CN115345846A - Intelligent grading method and system for grain size of medium and low carbon steel - Google Patents
Intelligent grading method and system for grain size of medium and low carbon steel Download PDFInfo
- Publication number
- CN115345846A CN115345846A CN202210967339.3A CN202210967339A CN115345846A CN 115345846 A CN115345846 A CN 115345846A CN 202210967339 A CN202210967339 A CN 202210967339A CN 115345846 A CN115345846 A CN 115345846A
- Authority
- CN
- China
- Prior art keywords
- grain size
- grain
- grain boundary
- picture
- carbon steel
- 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
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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- 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
- G06T2207/30136—Metal
Abstract
The invention discloses a method and a system for intelligently evaluating the grain size of medium-low carbon steel, wherein the method is carried out according to the following steps: step 1, collecting medium and low carbon steel metallographic images, cleaning the collected data, marking identifiable crystal boundaries, and establishing a training image and label database. And 2, extracting the primary grain boundary by using a deep learning model. And 3, performing primary intercept point statistics by using an intercept point method based on the extracted primary grain boundary. And 4, counting the missing crystal boundary intercept points through a mathematical statistical model, and calculating the grain size of the image by combining the result of the step 3 as the final intercept point. The invention can effectively solve the problems that the manual detection of the grain size of the medium-low carbon steel is low in efficiency and high in labor intensity, and the existing image analysis algorithm cannot solve the problems of fuzzy grain boundary extraction, missing grain boundary extraction and the like. The invention can effectively improve the detection efficiency and reduce the detection error, and has important pioneering significance in practical application.
Description
Technical Field
The invention relates to the technical field of metal average grain size calculation, in particular to a method and a system for intelligently determining the grain size of medium and low carbon steel based on deep learning and mathematical statistics, which are suitable for grading the grain size of medium and low carbon hot rolling, annealing and normalizing steel products.
Background
The microstructure is called the genetic code of the metallic material. Metallographic analysis is a science of studying the microstructure of metallic materials and determining the relationship between such microstructure and macroscopic properties. The characteristics, the number, the appearance, the size, the distribution, the orientation, the spatial arrangement state and other microstructure structural characteristics of the microstructure have corresponding functional relations with the macroscopic mechanical and physical properties of the material. Grain size is a parameter describing the grain size of polycrystalline materials and is one of the most important indicators in the microstructure of materials. The grain size has important influence on the performance of the metal material, and the smaller the general grain is, the higher the strength, plasticity and toughness of the metal material is, and the better the comprehensive mechanical property of the material is. Therefore, the grain size is a technical index which needs to be strictly controlled in the metal material processing, particularly the hot working process, and is also a necessary item for the analysis of the gold phase in each metal material standard.
According to the national standard GB6394-2017 "Metal average grain size rating method", common rating methods for the average grain size of metal materials include a comparison method, an intercept method and an area method. The comparison method is to give the closest grade by comparing with a standard series rating chart; the intercept method is to determine the grain size grade by calculating the intercept number of the intersection part of an experimental line segment with known length and a grain interface and using the intercept number of unit length; the area rule is to count the number of crystal grains with a given area, and then calculate the average area of the crystal grains to determine the grain size level.
In the existing rating method, the accuracy of an intercept method and an area method is high. At present, the two methods are mainly through manual measurement, calculation and grading. The traditional rating method is long in time consumption and strong in subjectivity, has the problems of poor measurement precision, non-uniform statistical results and low efficiency, and seriously influences the research and development and application of metal materials in production and life.
With the development of technologies such as machine learning and image processing, intelligent detection gradually leaves the head and corners in various fields, and the analysis of metal microstructures by using an image processing technology is widely applied in industry, so that the automation degree is high, and accurate measurement results can be obtained in a short time. However, the algorithm applied to average grain size rating at present has the problems of low detection precision, incapability of solving the problem of missing grain boundaries and the like, and further research is still needed to improve the precision of intelligent detection.
Disclosure of Invention
The invention aims to solve the problems that the grain size solving in the prior art depends on manual operation, the working strength is high, the detection efficiency is low, the grain boundary identification and statistical errors are large, the result consistency is low and the like, and particularly, the grain size calculation under the condition of fuzzy grain boundary and missing grain boundary interference exists in part of metallographic pictures.
