CN109916912A - A kind of railway rail clip Defect inspection method and system - Google Patents
A kind of railway rail clip Defect inspection method and system Download PDFInfo
- Publication number
- CN109916912A CN109916912A CN201910255993.XA CN201910255993A CN109916912A CN 109916912 A CN109916912 A CN 109916912A CN 201910255993 A CN201910255993 A CN 201910255993A CN 109916912 A CN109916912 A CN 109916912A
- Authority
- CN
- China
- Prior art keywords
- railway rail
- rail clip
- convolutional neural
- neural networks
- 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
Abstract
The embodiment of the invention provides a kind of railway rail clip Defect inspection method and system, method includes: acquisition railway rail clip image to be detected；By in the convolutional neural networks after the railway rail clip image input training to be detected, output obtains the original output result of the convolutional neural networks；Based on preset decision criterion, detection differentiation is carried out to the original output result, determines that there are the target railway rail clip images of disease.A kind of railway rail clip Defect inspection method and system provided in an embodiment of the present invention automatically learn fastener Disease Characters using the powerful adaptive learning ability of convolutional neural networks, to realize the function of automatic detection fastener disease, detection speed is fast and accurate.
Description
Technical field
The present embodiments relate to rail traffic security technology area more particularly to a kind of railway rail clip Defect inspection methods
And system.
Background technique
Currently, public transport just turns to curing period from the construction period, and rail is as railway with the high speed development of road traffic
The basis of operation ensures that rail is safe and reliable, is the important process of railway routine servicing.Railway rail clip is connection rail and sleeper
Centre part, in the operation of railway, it may appear that fastener setup error, fastener lose, tripping etc. diseases, will cause rail pine
It is dynamic, railway normal pass is influenced, or even cause safety accident.It is to guarantee railway peace how where discovery disease promptly and accurately
The important process of battalion for the national games.
In the prior art, artificial eye is mostly used to check fastener, this detection method can spend a large amount of manpower and time.
Also have through computer vision come by way of being detected, such as led to using gray scale and to carry out Canny edge detection and carry out contacting piece
Location determination is carried out, but this mode detection effect is poor, robustness is low.
Therefore, the new railway rail clip Defect inspection method of one kind is needed now to solve the above problems.
Summary of the invention
To solve the above-mentioned problems, the embodiment of the present invention provides one kind and overcomes the above problem or at least be partially solved
State a kind of railway rail clip Defect inspection method and system of problem.
The first aspect embodiment of the present invention provides a kind of railway rail clip Defect inspection method, comprising:
Acquire railway rail clip image to be detected；
By in the convolutional neural networks after the railway rail clip image input training to be detected, output obtains the convolution
The original output result of neural network；
Based on preset decision criterion, detection differentiation is carried out to the original output result, determines that there are diseases
Target railway rail clip image.
The embodiment of the invention provides a kind of railway rail clip Defect inspection systems for second aspect, comprising:
Acquisition module, for acquiring railway rail clip image to be detected；
Neural network output module, for the convolutional Neural net after training the railway rail clip image input to be detected
In network, output obtains the original output result of the convolutional neural networks；
Decision discrimination module carries out detection to the original output result and sentences for being based on preset decision criterion
Not, determine that there are the target railway rail clip images of disease.
The embodiment of the invention provides a kind of electronic equipment for the third aspect, comprising:
Processor, memory, communication interface and bus；Wherein, the processor, memory, communication interface pass through described
Bus completes mutual communication；The memory is stored with the program instruction that can be executed by the processor, the processor
Described program instruction is called to be able to carry out above-mentioned railway rail clip Defect inspection method.
The embodiment of the invention provides a kind of non-transient computer readable storage medium, the non-transient calculating for fourth aspect
Machine readable storage medium storing program for executing stores computer instruction, and the computer instruction makes the computer execute above-mentioned railway rail clip disease inspection
Survey method.
