CN109816253A - A kind of defect analysis method based on bar number identification - Google Patents
A kind of defect analysis method based on bar number identification Download PDFInfo
- Publication number
- CN109816253A CN109816253A CN201910085998.2A CN201910085998A CN109816253A CN 109816253 A CN109816253 A CN 109816253A CN 201910085998 A CN201910085998 A CN 201910085998A CN 109816253 A CN109816253 A CN 109816253A
- Authority
- CN
- China
- Prior art keywords
- defect
- pillar
- bar
- image
- method based
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 64
- 238000004458 analytical method Methods 0.000 title claims abstract description 46
- 238000001514 detection method Methods 0.000 claims abstract description 41
- 238000007689 inspection Methods 0.000 claims abstract description 4
- 238000011161 development Methods 0.000 claims description 6
- 238000012423 maintenance Methods 0.000 claims description 6
- 241001269238 Data Species 0.000 claims description 5
- 238000012552 review Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012790 confirmation Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims 1
- 238000012360 testing method Methods 0.000 abstract description 7
- 238000000034 method Methods 0.000 abstract description 6
- 238000010835 comparative analysis Methods 0.000 abstract description 4
- 238000013135 deep learning Methods 0.000 abstract description 3
- 230000018109 developmental process Effects 0.000 description 5
- 230000028473 proximal convoluted tubule segment 2 development Effects 0.000 description 2
- RTAQQCXQSZGOHL-UHFFFAOYSA-N Titanium Chemical compound [Ti] RTAQQCXQSZGOHL-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Landscapes
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a kind of defect analysis methods based on bar number identification, including S1: alert data, inspection data loading;S2: pillar unique positions calculate;S3: defect Multidimensional Comprehensive analysis.Image can be aligned using the bar recognition methods based on deep learning based on bar number, correct the position location 1C, 3C, 4C, 5C, 6C and identification 2C image bar number.Recognition efficiency is very high, can accurately identify the number in rod board, and bar board is avoided to fail to judge, and increases recall rate, improves efficiency and combine accuracy.Support the testing result comparative analysis of single detection device history dimension and different detection type (1C, 2C, 3C, 4C, 5C, 6C) all testing results of history when front standing pillar are associated check, the ability of comparative analysis, there is very high defect dipoles analysis validity and efficiency.
Description
Technical field
The present invention relates to the comprehensive analysis field of railway 6C data center more particularly to a kind of synthesis based on bar number identification
Defect analysis method.
Background technique
The defect of railway 6C data center is analyzed, and is needed 1C, 2C, 3C, 4C, 5C, 6C alarm data and hard disk number
According to based on pillar alignment, that is to say, that 1C-6C alarm data and hard disc data near some pillar will be put into the branch
Column correspondence database, main purpose be to provide various dimensions defect analysis, some power supply unit (any one equipment in 1C-6C)
It was found that defect, but since weather, camera angle or clarity etc. influence, it needs to provide detection data (figure by other C equipment
Picture, detected value), the authenticity of comprehensive analysis and judgement defect, influencing factor.It is also required to check the history testing number of same pillar simultaneously
According to, the variation tendency of defect is analyzed, to maintenance result check.
Comprehensive analysis is that 1C-6C image data needs to be aligned according to pillar there are difficult point, only same pillar image alignment
It just can be carried out comprehensive analysis, if comprehensive analysis can not be carried out with pillar image offset;When 1C, 3C, 4C, 5C, 6C acquire image
Image can be aligned based on pillar by GPS, LKJ equipment etc., if there is certain dislocation images (being based on pillar) need manually
Adjustment, and 2C equipment is not installed GPS, LKJ therefore can not be positioned, and needs manually to be aligned, consumes a large amount of man power and material.
Common bar recognition methods is similar to Car license recognition: the first step identifies rod board by target detection from image;
Second step identifies number by bar board.Since railway detection background is extremely complex, camera shooting area is big in addition, bar board
It is too small relative to whole image target, increase identification difficulty, often first step bar board identifies mistake, therefore can not
Rod number is accurately identified from image.
1C-6C bar number alignment after, carry out defect analysis, obtain three aspect contents: find positive true defect reason,
Development of defects trend is observed, is checked defect has been repaired.
Summary of the invention
To solve the above-mentioned problems, the present invention proposes a kind of railway bar recognition methods based on deep learning, using base
Image can be aligned in the bar number identification of deep learning based on bar number, correct the position location 1C, 3C, 4C, 5C, 6C and identification 2C
Image bar number.
A kind of defect analysis method based on pillar number identification, comprising the following steps:
S1: alert data, inspection data loading;
S2: pillar unique positions calculate;
S3: defect Multidimensional Comprehensive analysis.
