CN113111875A - Seamless steel rail weld defect identification device and method based on deep learning - Google Patents
Seamless steel rail weld defect identification device and method based on deep learning Download PDFInfo
- Publication number
- CN113111875A CN113111875A CN202110362862.9A CN202110362862A CN113111875A CN 113111875 A CN113111875 A CN 113111875A CN 202110362862 A CN202110362862 A CN 202110362862A CN 113111875 A CN113111875 A CN 113111875A
- Authority
- CN
- China
- Prior art keywords
- steel rail
- seamless steel
- network
- image
- deep learning
- 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
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 60
- 239000010959 steel Substances 0.000 title claims abstract description 60
- 230000007547 defect Effects 0.000 title claims abstract description 27
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000013135 deep learning Methods 0.000 title claims abstract description 19
- 238000003466 welding Methods 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 238000007689 inspection Methods 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 9
- 230000001502 supplementing effect Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000005498 polishing Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 241001669679 Eleotris Species 0.000 description 1
- 208000037656 Respiratory Sounds Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a seamless steel rail weld defect identification device and method based on deep learning, wherein the method comprises the following steps: acquiring image data, namely acquiring a sample image of the rail web part of the seamless steel rail at real time without intervals by using an image acquisition unit and transmitting the sample image to a central computer; image identification calculation, wherein a central computer screens pictures containing seamless steel rail welding seams in a sample image, and marks the welding seam area of the screened pictures; and (3) data training, namely performing target recognition training on the marked picture by using the initialized deep learning network to obtain a model file, performing multi-layer recursive network training on the marked sample image by using the model file, and outputting a final seamless steel rail weld joint defect recognition result. According to the method, the image data acquired in real time on site at the early stage is screened, labeled and analyzed, deep learning training is carried out on the image data to obtain a model base, and the defect of the rail weld joint can be efficiently identified in real time by utilizing the model base for detection.
Description
Technical Field
The invention relates to the field of rail flaw detection, in particular to a seamless rail weld defect identification device and method based on deep learning.
Background
With the development of the high-speed rail industry, the trend is to use industrial robots to replace manual work for polishing the welding seams of the seamless high-speed rail. Aiming at the problems of complex shape characteristic information, low manual polishing efficiency, bad manual operation environment and the like of the seamless steel rail welding seam, a set of welding seam polishing system is established to replace manual polishing of the seamless steel rail welding seam, so that the seamless steel rail welding seam polishing efficiency is improved, and the labor cost is reduced. The system is based on an industrial robot, and combines a machine vision technology and a point cloud processing technology to perform position identification and characteristic information extraction on a steel rail welding seam to generate a seamless steel rail welding seam polishing path. The welding seam between the existing seamless steel rails is influenced by a welding method, a welding process and a polishing mode, the position of the welding seam between the seamless steel rails is difficult to identify during manual track inspection, and the problem of missing inspection exists.
For example, patent No. CN109490416A discloses a weld joint identification method applied to double rail type steel rail flaw detection, which includes: s0, initializing a flaw detection and rail surface monitoring integrated system; s1, acquiring damage data inside the steel rail by using a double-rail type steel rail ultrasonic flaw detector; s2, converting the ultrasonic analog signal of the damage data in the steel rail into a digital signal, and sending the digital signal to a PC (personal computer) end through Ethernet communication; s3, synchronously executing the step S1, and collecting the image data of the surface of the steel rail by a camera; s4, synthesizing the acquired steel rail surface image data into a steel rail surface image; s5, identifying the synthesized rail surface image, and judging whether the image has a welding seam; s6, calling the coded value of the steel rail surface image data during acquisition, and converting the coded value into mileage information; and S7, displaying the flaw detection data and automatically marking the welding seam in the flaw detection B display screen rolling area. Although this scheme can automatic identification rail face image, and it marks to detect a flaw B shows that the interface automatic identification, but the recognition effect to the welding seam is not good, and is not accurate enough, and welding seam recognition efficiency is not high.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a seamless steel rail weld defect identification device and method based on deep learning.
