CN110254468A - A kind of raceway surface defect intelligent online detection device and detection method - Google Patents
A kind of raceway surface defect intelligent online detection device and detection method Download PDFInfo
- Publication number
- CN110254468A CN110254468A CN201910589600.9A CN201910589600A CN110254468A CN 110254468 A CN110254468 A CN 110254468A CN 201910589600 A CN201910589600 A CN 201910589600A CN 110254468 A CN110254468 A CN 110254468A
- Authority
- CN
- China
- Prior art keywords
- image
- module
- track
- raceway surface
- detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 85
- 230000007547 defect Effects 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 claims abstract description 41
- 238000013016 damping Methods 0.000 claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 22
- 230000005856 abnormality Effects 0.000 claims abstract description 13
- 230000003993 interaction Effects 0.000 claims abstract description 10
- 238000003062 neural network model Methods 0.000 claims abstract description 8
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 4
- 238000009434 installation Methods 0.000 claims abstract description 4
- 238000001914 filtration Methods 0.000 claims description 15
- 230000002159 abnormal effect Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 12
- 230000001681 protective effect Effects 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 11
- 230000004927 fusion Effects 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 10
- 238000010586 diagram Methods 0.000 claims description 9
- 239000000203 mixture Substances 0.000 claims description 8
- 230000001537 neural effect Effects 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- NJPPVKZQTLUDBO-UHFFFAOYSA-N novaluron Chemical compound C1=C(Cl)C(OC(F)(F)C(OC(F)(F)F)F)=CC=C1NC(=O)NC(=O)C1=C(F)C=CC=C1F NJPPVKZQTLUDBO-UHFFFAOYSA-N 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 6
- 230000006978 adaptation Effects 0.000 claims description 6
- 230000002146 bilateral effect Effects 0.000 claims description 6
- 230000002452 interceptive effect Effects 0.000 claims description 6
- 230000004807 localization Effects 0.000 claims description 6
- 238000003708 edge detection Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 230000000877 morphologic effect Effects 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000012790 confirmation Methods 0.000 claims description 3
- 238000005530 etching Methods 0.000 claims description 3
- 238000005192 partition Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 2
- 238000012986 modification Methods 0.000 claims description 2
- 230000004048 modification Effects 0.000 claims description 2
- 230000000717 retained effect Effects 0.000 claims description 2
- 230000001419 dependent effect Effects 0.000 claims 1
- 238000005562 fading Methods 0.000 claims 1
- 241001494479 Pecora Species 0.000 abstract 1
- 238000013480 data collection Methods 0.000 abstract 1
- 238000005259 measurement Methods 0.000 abstract 1
- 238000003672 processing method Methods 0.000 abstract 1
- 238000011160 research Methods 0.000 abstract 1
- 230000002950 deficient Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 206010033799 Paralysis Diseases 0.000 description 4
- 230000006378 damage Effects 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000011897 real-time detection Methods 0.000 description 2
- 238000002604 ultrasonography Methods 0.000 description 2
- 241001062009 Indigofera Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000001678 irradiating effect Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
- B61K9/00—Railway vehicle profile gauges; Detecting or indicating overheating of components; Apparatus on locomotives or cars to indicate bad track sections; General design of track recording vehicles
- B61K9/08—Measuring installations for surveying permanent way
- B61K9/10—Measuring installations for surveying permanent way for detecting cracks in rails or welds thereof
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/89—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles
- G01N21/8914—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles characterised by the material examined
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
-
- G06T5/77—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/89—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles
- G01N21/8914—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles characterised by the material examined
- G01N2021/8918—Metal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
Abstract
The present invention relates to a kind of raceway surface defect intelligent online detection device and detection method, device is made of the true defect detecting system of raceway surface, raceway surface defects detection model online updating system, train groups data interaction system, railroad flaws positioning system, abnormality processing system and power supply system.The consecutive image of image acquisition device shooting track of the method including the use of installation ON TRAINS；The track obtained in image is identified and extracted: track preliminary treatment and positioning are carried out to the image for extracting errant；Defects detection is carried out to the track being accurately positioned out.The research of present invention combination convolutional neural networks, realizes lossless, Poul Dorset Sheep.The raceway surface image of acquisition is made into the intelligent measurement that orbital data collection is sent into progress raceway surface defect in neural network model through volume of data image processing method；Involved damping device is used to prevent from jolting because of train, shakes and causes the image of acquisition fuzzy, achieve the purpose that high-precision detect, reduction manual intervention.
Description
Technical field
The invention belongs to raceway surface defect intelligent online detection technique fields, and in particular to a kind of raceway surface defect intelligence
It can on-line measuring device and detection method.
Background technique
Currently, rail track is the most important infrastructure of driving, due to being in extreme natural environment throughout the year and constantly
It is subjected to the effect of train load, so that train rail surface state is among variation always, is constantly occurring deformation and damage
Wound.For the safety by bus for guaranteeing passenger, it is necessary to which the health status of track is frequently detected and confirmed.Traditional track inspection
The mode of looking into is mainly artificial detection method, eddy current coil detection and ultrasound examination etc..Artificial detection method is mainly complete by the worker that inspects the road
At detection efficiency is lower, and is easy by environmental factor and thinks that factor is influenced；Ultrasound examination and eddy current coil detect
It can be generated with rail surface defects and contact the variation that physics and chemistry may occur, the region of further expansion defect.
Summary of the invention
The object of the invention is that in view of the above shortcomings of the prior art, providing a kind of raceway surface defect intelligent online
Detection device, also offer raceway surface defect intelligent online detecting method.For the defect state of real-time detection raceway surface, thus
The safety of railway is improved, and reduces the maintenance cost of railway；By the positioning system of the invention, can be sent to railway center
The exact position of raceway surface defect can be overhauled in time with guaranteeing to destroy track.