In order to realize the purpose, the technical scheme adopted by the invention is as follows:
on one hand, the invention provides a medium and low carbon steel grain size intelligent rating method based on deep learning and mathematical statistics, which comprises the following steps:
step 1: collecting a metallographic picture, carrying out data set annotation on the picture, wherein the data set annotation comprises a clear crystal boundary, a fuzzy crystal boundary and a discontinuous (partial deletion) crystal boundary, and establishing an image-label-magnification-crystal grain size level database;
step 2: performing grain boundary extraction by using an improved neural network based on CNN, dividing an input picture into sub-pictures with the size of 256 multiplied by 256 by using equal step length, obtaining a grain boundary extraction model through training, and obtaining a complete grain boundary extraction picture with the original input size through testing;
and step 3: carrying out preliminary intercept point statistics in the graph by using a four-circle intercept point method;
and 4, step 4: carrying out numerical value speculation on the missing crystal boundary by a mathematical statistical method, and predicting the intercept point number of the large carbon block area based on confidence coefficient;
and 5: using the formula
N＝N 1 +N 2
Solving the grain size corresponding to each circle; wherein G is the number of crystal grades, M is the magnification, L is the perimeter of the circular grid, N is the number of nodes on the circle, N1 is the number of primary statistical cut points, and N2 is the number of mathematical statistical guessed hidden cut points.
And 6: and averaging the grain sizes corresponding to the four circles to obtain the final grain size.
Furthermore, in the step 1, labeling of the data set is completed by using Labelme, and fuzzy and missing crystal boundaries are manually labeled except for visible crystal boundaries.
Further, in the database of the step 1, the image is an original picture of a metallographic picture, the label is a picture obtained after manually marking a grain boundary (including a clear grain boundary, a fuzzy grain boundary and a discontinuous missing grain boundary), and the magnification factor is the microscope magnification factor when the original picture is shot.
Further, in the step 2, during training, a pyrrch frame is used for realizing a neural network structure, and GPU resources are used for network training to obtain a grain boundary extraction model; during testing, performing grain boundary extraction on the partitioned pictures, fusing the partitioned grain boundary extraction pictures based on a Boolean addition method, and finally obtaining a complete grain boundary extraction picture with the original input size;
further, in the step 2, the base network uses the CNN as a framework, and the CNN network is optimized by modifying the number of network intermediate layers and the number of neurons in each layer.
Further, in the step S2, the input picture is subjected to equal-length and width segmentation based on step length in the grain boundary extraction, and the results are merged after the extraction is performed respectively.
Furthermore, the centers of the four concentric circles are located in the center of the input picture, and the number of the visible intercept points of each circle is preliminarily counted by using an intercept point counting module.
Further, four circle radii of 84.85mm,60.6mm,40.37mm and 20.12mm are used.
On the other hand, the invention provides a medium-low carbon steel grain size intelligent determination system based on deep learning and mathematical statistics, which comprises a metallographic image acquisition module, a grain boundary extraction network module, an intercept point presumption module based on mathematical statistics and an automatic rating module, wherein the system acquires a large amount of metallographic image data for marking and training, gradually optimizes the network, and simultaneously optimizes the intercept point statistics of the missing grain boundary by combining the mathematical statistics to obtain the grain size.
Further, the metallographic picture collecting module is used for collecting a metallographic picture, labeling a data set of the picture and establishing an image-label-magnification-grain size level database;
the grain boundary extraction network module adopts a pyrrch frame to build and train a model, stores a trained model file, and tests to finally obtain a complete grain boundary extraction diagram after the model is trained;
the intercept point presumption module comprises a visible intercept point number counting module and a hidden intercept point number counting module; the visible intercept point number counting module extracts an edge grain boundary binary image output by the network through an input grain boundary, and carries out intercept point preliminary counting based on a four-circle intercept point method;
and the automatic rating module accumulates the obtained visible cut points and the hidden cut points to obtain final cut points, and then calculates the grain size by using a formula.
The invention has the following beneficial effects:
the invention can effectively solve the problems that the manual detection efficiency of the medium and low carbon steel grain size is low, the labor intensity is high, the existing image analysis algorithm cannot solve the problems of fuzzy grain boundary extraction, missing grain boundary extraction and the like, the detection efficiency is improved, the detection error is reduced, and the method has important pioneering significance in practical application.
The model training and prediction of the invention are carried out in an end-to-end mode, the extraction of the crystal boundary utilizes the blocking idea to greatly reduce the training complexity of the model and improve the running speed of the model. Meanwhile, the whole using process is automatically carried out by software, and only the path of the picture needing to be rated is required to be input, so that the automatic processing can be carried out, and finally an evaluation report is obtained. Under the condition of high-quality data set training, the model can run stably, errors of manual rating are greatly reduced, and labor cost and time cost are saved.
