CN110044905A - A kind of crack detecting method of double-block type sleeper - Google Patents
A kind of crack detecting method of double-block type sleeper Download PDFInfo
- Publication number
- CN110044905A CN110044905A CN201910237937.3A CN201910237937A CN110044905A CN 110044905 A CN110044905 A CN 110044905A CN 201910237937 A CN201910237937 A CN 201910237937A CN 110044905 A CN110044905 A CN 110044905A
- Authority
- CN
- China
- Prior art keywords
- picture
- double
- crackle
- block
- probability
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- 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
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Chemical & Material Sciences (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of crack detecting methods of double-block type sleeper, comprising: training deep neural network model;Extract the hough transform picture of double-block type sleeper to be detected;The low-level image feature for extracting above-mentioned hough transform picture, obtains weighting coefficient w;Gridding is carried out to hough transform picture to divide to obtain picture block, picture block is inputted into trained deep neural network model, obtaining each picture block, there are the Probability ps of crackle;By Probability p multiplied by weighting coefficient w, obtaining final each picture block, there are the final probability of crackle, determine whether picture block has crackle.The present invention provides a kind of sleeper crack recognition methods based on underlying image feature and deep neural network weighted discrimination, this method combines the characteristics of the quick of underlying image feature, robustness, and the strong identification of depth convolutional neural networks, make it possible to accurately differentiate the crackle occurred in double-block type sleeper production process, and effectively avoids missing discrimination caused by because of water stain stain.
Description
Technical field
The present invention relates to sleeper manufacturing technology field more particularly to a kind of crack detecting methods of double-block type sleeper.
Background technique
Currently, in double-block type sleeper detection process, to guarantee that sleeper crack becomes apparent from accurate recognition detection, and double
After the completion of block type sleeper demoulding, generallys use and alcohol brushing naked eyes identification and artificial manually is carried out to double-block type sleeper support rail groove
It affixes one's seal mode, the mode of this traditional artificial detection can make personnel's visual fatigue that can not carry out sleeper crack accurate intelligence to know
It does not detect, extends detection time, there is detection missed detection risk influences sleeper production.Simultaneously as artificial by visually to sleeper
Recognition detection is carried out, double-block type sleeper crackle can not be rationalized, intelligent tracing record and later period trace.To sum up, by
Pass through undesirable element brought by naked eyes recognition detection crackle in artificial;On the other hand, manually affixing one's seal makes the production time, examines knot
By lack of standardization, waste time more slowly, thus double-block type sleeper crackle can not recognition detection, tracing record and automatic code-spraying, influence
The practical problems such as sleeper production.
And although the method developed in recent years based on image procossing detection is able to achieve quick image crack detection,
But it is easy to be easy the water stain of surface, stain etc. in Double-block Track of Express Railway production process, cause figure by such environmental effects
As processing method failure, crackle can not be effectively detected.
Summary of the invention
The object of the present invention is to provide a kind of sleepers based on underlying image feature and deep neural network weighted discrimination to split
Line recognition methods can effectively avoid missing discrimination caused by because of water stain stain.
For achieving the above object, the technical scheme is that a kind of crack detecting method of double-block type sleeper, packet
Include following steps:
Step 1) is trained deep neural network on sleeper image data collection, obtains splitting for differentiating that sleeper whether there is
The deep neural network model of line;
Step 2 positions its recess region using double-block type sleeper self poisoning to be detected hole, with the recess region of double-block type sleeper
For area-of-interest, hough transform picture is extracted;
Step 3) extracts the low-level image feature of above-mentioned hough transform picture, with the point quantity of non-zero pixels present in low-level image feature divided by
Its sum of all pixels obtains weighting coefficient w;
Step 4) carries out gridding to hough transform picture and divides to obtain picture block, and picture block is inputted trained depth nerve
Network model, obtaining each picture block, there are the Probability ps of crackle;
Step 5) sets probability threshold value, and by obtained each picture block, there are the Probability ps of crackle multiplied by weighting coefficient, obtains final
There are the final probability of crackle for each picture block, are judged to the picture block that final probability is greater than probability threshold value to have crackle.
Further, the step 3) specifically: hough transform picture is converted into grayscale image, after automatic white balance,
Impurity noise is removed, is obtained by region growing and gaussian filtering using the fringe region in Canny operator extraction picture
Bianry image, counts non-zero pixels point quantity present in bianry image, and non-zero pixels point quantity is obtained divided by its sum of all pixels
Weighting coefficient w.
Sentence the beneficial effects of the present invention are: the present invention provides a kind of weight based on underlying image feature and deep neural network
The characteristics of other sleeper crack recognition methods, this method combines the quick of underlying image feature, robustness and depth convolution
The strong identification of neural network makes it possible to accurately differentiate the crackle occurred in double-block type sleeper production process, and effective
It avoids missing discrimination caused by because of water stain stain.
Detailed description of the invention
Fig. 1 is that hough transform picture of the present invention extracts schematic diagram;
Fig. 2 is the weighting coefficient distribution schematic diagram of hough transform picture of the present invention;
Fig. 3 formats for hough transform picture network of the present invention and divides schematic diagram;
Fig. 4 is to determine the judgement obtained later for the window of crackle by deep neural network model.
Specific embodiment
Below in conjunction with attached drawing, technical scheme in the embodiment of the invention is clearly and completely described.
A kind of crack detecting method of double-block type sleeper, comprising the following steps:
Step 1) is trained deep neural network on sleeper image data collection, obtains splitting for differentiating that sleeper whether there is
The deep neural network model of line.
Step 2 positions its recess region using double-block type sleeper self poisoning to be detected hole, with the groove of double-block type sleeper
Region is area-of-interest, extracts hough transform picture, as shown in Figure 1.
Step 3) extracts the low-level image feature of above-mentioned hough transform picture, with the point quantity of non-zero pixels present in low-level image feature
Weighting coefficient w is obtained divided by its sum of all pixels;Specifically: hough transform picture is converted into grayscale image, after automatic white balance,
Impurity noise is removed, is obtained by region growing and gaussian filtering using the fringe region in Canny operator extraction picture
Bianry image, counts non-zero pixels point quantity present in bianry image, and non-zero pixels point quantity is obtained divided by its sum of all pixels
Weighting coefficient w.
In the present embodiment, fragmented parts have been partitioned into 6 segments, as shown in Fig. 2, therefore obtaining the different positions of 6 correspondences
The weighting coefficient set.
Step 4) carries out gridding to hough transform picture and divides to obtain picture block, instructs as shown in figure 3, picture block is inputted
The deep neural network model perfected, obtaining each picture block, there are the Probability ps of crackle.
As shown in figure 4, to determine the judgement obtained later for the window of crackle, wherein counting by deep neural network model
The probability that the neural network that word provides determines, for the ease of checking, the window of small probability does not carry out rendering expression.
Step 5) sets probability threshold value, and by obtained each picture block, there are the Probability ps of crackle multiplied by weighting coefficient, obtains
There are the final probability of crackle for final each picture block, and the picture block for determining that final probability is greater than probability threshold value is to have crackle.
As shown in figure 4, neural network output probability is 1.00 by taking leftmost detection window as an example, low-level image feature adds
Weight coefficient is 2.91, then the final probability of the window is 1.00 * 2.91=2.91.By taking window at black patch as an example, output probability
0.94, weighting coefficient 4.46, final probability is 0.94 * 4.46=4.1924.In this example, whole probability threshold value is set as
2.00, then these windows are all judged as slit region.
The above method combines the characteristics of the quick of underlying image feature, robustness and depth convolutional neural networks
Strong identification makes it possible to accurately differentiate the crackle occurred in double-block type sleeper production process, and effectively avoids because water stain
Discrimination is missed caused by stain.
Described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on the present invention
In embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, belongs to the scope of the present invention.
Claims (2)
1. a kind of crack detecting method of double-block type sleeper, which comprises the following steps:
Step 1) is trained deep neural network on sleeper image data collection, obtains splitting for differentiating that sleeper whether there is
The deep neural network model of line;
Step 2 positions its recess region using double-block type sleeper self poisoning to be detected hole, with the recess region of double-block type sleeper
For area-of-interest, hough transform picture is extracted;
Step 3) extracts the low-level image feature of above-mentioned hough transform picture, with the point quantity of non-zero pixels present in low-level image feature divided by
Its sum of all pixels obtains weighting coefficient w;
Step 4) carries out gridding to hough transform picture and divides to obtain picture block, and picture block is inputted trained depth nerve
Network model, obtaining each picture block, there are the Probability ps of crackle;
Step 5) sets probability threshold value, and by obtained each picture block, there are the Probability ps of crackle multiplied by weighting coefficient, obtains final
There are the final probability of crackle for each picture block, are judged to the picture block that final probability is greater than probability threshold value to have crackle.
2. a kind of crack detecting method of double-block type sleeper as described in claim 1, which is characterized in that the step 3) is specific
Are as follows: hough transform picture is converted into grayscale image, after automatic white balance, utilizes the marginal zone in Canny operator extraction picture
Domain removes impurity noise by region growing and gaussian filtering, obtains bianry image, counts non-present in bianry image
Zero pixel quantity, non-zero pixels point quantity obtain weighting coefficient w divided by its sum of all pixels.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910237937.3A CN110044905A (en) | 2019-03-27 | 2019-03-27 | A kind of crack detecting method of double-block type sleeper |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910237937.3A CN110044905A (en) | 2019-03-27 | 2019-03-27 | A kind of crack detecting method of double-block type sleeper |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110044905A true CN110044905A (en) | 2019-07-23 |
Family
ID=67275404
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910237937.3A Pending CN110044905A (en) | 2019-03-27 | 2019-03-27 | A kind of crack detecting method of double-block type sleeper |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110044905A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112304960A (en) * | 2020-12-30 | 2021-02-02 | 中国人民解放军国防科技大学 | High-resolution image object surface defect detection method based on deep learning |
CN112686888A (en) * | 2021-01-27 | 2021-04-20 | 上海电气集团股份有限公司 | Method, system, equipment and medium for detecting cracks of concrete sleeper |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104392224A (en) * | 2014-12-04 | 2015-03-04 | 西南交通大学 | Crack detection method for road surface |
CN106548182A (en) * | 2016-11-02 | 2017-03-29 | 武汉理工大学 | Based on deep learning and the causal analytic pavement distress survey method and device of master |
CN108562589A (en) * | 2018-03-30 | 2018-09-21 | 慧泉智能科技(苏州)有限公司 | A method of magnetic circuit material surface defect is detected |
CN108665450A (en) * | 2018-04-28 | 2018-10-16 | 中国农业大学 | A kind of corn ear mechanical damage area recognizing method |
CN109099840A (en) * | 2018-09-12 | 2018-12-28 | 北京好运达智创科技有限公司 | Ferric-cement sleeper vision-based detection and judge system |
CN109389116A (en) * | 2017-08-14 | 2019-02-26 | 高德软件有限公司 | A kind of character detection method and device |
-
2019
- 2019-03-27 CN CN201910237937.3A patent/CN110044905A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104392224A (en) * | 2014-12-04 | 2015-03-04 | 西南交通大学 | Crack detection method for road surface |
CN106548182A (en) * | 2016-11-02 | 2017-03-29 | 武汉理工大学 | Based on deep learning and the causal analytic pavement distress survey method and device of master |
CN109389116A (en) * | 2017-08-14 | 2019-02-26 | 高德软件有限公司 | A kind of character detection method and device |
CN108562589A (en) * | 2018-03-30 | 2018-09-21 | 慧泉智能科技(苏州)有限公司 | A method of magnetic circuit material surface defect is detected |
CN108665450A (en) * | 2018-04-28 | 2018-10-16 | 中国农业大学 | A kind of corn ear mechanical damage area recognizing method |
CN109099840A (en) * | 2018-09-12 | 2018-12-28 | 北京好运达智创科技有限公司 | Ferric-cement sleeper vision-based detection and judge system |
Non-Patent Citations (5)
Title |
---|
SATTAR DORAFSHAN 等: "Comparison of deep convolutional neural networks and edge detectors for image-based crack detection in concrete", 《CONSTRUCTION AND BUILDING MATERIALS》 * |
XIN ZHANG 等: "An improved method of rail health monitoring based on CNN and multiple acoustic emission events", 《2017 IEEE INTERNATIONAL INSTRUMENTATION AND MEASUREMENT TECHNOLOGY CONFERENCE (I2MTC)》 * |
ZHUN FAN 等: "Automatic Pavement Crack Detection Based on Structured Prediction with the Convolutional Neural Network", 《COMPUTER SCIENCE > COMPUTER VISION AND PATTERN RECOGNITION》 * |
崔磊: "基于图像处理技术的建筑物表面裂缝测量方法的应用研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
张辉 等: "钢轨缺陷无损检测与评估技术综述", 《仪器仪表学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112304960A (en) * | 2020-12-30 | 2021-02-02 | 中国人民解放军国防科技大学 | High-resolution image object surface defect detection method based on deep learning |
CN112686888A (en) * | 2021-01-27 | 2021-04-20 | 上海电气集团股份有限公司 | Method, system, equipment and medium for detecting cracks of concrete sleeper |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113592861B (en) | Bridge crack detection method based on dynamic threshold | |
CN109284758B (en) | Invoice seal eliminating method and device and computer storage medium | |
CN104535586B (en) | Strip steel edge defect detection identification method | |
CN105975972A (en) | Bridge crack detection and characteristic extraction method based on image | |
CN108898085A (en) | Intelligent road disease detection method based on mobile phone video | |
CN105447851A (en) | Glass panel sound hole defect detection method and system | |
CN105405142A (en) | Edge defect detection method and system for glass panel | |
CN110223282B (en) | Automatic identification method and system for organic pores and inorganic pores of shale | |
CN103442209A (en) | Video monitoring method of electric transmission line | |
CN110044905A (en) | A kind of crack detecting method of double-block type sleeper | |
CN111127448B (en) | Method for detecting air spring fault based on isolated forest | |
CN111080602B (en) | Method for detecting foreign matters in water leakage hole of railway wagon | |
CN111767874B (en) | Pavement disease detection method based on deep learning | |
CN107818321A (en) | A kind of watermark date recognition method for vehicle annual test | |
CN111489352A (en) | Tunnel gap detection and measurement method and device based on digital image processing | |
CN110838117A (en) | Rock face porosity identification method based on hole wall image | |
CN104168462B (en) | Camera scene change detection method based on image angle point set feature | |
CN112233111A (en) | Tunnel gap detection method based on digital image processing | |
CN107578414B (en) | Method for processing pavement crack image | |
CN107330440B (en) | Ocean state calculation method based on image recognition | |
CN111950606B (en) | Knife switch state identification method, device, equipment and storage medium | |
CN116152674A (en) | Dam unmanned aerial vehicle image crack intelligent recognition method based on improved U-Net model | |
CN105678795B (en) | A kind of field shoe watermark image method of inspection | |
CN105740831B (en) | A kind of stopping line detecting method applied to intelligent driving | |
CN112200782A (en) | Photovoltaic cell panel damage risk degree detection method based on artificial intelligence |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190723 |
|
RJ01 | Rejection of invention patent application after publication |