The step S2 includes following sub-step:
S21: pillar image is extracted from original image;
S22: bar board image is extracted from pillar image;
S23: digital information is extracted from bar board image;
S24: pillar position location correction.
Preferably, S1 uses YoloV3 target detection network algorithm, and pillar image is obtained from original image.
Preferably, S2 extracts bar board image from pillar image, if using YoloV3 target detection network algorithm from branch
Rod board is detected in column image, there are more wrong report, main cause is bar board relative to whole image very little, that is, is directed to
This Small object YoloV3 target detection network algorithm of bar board is lower in the accuracy of detection, it is therefore desirable to a kind of improved side
Method.
As shown in Fig. 2, the feature of bottom extracted also is used as to the object of target detection, this makes it possible to small mesh
Mark is detected, and YoloV3 is used as detection target (3 every layer) using 3 layers of 9 anchor, and the present invention extends 4 layer 12
A anchor thus is avoided that bar board is failed to judge, and increases recall rate.
Preferably, S3 combines accuracy to improve efficiency, using the VGG19 convolution mind for having used smaller convolution kernel
Through network, the number in rod board can be accurately identified.
Preferably, S4 corrects the position location 1C, 3C, 4C, 5C, 6C according to based on pillar number identification, for there is setting for GPS
It is standby, pillar is navigated to according to GPS and the conversion of foundation leg location information;Equipment for there is LKJ, obtains according to LKJ basic data
To " route, other, kilometer post of going ", pillar is navigated to further according to foundation leg location information.2C image pillar number is identified, by
In the equipment of no GPS, LKJ, match with foundation leg location information and navigate to branch by image recognition pillar number
Column has saved a large amount of manpowers.
The step S3 defect Multidimensional Comprehensive analysis includes following sub-step:
S31: Flaw discrimination analysis;
S32: development of defects trend analysis;
S33: defect overhauls result review.
Preferably, S1 Flaw discrimination is analyzed, and after some C has found defect, default association shows other 5 C same
The last detection data of pillar, including a variety of images, detected value.Since the detection focus of each C is different, have certainly
The data characteristics of body, such as: 2C pays close attention to entire pillar panorama, and 4C pays close attention to the details of pillar part, by different detection datas
Whether comprehensive analysis of the personnel to defect is analyzed in comprehensive analysis, help, to being that defect carries out qualitative confirmation, and passes through other C
Detection data can excavate occur defect the reason of.
Preferably, S2 development of defects trend analysis provides 6 C in all of same pillar after some C has found defect
The inquiry of history detection data is shown, passes through the comprehensive analysis of the different detection datas to different time points.Personnel's palm is analyzed in help
Historical development trend of the defect from rudiment to generation is held, summing up experience prevents trouble before it happens.
Preferably, S3 defect maintenance result review uploads the detection data with shore position after defect maintenance.It can be with
It checks analyzing defect treated state, testing result is checked.
The beneficial effects of the present invention are:
(1) recognition efficiency of the present invention is very high, can identify that 16 frame image accuracys rate can reach using the GPU of TITAN XP is per second
99.6% or so, the data in rod board can be accurately identified, bar board is avoided to fail to judge, increases recall rate, improves efficiency and combine
Accuracy;
(2) not only support single detection device history testing result comparative analysis, but also the different detection types of support (1C, 2C,
3C, 4C, 5C, 6C) the crutched all testing results of history of institute are associated check, comparative analysis, there is very high defect
Discriminatory analysis efficiency and use value.
Detailed description of the invention
Fig. 1 is a kind of defect analysis method flow chart based on bar number identification;
Fig. 2 is identification step figure;
Fig. 3 is that S2 step improves schematic diagram.
Specific embodiment
In order to make those skilled in the art more fully understand technical solution of the present invention, with reference to the accompanying drawing to the present invention
It is described in further detail.
As shown in Figure 1, a kind of defect analysis method based on bar number identification, comprising the following steps:
S1: alert data, inspection data loading;
S2: pillar unique positions calculate;
S3: defect Multidimensional Comprehensive analysis.
The step 2 includes following sub-step:
S21: pillar image is extracted from original image;
S22: bar board image is extracted from pillar image;
S23: digital information is extracted from bar board image;
S24: pillar position location correction.
Preferably, S1 uses YoloV3 target detection network algorithm, and pillar image is obtained from original image.
Preferably, S2 extracts bar board image from pillar image, if using YoloV3 target detection network algorithm from branch
Rod board is detected in column image, there are more wrong report, main cause is bar board relative to whole image very little, that is, is directed to
This Small object YoloV3 target detection network algorithm of bar board is lower in the accuracy of detection, it is therefore desirable to a kind of improved side
Method.