The purpose of the invention is realized by the following technical scheme:
the seamless steel rail weld defect identification method based on deep learning comprises the following steps:
s1, acquiring image data, namely acquiring a sample image of the rail web part of the seamless steel rail in real time at intervals by using an image acquisition unit and sending the sample image to a central computer;
s2, image recognition and calculation, wherein the central computer screens the pictures containing the seamless steel rail welding seams in the sample images, and marks the welding seam areas of the screened pictures;
and S3, performing data training, namely performing target recognition training on the marked picture by using the initialized deep learning network to obtain a model file, performing multi-layer recursive network training on the marked sample image by using the model file, and outputting a final seamless steel rail weld joint defect recognition result.
Specifically, step S2 specifically includes the following sub-steps:
s201, the central computer performs picture screening on a large number of received sample images, and extracts pictures containing seamless steel rail welding seams from the sample images as positive and negative sample pictures;
s202, identifying and marking the welding seam area of the picture containing the seamless steel rail welding seam by using marking software, and taking the marked sample picture as a sample training set.
Specifically, the step S3 specifically includes the following sub-steps:
s301, improving a yolov3 algorithm, constructing a multistage cascade feature map network Lyolo according to the improved yolov3 algorithm, and initializing the multistage cascade feature map network;
s302, importing a sample training set into a multi-level cascade feature map network for multi-level up-sampling cascade, and integrating multi-scale output results to form a target identification model file, namely a multi-level cascade feature pyramid;
and S303, importing the image data acquired in real time into a target identification model file for multi-layer recursive network training, and outputting a weld joint identification result in the image data in real time.
Specifically, the network structure of the multilevel cascade feature graph network Lyolo includes a first convolution layer, a second convolution layer, and a residual error network, where an output of the first convolution layer is connected to an input of the second convolution layer, and an output of the second convolution layer is connected to an input of the residual error network.
Specifically, the initializing the multistage cascade feature map network process in the substep S301 specifically includes: initializing a first convolution layer, the first convolution layer comprising 16 filters of 3 x 3 with a step size of 2; performing convolution operation on the image with the input size of 384 by 384 to output a characteristic map with the size of 192 by 16; and then inputting the output 192 × 16 feature map into an initialized second convolution layer, wherein the second convolution layer comprises 32 3 × 3 filters and has the step size of 2, and outputting 96 × 32 feature maps after convolution to finish the initialization of the multistage cascade feature map network.
A recognition device of a seamless steel rail weld defect recognition method based on deep learning comprises a central calculation processing unit, a steel rail inspection vehicle and an image acquisition unit arranged on the steel rail inspection vehicle; the central computing processing unit comprises an industrial computer and a power supply, and the industrial computer is connected with the power supply; the image acquisition unit comprises a high-definition industrial digital camera, a lens, a light supplementing light source, a speed encoder and a detection box, wherein the high-definition industrial digital camera and the light supplementing light source are respectively arranged in the detection box, and the detection box is fixed at the bottom of the steel rail inspection vehicle; and the industrial computer is respectively connected with the high-definition industrial digital camera, the lens, the light supplementing light source and the speed encoder.
The invention has the beneficial effects that: according to the method, the pictures containing the welding seams between the seamless steel rails are screened out through image data acquired in real time on site at the early stage, the marking software is used for marking the welding seam area between the seamless steel rails, a rough model identification is trained on the marking welding seam pictures, the pictures of the mistakenly identified welding seams are screened out, and then a part of the pictures of the mistakenly identified welding seams is randomly selected for retraining. Through continuous strengthening and optimizing training, the balance and diversity of the training library samples are finally guaranteed, and the defects of the rail welding seam can be efficiently identified in real time by utilizing the obtained model library.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a network structure diagram of the multilevel cascade feature map network Lyolo of the present invention;
fig. 3 is a diagram of the Lyolo network mapping and tensor relationship of the present invention.
FIG. 4 is a comparison of recognition accuracy results of the present invention;
fig. 5 is an electrical schematic diagram of the identification device of the present invention.
Detailed Description
In order to more clearly understand the technical features, objects, and effects of the present invention, embodiments of the present invention will now be described with reference to the accompanying drawings.