The purpose of the present invention is what is be achieved through the following technical solutions:
A kind of raceway surface defect intelligent online detection device, it is characterised in that: by the true defect detecting system of raceway surface,
Raceway surface defects detection model online updating system, train groups data interaction system, railroad flaws positioning system, abnormality processing
System and power supply system are constituted；
Wherein, the raceway surface defect detecting system include image acquisition subsystem and track detecting classification subsystem,
Described image obtains system and is acquired by pedestal 17, the damping device connecting with pedestal 17 20, calibration component 40, light source 50 and figure
Device 30 forms；The track classification and Detection subsystem identifies mould by image processor 60, the track connecting with image acquisition device 30
Block 70 and track categorization module 80 form；
The raceway surface defects detection model online updating system is by man-machine interactive interface 90, image tagged and nerve net
Network training centre 100 and connect with track identification module 70 and track categorization module 80 110 groups of model parameter update module
At man-machine interactive interface 90 is connected by image tagged with neural metwork training center 100 with model parameter update module 110；
The train groups data interaction system is by signal transmitting and receiving module 120, data simultaneous module 130, data normalization mould
Block 140 and central database 150 form, and signal transmitting and receiving module 120 passes through data simultaneous module 130, data normalization module
140 are connected with central database 150；
The railroad flaws positioning system calculates module by, the relative position that is connected with speed/acceleration sensor 160
170, GPS/ Beidou 180, the absolute position being connected with GPS/ Beidou 180 calculate module 190 and calculate module with relative position
170 and absolute position calculate module 190 connect fusion locating module 200 form；
The abnormality processing system is by the abnormality detection module 250, anomaly classification module 260, abnormal that connect with each system
Alarm 270 and 280 composition of abnormal removing being connected with abnormal alarm device 270, abnormality detection module 250 pass through anomaly classification
Module 260 is connected with abnormal alarm device 270；
The power supply system is by Train electrical system module 210, conventional electron module 220, emergency service module 230 and electricity
Source adaptation module 240 forms, and Train electrical system module 210 passes through conventional electron module 220 and emergency service module 230 respectively
It is connected with power adaptation module 240.
Further, the damping device 20 is by the protective cover 1, hydraulic damper hydraulic cylinder I 2, vertical fixed with train
To damping spring I 3, piston rod I 4, image pick-up device 5, image pick-up device fixed device 6, horizontal shock-absorbing spring I 9, light source fixing part 10, light
Source 11, longitudinal damping spring II 12, horizontal shock-absorbing spring II 13, axial damping spring 15 are constituted；
The image pick-up device 5 is fixed in the fixed device 6 of image pick-up device, the fixed device 6 of image pick-up device by connector 7, pin member 8 with
And hydraulic damper is connect with protective cover 1；The light source 11 is fixed by light source fixing part 10 and protective cover 1.
The detection method of above-mentioned raceway surface defect intelligent online detection device, comprising the following steps:
A, the consecutive image of the image acquisition device shooting track using installation ON TRAINS；
B, the track obtained in image is identified and is extracted:
Image is pre-processed, enhances useful real information while removing image garbage, enhances target
Detectability and reduce data throughout in follow-up process, to improve the reliability of data；In order to complete to railroad flaws
Accurate detection is carried out, needs to carry out preliminary treatment to the image that image acquisition device transmits using image processor；
C, track preliminary treatment and positioning are carried out to the image for extracting errant；
D, defects detection is carried out to the track being accurately positioned out: after the accurate positionin of track, selects inception-v3
As basic network structure, orbital image classification is carried out using convolutional neural networks；
Transfer training is carried out using data of the invention and lesser learning rate and obtains comparatively ideal model, so that this be sent out
Bright data set is applied in large-scale neural network, realizes that the model by big data training is applied to the model of oneself.
Further, as follows to the step of track identification of image is with extraction is obtained in the step B:
Gray processing processing is carried out to the image of image acquisition device transmission: utilizing gray weighted method by original RGB orbital image
Carry out gray processing processing.Picture gray value f (i, j) calculation formula such as (1) is shown after weighting.
F (i, j)=α R (i, j)+β G (i, j)+γ B (i, j) (1)
In formula (1), α, β, γ are gray value calculation formula coefficient, and the rower of image where i indicates the pixel, j is indicated should
The column mark of picture where pixel, R (i, j) indicate that the red pixel component of the original image pixel, G (i, j) indicate original image
The green component of the pixel, B (i, j) indicate the blue component of the original image pixel.
Region of interest regional partition is carried out to image obtained by aforesaid operations: in view of there is larger color difference on track and road surface, in order to
Efficiently raceway surface can be extracted from image, the present invention uses the adaptive threshold fuzziness based on Gaussian distribution model
Method；In ROI region, the mean value and variance of each pixel gray level are calculated, segmentation threshold is calculated further according to formula (2), with this
To obtain bianry image according to being split to grayscale image.
IT=μI+3σI (2)
In formula (2), μfIndicate the mean value of vegetarian refreshments gray scale, σfIndicate the variance of vegetarian refreshments gray scale, ITIndicate segmentation threshold.
Carry out Morphological scale-space to image obtained by aforesaid operations: the high bright part in order to eliminate noise and prominent track is sought
Look for the obvious maximum region in orbital image, handled with the structural element bianry image of T × T pixel, to image into
Row expansion, etching operation, two images are made the difference to obtain image edge information, gained image is as shown in Figure 5；To identify bat
Take the photograph the track in image.
Further, the following institute of the step of track preliminary treatment and positioning being carried out to the image extracted in the step C
Show:
The filtering processing of track picture: due to the uncertainty of orbital environment, in picture often occurrence it is very big from
Noise is dissipated, for the protection to image border, the present invention uses median filtering first, then carries out bilateral filtering；Bilateral filtering
In, the output valve of pixel depends on the weighted array of neighborhood territory pixel value, shown in calculation method such as formula (3).
In formula (3), weighting coefficient h (i, j, k, l) determines in the product for defining core and codomain core.
Wherein, core field is defined to be expressed as shown in formula (4).
Codomain kernel representation is formula (5).
Weighting function is expressed as formula (6).
H (i, j, k, l)=dv (6)
Edge detection is carried out to orbital image using canny operator；
Preliminary pseudo-edge removal is carried out to the image after aforesaid operations: carrying out preliminary pseudo-edge removal using threshold method,
In, threshold value STCalculation method such as formula (7) shown in.
In formula (7), n is the total columns of picture, ImaxFor max pixel value, IsFor pixel minification.
Carry out further pseudo-edge removal to the picture after aforesaid operations: the present invention proposes that constraint condition (8) removal is pseudo-
Edge.
