electric pole in the step 2
Normal morphology is labeled as (1,0,0,0)；Inclined configuration is labeled as (0,1,0,0)；The state of falling rod is labeled as (0,0,1,0)；
Disconnected rod state is labeled as (0,0,0,1).
Further, 5% in the training material in the step 1 be yolo detection model accuracy rate verify data, 5%
It is the training data of yolo detection model for the test data of yolo detection model, 90%.
Further, in the step 3 the following steps are included:
Step 31, based on the considerations of training material quantity do not reach training requirement, using the technology of data augmentation, to training
The data of material are extended, the training material and original trained material that are obtained after extension input together yolo detection model into
Row training；
Step 32 enters data into after yolo detection model, using dropout technology, in repetitive exercise each time, with
The probability being set in advance eliminates the neuron in a part of neural network, reduces effect of the single neuron for training effect；
The advantages of step 33 is advanced optimized using Adam optimization algorithm, comprehensive Momentum algorithm and RMSprop algorithm, with
This carrys out the convergence process of Accelerated iteration calculating, more quickly approaches recognition methods globally optimal solution, when validation error gradually declines,
The deconditioning in validation error curve minimum point.
Further, the Data expansion mode in the step 31 includes picture overturning, image cropping, picture hue adjustment
And brightness adjustment.
Compared with the prior art, the extensive distribution line the condition of a disaster of the invention based on deep learning repairs image recognition side
Method is labeled data using the marking software of profession, and the data set that mark is finished inputs in specific model system repeatedly
Generation training, when system meets to data set precision of prediction, then training finishes preservation system parameter；During labeled data,
In order to meet requirement of the model to data volume, initial data is extended using Data expansion technology, and in training pattern
During, over-fitting caused by model excessively complexity is prevented using dropout and L2 regularization；The yolo that the present invention uses is calculated
Method processing data speed is fast, has reached the engine request of unmanned plane captured in real-time detection, while model accuracy rate is high, can accurately mark
Outpour the position of electric pole in captured in real-time photo；This method is had any different in traditional computer processing technology, is realized end and is arrived
The treatment effect at end, improves work efficiency, while feature extraction and calculation complicated inside model, effectively increases the Shandong of model
Stick.
Detailed description of the invention
Fig. 1 is that the process that the extensive distribution line the condition of a disaster of the invention based on deep learning repairs image-recognizing method is shown
It is intended to.
Fig. 2 is model for the error curve of training data and the error curve schematic diagram of verify data.
Specific embodiment
Specific implementation of the invention is described further below in conjunction with attached drawing and specific embodiment.It may be noted that
It is that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.
Embodiment 1
As shown in Figure 1, repairing the stream of image-recognizing method for the extensive distribution line the condition of a disaster of the invention based on deep learning
Journey schematic diagram, comprising the following steps:
Step S100, using unmanned plane in the overhead for being suitble to suitable height, the electric pole photograph of shooting, collecting various forms and classification
Piece.It is clear that picture requires, and electric pole size is suitable, and electric wire rod includes four kinds normal, and inclination, break bar, bar, each form
Requiring can be differentiated with human eye, while more data in order to obtain, manipulate unmanned plane to the same electric pole from different perspectives
It is shot, collected photo is finally divided into training set and verifying collection according to 9 to 1 ratio；
Step S200 carries out the artificial mark of electric pole position and classification to the obtained training set of step S100 and verifying collection photo
Note, comprises the steps that
Step S201, to training set and verifying collection photo building identification label, the last layer output of deep neural network belongs to
Softmax output belongs to probability distribution output, so during constructing picture tag, to normal electric pole, be labeled as (1,
0,0,0)；To inclination electric pole, it is labeled as (0,1,0,0)；To the electric pole that falls down to the ground, it is labeled as (0,0,1,0)；To power-off line bar,
It is labeled as (0,0,0,1)；
Step S202, to the data set that step S100 is obtained, by the marking software labelImg of profession to the electric wire in photo
Bar carries out the mark of position, and specific mark file is generated after mark, and file content includes the class categories of electric pole, and
Position coordinates.