In this embodiment, as shown in fig. 1, a method for identifying a weld defect of a seamless steel rail based on deep learning includes the following steps:
s1, acquiring image data, namely acquiring a sample image of the rail web part of the seamless steel rail in real time at intervals by using an image acquisition unit and sending the sample image to a central computer;
s2, image recognition and calculation, wherein the central computer screens the pictures containing the seamless steel rail welding seams in the sample images, and marks the welding seam areas of the screened pictures;
and S3, performing data training, namely performing target recognition training on the marked picture by using the initialized deep learning network to obtain a model file, performing multi-layer recursive network training on the marked sample image by using the model file, and outputting a final seamless steel rail weld joint defect recognition result.
Specifically, step S2 specifically includes the following sub-steps: s201, the central computer performs picture screening on a large number of received sample images, and extracts pictures containing seamless steel rail welding seams from the sample images as positive and negative sample pictures; s202, identifying and marking the welding seam area of the picture containing the seamless steel rail welding seam by using marking software, and taking the marked sample picture as a sample training set.
Specifically, the step S3 specifically includes the following sub-steps: s301, improving a yolov3 algorithm, constructing a multistage cascade feature map network Lyolo according to the improved yolov3 algorithm, and initializing the multistage cascade feature map network; s302, importing a sample training set into a multi-level cascade feature map network for multi-level up-sampling cascade, and integrating multi-scale output results to form a target identification model file, namely a multi-level cascade feature pyramid; and S303, importing the image data acquired in real time into a target identification model file for multi-layer recursive network training, and outputting a weld joint identification result in the image data in real time.
In the embodiment of the invention, in terms of the defect identification actual scene of the track inspection, the target objects to be detected are as follows: objects with smaller dimensions such as bolts, track bed foreign bodies, rail cracks, track bed cracks, welding seams and the like also have objects with common dimensions such as fasteners, clamping plates and the like, and more have objects with larger dimensions such as sleepers and the like. How to realize higher precision performance of the detector on multiple scales is a problem of key research, a yolov3 three-level feature pyramid is used for reasoning and predicting a target of each scale to realize a better effect, but the degree of refining the feature map particles by the structure of the three-level features is not enough, and the loss of detail information of the upper-level and lower-level features is caused. Therefore, based on the condition, the invention provides an improved algorithm based on yolov3, namely a multilevel cascaded feature diagram network Lyolo, so as to meet the requirement of detailed identification of different scale targets of the track inspection task; meanwhile, in order to meet the requirement of vehicle-mounted real-time identification under the condition that the electric passenger car runs at a high speed of 120km/h, the network characteristic diagram is added, and the network convolution layer is cut to a certain degree, so that the performance improvement on the identification speed is realized.
Specifically, the network structure of the multilevel cascade feature graph network Lyolo includes a first convolution layer, a second convolution layer, and a residual error network, where an output of the first convolution layer is connected to an input of the second convolution layer, and an output of the second convolution layer is connected to an input of the residual error network.
The network structure of the multilevel cascade characteristic diagram network Lyolo constructed by the invention is shown in figure 2, a first convolution layer is initialized, the first convolution layer comprises 16 filters of 3 x 3, and the step length is 2; performing convolution operation on the image with the input size of 384 by 384 to output a characteristic map with the size of 192 by 16; and then inputting the output 192 × 16 feature map into an initialized second convolution layer, wherein the second convolution layer comprises 32 3 × 3 filters and has the step size of 2, and outputting 96 × 32 feature maps after convolution to finish the initialization of the multistage cascade feature map network. After the initialized dimensionality reduction is completed, the method enters a plurality of residual error networks, each residual error network comprises 1 convolution of 1 × 1 and one convolution of 3 × 3, after the initialized 96 × 32 feature maps pass through the residual error network groups of 1 group, 2 groups, 4 groups and 4 groups in the graph 2, feature maps with the sizes of 96 × 96, 48, 24 × 24, 12, 6 × 6 and 3 × 3 are obtained respectively, and after the feature maps with the sizes of 48 × 48, 24 × 24, 12 × 12, 6 × 6 and 3 × 3 are obtained and cascaded through multi-level upsampling, the feature maps are output as final multi-scale, and a multi-level cascaded feature pyramid is formed to be used for predicting the target.
