CN113077426B - Method for detecting defects of clamp plate bolt on line in real time - Google Patents
Method for detecting defects of clamp plate bolt on line in real time Download PDFInfo
- Publication number
- CN113077426B CN113077426B CN202110317573.7A CN202110317573A CN113077426B CN 113077426 B CN113077426 B CN 113077426B CN 202110317573 A CN202110317573 A CN 202110317573A CN 113077426 B CN113077426 B CN 113077426B
- Authority
- CN
- China
- Prior art keywords
- image
- bolt
- splint
- bolts
- images
- 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.)
- Active
Links
Images
Classifications
-
- 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
- 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
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method for detecting the defects of a splint bolt on line in real time, which comprises the steps of collecting and merging images to obtain a merged atlas; selecting n merging pictures from the merging picture set to form a sample set, and marking a splint area in the merging pictures in the sample set to form a marked picture; all the labeled graphs are sent to a yolov3 network for training to obtain the network weight of yolov 3; sending a detection defect map in the image set to be detected into a deep learning network of yolov3, and predicting an ROI (region of interest) of the splint in the image; cutting out the splint image to obtain a splint image set, and selecting n1 images from the splint image set to obtain a bolt detection image sample set; marking bolt components in the images to obtain a bolt marking image training set, putting the images of the training set into a yolov3 network for training to obtain yolov3 target detection network weight, inputting the marking image set of the bolt to be detected into a yolov3 target detection network, and predicting the ROI of the bolt; and judging the defects of the bolts according to the specific coordinates (x, y) of the bolts.
Description
Technical Field
The invention relates to the field of rail transit image processing, in particular to a method for detecting a clamp plate bolt defect on line in real time.
Background
The importance of the cleat bolts as the primary means of ensuring a smooth connection between the rails is also apparent. However, at present, the control on the safety of the rail is mainly manual detection, and the difficulty of detecting the defects of the clamp bolt is greatly increased due to the particularity of the installation position of the clamp plate, so how to realize the efficient and safe detection of the defects of the clamp bolt is a subject which needs to be researched at present.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a method for detecting the defects of a splint bolt on line in real time, which comprises the following steps:
step one, merging 4 images into one image according to the row direction of the acquired image to obtain a merged atlas:
selecting n merging images from the merging image set to form a sample set, marking a splint area in the merging images in the sample set to form a marked image, and forming a to-be-detected image set by the rest merging images;
step three, all the labeled graphs are sent into a yolov3 network for training to obtain a network weight I of yolov 3;
step four, according to the weight one, the detection defect map in the image set to be detected is sent to a deep learning network of yolov3, and the ROI area (x, y, w, h) of the splint in the image to be detected is predicted;
step five, intercepting a splint image according to the position of the splint in the image in the defect image to be detected, obtaining a splint image set, and selecting n1 images from the splint image set to obtain a bolt detection image sample set;
step six, marking bolts to detect bolt components in the images in the image sample set to obtain a bolt mark map, obtaining a bolt mark map training set from the bolt mark map according to a set bolt mark pattern proportion, forming a bolt mark map set to be detected by the rest bolt mark maps, putting images of the training set into a yolov3 network for training to obtain a yolov3 target detection network weight two, inputting the bolt mark map set to be detected into a yolov3 target detection network, and predicting the ROI area of the bolts; and judging the defects of the bolts according to the specific coordinates (x, y) of the bolts.
Furthermore, coordinates P (x, y) of the splints in the image in the defect map to be detected are predicted, the images with 150 x 850 pixels are intercepted by taking P as an initial coordinate, a splint image set is obtained, and n1 images are selected from the splint image set, so that a bolt detection image sample set is obtained.
Further, the detected defect map in the image set to be detected is sent to a deep learning network of yolov3, and the ROI (x, y, w, h) of the splint in the image in the detected defect map is predicted.
Further, the method for judging the defects of the bolts according to the specific positions of the bolts comprises the following steps that 6 bolts are arranged on the clamping plate and are installed at positive and negative intervals, and if the number of the bolts on one side of the clamping plate is less than 3 or the distance between the bolts is not uniform, the bolts are lost on the clamping plate.
The invention has the beneficial effects that: the detection method can accurately detect the splint bolt in real time and judge the defects.
Drawings
FIG. 1 is a schematic diagram of a method for real-time online detection of a splint bolt defect;
FIG. 2 is a schematic diagram of a defect detection process;
fig. 3 is a schematic view of image merging.
Detailed Description
The technical solutions of the present invention are further described in detail below with reference to the accompanying drawings, but the scope of the present invention is not limited to the following.