A kind of railway rail clip Defect inspection method and system provided in an embodiment of the present invention are powerful using convolutional neural networks
Adaptive learning ability, automatically learn fastener Disease Characters, to realize the function of automatic detection fastener disease, detection speed
Degree is fast and accurate.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is a kind of railway rail clip Defect inspection method flow schematic diagram provided in an embodiment of the present invention；
Fig. 2 is detection image result schematic diagram provided in an embodiment of the present invention；
Fig. 3 is the convolutional neural networks training process signal provided in an embodiment of the present invention for detecting railway rail clip disease
Figure；
Fig. 4 is a kind of railway rail clip Defect inspection system structure diagram provided in an embodiment of the present invention；
Fig. 5 is the structural block diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical solution in the embodiment of the present invention is explicitly described, it is clear that described embodiment is the present invention
A part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not having
Every other embodiment obtained under the premise of creative work is made, shall fall within the protection scope of the present invention.
Fig. 1 is a kind of railway rail clip Defect inspection method flow schematic diagram provided in an embodiment of the present invention, as shown in Figure 1,
Include:
101, railway rail clip image to be detected is acquired；
102, by the convolutional neural networks after the railway rail clip image input training to be detected, output obtains described
The original output result of convolutional neural networks；
103, it is based on preset decision criterion, detection differentiation is carried out to the original output result, determines there is disease
Harmful target railway rail clip image.
It should be noted that the application scenarios of the embodiment of the present invention are the fastener Defect inspections for being directed to rail safety.Iron
Rail fastener disease it is common such as losing, button, anti-loaded, loosening, foreign matter, there are crack and defect.Losing button refers to railway rail clip de-
Loss is fallen, anti-loaded to refer to fastener installation direction mistake, loosening refers to fastener loosened screw, causes fastener to contact with rail not tight,
Foreign matter refers to that railway rail clip nearby has extra object.It is understood that there are multiple types, any one types for fastener disease
The disease of type occurs will affect the safe operation of rail traffic.
For above-mentioned scene, the embodiment of the invention provides a kind of railway rail clip Defect inspection based on convolutional neural networks
Method, specifically, in a step 101, the embodiment of the present invention can carry out image taking to the position that rail needs to detect first, it can
It include image fastener in the image of shooting, there may be diseases may also not have disease for the fastener, and the present invention is real with understanding
It applies example and the picture of shooting is known as railway rail clip image to be detected.Preferably, conventional number used in the embodiment of the present invention
Image acquisition mode can spend the acquisition of smaller and data to be relatively easy to.
Further, in a step 102, the embodiment of the present invention trains railway rail clip image to be detected input in advance
Convolutional neural networks in carry out image detection, output obtains the result of feature detection.Preferably, the embodiment of the present invention is exported
Target railway rail clip image in the fastener position for disease occur can be labeled automatically, notation methods can be used on the image
It loops, the modes such as picture frame, this embodiment of the present invention is not especially limited.
It should be noted that convolutional neural networks provided in an embodiment of the present invention are previously according to history railway rail clip image
What training obtained, by having carried out supervision to preset convolutional neural networks using history railway rail clip image as training sample set
Study, until obtained model parameter meets the condition of convergence.Then last model parameter reconstruction model is saved, is finally used
Convolutional neural networks after the training of detection.
The classification mark of the target area to be matched can be so exported by the convolutional neural networks after step 102 training
Label and confidence level, i.e., the original output in the embodiment of the present invention as a result, finally in step 103, mention according to embodiments of the present invention
The default decision criterion supplied evaluates the target railway rail clip image of disease.Fig. 2 is inspection provided in an embodiment of the present invention
Altimetric image result schematic diagram, in Fig. 2, R represents rail, and A represents left-hand fastener, and B represents dextrad fastener, and C represents nuts, D
The retaining ring position for having defect is represented, E represents reversed fastener, it should be noted that convolution is refreshing in a step 102 for the embodiment of the present invention
Original output result through network will export directly in the form of character class and character coordinates.Such as: this anti-loaded disease of detection
When evil, the target area that the target category comprising high confidence level is " counter to fill fastener ", ginseng directly are extracted in original output result
There is " the reversed installation " of E classification in the result so exported according to Fig. 2, output result is E6.When detecting fastener missing, equally
It is direct extract the target area that the target category comprising high confidence level is " missing fastener " in original output result, referring to figure
Exist in 2 results so exported " the retaining ring position for having defect " of D classification, output result is D4.Detection fastener loosens, bullet
When spring part does not achieve the effect that trapped orbit, according to primitive network export in fastener, fixing bolt and track position sit
Whether the relative positional relationship comprehensive descision of mark three, which belongs to fastener, loosens, and spring section does not achieve the effect that trapped orbit,
The position of each retaining ring, track, nut cap relative positional relationship can be assessed in the result so exported referring to Fig. 2, if
Any combination form, such as the combination of A3, C3, R1 exceed preset threshold range, then judge that retaining ring exists and loosen disease, differentiate
As a result it is loosened for A3.
