CN109591846A - A kind of wheel tread online test method - Google Patents
A kind of wheel tread online test method Download PDFInfo
- Publication number
- CN109591846A CN109591846A CN201811590768.3A CN201811590768A CN109591846A CN 109591846 A CN109591846 A CN 109591846A CN 201811590768 A CN201811590768 A CN 201811590768A CN 109591846 A CN109591846 A CN 109591846A
- Authority
- CN
- China
- Prior art keywords
- image
- gabor
- wheel tread
- fft
- directions
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
- B61K9/12—Measuring or surveying wheel-rims
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Image Processing (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of wheel tread online test methods, are related to technical field of nondestructive testing.The present invention uses FFT-GABOR Fast transforms, it is realized in conjunction with SVM classifier and defect recognition and classification is carried out to wheel tread, it can overcome the problems, such as that the false recognition rate of existing detection method is high, slow-footed, the wheel tread online test method false recognition rate of the application is lower than 7%, it can identify the main 3 class defect of wheel tread, the defect type of 90% or more covering, detection time are short.
Description
Technical field
The present invention relates to technical field of nondestructive testing, more specifically to a kind of wheel tread online test method.
Background technique
Wheel is fitted in same root vehicle to being the part being in contact on rolling stock with rail, by two wheels in left and right securely
It is formed on axis.The effect of wheel pair is the operation and steering for guaranteeing rolling stock on rail, is born from the complete of rolling stock
Portion is quiet, dynamic loading, passes it on rail, and give rolling stock each components the load transmission generated by guideway irregularity.
In recent years, as the fast development of China railways, route are continuously increased, speed is stepped up, and carrying capacity is increasingly
Greatly, vehicle is in the growth of explosion type, and at the same time, the accident as caused by wheel rail relation is also increasing, the inspection in terms of simple track
The needs for being no longer satisfied safe operation are surveyed, topical railway wheel problems are the important failure of rolling stock, tread damage form packet
Scratch, removing, chip off-falling etc. are included, tread damage relies primarily on artificial detection at present, and inefficiency, false detection rate is low, great work intensity.
With the development of artificial intelligence, has ready conditions and more reasonable, more intelligent detection means is applied on Wheel set detecting, it is contactless
The application of the key technologies such as machine vision technique combination deep learning has become certainty.
Currently, the defects detection of wheel tread is broadly divided into following several methods, a kind of tyre tread data are Wave data, are led to
The feature for crossing digital signal processing method analysis thread defect waveform, obtains the defect of tyre tread, and such method is the disadvantage is that misrecognition
Rate is higher；Another kind of tyre tread data be face battle array image data, by image processing method differentiate defect type, such method in addition to
False recognition rate is higher outer, and detection time is longer.
Summary of the invention
In order to overcome above-mentioned defect existing in the prior art and deficiency, the present invention provides a kind of wheel treads to examine online
Survey method, present invention purpose are that the false recognition rate height to online test method is taken turns in solution in the prior art, and speed is slow
The problems such as, the application use FFT-GABOR Fast transforms, in conjunction with SVM classifier realize to wheel tread carry out defect recognition and
Classification can overcome the problems, such as that the false recognition rate of existing detection method is high, slow-footed, the wheel tread on-line checking side of the application
Method false recognition rate is lower than 7%, can identify the main 3 class defect of wheel tread, the defect type of 90% or more covering, detection time
It is short.
In order to solve above-mentioned problems of the prior art, the application is achieved through the following technical solutions:
A kind of wheel tread online test method, including wheel tread image acquisition step and image preprocessing step, feature
It is: further includes FFT-GABOR shift step, feature calculation step and svm classifier step；
Described image pre-treatment step specifically refers to, by image scaling and approach of mean filter to collected original wheel to stepping on
Face image is pre-processed；
The FFT-GABOR shift step specifically refers to: by GABOR filter to wheel tread image after pretreatment into
The extraction of row profile details, GABOR filter take turns tyre tread image by 5 different directions and 4 different scales
Wide detail extraction first carries out pretreated wheel tread image using GABOR filter when carrying out profile details extraction
Then FFT transform carries out IFFT inverse transformation again and obtains temporal profile information；
The feature calculation step specifically refers to: after FFT-GABOR shift step, obtaining 4*5=20 image, Mei Getu
As tri- kinds of grid of each image segmentation 1*1,2*2 and 4*4 are calculated separately each lattice by the edge contour information of all characterization wheels pair
Then all mean value and variance are conspired to create a feature by mean value and variance in son；
The svm classifier step specifically refers to: the characteristic value being calculated in feature calculation step inputted in SVN classifier,
It is identified and is classified；The training parameter of the SVM classifier is provided that
Svm-type:c-SVM
kernel_type:linear。
In the FFT-GABOR shift step, wheel tread image after pretreatment is carried out by GABOR filter
The extraction of profile details, the window size for the Gaussian filter that selected scale specifically refers to, 4 window sizes be W=9, W=
15, W=21 and W=33.
In the FFT-GABOR shift step, wheel tread image after pretreatment is carried out by GABOR filter
The extraction of profile details, selected 5 directions are 0 ° of direction, 45 ° of directions, 75 ° of directions, 90 ° of directions and 135 ° of directions respectively.
In described image acquisition step, for wheel tread removing, scratch and chip off-falling three classes defect, every class defect acquisition is not
Lower than 300 images carry out image preprocessing, FFT-GABOR transformation, feature calculation and SVM respectively and divide to acquired image
Class.
Compared with prior art, technical effect beneficial brought by the application is shown:
1, the present invention can identify the main 3 class defect of wheel tread, and the defect type of 90% or more covering can be accomplished to know online
Not, false recognition rate is lower than 7%, and can effectively shorten detection time.
2, in the present invention, GABOR filter group main purpose is to extract the profile of the different scale different directions of image
Details, 4 scales of specific selection, 5 directions carry out two dimension to image according to FFT (Fourier transform) theory in the present invention
Convolution, which is equal to, seeks multiplication in frequency domain to the two-dimension fourier transform of image and the two-dimension fourier transform of kernel function.Pass through two dimension
Fourier transform can effectively improve the operation efficiency of convolution, while GABOR filter itself is windowed FFT, meet
FFT, in order to accelerate calculating speed, the present invention first uses GABOR filter and image to carry out FFT transform, then carries out IFFT again
Inverse transformation obtains temporal profile information.The present invention can effectively shorten detection time.
3,4 window sizes of the specific selection filter of the present invention and 5 directions carry out FFT-GABOR transformation to image,
The claimed present invention is a kind of online test method, is detected in the process of running, travelling speed is based on, to inspection
It is very high to survey time requirement, selects 4 window sizes and 5 directions that can improve Detection accuracy with the balance detection time.
4, specific setting SVM classifier training parameter in the present invention, by integration test, the SVM set in the present invention divides
The training parameter of class device can satisfy the actual needs of detection device, especially may insure Detection accuracy.
5, in FFT-GABOR shift step, by GABOR filter group to wheel tread image after pretreatment into
The extraction of row profile details, the window size of 4 Gaussian filters of specific selection, respectively W=9, W=15, W=21 and W=33, is adopted
There is preferable effect with this 4 scales, it is ensured that the accuracy rate of detection.
6, in FFT-GABOR shift step, by GABOR filter group to wheel tread image after pretreatment into
The extraction of row profile details, selected 5 directions are 0 ° of direction, 45 ° of directions, 75 ° of directions, 90 ° of directions and 135 ° of sides respectively
To.This 5 directions can cover all marginal informations of thread defect, can judge defect very well from the direction at edge.
7,300 images are not less than to the acquisition of every class defect in the present invention, it is ensured that SVM classifier can restrain, it is ensured that inspection
Survey accuracy rate.
Detailed description of the invention
Fig. 1 is flowage structure schematic diagram of the present invention.
Specific embodiment
Embodiment 1
As a preferred embodiment of the present invention, present embodiment discloses:
A kind of wheel tread online test method, including wheel tread image acquisition step and image preprocessing step further include
FFT-GABOR shift step, feature calculation step and svm classifier step；
Described image pre-treatment step specifically refers to, by image scaling and approach of mean filter to collected original wheel to stepping on
Face image is pre-processed；
The FFT-GABOR shift step specifically refers to: by GABOR filter to wheel tread image after pretreatment into
The extraction of row profile details, GABOR filter take turns tyre tread image by 5 different directions and 4 different scales
Wide detail extraction, when carrying out profile details extraction, first using GABOR filter group to pretreated wheel tread image into
Then row FFT transform carries out IFFT inverse transformation again and obtains temporal profile information；
The feature calculation step specifically refers to: after FFT-GABOR shift step, obtaining 4*5=20 image, Mei Getu
As tri- kinds of grid of each image segmentation 1*1,2*2 and 4*4 are calculated separately each lattice by the edge contour information of all characterization wheels pair
Then all mean value and variance are conspired to create a feature by mean value and variance in son；
The svm classifier step specifically refers to: the characteristic value being calculated in feature calculation step inputted in SVN classifier,
It is identified and is classified；The training parameter of the SVM classifier is provided that
Svm-type:c-SVM
kernel_type:linear。
Embodiment 2
As another preferred embodiment of the present invention, present embodiment discloses:
A kind of wheel tread online test method, including wheel tread image acquisition step and image preprocessing step further include
FFT-GABOR shift step, feature calculation step and svm classifier step；
Described image pre-treatment step specifically refers to, by image scaling and approach of mean filter to collected original wheel to stepping on
Face image is pre-processed；
The FFT-GABOR shift step specifically refers to: by GABOR filter to wheel tread image after pretreatment into
The extraction of row profile details, GABOR filter take turns tyre tread image by 5 different directions and 4 different scales
Wide detail extraction, when carrying out profile details extraction, first using GABOR filter group to pretreated wheel tread image into
Then row FFT transform carries out IFFT inverse transformation again and obtains temporal profile information；In the FFT-GABOR shift step, pass through
GABOR filter carries out the extraction of profile details to wheel tread image after pretreatment, and selected scale specifically refers to
Gaussian filter window size, 4 window sizes are W=9, W=15, W=21 and W=33.The FFT-GABOR shift step
In, the extraction of profile details, selected 5 sides are carried out to wheel tread image after pretreatment by GABOR filter
To being 0 ° of direction, 45 ° of directions, 75 ° of directions, 90 ° of directions and 135 ° of directions respectively.
The feature calculation step specifically refers to: after FFT-GABOR shift step, obtaining 4*5=20 image, often
A image all characterizes the edge contour information of wheel pair, tri- kinds of grid of each image segmentation 1*1,2*2 and 4*4, calculates separately every
Then all mean value and variance are conspired to create a feature by mean value and variance in a grid；
The svm classifier step specifically refers to: the characteristic value being calculated in feature calculation step inputted in SVN classifier,
It is identified and is classified；The training parameter of the SVM classifier is provided that
Svm-type:c-SVM
kernel_type:linear。
In described image acquisition step, for wheel tread removing, scratch and chip off-falling three classes defect, every class defect acquisition is not
Lower than 300 images carry out image preprocessing, FFT-GABOR transformation, feature calculation and SVM respectively and divide to acquired image
Class.
Embodiment 3
As another preferred embodiment of the present invention, referring to Figure of description 1, present embodiment discloses:
A kind of wheel tread online test method, including wheel tread image acquisition step and image preprocessing step, feature
It is: further includes FFT-GABOR shift step, feature calculation step and svm classifier step；
Described image pre-treatment step specifically refers to, by image scaling and approach of mean filter to collected original wheel to stepping on
Face image is pre-processed, and specific for original graph image width height is all scaled 1/4, mean filter window is 3*3；
The FFT-GABOR shift step specifically refers to: by GABOR filter to wheel tread image after pretreatment into
The extraction of row profile details, GABOR filter take turns tyre tread image by 5 different directions and 4 different scales
Wide detail extraction first carries out pretreated wheel tread image using GABOR filter when carrying out profile details extraction
Then FFT transform carries out IFFT inverse transformation again and obtains temporal profile information；In the present embodiment, Gabor filter is general
Filtering method, it may be assumed that
S (x, y) indicates that image, w (x, y) indicate that gabor core, g (x, y) are filtered image.It is of the invention why use
For gabor there are two reason, the first gabor itself can extract the profile of image, this profile can describe the deemed-to-satisfy4 of edge, and second,
We can accelerate the runing time of gabor transformation using FFT transform.
The feature calculation step specifically refers to: after FFT-GABOR shift step, obtaining 4*5=20 image, often
A image all characterizes the edge contour information of wheel pair, tri- kinds of grid of each image segmentation 1*1,2*2 and 4*4, calculates separately every
Then all mean value and variance are conspired to create a feature by mean value and variance in a grid；Such as: become by FFT-GABOR
After changing step, obtain 4*5=20 image, the edge contour information of each image characterization wheel pair, each image segmentation 1*1,
Tri- kinds of grid of 2*2 and 4*4, calculate separately the mean value and variance in each grid, all mean value and variance are then conspired to create one
A feature is sent into SVM such as feature=[mean1, var1, mean2, var2......].
The svm classifier step specifically refers to: the characteristic value being calculated in feature calculation step is inputted SVN classifier
In, it is identified and is classified；The training parameter of the SVM classifier is provided that
Svm-type:c-SVM
kernel_type:linear。
SVM classifier uses general SVM classifier in the present embodiment.
Claims (4)
1. a kind of wheel tread online test method, including wheel tread image acquisition step and image preprocessing step, special
Sign is: further including FFT-GABOR shift step, feature calculation step and svm classifier step；
Described image pre-treatment step specifically refers to, by image scaling and approach of mean filter to collected original wheel to stepping on
Face image is pre-processed；
The FFT-GABOR shift step specifically refers to: by GABOR filter to wheel tread image after pretreatment into
The extraction of row profile details, GABOR filter take turns tyre tread image by 5 different directions and 4 different scales
Wide detail extraction first carries out pretreated wheel tread image using GABOR filter when carrying out profile details extraction
Then FFT transform carries out IFFT inverse transformation again and obtains temporal profile information；
The feature calculation step specifically refers to: after FFT-GABOR shift step, obtaining 4*5=20 image, Mei Getu
As tri- kinds of grid of each image segmentation 1*1,2*2 and 4*4 are calculated separately each lattice by the edge contour information of all characterization wheels pair
Then all mean value and variance are conspired to create a feature by mean value and variance in son；
The svm classifier step specifically refers to: the characteristic value being calculated in feature calculation step inputted in SVN classifier,
It is identified and is classified；The training parameter of the SVM classifier is provided that
Svm-type:c-SVM
kernel_type:linear。
2. a kind of wheel tread online test method as described in claim 1, it is characterised in that: the FFT-GABOR transformation
In step, the extraction of profile details, selected ruler are carried out to wheel tread image after pretreatment by GABOR filter
The window size of the Gaussian filter specifically referred to is spent, 4 window sizes are W=9, W=15, W=21 and W=33.
3. a kind of wheel tread online test method as described in claim 1, it is characterised in that: the FFT-GABOR transformation
In step, by GABOR filter to wheel tread image after pretreatment carry out profile details extraction, selected 5
A direction is 0 ° of direction, 45 ° of directions, 75 ° of directions, 90 ° of directions and 135 ° of directions respectively.
4. a kind of wheel tread online test method as described in claim 1, it is characterised in that: described image acquisition step
In, for wheel tread removing, scratch and chip off-falling three classes defect, every class defect acquisition is not less than 300 images, to collected
Image carries out image preprocessing, FFT-GABOR transformation, feature calculation and svm classifier respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811590768.3A CN109591846A (en) | 2018-12-20 | 2018-12-20 | A kind of wheel tread online test method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811590768.3A CN109591846A (en) | 2018-12-20 | 2018-12-20 | A kind of wheel tread online test method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109591846A true CN109591846A (en) | 2019-04-09 |
Family
ID=65964327
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811590768.3A Pending CN109591846A (en) | 2018-12-20 | 2018-12-20 | A kind of wheel tread online test method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109591846A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110239588A (en) * | 2019-06-12 | 2019-09-17 | 中国神华能源股份有限公司 | Wheel tread wear determines method and apparatus |
| CN110261139A (en) * | 2019-06-12 | 2019-09-20 | 中国神华能源股份有限公司 | Wheel tread flat recognition methods and identification device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN106204543A (en) * | 2016-06-29 | 2016-12-07 | 江南大学 | Fabric defect detection method based on single category support vector machines |
| CN108596872A (en) * | 2018-03-08 | 2018-09-28 | 北京交通大学 | The detection method of rail disease based on Gabor wavelet and SVM |
-
2018
- 2018-12-20 CN CN201811590768.3A patent/CN109591846A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN106204543A (en) * | 2016-06-29 | 2016-12-07 | 江南大学 | Fabric defect detection method based on single category support vector machines |
| CN108596872A (en) * | 2018-03-08 | 2018-09-28 | 北京交通大学 | The detection method of rail disease based on Gabor wavelet and SVM |
Non-Patent Citations (3)
| Title |
|---|
| 田启川: "《虹膜识别原理及算法》", 30 June 2010, 国防工业出版社 * |
| 薛乐: "基于傅里叶变换和Gabor变换的机织物纹理分析方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 闫亚娣: "基于Gabor滤波器的织物纹理图像分类", 《湖北工程学院学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110239588A (en) * | 2019-06-12 | 2019-09-17 | 中国神华能源股份有限公司 | Wheel tread wear determines method and apparatus |
| CN110261139A (en) * | 2019-06-12 | 2019-09-20 | 中国神华能源股份有限公司 | Wheel tread flat recognition methods and identification device |
| CN110239588B (en) * | 2019-06-12 | 2020-08-18 | 中国神华能源股份有限公司 | Wheel set tread wear determination method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105069415B (en) | Method for detecting lane lines and device | |
| CN107808378B (en) | Method for detecting potential defects of complex-structure casting based on vertical longitudinal and transverse line profile features | |
| CN106251361B (en) | A kind of rail surface defects image self-adapting division method | |
| CN105373779B (en) | A kind of vehicle safety belt intelligent detecting method and intelligent checking system | |
| CN105426905B (en) | Robot barrier object recognition methods based on histogram of gradients and support vector machines | |
| CN112767326B (en) | Rail surface defect detection method, device, equipment and storage medium | |
| CN103871053A (en) | Vision conspicuousness-based cloth flaw detection method | |
| CN102737252B (en) | Method for detecting faults caused by foreign body pollution between electrified railway insulator plates based on affine invariant moment | |
| CN106485694B (en) | A kind of high iron catenary double-jacket tube connector six-sided nut based on cascade classifier falls off defective mode detection method | |
| CN105335743A (en) | Vehicle license plate recognition method | |
| CN107665348B (en) | Digital identification method and device for digital instrument of transformer substation | |
| CN106372667A (en) | Method for detecting adverse state of inclined sleeve part screws of high-speed train overhead line system | |
| CN105447512A (en) | Coarse-fine optical surface defect detection method and coarse-fine optical surface defect detection device | |
| CN111311567A (en) | Method for identifying fastener and steel rail diseases of track line image | |
| CN103729842A (en) | Fabric defect detection method based on local statistical characteristics and overall significance analysis | |
| CN110223266A (en) | A kind of Railway wheelset tread damage method for diagnosing faults based on depth convolutional neural networks | |
| CN104318582A (en) | Detection method for bad state of rotating double-lug component pin of high-speed rail contact network on basis of image invariance target positioning | |
| CN104424632B (en) | A kind of video contrast's method for detecting abnormality and device | |
| CN107240086B (en) | A kind of fabric defects detection method based on integral nomography | |
| CN104751474A (en) | Cascade quick image defect segmentation method | |
| CN103679205A (en) | Preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification | |
| CN104268588A (en) | Automatic detection method for brake shoe borer loss fault of railway wagon | |
| CN109785301A (en) | A kind of rail corrugation automatic identifying method and appraisal procedure based on image procossing | |
| CN109591846A (en) | A kind of wheel tread online test method | |
| CN113689392A (en) | Railway fastener defect detection method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190409 |