As shown in fig. 1, a method for real-time online detection of a splint bolt defect includes:
step one, merging 4 images into one image according to the row direction of the acquired image to obtain a merged atlas:
selecting n merging images from the merging image set to form a sample set, marking a splint area in the merging images in the sample set to form a marked image, and forming a to-be-detected image set by the rest merging images;
step three, all the labeled graphs are sent to a yolov3 network for training to obtain a network weight one of yolov 3;
step four, according to the weight I, sending the detected defect map in the image set to be detected into a deep learning network of yolov3, and predicting an ROI (region of interest) of a splint in the image to be detected in the defect map, wherein w is the width of the ROI and h is the height of the ROI;
step five, intercepting a splint image according to the position of the splint in the image in the defect image to be detected, obtaining a splint image set, and selecting n1 images from the splint image set to obtain a bolt detection image sample set;
step six, marking bolts to detect bolt components in the images in the image sample set to obtain a bolt mark map, obtaining a bolt mark map training set from the bolt mark map according to a set bolt mark pattern proportion, forming a bolt mark map set to be detected by the rest bolt mark maps, putting images of the training set into a yolov3 network for training to obtain a yolov3 target detection network weight two, inputting the bolt mark map set to be detected into a yolov3 target detection network, and predicting the ROI area of the bolts; and judging the defects of the bolts according to the specific coordinates (x, y) of the bolts.
And predicting coordinates P (x, y) of the splint in the defect image to be detected, taking P as an initial coordinate, taking images of 150-850 pixels to obtain a splint image set, and selecting n1 images from the splint image set to obtain a bolt detection image sample set.
And (3) sending the detection defect map in the image set to be detected into a deep learning network of yolov3, and predicting the ROI (x, y, w, h) of the splint in the image in the defect map to be detected.
The method comprises the following steps that 6 bolts are arranged on the clamping plate and are installed at positive and negative intervals, and if the number of the bolts on one side of the clamping plate is less than 3 or the distance between the bolts is not uniform, the clamping plate is lost.
Specifically, merging of images: taking 4 images and combining the images into one image according to the row direction
1. Selecting 1000 merged pictures as a sample set
2. The area where the merged image sample collected the splint was marked. All the marked images are sent into yolov3 network for training to obtain the network weight of yolov3,
3. the merged image to be detected is fed into the deep learning network of yolov3 and the specific position of each splint in the image is predicted.
4. And (5) cutting 150-850 images at the position of the clamping plate, and screening out 1000 images to serve as a bolt detection sample set.
5. The bolt features in the 150 x 850 image are labeled. And 1000 images marked with 150 × 850 are put into yolov3 network for training, and the target detection network weight of yolov3 is obtained.
6. Inputting the image of 150 x 850 to be detected into yolov3 target detection network, and predicting the concrete position of the bolt.
According to there are 6 bolts on present splint to according to positive and negative spaced installation side, the inference: if there are fewer than 3 bolts on one side of the clamping plate or the spacing of the bolts is not uniform, there is a bolt missing at the clamping plate.
The foregoing is illustrative of the preferred embodiments of this invention, and it is to be understood that the invention is not limited to the precise form disclosed herein and that various other combinations, modifications, and environments may be resorted to, falling within the scope of the concept as disclosed herein, either as described above or as apparent to those skilled in the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (2)
1. A method for detecting the defects of a splint bolt on line in real time is characterized by comprising the following steps:
step one, merging 4 images into one image according to the row direction of the acquired image to obtain a merged atlas:
selecting n merging images from the merging image set to form a sample set, marking a splint area in the merging images in the sample set to form a marked image, and forming a to-be-detected image set by the rest merging images;
step three, all the labeled graphs are sent to a yolov3 network for training to obtain a network weight one of yolov 3;
step four, according to the weight one, the detection defect map in the image set to be detected is sent to a deep learning network of yolov3, and the ROI area (x, y, w, h) of the splint in the image to be detected is predicted;
step five, intercepting a splint image according to the position of the splint in the image in the predicted defect map to be detected to obtain a splint image set, and selecting n1 images from the splint image set to obtain a bolt detection image sample set;
step six, marking bolts to detect bolt components in the images in the image sample set to obtain a bolt mark map, obtaining a bolt mark map training set from the bolt mark map according to a set bolt mark pattern proportion, forming a bolt mark map set to be detected by the rest bolt mark maps, putting images of the training set into a yolov3 network for training to obtain a yolov3 target detection network weight two, inputting the bolt mark map set to be detected into a yolov3 target detection network, and predicting the ROI area of the bolts; and judging the defects of the bolts according to the coordinates (x, y) of the bolts.