STD in formula (8) formulaiFor the pixel value standard deviation of the i-th row of orbital image, wherein i=(0,1,2 ..., m-1), n generation
Total columns of table input picture, total line number of m representing input images, c are adjustable dynamic factor.
Track fitting and positioning: the operation due to removing pseudo-edge may remove true edge, to need into one
Step is restored；Linear fit is carried out to two groups of marginal points retaining after pseudo-edge cleaning herein, in the intercept for obtaining two straight lines and
Approximate it can restore actual edge after slope.
Further, on the basis of the above-mentioned railroad flaws detection method based on neural network and machine vision, this hair
The method that the bright global position for also proposing a kind of pair of defect track is positioned: remember by using based on Kalman filtering, decaying
The sensor fusion techniques of filtering etc. are recalled to merge GPS/ Beidou etc. and inertial sensor data, realize train more accurately
Positioning；In addition to can be realized train pinpoint accuracy positioning, the present invention also utilize high-precision map localization method with it is above-mentioned
GPS/ Beidou and the localization method of IMU fusion correct the technology of position error mutually to realize the accurate positioning of train.
Further, in order to improve the accuracy of raceway surface defects detection result, the present invention also proposes a kind of right
The method of this neural network model realization real-time update: it is sent to using the known testing result and the image that is marked of confirmation
Neural metwork training center carries out supervised training again to the neural network model；Finally utilize the model for completing to train again
Raceway surface defect detecting system is updated, to realize the accuracy that detection system is continuously improved.
Further, in order to improve detection device work accuracy and prevent because image acquisition equipment damage due to cause to examine
The paralysis of device is surveyed, the present invention proposes a kind of method for improving detection device work accuracy and high efficiency, this method comprises:
A set of image acquisition subsystem is respectively installed in train front end and train rear end；
The accuracy that verification improves detection system is carried out to recognition result using two system of front and back；If also, wherein certain
One image acquisition subsystem failure and orbital image can not be obtained, at this time another set of image acquisition subsystem can work on and
Whole system will not be made to paralyse；
If be the train rail defect detecting system by the data that data interaction center between the train on previous train is sent,
Device is that raceway surface is defective at this in the result of a certain position detection, then this train will be detected carefully when by herein
The defect situation of track at this, and result is compared analysis with the testing result of previous train, to improve system defect
The accuracy of detection.
Further, in order to guarantee that each system of the present invention, device can work normally, present invention proposition be provided as each system,
The power supply system of device offer electric energy.
Further, to prevent the image for leading to acquisition because of the vibration in train driving fuzzy, the present invention also provides one kind
The damping device of image capture module.
Compared with prior art, the beneficial effects of the present invention are:
1, the data set as used in the neural metwork training stage of the invention contains each under the conditions of varying environment
Kind of defect form target, trained data are relatively comprehensively and balanced, so algorithm has stronger robustness, have certain anti-interference
Ability；2, the system updated to neural network model real-time online is designed for testing result of the invention, improved
The validity of neural network model；3, it is had in the process of running for train and jolts, shakes, so as to cause image mould is obtained
Paste, reduces the influence of detection accuracy, devises a damping device to image capturing system；4, the present invention comprehensively improves
The accuracy and recall ratio of detection can substantially reduce the cost of railway maintenance, improve detection efficiency.
Detailed description of the invention
Fig. 1 is the embodiment block diagram of a kind of raceway surface defect intelligent on-line detection method of the present invention, apparatus and system；
Fig. 2 is a kind of flow diagram of raceway surface defect intelligent on-line detection method embodiment of the present invention；
Fig. 3 is the original track image of step S100 actual acquisition；
Fig. 4 is the flow diagram of the implementation procedure preferred embodiment of step S110；
Fig. 5 is the marginal information figure in the embodiment of the present invention to original track morphological image process；
Fig. 6 is the effect picture that coarse orbit is extracted in the embodiment of the present invention；
Fig. 7 is the flow diagram of the implementation procedure preferred embodiment of step S120；
Fig. 8 is the effect picture in the embodiment of the present invention to orbital image edge detection；
Fig. 9 is the effect picture in the embodiment of the present invention to track fitting and positioning；
Figure 10 is the testing result of actual track defect image in the embodiment of the present invention；
Figure 11 is tentatively to remove pseudo-edge program flow diagram in the embodiment of the present invention；
Figure 12 is the flow diagram of the implementation procedure preferred embodiment of step 140；
Figure 13 is neural network model real-time update flow diagram in the embodiment of the present invention
Figure 14, Figure 15 are the structural schematic diagram of the damping device of system in the embodiment of the present invention.
In figure, the camera shooting of 1. protective cover, 2. hydraulic damper hydraulic cylinder, I 3. longitudinal damping spring, I 4. piston rod I 5.
6. image pick-up device of device fixes 7. connector of device, 8. pin member, 9. horizontal shock-absorbing spring, I 10. light source fixing part, 11. light source 12.
II 13. horizontal shock-absorbing spring of longitudinal damping spring, II 14. hydraulic damper hydraulic cylinder, II 15. axial damping spring 16.
II 17. pedestal of piston rod, 20. damping device, 30. image acquisition device 40. demarcates 50. light source of component, 60. image processor
70. 100. neural metwork training center of track identification module 80. track categorization module, 90. machine interactive interface, 110. model is joined
Number 120. signal transmitting and receiving mould of update module, 130. data simultaneous module, 140. data normalization module, 150. central database
160. speed/acceleration sensor, 170. relative position calculates that 190. absolute position of module 180.GPS/ Beidou calculates module
200. fusion 210. Train electrical system module of locating module, 220. conventional electron module, 230. emergency service modules 240.
270. abnormal alarm device of power adaptation module 250. abnormality detection mould, 260. anomaly classification module 280. is abnormal to be removed.
Specific embodiment
The embodiment of the present invention is described more fully below, the example of applying of embodiment is shown in the accompanying drawings, wherein phase from beginning to end
Identical element or element with the same function are indicated with label.Embodiment below with reference to attached drawing description is exemplary
, it is intended to it is used to explain the present invention, and is not considered as limiting the invention, based on the embodiments of the present invention, this field
In those of ordinary skill's other embodiments obtained without making creative work, belong to guarantor of the invention
Protect range.