Step S300, the labeled data that step S200 is obtained input the object detection model yolo based on deep learning,
It is iterated to calculate by gradient descent method, when verifying collection error and training set error are all met the requirements, then deconditioning completes task,
It comprises the steps that
Step S301 does not reach training requirement based on the considerations of the quantity of data set, using the technology of data augmentation, to training
The data of collection carry out a series of extension, including picture overturning, image cropping, picture hue adjustment and brightness adjustment, extension
To data and initial data input model is trained together；
Step S302, after step S301 is obtained data input model, using dropout technology, in iteration instruction each time
In white silk, to shift to an earlier date scheduled probability, the neuron in a part of neural network is eliminated, single neuron is reduced and training is imitated
The effect of fruit, while L2 regularization is used, so that weight is realized decaying, avoids over-fitting caused by excessively complexity；
Step S303 advanced optimizes algorithm under the premise of step S302, using Adam optimization algorithm, combines
The advantages of Momentum algorithm and RMSprop algorithm, is carried out the convergence process of Accelerated iteration calculating with this, more quickly approaches the overall situation
Optimal solution collects error curve minimum point deconditioning in verifying, as shown in Figure 2 when the error of training set gradually declines.
Step S400, the model that step S300 is obtained is detected for picture electric pole, picture according to certain size
Trained yolo model marks out the electric pole position in picture to come by the calculating inside model for input, and
Mark classification.
In conclusion the extensive distribution line the condition of a disaster repairing image-recognizing method of the invention based on deep learning uses
The marking software of profession is labeled data, and the data set that mark is finished inputs repetitive exercise in specific model, when
Meet to data set precision of prediction, then training finishes preservation model parameter；During labeled data, in order to meet model pair
The requirement of data volume is extended initial data using Data expansion technology, and during training pattern, uses
Dropout and L2 regularization prevents over-fitting caused by model excessively complexity；The yolo algorithm process data speed that the present invention uses
Degree is fast, has reached the engine request of unmanned plane captured in real-time detection, while model accuracy rate is high, can accurately mark out captured in real-time
The position of electric pole in photo；This method is had any different in traditional computer processing technology, and treatment effect end to end is realized,
It improves work efficiency, while feature extraction and calculation complicated inside model, effectively increases the robustness of model.
Claims (7)
1. the extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method, it is characterised in that: including following step
It is rapid:
Step 1 carries out tilt and laser scanning progress 3D modeling to distribution line, collects electric pole photo as the knowledge
The training material of other method；
Step 2, in professional annotation tool labelImg, the electric pole station location marker in picture is gone out using the rectangle tool
Come, and manual sort is carried out to electric pole, corresponding location information and class code are generated in annotation tool labelImg；
Step 3, will generate corresponding location information and class code input yolo detection model in be trained, in training process
In, using L2 regularization make neuron weight decay, achieve the effect that prevent and treat over-fitting, while calculate yolo detection model for
The training error of training data and validation error for verify data, rendering error curve；
Step 4, analytical error curve, in training process early period, training error can keep synchronous decline with validation error, rear
In phase training process, training error can tend to reach a limit, and validation error can slowly rise, so when training error and testing
When card error respectively reaches setting ratio, deconditioning system preserves the model parameter that training obtains；
Step 5, using the model parameter preserved as archetype, and using test data as the defeated of the recognition methods
Enter, input is pre-processed, form multiple test objects, which carries out convolutional calculation to multiple test objects respectively
Calculate with pondization, finally predicted, then take the average value of a variety of results, the location information of output distribution line electricity line bar and
Classification information.
2. the extensive distribution line the condition of a disaster according to claim 1 based on deep learning repairs image-recognizing method,
Be characterized in that, specifically include in the step 1: using fire balloon and dirigible carries out tilt to distribution line and laser is swept
Carry out 3D modeling is retouched, collects 10,000, training element of the electric pole photo that resolution ratio is 3840*2160 as the recognition methods
Material.