2. The method for detecting defects of clamp bolts on line in real time according to claim 1, wherein the step of judging the defects of the bolts according to the specific positions of the bolts comprises the following steps that 6 bolts are arranged on the clamp plate and are arranged in a positive and negative spacing mode, and if the number of the bolts on one side of the clamp plate is less than 3 or the distance between the bolts is not uniform, the bolts on the clamp plate are missing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110317573.7A CN113077426B (en) | 2021-03-23 | 2021-03-23 | Method for detecting defects of clamp plate bolt on line in real time |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110317573.7A CN113077426B (en) | 2021-03-23 | 2021-03-23 | Method for detecting defects of clamp plate bolt on line in real time |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113077426A CN113077426A (en) | 2021-07-06 |
| CN113077426B true CN113077426B (en) | 2022-08-23 |
Family
ID=76611692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110317573.7A Active CN113077426B (en) | 2021-03-23 | 2021-03-23 | Method for detecting defects of clamp plate bolt on line in real time |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113077426B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886102A (en) * | 2019-01-14 | 2019-06-14 | 华中科技大学 | A kind of tumble behavior Spatio-temporal domain detection method based on depth image |
| CN112149665A (en) * | 2020-09-04 | 2020-12-29 | 浙江工业大学 | High-performance multi-scale target detection method based on deep learning |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210073692A1 (en) * | 2016-06-12 | 2021-03-11 | Green Grid Inc. | Method and system for utility infrastructure condition monitoring, detection and response |
| US20200090314A1 (en) * | 2018-09-13 | 2020-03-19 | Virtual Control Limited | System and method for determining a condition of an object |
| CN109977817B (en) * | 2019-03-14 | 2021-04-27 | 南京邮电大学 | Motor train unit bottom plate bolt fault detection method based on deep learning |
| CN110097536B (en) * | 2019-04-10 | 2023-04-18 | 东南大学 | Hexagonal bolt looseness detection method based on deep learning and Hough transform |
| CN111080597A (en) * | 2019-12-12 | 2020-04-28 | 西南交通大学 | Track fastener defect identification algorithm based on deep learning |
-
2021
- 2021-03-23 CN CN202110317573.7A patent/CN113077426B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109886102A (en) * | 2019-01-14 | 2019-06-14 | 华中科技大学 | A kind of tumble behavior Spatio-temporal domain detection method based on depth image |
| CN112149665A (en) * | 2020-09-04 | 2020-12-29 | 浙江工业大学 | High-performance multi-scale target detection method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113077426A (en) | 2021-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110567680B (en) | Track fastener looseness detection method based on angle comparison | |
| CN110310255B (en) | Point switch notch detection method based on target detection and image processing | |
| CN111564015B (en) | Method and device for monitoring perimeter intrusion of rail transit | |
| CN106679567A (en) | Contact net and strut geometric parameter detecting measuring system based on binocular stereoscopic vision | |
| CN106934800A (en) | A kind of metal plate and belt detection method of surface flaw and device based on YOLO9000 networks | |
| CN108174111B (en) | Crusing robot target image grasping means | |
| JP2023139099A (en) | Learning data collecting apparatus, learning data collecting method, and program | |
| CN106950952A (en) | For the unpiloted farm environment cognitive method of agricultural machinery | |
| CN106546263A (en) | A kind of laser leveler shoot laser line detecting method based on machine vision | |
| CN104240239A (en) | Method for detecting local road segment hazy weather based on road image | |
| CN111597904B (en) | Identification method for inclination of tunnel cable bracket | |
| CN110610516B (en) | Railway fastener nut center positioning method | |
| CN111126802A (en) | Highway inspection and evaluation method and system based on artificial intelligence | |
| CN107292926A (en) | Crusing robot movement locus verticality measuring method based on many image sequences | |
| CN108797241B (en) | Track fastener nut looseness detection method based on height comparison | |
| CN107798301A (en) | A kind of signature detection system and method for vehicle annual test | |
| CN113781537A (en) | Track elastic strip fastener defect identification method and device and computer equipment | |
| CN115018872B (en) | Intelligent control method of dust collection equipment for municipal construction | |
| CN111452840B (en) | Railway steel rail crawling displacement detection method based on monocular vision measurement technology | |
| CN113077426B (en) | Method for detecting defects of clamp plate bolt on line in real time | |
| Murao et al. | Concrete crack detection using uav and deep learning | |
| CN115713654A (en) | Track fastener bolt looseness automatic detection method based on 2D and 3D laser images | |
| CN111127409A (en) | Train component detection method based on SIFT image registration and cosine similarity | |
| CN107578001B (en) | Method and device for testing resolution of fingerprint acquisition equipment | |
| CN111380475A (en) | Truss hanging rail inspection method based on three-dimensional scanner technology |
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 |