CN115908275A - Hot ring rolling deformation geometric state online measurement method based on deep learning - Google Patents
Hot ring rolling deformation geometric state online measurement method based on deep learning Download PDFInfo
- Publication number
- CN115908275A CN115908275A CN202211326943.4A CN202211326943A CN115908275A CN 115908275 A CN115908275 A CN 115908275A CN 202211326943 A CN202211326943 A CN 202211326943A CN 115908275 A CN115908275 A CN 115908275A
- Authority
- CN
- China
- Prior art keywords
- image
- rolling
- ring
- deep learning
- hot
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a hot ring rolling deformation geometric state online measurement method based on deep learning. The method solves the problem that the geometric state characteristics of the ring cannot be rapidly and accurately extracted due to the fact that a roller shields the target ring in the rolling process of the hot ring and the situations of splashing oxide skin, aerial fog and the like, and accelerates the target matching operation efficiency and improves the overall robustness of the algorithm while ensuring the matching accuracy.
Description
Technical Field
The invention belongs to the technical field of machine vision, and particularly relates to a hot ring rolling deformation geometric state online measurement method based on deep learning.
Background
The ring piece is used as a core basic part in the fields of automobiles, ships, metallurgy, chemical industry, aerospace and the like, and the quality and performance of the ring piece are focused. Ring rolling is an advanced forming manufacturing technology of high-performance seamless rings. The large ring is usually rolled in a hot state, and the ring is in a high-temperature state in the initial stage of reheating rolling, so that the existing contact measurement technology is difficult to be applied to the high-temperature occasion.
Because the large ring is in a 1250-degree high-temperature state in the initial stage of hot rolling, the visible light radiated by the hot ring body can interfere with optical measurement, and the measurement error is increased. In different rolling stages, the roller can shield the hot ring to a certain extent, and simultaneously the outer layer of the ring peels off and splashes, aerial fog and the like, so that the non-contact measurement represented by visual measurement is difficult to ensure the measurement precision and efficiency in the real-time monitoring process.
The machine vision has the advantages of non-contact, high efficiency, comprehensive information acquisition and the like, and has remarkable advantages in measuring the geometric state of the hot ring. However, the visible light radiated by the hot ring body can interfere with the optical measurement, and the measurement error is increased. In addition, the deformation state of the ring is severely changed in time during the rolling process, so that the requirement on the response of a measuring system is high. The existing vision measurement method generally has the problems of low efficiency, poor robustness, low precision and the like, and cannot realize the vision measurement of an object to be measured more accurately.
Therefore, a need exists for a vision measuring method and apparatus to solve the above problems.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method is used for measuring the deformation geometric state parameters of the ring in the hot rolling process under the interference on line.
The technical scheme adopted by the invention for solving the technical problems is as follows: the hot ring rolling deformation geometric state online measurement method based on deep learning comprises the following steps:
s1: acquiring a rolling image of the hot-state ring in the hot rolling experiment through an image acquisition device, carrying out target detection on the rolling image of the hot-state ring through a deep learning algorithm, and determining a corresponding template image and a corresponding characteristic image;
s2: matching the template image and the rolling image by adopting an image edge matching algorithm;
s3: and fitting the boundary contour of the hot ring by using a weighting function, and calculating the outer diameter size of the hot ring.
According to the scheme, in the step S1, the specific steps are as follows:
s11: acquiring a real-time rolling image of the hot ring in the hot rolling experiment through an image acquisition device;
s12: taking the rolled image as an image sample, randomly dividing the image sample into a training set, a testing set and a verification set according to a preset proportion, training by adopting the training set through a deep learning algorithm, updating weight parameters of an iterative convolution neural network, and carrying out target rough identification on the rolled image acquired by an image acquisition device to obtain a template image corresponding to the rolled image; a target area is selected on the template image and used for capturing a target thermal ring piece;
s13: carrying out slicing operation on the image sample with the original size to obtain a first characteristic diagram; and performing convolution kernel convolution on the first characteristic image to obtain a characteristic image.
Further, the step S1 further includes the following steps:
and calibrating the actual size of the hot ring piece by adopting a correction method and a calibration plate, and correcting the distortion of the lens.
Further, in the step S2, the specific steps are:
s21: inputting a template image, carrying out affine transformation of the template image at different angles or different scales including translation, rotation and scaling, and calculating feature points and feature vectors of the template image;
s22: inputting a rolling image, calculating the position of a pixel point, and constructing a gradient corresponding graph;
s23: and calculating the similarity of the rolled image and the template image by adopting image edge matching to complete template matching.
Further, in the step S3, the specific steps are as follows:
s31: enhancing the image obtained in the step S2, performing linear transformation on the pixels, highlighting a bright color area, inhibiting a dark color area, improving the edge contrast, and capturing the edge of the high-precision hot ring piece;
s32: a caliper rectangular frame is equidistantly generated on the circle contour fitted by the algorithm by adopting a caliper tool, edge point detection is carried out, and the actual edge point with larger gradient amplitude in the caliper is screened out;
s33: and introducing a weight function to carry out iterative weighted fitting on the actual edge points, fitting and inhibiting outliers by using a weighted least square method, and calculating the geometric parameters of the thermal ring.
Further, in step S22, the specific steps of constructing the gradient map include:
dispersing the gradient into a plurality of directions with equal angular intervals, representing the quantized gradient direction by binary bits, and simplifying the information content represented by the gradient;
establishing a pre-response table through the quantization gradient, and converting a calculation process of comparing the template image with the search images at all positions into a lookup table;
and searching corresponding data in the linear storage table according to the characteristic points of the template image, and adding the corresponding data of the characteristic points to obtain template matching similarity measurement.
Further, in step S31, the specific steps of enhancing the image are:
and GMin is a gray value, g is a current gray value, mult is a multiplied coefficient, and Add is an additional offset value, and image enhancement is performed according to the following formulas (1) and (2):
Add＝-Mult*GMin (1)，
g'＝g*Mult+Add (2)。
a hot ring rolling deformation geometric state online measurement system based on deep learning comprises an image acquisition device, an industrial personal computer and a light source; the image acquisition device adopts an industrial camera and is used for acquiring the ring piece image; the industrial personal computer is used for processing the collected images and calculating the geometric state parameters of the thermal ring piece; the light source is used for providing illumination or supplementary lighting.
A computer storage medium having stored therein a computer program executable by a computer processor, the computer program performing a deep learning based hot ring rolling deformation geometry online measurement method.
The invention has the beneficial effects that:
1. the hot ring rolling deformation geometric state online measurement method based on deep learning comprises the steps of collecting hot ring images, training by using collected image samples to obtain template images, matching the test images with input images by using a template matching algorithm, enhancing real-time images, capturing the edges of the ring with higher precision, detecting actual edge points of the ring, and realizing the function of online measurement of the hot ring rolling deformation geometric state under the interference of non-uniform temperature chromatic aberration, flying chips, roller shielding and the like through fitting.
2. The method adopts a mode of combining rough target identification and fine target matching, utilizes a deep learning algorithm and image edge template matching to obtain a target high-precision fitting profile, and utilizes an introduced weighting function to correct the fitting effect, so as to obtain a real-time geometric state in the rolling process of the hot ring piece; by utilizing a deep learning algorithm, rolling images acquired by a camera are trained through a data set to intelligently capture the target ring for measuring the geometric state of the ring, so that the problem that the target images cannot be accurately acquired is solved.
3. The method solves the problem that the geometric state characteristics of the ring cannot be rapidly and accurately extracted due to the fact that a roller shields the target ring in the rolling process of the hot ring and the situations of splashing oxide skin, aerial fog and the like, and accelerates the target matching operation efficiency and improves the overall robustness of the algorithm while ensuring the matching accuracy.
4. The method solves the problems of time consumption and potential safety hazard of manual contact type measurement, solves the problem of low target extraction accuracy caused by the falling and splashing of the outer layer skin of the ring in the rolling process, improves the speed of the image matching process based on edge matching optimization, improves the fitting accuracy based on iterative weighted fitting, and ensures the online measurement effect and accuracy requirement in the hot ring rolling process.
5. The method comprises the steps of carrying out target detection on a hot ring image through a deep learning algorithm, carrying out template matching on a ring template image and a rolled ring image through an image edge matching algorithm, fitting a ring boundary outline by referring to a weight function, and calculating the outer diameter size of the ring. The calculation process of comparing the template model with the search images at all positions is converted into the process of searching the table, so that the searching speed of the matching process and the robustness to tiny positions and rotation are improved.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention.
Fig. 2 is a system layout diagram of an embodiment of the present invention.
Fig. 3 is a diagram illustrating measurement results according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1, an embodiment of the present invention includes an image acquisition device and an industrial personal computer; the image acquisition device adopts an industrial camera and is used for acquiring a ring piece image; the industrial personal computer is used for processing the collected images and calculating the geometric state parameters of the thermal ring piece; the light source is used for providing illumination or supplementary lighting.
The invention discloses a hot ring rolling deformation geometric state online measurement method based on deep learning, which comprises the following steps of:
(1) Acquiring a real-time image of the thermal ring to be detected, and determining a template image corresponding to the image, wherein an interested area is framed and selected on the template image.
(2) And calibrating the actual size of the hot ring by using the calibration plate, and correcting the distortion of a lens to ensure that the accurate size of the rolled ring is obtained. Wherein the calibration method adopts Zhangzhengyou calibration method.
(3) A hot ring image sample is collected from a ring hot rolling experiment, and the hot ring image is divided into a training set, a testing set and a verification set according to a certain proportion for training to obtain a template image. The template image is trained by a convolutional neural network, and has characteristic points and characteristic vectors which are easy to identify.
(4) And further, inputting the template information and the test image into a system, and calculating the similarity of the feature points to complete template matching. Before the template information is input, scaling training of different angles and scales is carried out.
(5) Further, the image is enhanced, pixels are linearly transformed, edge contrast is improved, and the edge of the ring piece with high precision is captured. And detecting actual edge points with larger gradient amplitude in the caliper by using a caliper tool, introducing a weight function to weight the data, and performing circle fitting.
(6) And transmitting the hot ring rolling image acquired by the industrial camera in real time to an industrial personal computer, synchronously processing the image by the industrial personal computer, and calculating the fitting circle to obtain the size of the hot ring.
The industrial personal computer controls the industrial camera to collect images of the thermal ring, trains collected image samples to obtain template images, matches test images with input images by using a template matching algorithm, enhances real-time images obtained by the industrial camera, captures the edge of the ring with higher precision, detects actual edge points of the ring, and realizes the online measurement of the size of the thermal ring through fitting, and the method specifically comprises the following operations:
and taking a part of hot ring images obtained from a ring hot rolling experiment as a sample, and immediately dividing the hot ring images into a training set, a testing set and a verification set according to a proportion. And training the hot ring image. The training content comprises the steps of carrying out slicing operation on the input image to obtain a feature map, and carrying out convolution kernel convolution on the obtained feature map to obtain the feature map. And outputting a prediction frame on the basis of the initial test anchor frame, and updating iterative network parameters after calculating the difference.
And importing the test image acquired by the industrial camera and the template characteristic information into an image edge matching algorithm to calculate the similarity of the characteristic points. Before template matching, carrying out scaling training on template information at different angles and different scales to obtain characteristic points and characteristic vectors of a template image. And calculating the position of a pixel point of the input test image to construct a corresponding gradient map.
The corresponding diagram of the constructed gradient is to disperse the gradient into a plurality of directions with equal angular intervals, and the quantized gradient direction is represented by binary bits, thereby simplifying the information quantity represented by the gradient. And a calculation process of comparing the template model with the search images of all the positions, and a process of converting the template model into a lookup table. And searching corresponding data in the linear storage table according to the characteristic points of the template, and adding the corresponding data of the characteristic points to obtain template matching similarity measurement.
Image enhancement is performed according to equations (1) (2):
Add＝-Mult*GMin (1)
g'＝g*Mult+Add (2)
and the pixels are subjected to linear transformation, so that the edge contrast is improved, and the edge of the ring piece with higher precision is captured.
And performing circle fitting on the contour of the thermal ring piece through a fitting algorithm, and detecting an actual edge point with a larger gradient amplitude value in the caliper through a caliper tool. And fitting and inhibiting outliers by using a weighted least square method to obtain a measurement result of the size of the thermal ring with a good effect.
The above embodiments are only used for illustrating the design idea and features of the present invention, and the purpose of the present invention is to enable those skilled in the art to understand the content of the present invention and implement the present invention accordingly, and the protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications based on the principles and design concepts disclosed herein are intended to be included within the scope of the present invention.
Claims (9)
1. The hot ring rolling deformation geometric state on-line measuring method based on deep learning is characterized by comprising the following steps of: the method comprises the following steps:
s1: acquiring a rolling image of the hot ring in the hot rolling experiment through an image acquisition device, carrying out target detection on the rolling image of the hot ring through a deep learning algorithm, and determining a corresponding template image and a corresponding characteristic image;
s2: matching the template image and the rolling image by adopting an image edge matching algorithm;
s3: and fitting the boundary contour of the hot ring by using a weighting function, and calculating the outer diameter size of the hot ring.
2. The on-line measuring method for the rolling deformation geometrical state of the hot ring based on the deep learning as claimed in claim 1, wherein: in the step S1, the specific steps are:
s11: acquiring a real-time rolling image of the hot ring in the hot rolling experiment through an image acquisition device;
s12: taking a rolling image as an image sample, randomly dividing the rolling image into a training set, a testing set and a verification set according to a preset proportion, training by adopting the training set through a deep learning algorithm, updating weight parameters of an iterative convolution neural network, and carrying out target rough identification on the rolling image acquired by an image acquisition device to obtain a template image corresponding to the rolling image; a target area is selected on the template image for capturing a target thermal ring piece;
s13: carrying out slicing operation on the image sample with the original size to obtain a first characteristic diagram; and performing convolution kernel convolution on the first characteristic image to obtain a characteristic image.
3. The hot ring rolling deformation geometric state online measurement method based on deep learning of claim 2, wherein: in the step S1, the method further includes the following steps:
and calibrating the actual size of the hot ring piece by adopting a Zhang friend calibration method and a calibration plate, and correcting the distortion of the lens.
4. The hot ring rolling deformation geometric state online measurement method based on deep learning of claim 2, wherein: in the step S2, the specific steps are as follows:
s21: inputting a template image, carrying out affine transformation of the template image at different angles or different scales including translation, rotation and scaling, and calculating feature points and feature vectors of the template image;
s22: inputting a rolling image, calculating the position of a pixel point, and constructing a gradient corresponding graph;
s23: and calculating the similarity of the rolled image and the template image by adopting image edge matching to complete template matching.
5. The hot ring rolling deformation geometric state online measurement method based on deep learning of claim 4, wherein: in the step S3, the specific steps are:
s31: enhancing the image obtained in the step S2, carrying out linear transformation on pixels, highlighting a bright color area, inhibiting a dark color area, improving the edge contrast and capturing the edge of the high-precision thermal ring;
s32: a caliper rectangular frame is equidistantly generated on the circle contour fitted by the algorithm by adopting a caliper tool, edge point detection is carried out, and the actual edge point with larger gradient amplitude in the caliper is screened out;
s33: and introducing a weight function to carry out iterative weighted fitting on the actual edge points, fitting and inhibiting outliers by using a weighted least square method, and calculating the geometric parameters of the thermal ring.
6. The hot ring rolling deformation geometric state online measurement method based on deep learning of claim 4, wherein: in the step S22, the specific steps of constructing the gradient map are as follows:
dispersing the gradient into a plurality of directions with equal angular intervals, representing the quantized gradient direction by binary bits, and simplifying the information content represented by the gradient;
establishing a pre-response table through the quantization gradient, and converting a calculation process of comparing the template image with the search images at all positions into a lookup table;
and searching corresponding data in the linear storage table according to the characteristic points of the template image, and adding the corresponding data of the characteristic points to obtain template matching similarity measurement.
7. The on-line measuring method for the rolling deformation geometrical state of the hot ring based on the deep learning as claimed in claim 5, wherein: in step S31, the specific steps of enhancing the image are:
and GMin is set as a gray value, g is a current gray value, mult is a multiplied coefficient, and Add is an additional offset value, and image enhancement is performed according to the following equations (1) and (2):
Add＝-Mult*GMin (1)，
g'＝g*Mult+Add (2)。
8. a measuring system for the deep learning based hot ring rolling deformation geometric state online measuring method according to any one of claims 1 to 7, characterized in that: the system comprises an image acquisition device, an industrial personal computer and a light source;
the image acquisition device adopts an industrial camera and is used for acquiring a ring piece image;
the industrial personal computer is used for processing the collected images and calculating the geometric state parameters of the thermal ring piece;
the light source is used for providing illumination or supplementary lighting.
9. A computer storage medium, characterized in that: stored therein is a computer program executable by a computer processor, the computer program executing the method for on-line measurement of rolling deformation geometry of hot ring based on deep learning according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211326943.4A CN115908275A (en) | 2022-10-25 | 2022-10-25 | Hot ring rolling deformation geometric state online measurement method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211326943.4A CN115908275A (en) | 2022-10-25 | 2022-10-25 | Hot ring rolling deformation geometric state online measurement method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115908275A true CN115908275A (en) | 2023-04-04 |
Family
ID=86490272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211326943.4A Pending CN115908275A (en) | 2022-10-25 | 2022-10-25 | Hot ring rolling deformation geometric state online measurement method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908275A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117876429A (en) * | 2024-03-12 | 2024-04-12 | 潍坊海之晨人工智能有限公司 | Real standard platform of sports type industry vision |
-
2022
- 2022-10-25 CN CN202211326943.4A patent/CN115908275A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117876429A (en) * | 2024-03-12 | 2024-04-12 | 潍坊海之晨人工智能有限公司 | Real standard platform of sports type industry vision |
| CN117876429B (en) * | 2024-03-12 | 2024-06-07 | 潍坊海之晨人工智能有限公司 | Real standard system of sports type industry vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109141232B (en) | Online detection method for disc castings based on machine vision | |
| CN106934800B (en) | Metal plate strip surface defect detection method and device based on YOLO9000 network | |
| CN107341802B (en) | Corner sub-pixel positioning method based on curvature and gray scale compounding | |
| CN102521560B (en) | Instrument pointer image identification method of high-robustness rod | |
| CN111862037A (en) | Method and system for detecting geometric characteristics of precision hole type part based on machine vision | |
| CN108846397B (en) | Automatic detection method for cable semi-conducting layer based on image processing | |
| CN103759676A (en) | Non-contact type workpiece surface roughness detecting method | |
| CN110706293B (en) | SURF feature matching-based electronic component positioning and detecting method | |
| CN103149087B (en) | Follow-up window and digital image-based non-contact real-time strain measurement method | |
| CN111415376B (en) | Automobile glass subpixel contour extraction method and automobile glass detection method | |
| Tao et al. | Rolling surface defect inspection for drum-shaped rollers based on deep learning | |
| CN111161260A (en) | Hot-rolled strip steel surface defect detection method and device based on deep learning | |
| CN110738644A (en) | automobile coating surface defect detection method and system based on deep learning | |
| CN112014407A (en) | Method for detecting surface defects of integrated circuit wafer | |
| CN115908275A (en) | Hot ring rolling deformation geometric state online measurement method based on deep learning | |
| JP4003465B2 (en) | Specific pattern recognition method, specific pattern recognition program, specific pattern recognition program recording medium, and specific pattern recognition apparatus | |
| CN116678826A (en) | Appearance defect detection system and method based on rapid three-dimensional reconstruction | |
| CN113670931A (en) | Steel plate surface defect detection method and system based on neural network | |
| CN113822810A (en) | Method for positioning workpiece in three-dimensional space based on machine vision | |
| CN113705564B (en) | Pointer type instrument identification reading method | |
| Ye et al. | Automatic Value Identification of Pointer-Type Pressure Gauge Based on Machine Vision. | |
| CN111862178B (en) | Improved LBP feature extraction method | |
| CN111553345B (en) | Method for realizing meter pointer reading identification processing based on Mask RCNN and orthogonal linear regression | |
| Zhou et al. | An adaptive clustering method detecting the surface defects on linear guide rails | |
| CN116091818B (en) | Pointer type instrument reading identification method based on multi-neural network cascading model |
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 |