In the multi-level upsampling cascade process, the relationship between the Lyolo network mapping and the tensor is as shown in fig. 3, an original image with the size of 384 × 384 is input during prediction, the original image is mapped to an output tensor with 5 scales through convolution and residual network processing, and the inside of the tensor shows the probability that various objects exist at each position in the image. The improved multi-stage cascade feature map network predicts (3 × 3 + 6 × 3 + 12 × 3 + 24 × 3 + 48 × 48) = 9207 prediction frames; each prediction is a (4 + 1 + 6) = 11 vector containing target bounding box coordinates (4 values: target abscissa x, target ordinate y, target frame width, target frame height), confidence of target bounding box (1 value), probability vector of target prediction class (6 values, total class, sum of probabilities equals 1).
After the target prediction is completed, the improved Lyolo multistage cascade network model can be used for well identifying multiple defect targets in track inspection at the same time. On the 10000 kinds of track defect target's of establishing data set by oneself down, carry out the test to Lyolo and yolov3 and compare, the contrast result is as shown in fig. 4, can see out that there is the promotion of certain range to fastener, sleeper, splint on the Lyolo algorithm recognition accuracy after the improvement, has great degree promotion to less targets such as foreign matter, crackle, welding seam.
In the invention, as shown in fig. 5, a seamless steel rail weld defect recognition device based on deep learning is also provided, which comprises a central calculation processing unit, a steel rail inspection vehicle and an image acquisition unit arranged on the steel rail inspection vehicle; the central computing processing unit comprises an industrial computer and a power supply, and the industrial computer is connected with the power supply; and the industrial computer is respectively connected with the high-definition industrial digital camera, the lens, the light supplementing light source and the speed encoder. The image acquisition unit mainly comprises a high-definition industrial digital camera, a lens, a light supplementing light source, a speed encoder and a detection box and is used for shooting the rail web image of the steel rail. High definition industry digital camera and light filling light source set up respectively in the detection case, and the detection case is fixed at the rail inspection car vehicle bottom.
The central processing unit comprises an industrial computer (high-performance GPU) and a power supply, and is mainly used for carrying out real-time intelligent identification on the steel rail web picture acquired by the acquisition unit and identifying whether a welding seam in the picture has a defect.
The recognition device of the present invention functions to include: 1) and acquiring images of the rail web part of the steel rail in real time without intervals. 2) And the central computer identifies the picture in real time and calculates whether the picture has weld defects. 3) Compressing the collected pictures into a video file and storing the video file into a local disk; the resulting picture is saved to the local disk.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (6)
1. The seamless steel rail weld defect identification method based on deep learning is characterized by comprising the following steps:
s1, acquiring image data, namely acquiring a sample image of the rail web part of the seamless steel rail in real time at intervals by using an image acquisition unit and sending the sample image to a central computer;
s2, image recognition and calculation, wherein the central computer screens the pictures containing the seamless steel rail welding seams in the sample images, and marks the welding seam areas of the screened pictures;
and S3, performing data training, namely performing target recognition training on the marked picture by using the initialized deep learning network to obtain a model file, performing multi-layer recursive network training on the marked sample image by using the model file, and outputting a final seamless steel rail weld joint defect recognition result.
2. The method for identifying the weld defects of the seamless steel rail based on the deep learning of claim 1, wherein the step S2 specifically comprises the following sub-steps:
s201, the central computer performs picture screening on a large number of received sample images, and extracts pictures containing seamless steel rail welding seams from the sample images as positive and negative sample pictures;
s202, identifying and marking the welding seam area of the picture containing the seamless steel rail welding seam by using marking software, and taking the marked sample picture as a sample training set.
3. The method for identifying the weld defect of the seamless steel rail based on the deep learning of claim 1, wherein the step S3 specifically comprises the following sub-steps:
s301, improving a yolov3 algorithm, constructing a multistage cascade feature map network Lyolo according to the improved yolov3 algorithm, and initializing the multistage cascade feature map network;
s302, importing a sample training set into a multi-level cascade feature map network for multi-level up-sampling cascade, and integrating multi-scale output results to form a target identification model file, namely a multi-level cascade feature pyramid;
and S303, importing the image data acquired in real time into a target identification model file for multi-layer recursive network training, and outputting a weld joint identification result in the image data in real time.