3. the extensive distribution line the condition of a disaster according to claim 1 based on deep learning repairs image-recognizing method,
Be characterized in that: the form of the electric pole in the step 1 includes normal, bar, disconnected bar and inclination；Pre- place in the step 5
Reason includes picture with 10 degree to 20 degree of range continuous overturning clockwise and the pictorial information for cutting out middle section.
4. the extensive distribution line the condition of a disaster according to claim 1 based on deep learning repairs image-recognizing method,
It is characterized in that, includes: to the mark of electric pole in the step 2
Normal morphology is labeled as (1,0,0,0)；Inclined configuration is labeled as (0,1,0,0)；The state of falling rod is labeled as (0,0,1,0)；
Disconnected rod state is labeled as (0,0,0,1).
5. the extensive distribution line the condition of a disaster according to claim 1 based on deep learning repairs image-recognizing method,
Be characterized in that: 5% in training material in the step 1 is the accuracy rate verify data of yolo detection model, and 5% examines for yolo
The test data of model is surveyed, 90% is the training data of yolo detection model.
6. the extensive distribution line the condition of a disaster according to claim 1 based on deep learning repairs image-recognizing method,
Be characterized in that, in the step 3 the following steps are included:
Step 31, based on the considerations of training material quantity do not reach training requirement, using the technology of data augmentation, to training
The data of material are extended, the training material and original trained material that are obtained after extension input together yolo detection model into
Row training；
Step 32 enters data into after yolo detection model, using dropout technology, in repetitive exercise each time, with
The probability being set in advance eliminates the neuron in a part of neural network, reduces effect of the single neuron for training effect；
The advantages of step 33 is advanced optimized using Adam optimization algorithm, comprehensive Momentum algorithm and RMSprop algorithm, with
This carrys out the convergence process of Accelerated iteration calculating, more quickly approaches globally optimal solution, when validation error gradually declines, in verifying mistake
Deconditioning when poor curve minimum point.
7. the extensive distribution line the condition of a disaster according to claim 5 based on deep learning repairs image-recognizing method,
Be characterized in that: the Data expansion mode in the step 31 includes picture overturning, image cropping, picture hue adjustment and brightness tune
It is whole.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811400394.4A CN109614969A (en) | 2018-11-22 | 2018-11-22 | Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811400394.4A CN109614969A (en) | 2018-11-22 | 2018-11-22 | Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109614969A true CN109614969A (en) | 2019-04-12 |
Family
ID=66004370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811400394.4A Pending CN109614969A (en) | 2018-11-22 | 2018-11-22 | Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109614969A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992307A (en) * | 2019-11-04 | 2020-04-10 | 华北电力大学（保定） | Insulator positioning and identifying method and device based on YOLO |
| CN111062925A (en) * | 2019-12-18 | 2020-04-24 | 华南理工大学 | Intelligent cloth defect identification method based on deep learning |
| CN112329870A (en) * | 2020-11-11 | 2021-02-05 | 国网山东省电力公司威海供电公司 | Method for identifying state of transformer substation pressure plate based on YOLO3 algorithm |
| CN112364878A (en) * | 2020-09-25 | 2021-02-12 | 江苏师范大学 | Power line classification method based on deep learning under complex background |
| CN112380944A (en) * | 2020-11-06 | 2021-02-19 | 中国电力科学研究院有限公司 | Method and system for evaluating structural state of transmission tower |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106405640A (en) * | 2016-08-26 | 2017-02-15 | 中国矿业大学(北京) | Automatic microseismic signal arrival time picking method based on depth belief neural network |
| CN108416299A (en) * | 2018-03-09 | 2018-08-17 | 重庆览辉信息技术有限公司 | A kind of power circuit intelligent inspection system |
| EP3376213A1 (en) * | 2017-03-15 | 2018-09-19 | Siemens Aktiengesellschaft | Method and assembly for monitoring the state of a system with operating elements |
-