Drawings
FIG. 1 is a diagram of a model architecture of the present invention;
FIG. 2 is a grain size diagram for an example of the present invention;
FIG. 3 is a cut-off statistical graph obtained in accordance with an embodiment of the present invention;
Detailed Description
The invention is described in detail below with reference to the figures and the specific embodiments.
Example one
The embodiment provides a medium and low carbon steel grain size intelligent rating method based on deep learning and mathematical statistics, which comprises the following steps of:
step 1: collecting a metallographic picture, carrying out data set labeling on the picture, wherein the data set labeling comprises a clear crystal boundary, a fuzzy crystal boundary and a discontinuous (partial missing) crystal boundary, carrying out the labeling of the crystal boundary by using Labelme, and establishing an image-label-magnification-grain size level database, wherein the image is an original picture of the metallographic picture, the label is a picture after the crystal boundary is manually marked, and the magnification is the microscope magnification when the original picture is shot;
and 2, step: performing grain boundary extraction by using a CNN-based improved neural network, dividing an input picture into sub-pictures with the size of 256 multiplied by 256 by using equal step length, during training, realizing a neural network structure by using a pyrrch frame, and performing network training by using GPU resources to obtain a grain boundary extraction model; during testing, performing grain boundary extraction on the partitioned pictures, fusing the partitioned grain boundary extraction pictures based on a Boolean addition method, and finally obtaining a complete grain boundary extraction picture with the original input size;
and 3, step 3: carrying out preliminary intercept point statistics on the image by using a four-circle intercept point method, wherein the centers of four concentric circles are positioned in the center of the input image, and carrying out preliminary statistics on the number of visible intercept points of each circle by using an intercept point statistics module;
and 4, step 4: and carrying out numerical value speculation on the missing crystal boundary by a mathematical statistics method, and predicting the intercept point number of the large carbon block area based on the confidence coefficient.
And 5: using the formula
N＝N 1 +N 2
Solving the grain size corresponding to each circle; wherein G is the number of crystal grades, M is the magnification, L is the perimeter of the circular grid, N is the number of nodes on the circle, N1 is the number of preliminary statistical cut points, and N2 is the number of presumed hidden cut points for mathematical statistics.
Step S6: and averaging the grain sizes corresponding to the four circles to obtain the final grain size.
In the step 1, the data set marking is finished by using Labelme, and besides visible crystal boundaries, missing crystal boundaries need to be marked manually.
In the step 2, the basic network takes the CNN as a framework, and the CNN network is optimized by modifying the number of network intermediate layers and the number of neurons in each layer.
In the step 2, the input picture is subjected to step length-based equal-length and width segmentation in the crystal boundary extraction, and results are merged after extraction.
In the step 3, the four circle radii used are 84.85mm,60.6mm,40.37mm and 20.12mm respectively.
In step 4, the hidden cutoff number is estimated by using a confidence-based method.
Example two
The embodiment provides a medium and low carbon steel grain size intelligent determination system based on deep learning and mathematical statistics, which comprises a metallographic image acquisition module, a grain boundary extraction network module, an intercept point presumption module based on mathematical statistics and an automatic rating module. The system gradually optimizes the network by marking and training a large amount of data, simultaneously optimizes the statistics of the number of the intercept points of the missing crystal boundary by combining mathematical statistics, has high automation of the whole process, reduces the manual participation and improves the efficiency of metallographic analysis. Wherein:
the metallographic picture acquisition module is used for acquiring a metallographic picture, labeling a data set of the picture, establishing an image-label-magnification-grain size level database, wherein the data set comprises a clear grain boundary, a fuzzy grain boundary and a discontinuous (partial deletion) grain boundary;
in the grain boundary extraction network module, training a model is the most critical one-step operation, and the specific training steps are as follows:
step 1: and calibrating the original data by using Labelme software, wherein the picture size is 2048 multiplied by 1536, marking visible crystal boundaries and filling the extendable crystal boundaries. At the same time, naming the pictures by id _ magnification _ level and png;
and 2, step: segmenting the marked picture into pictures with the size of 256 multiplied by 256 by taking 128 as step length, and simultaneously segmenting the label in the same way;
and step 3: inputting the data set into a neural network for training, adopting a pyrrch frame for model building and training, and storing the trained model files.
After the model is trained, testing can be started, metallographic pictures with the size of 2048 × 1536 are input, the picture is firstly segmented by a program, the pictures are segmented into 256 × 256 slices by taking 128 as a step length, a grain boundary extraction model is operated to extract grain boundaries, and finally a complete grain boundary extraction image is obtained.
The intercept point presumption module comprises a visible intercept point number counting module and a hidden intercept point number counting module; wherein:
and the visible intercept number counting module inputs an edge grain boundary binary image output by the grain boundary extraction network, and performs initial counting of the intercept based on a four-circle intercept method, wherein the used four circle radiuses are 84.85mm,60.6mm,40.37mm and 20.12mm respectively.
And the hidden intercept point number counting module is used for inputting the edge grain boundary binary image output by the grain boundary extraction network and carrying out hidden intercept point speculation based on a mathematical statistics method.
And the automatic rating module accumulates the obtained visible cut points and the hidden cut points to obtain final cut points, and calculates the grain size by using a formula. The grain size plots used in the examples were ranked with an artificial standard rating of 12.41, a test rating of 12.329475 using the method and an error of 0.080524.
The foregoing shows and describes the general principles, principal features and advantages of the invention. It should be understood by those skilled in the art that the above embodiments do not limit the scope of the present invention in any way, and all technical solutions obtained by using equivalent substitution methods fall within the scope of the present invention.
The parts not involved in the present invention are the same as or can be implemented using the prior art.
Claims (10)
1. An intelligent grading method for the grain size of medium and low carbon steel is characterized by comprising the following steps:
step 1: collecting a metallographic picture, carrying out data set labeling on the picture, wherein the data set labeling comprises a clear crystal boundary, a fuzzy crystal boundary and a discontinuous crystal boundary, and establishing an image-label-magnification-grain size level database;
step 2: performing grain boundary extraction by using an improved neural network based on CNN, dividing an input picture into sub-pictures with the size of 256 multiplied by 256 by using equal step length, obtaining a grain boundary extraction model through training, and obtaining a complete grain boundary extraction picture with the original input size through testing;
and step 3: carrying out preliminary intercept point statistics in the graph by using a four-circle intercept point method;
and 4, step 4: carrying out numerical value speculation on the missing crystal boundary by a mathematical statistical method, and predicting the intercept point number of the large carbon block area based on confidence coefficient;
and 5: using the formula
N＝N 1 +N 2
Solving the grain size corresponding to each circle; wherein G is the number of crystal grades, M is the magnification, L is the perimeter of the circular grid, N is the number of nodes on the circle, N1 is the number of primary statistical cut points, and N2 is the number of mathematical statistical guessed hidden cut points;
and 6: and averaging the grain sizes corresponding to the four circles to obtain the final grain size.
2. The intelligent grading method for the grain size of the medium-low carbon steel according to claim 1, wherein in the step 1, the data set marking is completed by Labelme, and fuzzy grain boundaries and missing grain boundaries are manually marked except for visible and clear grain boundaries.
3. The intelligent grading method for the grain size of medium and low carbon steel according to claim 1, wherein in the database of step 1, the image is an original picture of a metallographic picture, the label is a picture marked by a grain boundary manually, and the magnification is the microscope magnification when the original picture is shot.
4. The intelligent grading method for the grain size of medium and low carbon steel according to claim 1, wherein in the step 2, during training, a pitorch frame is used for realizing a neural network structure, and GPU resources are used for network training to obtain a grain boundary extraction model; during testing, grain boundary extraction is carried out on the partitioned pictures, and the partitioned grain boundary extraction pictures are fused based on Boolean addition, so that a complete grain boundary extraction picture with the original input size is finally obtained.
5. The intelligent grading method for medium and low carbon steel grain size according to claim 1 or 4, characterized in that in the step 2, the base network uses CNN as a skeleton, and the CNN network is optimized by modifying the number of network intermediate layers and the number of neurons in each layer.
6. The intelligent grading method for the grain size of medium and low carbon steel according to claim 1 or 4, wherein in the step S2, the input picture is subjected to equal-length and width segmentation based on step length in grain boundary extraction, and results are merged after extraction.
7. The intelligent grading method for the grain size of medium and low carbon steel according to claim 1, wherein in the step 3, the centers of four concentric circles are located at the center of the input picture, and the number of the visible cut points of each circle is primarily counted by using a cut point counting module.
8. The intelligent grading method for grain size of medium and low carbon steel according to claim 1 or 7, wherein the four circle radii used are 84.85mm,60.6mm,40.37mm and 20.12mm.