4. The method for identifying the defect of the seamless steel rail welding seam based on the deep learning of claim 3, wherein the network structure of the multilevel cascade feature map network Lyolo comprises a first convolution layer, a second convolution layer and a residual error network, wherein the output of the first convolution layer is connected with the input of the second convolution layer, and the output of the second convolution layer is connected with the input of the residual error network.
5. The method for identifying the weld defect of the seamless steel rail based on the deep learning of claim 3, wherein the initializing the multistage cascade feature map network process in the substep S301 specifically comprises: initializing a first convolution layer, the first convolution layer comprising 16 filters of 3 x 3 with a step size of 2; performing convolution operation on the image with the input size of 384 by 384 to output a characteristic map with the size of 192 by 16; and then inputting the output 192 × 16 feature map into an initialized second convolution layer, wherein the second convolution layer comprises 32 3 × 3 filters and has the step size of 2, and outputting 96 × 32 feature maps after convolution to finish the initialization of the multistage cascade feature map network.
6. The recognition device based on the seamless steel rail weld joint defect recognition method according to any one of claims 1 to 5 is characterized by comprising a central calculation processing unit, a steel rail inspection vehicle and an image acquisition unit arranged on the steel rail inspection vehicle; the central computing processing unit comprises an industrial computer and a power supply, and the industrial computer is connected with the power supply; the image acquisition unit comprises a high-definition industrial digital camera, a lens, a light supplementing light source, a speed encoder and a detection box, wherein the high-definition industrial digital camera and the light supplementing light source are respectively arranged in the detection box, and the detection box is fixed at the bottom of the steel rail inspection vehicle; and the industrial computer is respectively connected with the high-definition industrial digital camera, the lens, the light supplementing light source and the speed encoder.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110362862.9A CN113111875A (en) | 2021-04-02 | 2021-04-02 | Seamless steel rail weld defect identification device and method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110362862.9A CN113111875A (en) | 2021-04-02 | 2021-04-02 | Seamless steel rail weld defect identification device and method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113111875A true CN113111875A (en) | 2021-07-13 |
Family
ID=76713873
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110362862.9A Pending CN113111875A (en) | 2021-04-02 | 2021-04-02 | Seamless steel rail weld defect identification device and method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113111875A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114119461A (en) * | 2021-10-08 | 2022-03-01 | 厦门微亚智能科技有限公司 | Lithium battery module side weld appearance detection algorithm and system based on deep learning |
| CN116551263A (en) * | 2023-07-11 | 2023-08-08 | 苏州松德激光科技有限公司 | Visual control method and system for welding position selection |
| CN117092116A (en) * | 2023-10-20 | 2023-11-21 | 上海嘉朗实业南通智能科技有限公司 | Automobile aluminum alloy casting defect detection system and method based on machine vision |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101699273A (en) * | 2009-10-29 | 2010-04-28 | 北京交通大学 | Auxiliary detection device and method of image processing for on-line flaw detection of rails |
| CN110660052A (en) * | 2019-09-23 | 2020-01-07 | 武汉科技大学 | Hot-rolled strip steel surface defect detection method based on deep learning |
| CN110929795A (en) * | 2019-11-28 | 2020-03-27 | 桂林电子科技大学 | Method for quickly identifying and positioning welding spot of high-speed wire welding machine |
| CN111353413A (en) * | 2020-02-25 | 2020-06-30 | 武汉大学 | Low-missing-report-rate defect identification method for power transmission equipment |
| CN111402211A (en) * | 2020-03-04 | 2020-07-10 | 广西大学 | High-speed train bottom foreign matter identification method based on deep learning |
| CN111815605A (en) * | 2020-07-09 | 2020-10-23 | 成都协创信和科技有限公司 | Sleeper defect detection method based on step-by-step deep learning and storage medium |
| CN111899227A (en) * | 2020-07-06 | 2020-11-06 | 北京交通大学 | Automatic railway fastener defect acquisition and identification method based on unmanned aerial vehicle operation |
-
2021