In order to solve the above technical problems, the present invention proposes a kind of raceway surface defect intelligent online device and method, the invention
It is mainly used for the defect state of real-time detection raceway surface, to improve the safety of railway, and reduces the maintenance cost of railway.
By the positioning system of the invention, the exact position of raceway surface defect can be sent to railway center, to guarantee to destroy track
It can be overhauled in time.
As shown in Figure 1, raceway surface defect intelligent online detection device of the present invention by the true defect detecting system of raceway surface,
Raceway surface defects detection model online updating system, train groups data interaction system, railroad flaws positioning system, abnormality processing
System and power supply system are constituted.Wherein, which includes image acquisition subsystem and track detecting point
Class subsystem, the image-taking system include pedestal 17, the damping device connecting with pedestal 17 20, calibration component 40, light source 50
And figure collector 30；Track classification and Detection subsystem includes the image processor 60 connecting with image acquisition device 30, track knowledge
Other module 70, track categorization module 80；Raceway surface defects detection model online updating system includes man-machine interactive interface 90, figure
As label and neural metwork training center 100 and the model parameter being connect with track identification module 70 and track categorization module 80
Update module 110；Train groups data interaction system includes signal transmitting and receiving module 120, data simultaneous module 130, data normalization
Module 140, central database 150；Railroad flaws positioning system includes speed/acceleration sensor 160, relative position reckoning mould
Block 170, GPS/ Beidou 180, absolute position calculate module 190 and calculate that module 170, absolute position calculate with relative position
The fusion locating module 200 that module 190 connects；Power supply system include Train electrical system module 210, conventional electron module 220,
Emergency service module 230 and power adaptation module 240；In addition, abnormality processing system includes the abnormality detection connecting with each system
Module 250, anomaly classification module 260, abnormal alarm device 270 and abnormal removing 280；It can not be normal to prevent each system failure
Acquisition and detection image, need to back up equipment to take over the work of faulty equipment；In addition, also needing to data interaction center
All received data carry out data backup, prevent loss of data from causing system that can not be worked normally.
Hydraulic damper hydraulic cylinder II 14 and pressure damper hydraulic cylinder I and piston rod II 16 and I 4 size of piston rod
Specification has difference.
On the basis of above-mentioned raceway surface defect detecting system implementation, the present invention also proposes that a kind of raceway surface lacks
Intelligent detecting method is fallen into, as shown in Fig. 2, the railroad flaws detection method includes:
A, the consecutive image of the image acquisition device shooting track using installation ON TRAINS: in order to which quick obtaining is complete, clear
Clear, accurate orbital image devises orbital image acquisition system by image acquisition device 30, light source 50 and damping device first
20 compositions.Wherein, image acquisition device 30 uses the sufficiently high high speed face array CCD camera of resolution ratio, and installs determining for appropriate burnt section
Zoom lens；Light source 50 uses the secondary light source of brightness and the adjustable suitable shape of irradiating angle and suitable category.The figure that will acquire
As being transferred to image processor 60, the real image of acquisition is as shown in Figure 3.
B, the track obtained in image is identified and is extracted.
Image is pre-processed, enhances useful real information while removing image garbage, enhances target
Detectability and reduce data throughout in follow-up process, to improve the reliability of data；In order to complete to railroad flaws
Accurate detection is carried out, needs to carry out preliminary treatment, such as Fig. 4 to the image that image acquisition device 30 transmits using image processor 60
Shown, step B is specifically included:
B1, gray processing processing is carried out to the image that image acquisition device 30 transmits: utilizes gray weighted method by original RGB track
Image carries out gray processing processing.Picture gray value f (i, j) calculation formula such as (1) is shown after weighting.
F (i, j)=α R (i, j)+β G (i, j)+γ B (i, j) (1)
In formula (1), α, β, γ are gray value calculation formula coefficient, and the rower of image where i indicates the pixel, j is indicated should
The column mark of picture where pixel, R (i, j) indicate that the red pixel component of the original image pixel, G (i, j) indicate original image
The green component of the pixel, B (i, j) indicate the blue component of the original image pixel.
B2, region of interest regional partition is carried out to image obtained by aforesaid operations: in view of there is larger color difference on track and road surface, is
Efficiently raceway surface can be extracted from image, the present invention is using the adaptive threshold based on Gaussian distribution model point
Segmentation method；In ROI region, the mean value and variance of each pixel gray level are calculated, calculates segmentation threshold further according to formula (2), with
This is to obtain bianry image according to being split to grayscale image.
IT=μ I+3 σI (2)
In formula (2), μfIndicate the mean value of vegetarian refreshments gray scale, σfIndicate the variance of vegetarian refreshments gray scale, ITIndicate segmentation threshold.
B3, Morphological scale-space is carried out to image obtained by aforesaid operations: the high bright part in order to eliminate noise and prominent track,
The obvious maximum region in orbital image is found, is handled with the structural element bianry image of T × T pixel, to image
It is expanded, etching operation, two images is made the difference to obtain image edge information, gained image is as shown in Figure 5；To identify
Shoot the track in image.
In order to realize the accurate detection to railroad flaws, need first to extract the track identified roughly, institute
The image extracted is as shown in Figure 6.
C, track preliminary treatment and positioning are carried out to the image for extracting errant.
Further, it is accurately positioned to be realized to the track in Fig. 6, first has to carry out preliminary treatment to it, such as Fig. 7 institute
Show, step C is specifically included:
The filtering processing of C1, track picture: due to the uncertainty of orbital environment, often occurrence is very big in picture
Discrete noise, for the protection to image border, the present invention uses median filtering first, then carries out bilateral filtering；Bilateral filter
In wave, the output valve of pixel depends on the weighted array of neighborhood territory pixel value, shown in calculation method such as formula (3).
In formula (3), weighting coefficient h (i, j, k, l) determines in the product for defining core and codomain core.
Wherein, core field is defined to be expressed as shown in formula (4).
Codomain kernel representation is formula (5).
Weighting function is expressed as formula (6).
H (i, j, k, l)=dv (6)
C2, carry out edge detection to orbital image using canny operator: edge detection results are as shown in figure (9).
C3, preliminary pseudo-edge removal is carried out to the image after aforesaid operations: carries out preliminary pseudo-edge removal using threshold method,
Wherein, threshold value STCalculation method such as formula (7) shown in.