A kind of railway rail clip Defect inspection method provided in an embodiment of the present invention utilizes powerful adaptive of convolutional neural networks
Answer learning ability, automatically learn fastener Disease Characters, to realize the function of automatic detection fastener disease, detection speed it is fast and
Accurately.
On the basis of the above embodiments, the convolutional neural networks after the input training by the railway rail clip image
In before, the method also includes:
The convolutional neural networks are trained using the history railway rail clip image marked as training sample set.
Fig. 3 is the convolutional neural networks training process signal provided in an embodiment of the present invention for detecting railway rail clip disease
Figure, as shown in figure 3, model training successively includes data processing section and model training part, data processing section is mainly wrapped
It includes target mark and training sample set test sample collection divides.Model training process successively include input data, data enhancing,
Basic network, non-maximum restraining, calculates loss, evaluation index, model preservation according to prediction result at additional networks.It is understood that
, training process is the process of an iterative cycles, the continuous adjusting training parameter in training, etc. after the completion of iteration, save
Final model parameter, the convolutional neural networks after training needed for obtaining the embodiment of the present invention.
By the content of above-described embodiment it is found that the embodiment of the present invention is in order to realize the detection to railway rail clip disease, in advance
A convolutional neural networks are had trained to extract the Disease Characters in railway rail clip image, complete the detection of disease.
I.e. the embodiment of the present invention substantially before testing, needs in advance to be trained convolutional neural networks, makes its inspection
The accuracy rate meet demand of survey.The training process embodiment of the present invention uses the history railway rail clip image marked.It needs
Whether bright, history railway rail clip image is different from railway rail clip image to be detected, be manual identified and clear to have deposited
It is marked in the image of fastener disease, and in advance, marked content mainly includes Damage Types and the disease position of the image.Into one
Step, history rail image can be divided into according to a certain percentage training sample set and test sample collection by the embodiment of the present invention, instruction
Practice sample set and be used to be iterated the convolutional neural networks training, test sample collection is for detection model training effect when training
Fruit, effect are measured with the index of AP (average precision) and MAP (mean average precision).When repeatedly
After the completion of generation, the convolutional neural networks after obtaining the training in the embodiment of the present invention.
On the basis of the above embodiments, the history railway rail clip image that will have been marked is as training sample set to institute
Convolutional neural networks are stated to be trained, comprising:
Change of scale and data enhancing are carried out to the training sample set, obtained and the convolutional neural networks input format
The first data set to match；
Based on first data set, training is iterated to the convolutional neural networks, the model ginseng after being trained
Number.
By the content of above-described embodiment it is found that the embodiment of the present invention is using the history railway rail clip image marked as training
Sample set is trained convolutional neural networks.Specifically, the embodiment of the present invention first is become before being trained by scale
Change commanders training sample set Format adjusting for the benefit of convolutional neural networks read format, while by data enhancing acquirement it is most fast
Change of scale and the enhanced sample set of data are known as the first data set by optimal training effect, the embodiment of the present invention.
Further, it is based on first data set, carries out feature extraction by the feature extraction functions of convolutional neural networks.
The result of feature extraction is carried out classifier to predict, obtains final prediction result.
On the basis of the above embodiments, described that change of scale and data enhancing are carried out to the training sample set, it obtains
The first data set to match with the convolutional neural networks input format, comprising:
The training sample set is converted into lmdb formatted data；
Luminosity transformation and geometric transformation are carried out to the lmdb formatted data, obtain inputting lattice with the convolutional neural networks
The first data set that formula matches.
Convolutional Neural is met it is found that the embodiment of the present invention needs to convert training sample set to by the content of above-described embodiment
The format that network output requires.
Specifically, the embodiment of the present invention can convert the image data of training sample set to lmdb formatted data, lmdb lattice
Formula data are a kind of high performance data storage method, the reading conducive to convolutional neural networks to data.
Further, the embodiment of the present invention carries out luminosity transformation and geometric transformation to data.Wherein luminosity transformation include with
Machine adjusts the one or more of brightness, contrast, form and aspect and saturation degree.Geometric transformation include Stochastic propagation, random cropping and
Mirror image is one or more at random.It is converted by above-mentioned luminosity and geometric transformation, training sample set can be enhanced, improve convolution mind
Learning effect through network.
On the basis of the above embodiments, described to be based on first data set, it changes to the convolutional neural networks
Generation training, the model parameter after being trained, comprising:
An additional networks are connected, in the output of the convolutional neural networks with composing training model；
First data set is inputted in the training pattern, multiple prediction results are obtained；
Based on non-maximum restraining method, optimal result is chosen from the multiple prediction result.
By the content of above-described embodiment it is found that the embodiment of the present invention has carried out repetitive exercise to convolutional neural networks, obtain
Trained model parameter.
Specifically, to will use the good basic network of pre-training first in training additional plus one for the embodiment of the present invention
New network.The convolutional neural networks of basic network, that is, original, for the parameter in basic network, without random assignment, but
Using the good parameter of pre-training, so that model reduces operand while guaranteeing feature extraction.The output of basic network is made
For input, to the additional networks that a size is gradually successively decreased, which does the convolution algorithm as basic network, to constitute instruction
Practice model.It should be noted that additional networks initial value needs in building are randomly generated.
Further, after training sample set is inputted training pattern by the embodiment of the present invention, the feature of different scale can be chosen
Figure is predicted, and generates various sizes of default frame.Preferably, the embodiment of the present invention chooses six various sizes of features
Figure, while according to size difference, default frame of different sizes is generated as unit of pixel, and predict default frame is generated.
The last embodiment of the present invention rejects extra prediction result using non-maximum restraining method, retains optimal prediction.
It is on the basis of the above embodiments, described that training is iterated to the convolutional neural networks, further includes:
Difference based on predefined loss function, between predictive metrics result and legitimate reading.
The embodiment of the present invention defines a loss function in the training process and comes predictive metrics result and legitimate reading
Difference.
The formula of loss function are as follows:
Wherein, (Lconf(c, x)) it is that confidence is lost, (Lloc(x, l, g)) it is positioning loss, N is matched default frame
Number, x indicate whether matched frame belongs to classification P, and I is prediction block, and g is true frame, and c is that selected frame target belongs to setting for classification p
Reliability, α indicate weight.
Further, the formula of loss is positioned are as follows:
Wherein:
The formula of confidence loss are as follows:
So according to above-mentioned loss function can in repetitive exercise continuous adjusting parameter, constantly reduce prediction result and true
The distance between real result, to achieve the effect that fitting.After the completion of repetitive exercise, parameter and reconstruction model are saved, is obtained most
Convolutional neural networks after training eventually.
On the basis of the above embodiments, the method also includes:
Obtain history railway rail clip image；
Coordinate system is established in the history railway rail clip image and marks out the coordinate position and disease classification of fastener, is obtained
To the history railway rail clip image marked.
It is understood that need to be labeled the disease in history railway rail clip image can be into for the embodiment of the present invention
Row training.Marked content mainly includes the position mark of disease and the classification mark of disease, so that convolutional neural networks
It is able to carry out study.
Preferably, the embodiment of the present invention marks out the coordinate position of fastener as origin using the image upper left corner and provides disease
Classification, and corresponding xml document is generated, to obtain the history railway rail clip image marked.It should be noted that can select
Coordinate label is carried out with other origins, this embodiment of the present invention is not especially limited.
By the content of above-described embodiment it is found that the embodiment of the present invention is using the partial data of history railway rail clip image as instruction
Practice sample set, partial data is tested model performance as test sample collection.
Test result shows that railway rail clip Defect inspection method provided in an embodiment of the present invention can learn the height to target
Grade feature, has better robustness, and be able to carry out the study of multiple target, and taking leave of traditional detection method can only be to single mesh
Mark is differentiated.While providing classification, the coordinate position of classification has been also indicated that.Make classification judgement and positioning combination one
It rises, it is more preferable to help the daily O&M of rail.
Fig. 4 is a kind of railway rail clip Defect inspection system structure diagram provided in an embodiment of the present invention, as shown in figure 4,
It include: acquisition module 401, neural network output module 402 and decision discrimination module 403, in which:
Acquisition module 401 is for acquiring railway rail clip image to be detected；
Neural network output module 402 is used for the convolutional Neural after the railway rail clip image input training to be detected
In network, output obtains the original output result of the convolutional neural networks；
Decision discrimination module 403 is used to be based on preset decision criterion, detects to the original output result
Differentiate, determines that there are the target railway rail clip images of disease.
It is specific how to be used by acquisition module 401, neural network output module 402 and decision discrimination module 403
In the technical solution for executing railway rail clip Defect inspection embodiment of the method shown in FIG. 1, it is similar that the realization principle and technical effect are similar,
Details are not described herein again.
A kind of railway rail clip Defect inspection system provided in an embodiment of the present invention utilizes powerful adaptive of convolutional neural networks
Answer learning ability, automatically learn fastener Disease Characters, to realize the function of automatic detection fastener disease, detection speed it is fast and
Accurately.
On the basis of the above embodiments, the system also includes:
Training module, the history railway rail clip image for that will mark is as training sample set to the convolutional Neural net
Network is trained.
On the basis of the above embodiments, the training module includes:
Data pre-processing unit enhances for carrying out change of scale and data to the training sample set, obtain with it is described
The first data set that convolutional neural networks input format matches；
Repetitive exercise unit is iterated training to the convolutional neural networks, obtains for being based on first data set
Model parameter after to training.
On the basis of the above embodiments, the data pre-processing unit is specifically used for:
The training sample set is converted into lmdb formatted data；
Luminosity transformation and geometric transformation are carried out to the lmdb formatted data, obtain inputting lattice with the convolutional neural networks
The first data set that formula matches.
On the basis of the above embodiments, the repetitive exercise unit is specifically used for:
An additional networks are connected, in the output of the convolutional neural networks with composing training model；
First data set is inputted in the training pattern, multiple prediction results are obtained；
Based on non-maximum restraining method, optimal result is chosen from the multiple prediction result.
On the basis of the above embodiments, the repetitive exercise unit is also used to:
Difference based on predefined loss function, between predictive metrics result and legitimate reading.
On the basis of the above embodiments, the system also includes:
Labeling module, for obtaining history railway rail clip image；
And coordinate system is established in the history railway rail clip image and marks out the coordinate position and disease class of fastener
Not, the history railway rail clip image marked is obtained.
The embodiment of the present invention provides a kind of electronic equipment, comprising: at least one processor；And with the processor communication
At least one processor of connection, in which:
Fig. 5 is the structural block diagram of electronic equipment provided in an embodiment of the present invention, referring to Fig. 5, the electronic equipment, comprising:
Processor (processor) 501, communication interface (Communications Interface) 502, memory (memory) 503
With bus 504, wherein processor 501, communication interface 502, memory 503 complete mutual communication by bus 504.Place
Reason device 501 can call the logical order in memory 503, to execute following method: acquiring railway rail clip image to be detected；
By in the convolutional neural networks after the railway rail clip image input training to be detected, output obtains the convolutional neural networks
Original output result；Based on preset decision criterion, detection differentiation is carried out to the original output result, determines and exists
The target railway rail clip image of disease.
The embodiment of the present invention discloses a kind of computer program product, and the computer program product is non-transient including being stored in
Computer program on computer readable storage medium, the computer program include program instruction, when described program instructs quilt
When computer executes, computer is able to carry out method provided by above-mentioned each method embodiment, for example, acquisition is to be detected
Railway rail clip image；By in the convolutional neural networks after the railway rail clip image input training to be detected, output obtains institute
State the original output result of convolutional neural networks；Based on preset decision criterion, the original output result is examined
It surveys and differentiates, determine that there are the target railway rail clip images of disease.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, the non-transient computer readable storage
Medium storing computer instruction, the computer instruction make the computer execute side provided by above-mentioned each method embodiment
Method, for example, acquire railway rail clip image to be detected；By the volume after the railway rail clip image input training to be detected
In product neural network, output obtains the original output result of the convolutional neural networks；It is right based on preset decision criterion
The original output result carries out detection differentiation, determines that there are the target railway rail clip images of disease.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of railway rail clip Defect inspection method characterized by comprising
Acquire railway rail clip image to be detected；
By in the convolutional neural networks after the railway rail clip image input training to be detected, output obtains the convolutional Neural
The original output result of network；
Based on preset decision criterion, detection differentiation is carried out to the original output result, determines that there are the targets of disease
Railway rail clip image.
2. the method according to claim 1, wherein after the input training by the railway rail clip image
Before in convolutional neural networks, the method also includes:
The convolutional neural networks are trained using the history railway rail clip image marked as training sample set.
3. according to the method described in claim 2, it is characterized in that, the history railway rail clip image that will have been marked is as instruction
Practice sample set to be trained the convolutional neural networks, comprising:
Change of scale and data enhancing are carried out to the training sample set, obtained and the convolutional neural networks input format phase
The first data set matched；
Based on first data set, training is iterated to the convolutional neural networks, the model parameter after being trained.
4. according to the method described in claim 3, it is characterized in that, described carry out change of scale sum number to the training sample set
According to enhancing, the first data set to match with the convolutional neural networks input format is obtained, comprising:
The training sample set is converted into lmdb formatted data；
Luminosity transformation and geometric transformation are carried out to the lmdb formatted data, obtained and the convolutional neural networks input format phase
Matched first data set.
5. according to the method described in claim 3, it is characterized in that, described be based on first data set, to the convolution mind
It is iterated training through network, the model parameter after being trained, comprising:
An additional networks are connected, in the output of the convolutional neural networks with composing training model；
First data set is inputted in the training pattern, multiple prediction results are obtained；
Based on non-maximum restraining method, optimal result is chosen from the multiple prediction result.
6. according to the method described in claim 5, it is characterized in that, described be iterated training to the convolutional neural networks,
Further include:
Difference based on predefined loss function, between predictive metrics result and legitimate reading.
7. according to the method described in claim 2, it is characterized in that, the method also includes:
Obtain history railway rail clip image；
Coordinate system is established in the history railway rail clip image and marks out the coordinate position and disease classification of fastener, is marked
The history railway rail clip image being poured in.
8. a kind of railway rail clip Defect inspection system characterized by comprising
Acquisition module, for acquiring railway rail clip image to be detected；
Neural network output module, for the convolutional neural networks after training the railway rail clip image input to be detected
In, output obtains the original output result of the convolutional neural networks；
Decision discrimination module carries out detection differentiation to the original output result, really for being based on preset decision criterion
Surely there is the target railway rail clip image of disease.
9. a kind of electronic equipment, which is characterized in that including memory and processor, the processor and the memory pass through always
Line completes mutual communication；The memory is stored with the program instruction that can be executed by the processor, the processor tune
The method as described in claim 1 to 7 is any is able to carry out with described program instruction.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction makes the computer execute method as described in any one of claim 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910255993.XA CN109916912A (en) | 2019-04-01 | 2019-04-01 | A kind of railway rail clip Defect inspection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910255993.XA CN109916912A (en) | 2019-04-01 | 2019-04-01 | A kind of railway rail clip Defect inspection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109916912A true CN109916912A (en) | 2019-06-21 |
Family
ID=66967797
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910255993.XA Pending CN109916912A (en) | 2019-04-01 | 2019-04-01 | A kind of railway rail clip Defect inspection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109916912A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378885A (en) * | 2019-07-19 | 2019-10-25 | 王晓骁 | A kind of focal area WSI automatic marking method and system based on machine learning |
| CN110533640A (en) * | 2019-08-15 | 2019-12-03 | 北京交通大学 | Based on the track circuit disease discrimination method for improving YOLOv3 network model |
| CN110717464A (en) * | 2019-10-15 | 2020-01-21 | 中国矿业大学(北京) | Intelligent railway roadbed disease identification method based on radar data |
| CN112308073A (en) * | 2020-11-06 | 2021-02-02 | 中冶赛迪重庆信息技术有限公司 | Method, system, equipment and medium for identifying loading and unloading transshipment state of scrap steel train |
| CN115063416A (en) * | 2022-08-08 | 2022-09-16 | 苏州立创致恒电子科技有限公司 | Rail fastener state detection method and system |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105574872A (en) * | 2015-12-16 | 2016-05-11 | 南京铁道职业技术学院 | Track fastener detection method based on computer image and visual identity |
| CN105699381A (en) * | 2016-03-28 | 2016-06-22 | 成都唐源电气有限责任公司 | Rail fastener detection device based on machine vision |
| CN205555965U (en) * | 2016-04-25 | 2016-09-07 | 西安科技大学 | Portable folding dolly |
| CN106290378A (en) * | 2016-08-23 | 2017-01-04 | 东方晶源微电子科技（北京）有限公司 | Defect classification method and defect inspecting system |
| CN107563412A (en) * | 2017-08-09 | 2018-01-09 | 浙江大学 | A kind of infrared image power equipment real-time detection method based on deep learning |
| CN107576666A (en) * | 2017-10-13 | 2018-01-12 | 成都精工华耀机械制造有限公司 | A kind of double light spectrum image-forming rails and fastener method for detecting abnormality |
| CN107607207A (en) * | 2017-08-22 | 2018-01-19 | 中国科学院深圳先进技术研究院 | A kind of power equipment thermal fault detection method, system and electronic equipment |
| CN108537262A (en) * | 2018-03-29 | 2018-09-14 | 北京航空航天大学 | A kind of railway rail clip method for detecting abnormality based on multilayer neural network |
| US20180263731A1 (en) * | 2017-03-20 | 2018-09-20 | Align Technology, Inc. | Generating a virtual depiction of an orthodontic treatment of a patient |
| CN108846826A (en) * | 2018-04-24 | 2018-11-20 | 深圳大学 | Object detecting method, device, image processing equipment and storage medium |
| CN109532940A (en) * | 2018-11-06 | 2019-03-29 | 北京交通大学 | The fastener of high speed non-fragment orbit loosens degree detecting method |
-
2019
- 2019-04-01 CN CN201910255993.XA patent/CN109916912A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105574872A (en) * | 2015-12-16 | 2016-05-11 | 南京铁道职业技术学院 | Track fastener detection method based on computer image and visual identity |
| CN105699381A (en) * | 2016-03-28 | 2016-06-22 | 成都唐源电气有限责任公司 | Rail fastener detection device based on machine vision |
| CN205555965U (en) * | 2016-04-25 | 2016-09-07 | 西安科技大学 | Portable folding dolly |
| CN106290378A (en) * | 2016-08-23 | 2017-01-04 | 东方晶源微电子科技（北京）有限公司 | Defect classification method and defect inspecting system |
| US20180263731A1 (en) * | 2017-03-20 | 2018-09-20 | Align Technology, Inc. | Generating a virtual depiction of an orthodontic treatment of a patient |
| CN107563412A (en) * | 2017-08-09 | 2018-01-09 | 浙江大学 | A kind of infrared image power equipment real-time detection method based on deep learning |
| CN107607207A (en) * | 2017-08-22 | 2018-01-19 | 中国科学院深圳先进技术研究院 | A kind of power equipment thermal fault detection method, system and electronic equipment |
| CN107576666A (en) * | 2017-10-13 | 2018-01-12 | 成都精工华耀机械制造有限公司 | A kind of double light spectrum image-forming rails and fastener method for detecting abnormality |
| CN108537262A (en) * | 2018-03-29 | 2018-09-14 | 北京航空航天大学 | A kind of railway rail clip method for detecting abnormality based on multilayer neural network |
| CN108846826A (en) * | 2018-04-24 | 2018-11-20 | 深圳大学 | Object detecting method, device, image processing equipment and storage medium |
| CN109532940A (en) * | 2018-11-06 | 2019-03-29 | 北京交通大学 | The fastener of high speed non-fragment orbit loosens degree detecting method |
Non-Patent Citations (3)
| Title |
|---|
| 刘欣 等: ""基于卷积神经网络的轨道扣件状态检测"", 《实验室研究与探索》 * |
| 娄岩: "《智能医学概论 第1版》", 31 October 2018, 中国铁道出版社 * |
| 赵鑫欣: ""基于卷积神经网络的铁轨扣件异常状态检测研究"", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378885A (en) * | 2019-07-19 | 2019-10-25 | 王晓骁 | A kind of focal area WSI automatic marking method and system based on machine learning |
| CN110533640A (en) * | 2019-08-15 | 2019-12-03 | 北京交通大学 | Based on the track circuit disease discrimination method for improving YOLOv3 network model |
| CN110533640B (en) * | 2019-08-15 | 2022-03-01 | 北京交通大学 | Improved YOLOv3 network model-based track line defect identification method |
| CN110717464A (en) * | 2019-10-15 | 2020-01-21 | 中国矿业大学(北京) | Intelligent railway roadbed disease identification method based on radar data |
| CN110717464B (en) * | 2019-10-15 | 2022-03-22 | 中国矿业大学(北京) | Intelligent railway roadbed disease identification method based on radar data |
| CN112308073A (en) * | 2020-11-06 | 2021-02-02 | 中冶赛迪重庆信息技术有限公司 | Method, system, equipment and medium for identifying loading and unloading transshipment state of scrap steel train |
| CN112308073B (en) * | 2020-11-06 | 2023-08-25 | 中冶赛迪信息技术(重庆)有限公司 | Method, system, equipment and medium for identifying loading and unloading and transferring states of scrap steel train |
| CN115063416A (en) * | 2022-08-08 | 2022-09-16 | 苏州立创致恒电子科技有限公司 | Rail fastener state detection method and system |
| CN115063416B (en) * | 2022-08-08 | 2022-11-04 | 苏州立创致恒电子科技有限公司 | Rail fastener state detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109916912A (en) | A kind of railway rail clip Defect inspection method and system | |
| CN104156731B (en) | Vehicle License Plate Recognition System and method based on artificial neural network | |
| CN105654067A (en) | Vehicle detection method and device | |
| CN104268505A (en) | Automatic cloth defect point detection and recognition device and method based on machine vision | |
| CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
| CN103442209A (en) | Video monitoring method of electric transmission line | |
| CN113920107A (en) | Insulator damage detection method based on improved yolov5 algorithm | |
| CN109376740A (en) | A kind of water gauge reading detection method based on video | |
| CN112801972A (en) | Bridge defect detection method, device, system and storage medium | |
| CN110660049A (en) | Tire defect detection method based on deep learning | |
| CN110390673A (en) | Cigarette automatic testing method based on deep learning under a kind of monitoring scene | |
| CN114299060A (en) | Deep learning dam crack detection method suitable for complex underwater environment | |
| CN114841972A (en) | Power transmission line defect identification method based on saliency map and semantic embedded feature pyramid | |
| CN110866915A (en) | Circular inkstone quality detection method based on metric learning | |
| CN111079734A (en) | Method for detecting foreign matters in triangular holes of railway wagon | |
| CN109993758A (en) | Dividing method, segmenting device, computer equipment and storage medium | |
| CN112734732B (en) | Railway tunnel cable leakage clamp detection method based on improved SSD algorithm | |
| CN112329858B (en) | Image recognition method for breakage fault of anti-loosening iron wire of railway motor car | |
| CN109086781A (en) | A kind of cabinet lamp state identification method based on deep learning | |
| CN115830302B (en) | Multi-scale feature extraction fusion power distribution network equipment positioning identification method | |
| CN108681702A (en) | A kind of container handling stowage shellfish position information determines method and system | |
| CN111723656A (en) | Smoke detection method and device based on YOLO v3 and self-optimization | |
| CN110503092A (en) | The improvement SSD monitor video object detection method adapted to based on field | |
| CN116030013A (en) | Multi-target multi-category damage defect detection method, system and computer storage medium | |
| KR102585146B1 (en) | Pothole detection device and method based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190621 |
|
| RJ01 | Rejection of invention patent application after publication |