- 2021-04-02 CN CN202110362862.9A patent/CN113111875A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101699273A (en) * | 2009-10-29 | 2010-04-28 | 北京交通大学 | Auxiliary detection device and method of image processing for on-line flaw detection of rails |
| CN110660052A (en) * | 2019-09-23 | 2020-01-07 | 武汉科技大学 | Hot-rolled strip steel surface defect detection method based on deep learning |
| CN110929795A (en) * | 2019-11-28 | 2020-03-27 | 桂林电子科技大学 | Method for quickly identifying and positioning welding spot of high-speed wire welding machine |
| CN111353413A (en) * | 2020-02-25 | 2020-06-30 | 武汉大学 | Low-missing-report-rate defect identification method for power transmission equipment |
| CN111402211A (en) * | 2020-03-04 | 2020-07-10 | 广西大学 | High-speed train bottom foreign matter identification method based on deep learning |
| CN111899227A (en) * | 2020-07-06 | 2020-11-06 | 北京交通大学 | Automatic railway fastener defect acquisition and identification method based on unmanned aerial vehicle operation |
| CN111815605A (en) * | 2020-07-09 | 2020-10-23 | 成都协创信和科技有限公司 | Sleeper defect detection method based on step-by-step deep learning and storage medium |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114119461A (en) * | 2021-10-08 | 2022-03-01 | 厦门微亚智能科技有限公司 | Lithium battery module side weld appearance detection algorithm and system based on deep learning |
| CN114119461B (en) * | 2021-10-08 | 2022-11-29 | 厦门微亚智能科技有限公司 | Deep learning-based lithium battery module side weld appearance detection method and system |
| CN116551263A (en) * | 2023-07-11 | 2023-08-08 | 苏州松德激光科技有限公司 | Visual control method and system for welding position selection |
| CN116551263B (en) * | 2023-07-11 | 2023-10-31 | 苏州松德激光科技有限公司 | Visual control method and system for welding position selection |
| CN117092116A (en) * | 2023-10-20 | 2023-11-21 | 上海嘉朗实业南通智能科技有限公司 | Automobile aluminum alloy casting defect detection system and method based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113674247B (en) | X-ray weld defect detection method based on convolutional neural network | |
| Doshi et al. | Road damage detection using deep ensemble learning | |
| CN112967243B (en) | Deep learning chip packaging crack defect detection method based on YOLO | |
| CN113111875A (en) | Seamless steel rail weld defect identification device and method based on deep learning | |
| KR102166458B1 (en) | Defect inspection method and apparatus using image segmentation based on artificial neural network | |
| CN109859163A (en) | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks | |
| US11715190B2 (en) | Inspection system, image discrimination system, discrimination system, discriminator generation system, and learning data generation device | |
| Wan et al. | Ceramic tile surface defect detection based on deep learning | |
| JP2017049974A (en) | Discriminator generator, quality determine method, and program | |
| CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
| CN112488025B (en) | Double-temporal remote sensing image semantic change detection method based on multi-modal feature fusion | |
| CN114445366A (en) | Intelligent long-distance pipeline radiographic image defect identification method based on self-attention network | |
| Li et al. | An end-to-end defect detection method for mobile phone light guide plate via multitask learning | |
| CN112766110A (en) | Training method of object defect recognition model, object defect recognition method and device | |
| CN115439458A (en) | Industrial image defect target detection algorithm based on depth map attention | |
| CN105931246A (en) | Fabric flaw detection method based on wavelet transformation and genetic algorithm | |
| CN113436157A (en) | Vehicle-mounted image identification method for pantograph fault | |
| CN114723709A (en) | Tunnel disease detection method and device and electronic equipment | |
| CN115830004A (en) | Surface defect detection method, device, computer equipment and storage medium | |
| CN115546223A (en) | Method and system for detecting loss of fastening bolt of equipment under train | |
| CN115294541A (en) | Local feature enhanced Transformer road crack detection method | |
| Mohan et al. | Yolo v2 with bifold skip: a deep learning model for video based real time train bogie part identification and defect detection | |
| Feng et al. | Improved SOLOv2 detection method for shield tunnel lining water leakages | |
| Luo et al. | Waterdrop removal from hot-rolled steel strip surfaces based on progressive recurrent generative adversarial networks | |
| CN115909157A (en) | Machine vision-based identification detection method, device, equipment and medium |
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 |