As shown in Fig. 2, the feature of bottom extracted also is used as to the object of target detection, this makes it possible to small mesh
Mark is detected, and YoloV3 is used as detection target (3 every layer) using 3 layers of 9 anchor, and the present invention extends 4 layer 12
A anchor thus is avoided that bar board is failed to judge, and increases recall rate.
Preferably, S3 combines accuracy to improve efficiency, using the VGG19 convolution mind for having used smaller convolution kernel
Through network, the number in rod board can be accurately identified.
Preferably, S4 corrects the position location 1C, 3C, 4C, 5C, 6C according to based on pillar number identification, for there is setting for GPS
It is standby, pillar is navigated to according to GPS and the conversion of foundation leg location information;Equipment for there is LKJ, obtains according to LKJ basic data
To " route, other, kilometer post of going ", pillar is navigated to further according to foundation leg location information.2C image pillar number is identified, by
In the equipment of no GPS, LKJ, match with foundation leg location information and navigate to branch by image recognition pillar number
Column has saved a large amount of manpowers.
The step c defect Multidimensional Comprehensive analysis includes following sub-step:
S1: Flaw discrimination analysis;
S2: development of defects trend analysis;
S3: defect overhauls result review.
Preferably, S1 Flaw discrimination is analyzed, and after some C has found defect, default association shows other 5 C same
The last detection data of pillar, including a variety of images, detected value.Since the detection focus of each C is different, have certainly
The data characteristics of body, such as: 2C pays close attention to entire pillar panorama, and 4C pays close attention to the details of pillar part, by different detection datas
Whether comprehensive analysis of the personnel to defect is analyzed in comprehensive analysis, help, to being that defect carries out qualitative confirmation, and passes through other C
Detection data can excavate occur defect the reason of.
Preferably, S2 development of defects trend analysis provides 6 C in all of same pillar after some C has found defect
The inquiry of history detection data is shown, passes through the comprehensive analysis of the different detection datas to different time points.Personnel's palm is analyzed in help
Historical development trend of the defect from rudiment to generation is held, summing up experience prevents trouble before it happens.
Preferably, S3 defect maintenance result review uploads the detection data with shore position after defect maintenance.It can be with
It checks analyzing defect treated state, testing result is checked.
The above disclosure is only the preferred embodiments of the present invention, cannot limit the right model of the present invention with this certainly
It encloses, therefore equivalent changes made in accordance with the claims of the present invention, is still within the scope of the present invention.
Claims (10)
1. a kind of defect analysis method based on bar number identification, it is characterised in that: the following steps are included:
S1: alert data, inspection data loading;
S2: pillar unique positions calculate;
S3: defect Multidimensional Comprehensive analysis.
2. a kind of defect analysis method based on bar number identification according to claim 1, it is characterised in that: the step
Rapid S2 the following steps are included:
S21: pillar image is extracted from original image;
S22: bar board image is extracted from pillar image;
S23: digital bar information is extracted from bar board image;
S24: pillar position location correction.
3. a kind of defect analysis method based on bar number identification according to claim 2, it is characterised in that: the step
Rapid S21 extracts pillar image using YoloV3 algorithm of target detection from original image.
4. a kind of defect analysis method based on bar number identification according to claim 2, it is characterised in that: the step
Rapid S22 uses YoloV3 algorithm of target detection, and the feature that bottom is extracted haves three layers 9 as the object of target detection in original
4 layers of 12 anchor are extended on the basis of anchor to detect Small object as detection target.
5. a kind of defect analysis method based on bar number identification according to claim 2, it is characterised in that: the step
Rapid S23 extracts digital bar information using 19 algorithm of depth convolutional neural networks VGG from bar board image.
6. a kind of defect analysis method based on bar number identification according to claim 2, it is characterised in that: described
S24 navigates to pillar for there is the equipment of GPS, according to GPS and the conversion of foundation leg location information;Equipment for there is LKJ, root
Route, capable other and kilometer post are obtained according to LKJ basic data, navigates to pillar further according to foundation leg location information;For not having
The equipment of GPS, LKJ match with foundation leg location information navigating to pillar by image recognition pillar number.
7. a kind of defect analysis method based on bar number identification according to claim 1, it is characterised in that: the step
Rapid S3 is further comprising the steps of:
S31: Flaw discrimination analysis;
S32: development of defects trend analysis;
S33: defect overhauls result review.
8. a kind of defect analysis method based on bar number identification according to claim 7, it is characterised in that: the step
Rapid S31 is associated with remaining equipment in the last detection data of same pillar, carries out qualitative after some equipment finds defect
Confirmation.
9. a kind of defect analysis method based on bar number identification according to claim 7, it is characterised in that: the step
Rapid S32 provides remaining equipment in all history detection datas of same pillar after some equipment finds defect.