2018
- 2018-11-22 CN CN201811400394.4A patent/CN109614969A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106405640A (en) * | 2016-08-26 | 2017-02-15 | 中国矿业大学(北京) | Automatic microseismic signal arrival time picking method based on depth belief neural network |
| EP3376213A1 (en) * | 2017-03-15 | 2018-09-19 | Siemens Aktiengesellschaft | Method and assembly for monitoring the state of a system with operating elements |
| CN108416299A (en) * | 2018-03-09 | 2018-08-17 | 重庆览辉信息技术有限公司 | A kind of power circuit intelligent inspection system |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992307A (en) * | 2019-11-04 | 2020-04-10 | 华北电力大学（保定） | Insulator positioning and identifying method and device based on YOLO |
| CN111062925A (en) * | 2019-12-18 | 2020-04-24 | 华南理工大学 | Intelligent cloth defect identification method based on deep learning |
| CN112364878A (en) * | 2020-09-25 | 2021-02-12 | 江苏师范大学 | Power line classification method based on deep learning under complex background |
| CN112380944A (en) * | 2020-11-06 | 2021-02-19 | 中国电力科学研究院有限公司 | Method and system for evaluating structural state of transmission tower |
| CN112380944B (en) * | 2020-11-06 | 2021-12-21 | 中国电力科学研究院有限公司 | Method and system for evaluating structural state of transmission tower based on satellite remote sensing |
| CN112329870A (en) * | 2020-11-11 | 2021-02-05 | 国网山东省电力公司威海供电公司 | Method for identifying state of transformer substation pressure plate based on YOLO3 algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109614969A (en) | Extensive distribution line the condition of a disaster based on deep learning repairs image-recognizing method | |
| CN110543878B (en) | Pointer instrument reading identification method based on neural network | |
| CN111914813B (en) | Power transmission line inspection image naming method and system based on image classification | |
| CN110245644A (en) | A kind of unmanned plane image transmission tower lodging knowledge method for distinguishing based on deep learning | |
| CN104463117B (en) | A kind of recognition of face sample collection method and system based on video mode | |
| CN109902677A (en) | A kind of vehicle checking method based on deep learning | |
| CN109785337A (en) | Mammal counting method in a kind of column of Case-based Reasoning partitioning algorithm | |
| CN109858569A (en) | Multi-tag object detecting method, system, device based on target detection network | |
| CN105787466B (en) | A kind of fine recognition methods and system of type of vehicle | |
| CN111582234B (en) | Large-scale oil tea tree forest fruit intelligent detection and counting method based on UAV and deep learning | |
| CN110084165A (en) | The intelligent recognition and method for early warning of anomalous event under the open scene of power domain based on edge calculations | |
| CN108629288A (en) | A kind of gesture identification model training method, gesture identification method and system | |
| CN109376606A (en) | A kind of electric inspection process image pole and tower foundation fault detection method | |
| Ren et al. | A novel squeeze YOLO-based real-time people counting approach | |
| CN109766873A (en) | A kind of pedestrian mixing deformable convolution recognition methods again | |
| CN109214280A (en) | Shop recognition methods, device, electronic equipment and storage medium based on streetscape | |
| CN108664878A (en) | Pig personal identification method based on convolutional neural networks | |
| CN110781882A (en) | License plate positioning and identifying method based on YOLO model | |
| CN108229524A (en) | A kind of chimney and condensing tower detection method based on remote sensing images | |
| CN112016497A (en) | Single-view Taijiquan action analysis and assessment system based on artificial intelligence | |
| CN115719337A (en) | Wind turbine surface defect detection method | |
| CN113780200A (en) | Computer vision-based pavement multi-disease area detection and positioning method | |
| CN106803084A (en) | A kind of facial characteristics independent positioning method based on end-to-end recirculating network | |
| CN111833439A (en) | Artificial intelligence-based ammunition throwing analysis and mobile simulation training method | |
| Baoyuan et al. | Research on object detection method based on FF-YOLO for complex scenes |
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 |
Application publication date: 20190412 |
|
| RJ01 | Rejection of invention patent application after publication |