In formula (7), n is the total columns of picture, ImaxFor max pixel value, IsFor pixel minification.
Further, preliminary pseudo-edge removal algorithm flow chart is as shown in figure 11.
C4, further pseudo-edge removal is carried out to the picture after aforesaid operations, the present invention proposes constraint condition (8) removal
Pseudo-edge.
STD in formula (8) formulaiFor the pixel value standard deviation of the i-th row of orbital image, wherein i=(0,1,2 ..., m-1), n generation
Total columns of table input picture, total line number of m representing input images, c are adjustable dynamic factor.
C5, track fitting and positioning: the operation due to removing pseudo-edge may remove true edge, so that 0 needs
Further restore；Linear fit is carried out to the two groups of marginal points retained after pseudo-edge cleaning herein, is obtaining cutting for two straight lines
Away from after slope approximate actual edge can be restored；The final positioning result of track is as shown in Figure 9.
D, defects detection is carried out to the track being accurately positioned out: after the accurate positionin of track, selects inception-v3
As basic network structure, orbital image classification is carried out using convolutional neural networks.
Transfer training is carried out using data of the invention and lesser learning rate and obtains comparatively ideal model, so that this be sent out
Bright data set is applied in large-scale neural network, realizes that the model by big data training is applied to the model of oneself.
Wherein, recall ratio (P) and recall rate (P) are the important indicators of evaluation detection and recognition effect, are defined respectively such as
Shown in formula (9), (10).
In formula (9), (10), TP represents the number of tracks accurately detected, and FP represents the defect track number that error detection goes out
Amount, FN represent the practical quantity that intact track is but detected as defect track.
Wherein, rgb color mode is a kind of color standard of industry, is by red (R), green (G), three, indigo plant (B)
The variation of Color Channel and their mutual superpositions are to obtain miscellaneous color.
On the basis of the above-mentioned railroad flaws detection method based on neural network and machine vision, the present invention also proposes one
The method that kind positions the global position of defect track, as shown in figure 12, which includes:
In view of the particularity of railway, the present invention uses global positioning system (GPS/ Beidou etc.) and inertial sensor
(IMU) data blend and the method in conjunction with high-precision map；Since train driving environment is more complicated and has and wears often
The more signal shieldings such as tunnel, jungle section, so GPS/ Beidou etc. has obvious Multipath reflection problem and no signal is asked
Topic, the location informations such as GPS/ Beidou caused generate large error；IMU is a kind of inertial sensor detection using high frequency
The data such as speed, acceleration calculate the traveling displacement information of train in real time, but due to that can be produced during calculating train displacement
The raw integral error constantly accumulated, eventually leads to the effective position that cannot achieve train；By using based on Kalman filtering, decline
Subtract the sensor fusion techniques of memory filtedng etc. to merge GPS/ Beidou etc. and inertial sensor data, realizes that the ratio of train is calibrated
True positioning；In addition to can be realized train pinpoint accuracy positioning, the present invention also utilize high-precision map localization method with
Above-mentioned GPS/ Beidou and the localization method of IMU fusion correct the technology of position error mutually to realize the accurate positioning of train.
In order to improve the accuracy of raceway surface defects detection result, the present invention also proposes one kind to this neural network
The method of model realization real-time update, as shown in figure 13, this method principle are as follows:
The result detected by raceway surface defective system can will test image after railway maintenance personnel confirm testing result
Be divided into two classes: one kind is raceway surface really defective image, is " 1 " by such image tagged, and one kind is that raceway surface determines
Do not have defective image, is " 0 " by such image tagged；It is the image that testing result is mistake, such image warp there are also one kind
It crosses after other detection systems or desk checking and is also marked as " 0 " or " 1 " according to its practical railroad flaws that whether there is.It will be upper
State the known testing result of confirmation and the image being marked be sent to neural metwork training center, to the neural network model into
Capable supervised training again；Finally using model modification raceway surface defect detecting system trained again is completed, to realize not
The disconnected accuracy for improving detection system.
In order to improve the accuracy of detection device work and prevent from leading to detection device because image acquisition equipment damages
Paralysis, the method that the present invention proposes a kind of raising detection device work accuracy and high efficiency, this method comprises:
A set of image acquisition subsystem is respectively installed in train front end and train rear end；
The accuracy that verification improves detection system is carried out to recognition result using two system of front and back；If also, wherein certain
One image acquisition subsystem failure and orbital image can not be obtained, at this time another set of image acquisition subsystem can work on and
Whole system will not be made to paralyse；
If be the train rail defect detecting system by the data that data interaction center between the train on previous train is sent,
Device is that raceway surface is defective at this in the result of a certain position detection, then this train will be detected carefully when by herein
The defect situation of track at this, and result is compared analysis with the testing result of previous train, to improve system defect
The accuracy of detection.
In order to guarantee that each system of the present invention, device can work normally, present invention proposition is provided as each system, device provides electricity
The power supply system of energy, the power supply system method of supplying power to are as follows:
Electric energy is directly acquired from Train electrical system module 210, is charged for conventional electron module 220；
Simultaneously for prevent because conventional electron module 220 failure due to cause whole system power failure can not work, the present invention also matches
A standby emergency service module 220, the emergency service module are equally charged by Train electrical system module 210 for it；
In view of the inconsistent problem of voltage needed for all systems of the invention, device, conventional electron module 220 and emergency
Power supply module 220 also need to connect with power supply adaptor module 240 to export it is various needed for voltages, meet each system, device
It is required that.
Finally, the present invention also provides a kind of images to prevent the image for leading to acquisition because of the vibration in train driving fuzzy
The damping device of acquisition module, as shown in Figure 14, Figure 15, which includes:
Longitudinal damping spring I 3, longitudinal damping spring II 12, horizontal shock-absorbing spring I 9, horizontal shock-absorbing spring II 13, axial direction
Damping spring 15, hydraulic damper hydraulic cylinder 2, piston rod 4 are absorbed using hydraulic damper and consume the acceleration because of train, subtract
Speed the brings device such as jolts longitudinally, laterally and the energy of axial vibrations, to keep the stabilization of image capture module；
The image pick-up device 5 is fixed in the fixed device 6 of image pick-up device, the fixed device 6 of image pick-up device by connector 7, pin member 8 with
And hydraulic damper is connect with protective cover 1, to realize the damping of image capture module；
Light source 11, is fixed on protective cover 1 using light source fixing part 10, is prevented because of light by light source 11 and light source fixing part 10
It can not shooting clear image according to deficiency；
Protective cover 1, protective cover 1 are fixed on train, for placing, protecting damping spring and protection image acquisition device, prevent
Because weather reason leads to the destruction of image collecting device and damping spring.