9. The utility model provides a medium and low carbon steel grain size intelligent determination system based on degree of depth study and mathematical statistics, its characterized in that, the system includes metallographic image acquisition module, and the network module is drawed to the grain boundary to and intercept conjecture module and the automatic rating module based on mathematical statistics, the system marks and trains through gathering a large amount of metallographic image data, progressively optimizes the network, combines the statistics of the intercept number of the statistical optimization disappearance grain boundary of mathematical statistics simultaneously, reachs the grain size.
10. The system for intelligently measuring the grain size of the medium-low carbon steel according to claim 9, wherein the metallographic picture collecting module is used for collecting the metallographic picture, labeling a data set of the picture, and establishing a database of image-label-magnification-grain size grade;
the grain boundary extraction network module adopts a pyrrch frame to build and train a model, stores a trained model file, and tests to finally obtain a complete grain boundary extraction diagram after the model is trained;
the intercept point presumption module comprises a visible intercept point number counting module and a hidden intercept point number counting module; the visible intercept number counting module extracts an edge grain boundary binary image output by the network through an input grain boundary, and performs intercept preliminary counting based on a four-circle intercept method;
and the automatic rating module accumulates the obtained visible cut points and the hidden cut points to obtain final cut points, and then calculates the grain size by using a formula.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210967339.3A CN115345846A (en) | 2022-08-12 | 2022-08-12 | Intelligent grading method and system for grain size of medium and low carbon steel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210967339.3A CN115345846A (en) | 2022-08-12 | 2022-08-12 | Intelligent grading method and system for grain size of medium and low carbon steel |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115345846A true CN115345846A (en) | 2022-11-15 |
Family
ID=83952824
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210967339.3A Pending CN115345846A (en) | 2022-08-12 | 2022-08-12 | Intelligent grading method and system for grain size of medium and low carbon steel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115345846A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252804A (en) * | 2023-07-06 | 2023-12-19 | 南京航空航天大学 | Automatic analysis method for band-shaped carbide in bearing steel |
-
2022
- 2022-08-12 CN CN202210967339.3A patent/CN115345846A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252804A (en) * | 2023-07-06 | 2023-12-19 | 南京航空航天大学 | Automatic analysis method for band-shaped carbide in bearing steel |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110660052B (en) | Hot-rolled strip steel surface defect detection method based on deep learning | |
| CN111507990B (en) | Tunnel surface defect segmentation method based on deep learning | |
| CN109190442B (en) | Rapid face detection method based on deep cascade convolution neural network | |
| CN108257114A (en) | A kind of transmission facility defect inspection method based on deep learning | |
| CN108711148B (en) | Tire defect intelligent detection method based on deep learning | |
| CN111104981A (en) | Hydrological prediction precision evaluation method and system based on machine learning | |
| CN111860596A (en) | Unsupervised pavement crack classification method based on deep learning and model establishment method | |
| CN113516228B (en) | Network anomaly detection method based on deep neural network | |
| CN112991271A (en) | Aluminum profile surface defect visual detection method based on improved yolov3 | |
| CN115345846A (en) | Intelligent grading method and system for grain size of medium and low carbon steel | |
| CN113506269A (en) | Turnout and non-turnout rail fastener positioning method based on deep learning | |
| CN116953006A (en) | Casting material scanning electron microscope image defect identification and quantification method | |
| CN116883393B (en) | Metal surface defect detection method based on anchor frame-free target detection algorithm | |
| CN114565314A (en) | Hot rolled steel coil end face quality control system and method based on digital twinning | |
| CN111582270A (en) | Identification tracking method based on high-precision bridge region visual target feature points | |
| CN111091534A (en) | Target detection-based pcb defect detection and positioning method | |
| CN116052110B (en) | Intelligent positioning method and system for pavement marking defects | |
| CN115100173B (en) | Road pavement image crack geometric property detection and crack identification method based on deep convolutional neural network | |
| CN112084941A (en) | Target detection and identification method based on remote sensing image | |
| CN116342542A (en) | Lightweight neural network-based steel product surface defect detection method | |
| CN116434054A (en) | Intensive remote sensing ground object extraction method based on line-plane combination | |
| CN116934687A (en) | Injection molding product surface defect detection method based on semi-supervised aided learning semantic segmentation | |
| CN113641733B (en) | Real-time intelligent estimation method for river cross section flow | |
| CN114863211A (en) | Magnetic shoe defect detection and segmentation method based on deep learning | |
| CN114596296A (en) | High-sensitivity hot-rolled steel coil end surface defect identification system and method |
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 |