10. a kind of defect analysis method based on bar number identification according to claim 7, it is characterised in that: described
Step S33 uploads the detection data with shore position after defect maintenance, analyzing defect is checked treated state, to detection
As a result it is checked.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910085998.2A CN109816253A (en) | 2019-01-29 | 2019-01-29 | A kind of defect analysis method based on bar number identification |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910085998.2A CN109816253A (en) | 2019-01-29 | 2019-01-29 | A kind of defect analysis method based on bar number identification |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109816253A true CN109816253A (en) | 2019-05-28 |
Family
ID=66605632
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910085998.2A Pending CN109816253A (en) | 2019-01-29 | 2019-01-29 | A kind of defect analysis method based on bar number identification |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109816253A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100021049A1 (en) * | 2008-07-28 | 2010-01-28 | Nec Electronics Corporation | Method and apparatus for failure analysis of semiconductor integrated circuit devices |
CN104298993A (en) * | 2014-01-09 | 2015-01-21 | 郑州金惠计算机系统工程有限公司 | Pole number positioning and identification method suitable for railway line in complex scene |
CN105416097A (en) * | 2015-11-12 | 2016-03-23 | 北京天格高通科技有限公司 | Contact screen detecting and overhauling car and method based on laser radar |
CN108597053A (en) * | 2018-04-25 | 2018-09-28 | 北京御航智能科技有限公司 | Shaft tower and channel targets identification based on image data and neural network and defect diagnostic method |
-
2019
- 2019-01-29 CN CN201910085998.2A patent/CN109816253A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100021049A1 (en) * | 2008-07-28 | 2010-01-28 | Nec Electronics Corporation | Method and apparatus for failure analysis of semiconductor integrated circuit devices |
CN104298993A (en) * | 2014-01-09 | 2015-01-21 | 郑州金惠计算机系统工程有限公司 | Pole number positioning and identification method suitable for railway line in complex scene |
CN105416097A (en) * | 2015-11-12 | 2016-03-23 | 北京天格高通科技有限公司 | Contact screen detecting and overhauling car and method based on laser radar |
CN108597053A (en) * | 2018-04-25 | 2018-09-28 | 北京御航智能科技有限公司 | Shaft tower and channel targets identification based on image data and neural network and defect diagnostic method |
Non-Patent Citations (3)
Title |
---|
孔繁利: "基于视觉识别算法的接触网自动巡检系统设计", 《铁道机车车辆》 * |
张润宝等: "6C系统信息综合应用技术方案研究", 《中国铁路》 * |
陈忠革: "接触网安全巡检系统中的支柱号牌识别算法", 《电气化铁道》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110033489A (en) | A kind of appraisal procedure, device and the equipment of vehicle location accuracy | |
US10878556B2 (en) | Interactive semi-automated borescope video analysis and damage assessment system and method of use | |
CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
US20080300748A1 (en) | Gps enabled datalogging system for a non-destructive inspection instrument | |
CN107507174A (en) | Power plant's instrument equipment drawing based on hand-held intelligent inspection is as recognition methods and system | |
CN111260615A (en) | Laser and machine vision fusion-based method for detecting apparent diseases of unmanned aerial vehicle bridge | |
JP2016524334A (en) | Wafer inspection using free form protection area | |
CN109543665A (en) | Image position method and device | |
CN115423815B (en) | Tunnel detection system and method based on mobile laser scanning | |
KR101130284B1 (en) | System and method for checking accuracy of true-orthoimage using digital map | |
Lang et al. | Pavement cracking detection and classification based on 3d image using multiscale clustering model | |
CN110728269A (en) | High-speed rail contact net support pole number plate identification method | |
JP6908445B2 (en) | Maintenance management method for change detectors and railway equipment parts | |
CN116796044B (en) | Data processing method and device for POI data set | |
CN109816253A (en) | A kind of defect analysis method based on bar number identification | |
JP6960252B2 (en) | Maintenance management method for image analyzers and railway equipment parts | |
CN103606321B (en) | Based on video location and the road test judgment method of digital modeling combined technology | |
CN114529543B (en) | Installation detection method and device for peripheral screw gasket of aero-engine | |
CN111966777B (en) | Railway GIS map data processing method | |
CN111160420B (en) | Map-based fault diagnosis method, map-based fault diagnosis device, electronic equipment and storage medium | |
CN113124326B (en) | Oil and gas pipeline leakage line patrol detection method and system | |
CN111220173B (en) | POI (Point of interest) identification method and device | |
Besharatian et al. | Non-Contact Bridge Deck Evaluation Using Infrared Thermography, A Pipeline for Data Annotation | |
CN111080626B (en) | Detection method and electronic equipment | |
CN111489339B (en) | Method for detecting backup and main defects of high-speed railway locator bolts |
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: 20190528 |
|
RJ01 | Rejection of invention patent application after publication |