Above is only part or preferred embodiment of the invention, therefore either text or attached drawing cannot all limit this
It invents the protection scope of hair and utilizes description of the invention and accompanying drawing content institute under all designs with an entirety of the invention
The equivalent structure transformation of work, or directly/be used in other technologies related fields indirectly and be included in the scope of protection of the invention.
Claims (8)
1. a kind of raceway surface defect intelligent online detection device, it is characterised in that: by the true defect detecting system of raceway surface, rail
Road surface defects detection model online updating system, train groups data interaction system, railroad flaws positioning system, abnormality processing system
System and power supply system are constituted；
Wherein, the raceway surface defect detecting system include image acquisition subsystem and track detecting classification subsystem, it is described
Image-taking system by pedestal (17), the damping device (20) being connect with pedestal (17), calibration component (40), light source (50) and
Figure collector (30) composition；The track classification and Detection subsystem is by the image processor that connect with image acquisition device (30)
(60), track identification module (70) and track categorization module (80) composition；
The raceway surface defects detection model online updating system is by man-machine interactive interface (90), image tagged and neural network
Training centre (100) and the model parameter update module being connect with track identification module (70) and track categorization module (80)
(110) it forms, man-machine interactive interface (90) updates mould by image tagged and neural metwork training center (100) and model parameter
Block (110) is connected；
The train groups data interaction system is by signal transmitting and receiving module (120), data simultaneous module (130), data normalization mould
Block (140) and central database (150) composition, signal transmitting and receiving module (120) pass through data simultaneous module (130), data mark
Standardization module (140) is connected with central database (150)；
The railroad flaws positioning system calculates module by, the relative position that is connected with speed/acceleration sensor (160)
(170), GPS/ Beidou (180), the absolute position being connected with GPS/ Beidou (180) calculate module (190) and and relative position
Calculate that module (170) and absolute position calculate fusion locating module (200) composition of module (190) connection；
The abnormality processing system is by the abnormality detection module (250), anomaly classification module (260), abnormal that connect with each system
Alarm (270) and abnormal remove (280) being connected with abnormal alarm device (270) form, and abnormality detection module (250) passes through
Anomaly classification module (260) is connected with abnormal alarm device (270)；
The power supply system by Train electrical system module (210), conventional electron module (220), emergency service module (230) and
Power adaptation module (240) composition, Train electrical system module (210) are supplied by conventional electron module (220) and emergency respectively
Electric module (230) is connected with power adaptation module (240).
2. a kind of raceway surface defect intelligent online detection device according to claim 1, it is characterised in that: the damping
Device (20) is by the protective cover (1), hydraulic damper hydraulic cylinder I (2), longitudinal damping spring I (3) 3, piston fixed with train
Bar I (4), image pick-up device (5), image pick-up device fixed device (6), horizontal shock-absorbing spring I (9), light source fixing part (10), light source (11),
Longitudinal damping spring II (12), horizontal shock-absorbing spring II (13), axial damping spring (15) are constituted；
The image pick-up device (5) is fixed in image pick-up device fixed device (6), and image pick-up device fixed device (6) passes through connector (7), pin
Part (8) and hydraulic damper are connect with protective cover (1)；The light source (11) passes through light source fixing part (10) and protective cover (1)
It is fixed.
3. a kind of detection method of raceway surface defect intelligent online detection device according to claim 1, feature exist
In, comprising the following steps:
A, the consecutive image of image acquisition device (30) the shooting track using installation ON TRAINS；Described image collector (30) is adopted
With high speed face array CCD camera and tight shot is installed, the image transmitting that will acquire to image processor (60)；
B, track identification module (70) to obtain image in track identified and extracted: image processor (60) to image into
Row pretreatment, enhances useful real information while removing image garbage；
C, track preliminary treatment and positioning are carried out to the image for extracting errant；
D, defects detection is carried out to the track that is accurately positioned out: after the accurate positionin of track, select inception-v3 as
Infrastructure network carries out orbital image classification using convolutional neural networks；Required mould is obtained by image data transfer training
Type, and it is applied to large-scale neural network.
4. a kind of detection method of raceway surface defect intelligent online detection device according to claim 3, feature exist
In, the step B specifically includes the following steps:
B1, gray processing processing is carried out to the image of image acquisition device (30) transmission: utilizes gray weighted method by original RGB trajectory diagram
As carry out gray processing processing, after weighting picture gray value f (i, j) calculation formula such as (1),
F (i, j)=α R (i, j)+β G (i, j)+γ B (i, j) (1)
In formula (1), α, β, γ are gray value calculation formula coefficient, and the rower of image, j indicate the pixel where i indicates the pixel
The column mark of place picture, R (i, j) indicate that the red pixel component of the original image pixel, G (i, j) indicate the original image picture
The green component of element, B (i, j) indicate the blue component of the original image pixel；
B2, region of interest regional partition is carried out to image obtained by aforesaid operations: uses the adaptive threshold based on Gaussian distribution model
Dividing method；In ROI region, the mean value and variance of each pixel gray level are calculated, calculates segmentation threshold further according to formula (2),
As according to being split to grayscale image, bianry image is obtained；
Ir=μI3σI (2)
In formula (2), μfIndicate the mean value of vegetarian refreshments gray scale, σfIndicate the variance of vegetarian refreshments gray scale, ITIndicate segmentation threshold；
B3, Morphological scale-space is carried out to above-mentioned gained image: is handled with the structural element bianry image of T × T pixel, it is right
Image expanded, etching operation, and two images are made the difference to obtain image edge information, identifies the track in shooting image.
5. a kind of detection method of raceway surface defect intelligent online detection device according to claim 3, feature exist
In, the step C specifically includes the following steps:
The filtering processing of C1, track picture: median filtering is first used, then carries out bilateral filtering；In bilateral filtering, the output of pixel
Weighted array of the value dependent on neighborhood territory pixel value, calculation method such as formula (3),
In formula (3), weighting coefficient h (i, j, k, l) determines in the product for defining core and codomain core；
Wherein, it defines core field and is expressed as formula (4),
Codomain kernel representation is formula (5),
Weighting function is expressed as formula (6),
H (i, j, k, l)=dv； (6)
C2, edge detection is carried out to orbital image using canny operator；
C3, preliminary pseudo-edge removal is carried out to gained image: carries out preliminary pseudo-edge removal using threshold method, wherein threshold value ST's
Calculation method such as formula (7),
In formula (7), n is the total columns of picture, ImaxFor max pixel value, IsFor pixel minification；
C4, further pseudo-edge removal being carried out to gained picture, the present invention proposes that constraint condition (8) remove pseudo-edge,
STD in formula (8) formulaiFor the pixel value standard deviation of the i-th row of orbital image, wherein i=(0,1,2 ..., m-1), n represent input
Total columns of image, total line number of m representing input images, c are adjustable dynamic factor；
C5, track fitting and positioning: carrying out linear fit to the two groups of marginal points retained after pseudo-edge cleaning, straight obtaining two
Approximate it can restore actual edge after the intercept and slope of line.
6. a kind of detection method of raceway surface defect intelligent online detection device according to claim 3, feature exist
In: the step C be by using based on Kalman filtering, fading filter sensor fusion techniques to merge GPS/ Beidou
And inertial sensor data realizes the positioning of train, and utilizes high-precision map localization method and above-mentioned GPS/ Beidou and IMU
The localization method of fusion corrects mutually the accurate positioning that position error realizes train.
7. a kind of detection method of raceway surface defect intelligent online detection device according to claim 3, feature exist
In the step D, neural metwork training center is sent to using the known testing result and the image that is marked of confirmation, it is right
The neural network model carries out supervised training again；Finally utilize the model modification raceway surface defects detection for completing to train again
System improves the accuracy of detection system.
8. a kind of detection method of raceway surface defect intelligent online detection device according to claim 3, feature exist
In: step A respectively installs a set of image acquisition subsystem in train front end and train rear end, using two system of front and back to identification
As a result verification is carried out to improve the accuracy of detection system and prevent from paralysing.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2019105364168 | 2019-06-20 | ||
| CN201910536416 | 2019-06-20 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110254468A true CN110254468A (en) | 2019-09-20 |
| CN110254468B CN110254468B (en) | 2021-12-03 |
Family
ID=67923812
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910589600.9A Active CN110254468B (en) | 2019-06-20 | 2019-07-02 | Intelligent online detection device and detection method for track surface defects |
| CN201921015919.2U Withdrawn - After Issue CN211731391U (en) | 2019-06-20 | 2019-07-02 | Intelligent online detection device for track surface defects |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201921015919.2U Withdrawn - After Issue CN211731391U (en) | 2019-06-20 | 2019-07-02 | Intelligent online detection device for track surface defects |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN110254468B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110782443A (en) * | 2019-10-23 | 2020-02-11 | 四川大学 | Railway track defect detection method and system |
| CN110991667A (en) * | 2019-11-28 | 2020-04-10 | 中国铁道科学研究院集团有限公司 | Railway track facility abnormity identification method and system |
| CN111003018A (en) * | 2019-12-27 | 2020-04-14 | 郑州铁路职业技术学院 | System and method for dynamically detecting track condition of high-speed railway |
| CN111024728A (en) * | 2019-12-26 | 2020-04-17 | 江西交通职业技术学院 | Railway detection method and system based on computer vision and ultrasonic flaw detection |
| CN111080611A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon bolster spring fracture fault image identification method |
| CN111402250A (en) * | 2020-03-26 | 2020-07-10 | 中国联合网络通信集团有限公司 | Machine vision defect detection method and platform based on edge calculation |
| CN111859676A (en) * | 2020-07-23 | 2020-10-30 | 西南交通大学 | Intelligent detection method for damage of floating slab track steel spring |
| CN112150453A (en) * | 2020-09-30 | 2020-12-29 | 哈尔滨市科佳通用机电股份有限公司 | Automatic detection method for breakage fault of bolster spring of railway wagon |
| CN113011315A (en) * | 2021-03-16 | 2021-06-22 | 华南理工大学 | Subway track identification method based on ultra-fast structure perception depth network |
| CN114066890A (en) * | 2022-01-14 | 2022-02-18 | 成都数联云算科技有限公司 | Gear defect detection method and device, computer equipment and storage medium |
| CN114670899A (en) * | 2022-04-20 | 2022-06-28 | 北京运达华开科技有限公司 | Image acquisition device for track detection system |
| CN114972117A (en) * | 2022-06-30 | 2022-08-30 | 成都理工大学 | Track surface wear identification and classification method and system |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110254468B (en) * | 2019-06-20 | 2021-12-03 | 吉林大学 | Intelligent online detection device and detection method for track surface defects |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102854194A (en) * | 2012-09-11 | 2013-01-02 | 中南大学 | Object surface defect detection method and apparatus based on linear array CCD |
| CN102854191A (en) * | 2012-07-18 | 2013-01-02 | 湖南大学 | Real-time visual detection and identification method for high speed rail surface defect |
| US20130034298A1 (en) * | 2011-08-04 | 2013-02-07 | University Of Southern California | Image-based crack detection |
| CN104608799A (en) * | 2014-12-12 | 2015-05-13 | 郑州轻工业学院 | Information fusion technology based train wheel set tread damage online detection and recognition method |
| CN105891217A (en) * | 2016-04-27 | 2016-08-24 | 重庆大学 | System and method for detecting surface defects of steel rails based on intelligent trolley |
| CN106290379A (en) * | 2016-08-30 | 2017-01-04 | 哈尔滨工业大学（威海） | Rail surface defects based on Surface scan camera detection device and method |
| CN107782733A (en) * | 2017-09-30 | 2018-03-09 | 中国船舶重工集团公司第七〇九研究所 | Image recognition the cannot-harm-detection device and method of cracks of metal surface |
| CN109658397A (en) * | 2018-12-12 | 2019-04-19 | 广州地铁集团有限公司 | A kind of rail polling method and system |
| CN211731391U (en) * | 2019-06-20 | 2020-10-23 | 吉林大学 | Intelligent online detection device for track surface defects |
-
2019
- 2019-07-02 CN CN201910589600.9A patent/CN110254468B/en active Active
- 2019-07-02 CN CN201921015919.2U patent/CN211731391U/en not_active Withdrawn - After Issue
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130034298A1 (en) * | 2011-08-04 | 2013-02-07 | University Of Southern California | Image-based crack detection |
| CN102854191A (en) * | 2012-07-18 | 2013-01-02 | 湖南大学 | Real-time visual detection and identification method for high speed rail surface defect |
| CN102854194A (en) * | 2012-09-11 | 2013-01-02 | 中南大学 | Object surface defect detection method and apparatus based on linear array CCD |
| CN104608799A (en) * | 2014-12-12 | 2015-05-13 | 郑州轻工业学院 | Information fusion technology based train wheel set tread damage online detection and recognition method |
| CN105891217A (en) * | 2016-04-27 | 2016-08-24 | 重庆大学 | System and method for detecting surface defects of steel rails based on intelligent trolley |
| CN106290379A (en) * | 2016-08-30 | 2017-01-04 | 哈尔滨工业大学（威海） | Rail surface defects based on Surface scan camera detection device and method |
| CN107782733A (en) * | 2017-09-30 | 2018-03-09 | 中国船舶重工集团公司第七〇九研究所 | Image recognition the cannot-harm-detection device and method of cracks of metal surface |
| CN109658397A (en) * | 2018-12-12 | 2019-04-19 | 广州地铁集团有限公司 | A kind of rail polling method and system |
| CN211731391U (en) * | 2019-06-20 | 2020-10-23 | 吉林大学 | Intelligent online detection device for track surface defects |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110782443A (en) * | 2019-10-23 | 2020-02-11 | 四川大学 | Railway track defect detection method and system |
| CN110991667A (en) * | 2019-11-28 | 2020-04-10 | 中国铁道科学研究院集团有限公司 | Railway track facility abnormity identification method and system |
| CN111080611A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Railway wagon bolster spring fracture fault image identification method |
| CN111024728B (en) * | 2019-12-26 | 2022-04-26 | 江西交通职业技术学院 | Railway detection method and system based on computer vision and ultrasonic flaw detection |
| CN111024728A (en) * | 2019-12-26 | 2020-04-17 | 江西交通职业技术学院 | Railway detection method and system based on computer vision and ultrasonic flaw detection |
| CN111003018A (en) * | 2019-12-27 | 2020-04-14 | 郑州铁路职业技术学院 | System and method for dynamically detecting track condition of high-speed railway |
| CN111402250A (en) * | 2020-03-26 | 2020-07-10 | 中国联合网络通信集团有限公司 | Machine vision defect detection method and platform based on edge calculation |
| CN111859676A (en) * | 2020-07-23 | 2020-10-30 | 西南交通大学 | Intelligent detection method for damage of floating slab track steel spring |
| CN111859676B (en) * | 2020-07-23 | 2022-11-22 | 西南交通大学 | Intelligent detection method for damage of floating slab track steel spring |
| CN112150453A (en) * | 2020-09-30 | 2020-12-29 | 哈尔滨市科佳通用机电股份有限公司 | Automatic detection method for breakage fault of bolster spring of railway wagon |
| CN113011315A (en) * | 2021-03-16 | 2021-06-22 | 华南理工大学 | Subway track identification method based on ultra-fast structure perception depth network |
| CN114066890A (en) * | 2022-01-14 | 2022-02-18 | 成都数联云算科技有限公司 | Gear defect detection method and device, computer equipment and storage medium |
| CN114670899A (en) * | 2022-04-20 | 2022-06-28 | 北京运达华开科技有限公司 | Image acquisition device for track detection system |
| CN114972117A (en) * | 2022-06-30 | 2022-08-30 | 成都理工大学 | Track surface wear identification and classification method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110254468B (en) | 2021-12-03 |
| CN211731391U (en) | 2020-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110254468A (en) | A kind of raceway surface defect intelligent online detection device and detection method | |
| WO2020199538A1 (en) | Bridge key component disease early-warning system and method based on image monitoring data | |
| CN105373135B (en) | A kind of method and system of aircraft docking guidance and plane type recognition based on machine vision | |
| CN104599249B (en) | Cableway platform bridge floor car load is distributed real-time detection method | |
| CN109658397A (en) | A kind of rail polling method and system | |
| CN109489724B (en) | Tunnel train safe operation environment comprehensive detection device and detection method | |
| CN113947731B (en) | Foreign matter identification method and system based on contact net safety inspection | |
| CN102854191A (en) | Real-time visual detection and identification method for high speed rail surface defect | |
| CN111080612B (en) | Truck bearing damage detection method | |
| CN107507174A (en) | Power plant's instrument equipment drawing based on hand-held intelligent inspection is as recognition methods and system | |
| CN110046584B (en) | Road crack detection device and detection method based on unmanned aerial vehicle inspection | |
| CN106124512A (en) | Suspension type monorail box beam inspection device | |
| CN106803087A (en) | A kind of car number automatic identification method and system | |
| CN111079822A (en) | Method for identifying dislocation fault image of middle rubber and upper and lower plates of axle box rubber pad | |
| CN111605578A (en) | Railway track inspection method for carrying 3D equipment by using unmanned aerial vehicle | |
| CN113371028A (en) | Intelligent inspection system and method for electric bus-mounted track | |
| CN111091551A (en) | Method for detecting loss fault of brake beam strut opening pin of railway wagon | |
| CN114565845A (en) | Intelligent inspection system for underground tunnel | |
| CN112907532B (en) | Improved truck door falling detection method based on fast RCNN | |
| CN114119505A (en) | Method and device for detecting chip adhesion area defects | |
| CN117372677A (en) | Method for detecting health state of cotter pin of fastener of high-speed railway overhead contact system | |
| CN115857040A (en) | Dynamic visual detection device and method for foreign matters on locomotive roof | |
| CN107194923B (en) | Ultraviolet image diagnosis method for defect inspection of contact network power equipment | |
| CN104714068A (en) | Robot for outdoor electric energy meter to read meter | |
| CN113343782B (en) | Expressway sign label detection method based on unmanned aerial vehicle remote sensing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |