CN113935953A - Steel coil defect detection method based on image processing - Google Patents
Steel coil defect detection method based on image processing Download PDFInfo
- Publication number
- CN113935953A CN113935953A CN202111096944.XA CN202111096944A CN113935953A CN 113935953 A CN113935953 A CN 113935953A CN 202111096944 A CN202111096944 A CN 202111096944A CN 113935953 A CN113935953 A CN 113935953A
- Authority
- CN
- China
- Prior art keywords
- image
- circle
- steel coil
- clustering
- edge
- 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.)
- Withdrawn
Links
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20061—Hough transform
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a steel coil defect detection method based on image processing, which comprises the following steps: the method comprises the steps of obtaining a steel coil area image by performing semantic segmentation on an acquired image, extracting circular edges of steel coils on each layer in the steel coil area by Hough circle detection, and optimizing edge point statistical strategies of Hough circles to obtain the uncoiling degree of each layer of the steel coils; based on the method, the edge image of the normally coiled steel coil layer can be obtained through the Hough circle detection method, compared with the prior art, the method has the advantages that the missing steel coil layer edge image can be obtained, reference is provided for edge point statistical strategy optimization according to the size of the missing area, and optimization efficiency is improved.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a steel coil defect detection method based on image processing.
Background
At present, the defect types of steel coils are mainly as follows: the coil of strip degree is looser, and the coil of strip presents the turriform, rolls up the inclined to one side problem, and its main cause of production is: the hot rolled coil has poor coiling guide, poor centering, misaligned edges, deviation, or poor shape of the steel strip, and sickle curves on the surface of the steel strip. Aiming at the problems, the main defect detection method still detects the quality of the steel coil manually, and a large amount of manpower is needed for continuously detecting the quality of the steel coil in the detection process.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention adopts the following technical scheme:
a steel coil defect detection method based on image processing comprises the following steps:
step (ii) of
: semantic segmentation is carried out on the collected image to obtain a steel coil area image;
step (ii) of
: extracting circular edges of steel coils in each layer in a steel coil area through Hough circle detection;
step (ii) of
: and optimizing the edge point statistical strategy of the Hough circle to obtain the uncoiling degree of each layer of the steel coil.
Further, the steps
Comprises the following steps: the method comprises the following steps of carrying out necessary image data preprocessing on a front image of a steel coil acquired by a camera, carrying out gray processing on the acquired image, carrying out weighted gray processing on gray, inputting the processed image into a semantic segmentation network, and outputting a semantically segmented steel coil image, wherein the specific semantic segmentation network comprises the following steps: the semantic segmentation network structure is Encoder-Decoder, the output image is a binary image, pixel points of the image are marked manually, the pixel value of a steel coil area of a training set is marked as 1, pixels of other areas are marked as 0, and the marked image is input into the semantic segmentation network for training. The label is used for monitoring network training, the network loss function is a cross entropy loss function, and the binary image output by the trained semantic segmentation network is multiplied by the original image to obtain a steel coil image.
Further, the steps
Is as follows; extracting pixel points belonging to edges in the image through a Canny edge detection algorithm, and then converting the position coordinates of all edge pixel points into a three-dimensional polar coordinate system, wherein at the moment, each pixel point in the image is changed into a circle in a three-dimensional space (a, b, r), and an equation in a Cartesian coordinate system of the circle is as follows:
and is obtained after three-dimensional space mapping is carried out,
assuming a given point
We can draw all circles passing through it in a three-dimensional rectangular coordinate system, and if two different points do the above operation, the obtained curve is in space
Intersecting, i.e. they have a common set of (a, b, r), which means that they are on the same circle, more curves intersect at a point, which means that the circle represented by the intersection is composed of more points, by setting a threshold value, it is determined how many curves compare with a point before a circle is considered to be detected, when multiple sets of curves have a common set of (a, b, r), a circle is considered to be detected
When a plurality of circles exist in the image, the threshold value of the intersection of a plurality of curves at one point is repeatedly performed
Judging, then, just can detect a plurality of circles in the image, because the image of coil of strip is great relatively, the threshold value sets up also to great
。
Further, the steps
The method comprises the following specific steps: in curve intersection points in the three-dimensional space corresponding to edge pixel points of the image missing area, intersection points where a part of three-dimensional space curves are intersected may exist, and the curve number of the intersection points does not reach a threshold value
And the whole circular edge of partial unrolling is irregular circular, when the same circular edge point is mapped, the edge point is correspondent to several radiuses in a certain range of three-dimensional space, and several crossed points are crossed
,
Representing the total number of all three-dimensional space intersections that do not satisfy the threshold, each intersection
Representing a circular edge corresponding to a certain circle center, we will refer to such intersection points
Clustering is carried out to obtain the steel coil edge of the missing area, and the steps are as follows
Three-dimensional intersection point of standard circular edge obtained in (1)
Intersection point of nearest neighbors
The specific selection method of the initial central point as the initial central point of the cluster is as follows: first, to the threshold value
Making an adjustment to the threshold
Reducing by 10 times, obtaining the intersection point of the three-dimensional space of the edge image of a part of circles as an initial clustering center point, wherein the reduced proportion of each time is the initial radius of the clustering center
Clustering by taking the intersection point of the nearest neighbor standard circular edge area of the defect area as an initial central point, and obtaining the initial radius
All the intersection points in the steel coil are clustered into one class, the arcs where different intersection points in the same class are located are considered as edge pixel points on the same irregular circle, the irregular steel coil area nearest to the standard steel coil circle is obtained, and the clustered initial radius is
The initial radius is continuously enlarged, and when no new intersection point appears after the first initial radius is clustered, the initial radius
Clustering, then continuously expanding the initial clustering radius, and increasing the statistical point range of the nearest irregular circular edge until the number of the intersection points participating in clustering in the clustering area is greater than the threshold value
C, finishing the first clustering to obtain all edge pixel points of the nearest neighbor irregular circle, taking the ending clustering radius of the clustering center point of the nearest neighbor irregular circle as the starting radius of the second clustering radius, continuing to cluster the edge pixel points of the second irregular circle outside the nearest neighbor irregular circle, wherein the clustering mode is the same as the step c, and when the number of intersection points participating in clustering in the clustering region is the same as that of the intersection points participating in clusteringGreater than a threshold value
And finishing the second clustering to obtain all edge pixel points of a second irregular circle, then continuously carrying out iterative clustering to obtain irregular circles formed by the edges of the steel coils of all layers, and obtaining the number of times of radius iteration in the clustering process
And m represents: a total of m layers of irregular circular edges are used to estimate the size of the defect area, i.e. the degree of unrolling of each layer of irregular circles
And judging the current uncoiling defect through the change rule of the uncoiling degree of the coiled steel coils on different layers.
The invention has the beneficial effects that:
based on the method, the edge image of the normally coiled steel coil layer can be obtained through the Hough circle detection method, compared with the prior art, the method has the advantages that the missing steel coil layer edge image can be obtained, reference is provided for edge point statistical strategy optimization according to the size of the missing area, and optimization efficiency is improved. Based on the method, the statistical strategy of the Hough circle edge points corresponding to the steel coil layer is optimized through guiding the size of the missing area of the steel coil layer edge, and compared with the prior art, the method has the advantages that the edge images of all steel coils can be acquired in a self-adaptive mode, the detection capability of the Hough circle detection algorithm is improved, and the edge images close to the circle can be acquired.
Detailed Description
The invention is described in detail below with reference to the figures and examples.
The specific scenes aimed by the invention are as follows: and (5) in a steel strip processing scene, coiling the hot-rolled and cooled steel strip to obtain a steel coil. The camera is arranged at the steel coil coiling completion end to shoot images of the front side of the steel coil, and particularly, the front side of the steel coil refers to a plane with a plurality of steel coil coiling edges. The main defect problems that aim at are: the problem of loose steel coil is solved, and the problem of the protruding part of the steel coil is not explained and treated too much.
Step (ii) of
: and performing semantic segmentation on the acquired image to obtain a steel coil area image.
The purpose of this step is: and segmenting the acquired image through a semantic segmentation network to obtain a front image of the steel coil. The method has the advantages that the image of the non-steel coil area in the image can be segmented through the semantic segmentation network, and the efficiency of subsequent image detection is improved.
The input is as follows: performing semantic segmentation processing, and outputting: the segmented image.
The front image of the steel coil collected by the camera is subjected to necessary image data preprocessing, wherein the image data preprocessing comprises the following steps: image denoising (median filtering denoising), image enhancement (histogram equalization), and image preprocessing are conventional processing means in the field, and are not described in detail. And carrying out graying processing on the acquired image, wherein the graying adopts weighted graying, a specific graying method is not described, and a grayscale image is finally obtained.
Inputting the processed image into a semantic segmentation network, and outputting a steel coil image after semantic segmentation, wherein the specific semantic segmentation network comprises the following steps:
1. the semantic segmentation network structure is Encoder-Decoder, and the output image is a binary image.
2. And marking image pixel points manually, marking the pixel value of the steel coil area of the training set as 1, marking the pixels of other areas as 0, and inputting the marked image into a semantic segmentation network for training. The label serves as a supervision for network training.
3. The network loss function is a cross entropy loss function.
And multiplying the binary image output by the trained semantic segmentation network with the original image to obtain a steel coil image.
Step (ii) of
: and extracting the circular edges of the steel coils in each layer in the steel coil area through Hough circle detection.
The purpose of this step is: and carrying out Hough circle detection on all pixel points in the steel coil image to obtain the circular steel coil edge in the steel coil image. The method has the advantage that the edge of the steel coil at the normal coiling part in the steel coil image can be quickly found.
The input is as follows: and (3) carrying out Hough circle detection on the steel coil image, and outputting: the circular edge of the steel coil.
The traditional Hough circle transform detection principle:
because the image is the front image of the steel coil, the types of pixel points in the image are two, one is the edge pixel points of each layer of the steel coil layer, and the other is the image pixel points between the edges of each layer (the partial pixel points are non-edge pixel points). By means of a Canny edge detection algorithm (the Canny edge detection algorithm is a known algorithm and is not described in detail), pixel points belonging to edges in the image are extracted, then the position coordinates (Cartesian coordinate system) of all the edge pixel points are subjected to three-dimensional polar coordinate system conversion, and at the moment, each pixel point in the image is changed into a circle in a three-dimensional space (a, b and r). It can be understood that: the equation in the cartesian coordinate system of the circle is:
and is obtained after three-dimensional space mapping is carried out,
。
then, assume that a point is given
We can draw all circles passing through it in the three-dimensional rectangular coordinate system, finally we will get a three-dimensional curve, we can do the above operation to all points in the image, if the curve obtained after the above operation is done to two different points is in space
Intersect, i.e. they have a set of (a, b, r) in common, which means that they are on the same circle. More curves intersect at a point, which means that the circle represented by the intersection consists of more points, and a threshold is set to determine how many curves compare with a point to consider that a circle is detected.
When multiple sets of curves appear common
It is illustrated that a plurality of circles exist in the image. Then the threshold at which multiple curves intersect at a point is repeated
Judging, then, just can detect a plurality of circles in the image, because the image of coil of strip is great relatively, the threshold value sets up also to great
。
Thus, the edge of the steel coil area normally coiled in the image is obtained.
Step (ii) of
: and optimizing the edge point statistical strategy of the Hough circle to obtain the uncoiling degree of each layer of the steel coil.
The purpose of this step is: and performing edge point statistical strategy optimization on the image areas missing in the steel coil area image. The method has the advantages that the edge points of the irregular circle can be obtained, and the unrolling degree of the current missing area can be evaluated according to the size of the optimization range.
The input is as follows: and (3) carrying out statistical point strategy optimization processing on the image of the missing area of the edge of the steel coil, and outputting: the degree of unrolling of the defective area.
The specific process for obtaining the edge point uncoiling degree of the steel coil uncoiling area is as follows:
three-dimensional image missing region edge pixel point corresponding three-dimensional image missing region edge pixel pointIn the intersection points of the curves in the space, there may be some intersection points where the curves in the three-dimensional space intersect, but since part of the edge of the unwound roll is an irregular circular edge, the number of curves at the intersection points does not reach the threshold value
And the whole circular edge of the partial loose roll is irregular circular, so that when the same circular edge point is mapped, the edge point corresponds to a plurality of radiuses in a certain range of a three-dimensional space and intersects a plurality of intersection points
（
Representing the total number of all three-dimensional space intersection points which do not meet the threshold, wherein the intersection points are edge pixel points partially belonging to the same radius in practice), and each intersection point
Representing a circular edge corresponding to a certain circle center, we will refer to such intersection points
And clustering to obtain the steel coil edge of the missing area.
By the steps of
Three-dimensional intersection point of standard circular edge obtained in (1)
Intersection point of nearest neighbors
The specific selection method of the initial central point as the initial central point of the cluster is as follows: first, to the threshold value
Making an adjustment to the threshold
Reducing by 10 times (which is an empirical value and can be adjusted correspondingly) to obtain the intersection point of the three-dimensional space of the edge image of a part of circles as an initial clustering center point, wherein the reduced proportion of each time is the initial radius of the clustering center
. Clustering by taking the intersection point of the nearest neighbor standard circular edge area of the defect area as an initial central point, and obtaining the initial radius
All the intersection points in the same circle are grouped into one type, and the circular arcs where different intersection points in the same type are located are considered as edge pixel points on the same irregular circle. And obtaining the irregular steel coil area nearest to the standard steel coil circle.
In the coiling process of the steel coil, once coiling problems occur, the uncoiling degree of the irregular circle is constantly increased outwards from the edge of the standard circle, so that only the initial radius of the cluster is needed to be determined in the iteration process of the cluster center
The initial radius is continuously enlarged, and when no new intersection point appears after the first initial radius is clustered, the initial radius
Clustering, then continuously expanding the initial clustering radius, and increasing the statistical point range of the nearest irregular circular edge until the number of the intersection points participating in clustering in the clustering area is greater than the threshold value
And finishing the first clustering to obtain all edge pixel points of the nearest neighbor irregular circle.
Further, the cluster center point of the nearest neighbor irregular circle is used for terminating the clusteringC, the class radius is the initial radius of the second clustering radius, the edge pixel points of the second irregular circle outside the nearest irregular circle are continuously clustered, the clustering mode is the same as the step c, and when the number of the intersection points participating in clustering in the clustering area is larger than the threshold value
And finishing the second clustering to obtain all edge pixel points of the second irregular circle. And then continuously carrying out iterative clustering to obtain irregular circles formed by the edges of the steel coils of all layers.
By the number of iterations of the radius in the clustering process described above
(m denotes: m layers of irregular circular edges in total) to estimate the size of the defect region, that is, the degree of unrolling of each layer of irregular circles
Here, the formula of the degree of unrolling merely represents the nonlinear relationship between the number of iterations and the degree of unrolling, and does not represent the actual quantitative meaning. And finally, judging the current uncoiling defect through the change rule of the uncoiling degree of the coiled steel coils in different layers.
The above embodiments are merely illustrative of the present invention, and should not be construed as limiting the scope of the present invention, and all designs identical or similar to the present invention are within the scope of the present invention.
Claims (2)
1. A steel coil defect detection method based on image processing is characterized by comprising the following steps:
step (ii) of
: semantic segmentation is carried out on the collected image to obtain a steel coil area image;
step (ii) of
: extracting circular edges of steel coils in each layer in a steel coil area through Hough circle detection;
step (ii) of
: optimizing a statistical strategy of edge points of the Hough circle to obtain the uncoiling degree of each layer of the steel coil;
said step (c) is
Extracting pixel points belonging to edges in an image by a Canny edge detection algorithm, and then converting position coordinates of all edge pixel points into a three-dimensional polar coordinate system, wherein each pixel point in the image is a circle in a three-dimensional space (a, b, r), and an equation in a Cartesian coordinate system of the circle is as follows:
and is obtained after three-dimensional space mapping is carried out,
(ii) a Selecting any point
Drawing all passing points in a three-dimensional rectangular coordinate system
Obtaining a three-dimensional curve; when the two different points are subjected to the operation, the obtained curve is in a three-dimensional space
Intersecting, i.e. two different points being common to a group
Then the two different points are on the same circleThe above step (1); when multiple curves intersect at one point, the circle represented by the intersection point is composed of multiple points, and when the number of curves at the intersection point is greater than the threshold value
If the circle is a standard circle, otherwise, the circle is an irregular circle; when multiple sets of curves appear common
Then, there are multiple circles in the image; wherein
；
Said step (c) is
The method comprises the following specific steps: when the circle is irregular, the edge points correspond to a plurality of radiuses in a certain range of a three-dimensional space and intersect a plurality of intersection points when the same circle edge point is mapped
,
Representing the total number of all three-dimensional space intersections that do not satisfy the threshold, each intersection
Representing a circular edge corresponding to a certain circle center, and connecting the intersection points
Clustering to obtain the steel coil edge of the missing area; wherein the initial center point of the cluster is the step
Three-dimensional intersection point of the standard circle edge obtained in (1)
Intersection point of nearest neighbors
The specific selection method of the initial central point as the initial central point of the cluster is as follows: first, to the threshold value
Making an adjustment to the threshold
Reducing by 10 times, obtaining the intersection point of the three-dimensional space of the edge image of a part of circles as an initial clustering center point, wherein the reduced proportion of each time is the initial radius of the clustering center
Clustering by taking the intersection point of the nearest neighbor standard circular edge area of the defect area as an initial central point, and obtaining the initial radius
All the intersection points in the steel coil are clustered into one type, arcs where different intersection points in the same type are located are edge pixel points on the same irregular circle, the irregular steel coil area nearest to the standard steel coil circle is obtained, and the clustered initial radius is
Continuously enlarging, and when the first initial radius cluster is completed and no new intersection point appears, the initial radius
Clustering, and then continuously enlarging the initial radius until the number of the intersection points participating in clustering in the clustering area is greater than a threshold value
And after the first clustering is finished, obtaining all edge pixel points of the nearest neighbor irregular circle, taking the ending clustering radius of the clustering center point of the nearest neighbor irregular circle as the starting radius of the second clustering radius, continuing to cluster the edge pixel points of the second irregular circle outside the nearest neighbor irregular circle, and when the number of the intersection points participating in clustering in the clustering area is greater than the threshold value
And finishing the second clustering to obtain all edge pixel points of a second irregular circle, and then continuously carrying out iterative clustering to obtain the number of times of the iteration of the radius and the irregular circle formed by the edges of the steel coils of all layers
And m represents: there are m layers of irregular circular edges; the degree of unrolling of each layer of irregular circles
And judging the current uncoiling defect through the change rule of the uncoiling degree of the coiled steel coils on different layers.
2. The method for detecting the defects of the steel coil based on the image processing as claimed in claim 1, wherein the steps are as follows
Comprises the following steps: preprocessing image data through a front image of the steel coil acquired by a camera, and performing graying processing on the acquired image to obtain a grayscale image; the graying adopts weighted graying; inputting the processed image into a semantic segmentation network, and outputting a steel coil image subjected to semantic segmentation, wherein the semantic segmentation network comprises the following specific steps: the semantic segmentation network structure is Encoder-Decoder, the output image is a binary image, pixel points of the image are marked manually, the pixel value of a steel coil area of a training set is marked as 1, and the pixel value is used for identifying the steel coil area of the training setMarking the pixel of other areas as 0, and inputting the marked image into a semantic segmentation network for training; the label is used for monitoring network training, the network loss function is a cross entropy loss function, and the binary image output by the trained semantic segmentation network is multiplied by the original image to obtain a steel coil image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111096944.XA CN113935953A (en) | 2021-09-18 | 2021-09-18 | Steel coil defect detection method based on image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111096944.XA CN113935953A (en) | 2021-09-18 | 2021-09-18 | Steel coil defect detection method based on image processing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113935953A true CN113935953A (en) | 2022-01-14 |
Family
ID=79276092
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111096944.XA Withdrawn CN113935953A (en) | 2021-09-18 | 2021-09-18 | Steel coil defect detection method based on image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113935953A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445483A (en) * | 2022-01-28 | 2022-05-06 | 泗阳三江橡塑有限公司 | Injection molding part quality analysis method based on image pyramid |
| CN114943684A (en) * | 2022-04-15 | 2022-08-26 | 上海波士内智能科技有限公司 | Curved surface anomaly detection method by using confrontation to generate self-coding neural network |
| CN116309570A (en) * | 2023-05-18 | 2023-06-23 | 山东亮马新材料科技有限公司 | Titanium alloy bar quality detection method and system |
| CN116993721A (en) * | 2023-09-26 | 2023-11-03 | 江苏康意德科技有限公司 | Steel plate surface defect detection method based on weak supervision neural network |
| EP4336449A4 (en) * | 2022-07-25 | 2024-10-23 | Honor Device Co Ltd | Mark hole positioning method and apparatus, assembly device, and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109540918A (en) * | 2018-11-28 | 2019-03-29 | 鞍钢集团自动化有限公司 | A kind of hot rolled coil edge fault detection device and method |
| CN112712512A (en) * | 2021-01-05 | 2021-04-27 | 余波 | Hot-rolled strip steel scab defect detection method and system based on artificial intelligence |
| KR20210066411A (en) * | 2019-11-28 | 2021-06-07 | 동의대학교 산학협력단 | Determination of defects of electrolytic zinc-coated Circular Plugs by using Machine Learning Pretreatment |
-
2021
- 2021-09-18 CN CN202111096944.XA patent/CN113935953A/en not_active Withdrawn
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109540918A (en) * | 2018-11-28 | 2019-03-29 | 鞍钢集团自动化有限公司 | A kind of hot rolled coil edge fault detection device and method |
| KR20210066411A (en) * | 2019-11-28 | 2021-06-07 | 동의대학교 산학협력단 | Determination of defects of electrolytic zinc-coated Circular Plugs by using Machine Learning Pretreatment |
| CN112712512A (en) * | 2021-01-05 | 2021-04-27 | 余波 | Hot-rolled strip steel scab defect detection method and system based on artificial intelligence |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445483A (en) * | 2022-01-28 | 2022-05-06 | 泗阳三江橡塑有限公司 | Injection molding part quality analysis method based on image pyramid |
| CN114943684A (en) * | 2022-04-15 | 2022-08-26 | 上海波士内智能科技有限公司 | Curved surface anomaly detection method by using confrontation to generate self-coding neural network |
| CN114943684B (en) * | 2022-04-15 | 2023-04-07 | 上海波士内智能科技有限公司 | Curved surface anomaly detection method by using confrontation to generate self-coding neural network |
| EP4336449A4 (en) * | 2022-07-25 | 2024-10-23 | Honor Device Co Ltd | Mark hole positioning method and apparatus, assembly device, and storage medium |
| CN116309570A (en) * | 2023-05-18 | 2023-06-23 | 山东亮马新材料科技有限公司 | Titanium alloy bar quality detection method and system |
| CN116309570B (en) * | 2023-05-18 | 2023-08-04 | 山东亮马新材料科技有限公司 | Titanium alloy bar quality detection method and system |
| CN116993721A (en) * | 2023-09-26 | 2023-11-03 | 江苏康意德科技有限公司 | Steel plate surface defect detection method based on weak supervision neural network |
| CN116993721B (en) * | 2023-09-26 | 2023-11-28 | 江苏康意德科技有限公司 | Steel plate surface defect detection method based on weak supervision neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113935953A (en) | Steel coil defect detection method based on image processing | |
| CN110175982B (en) | Defect detection method based on target detection | |
| CN113658132B (en) | Computer vision-based structural part weld joint detection method | |
| CN115330645B (en) | Welding image enhancement method | |
| CN106803257B (en) | Method for segmenting disease spots in crop disease leaf image | |
| CN110443778B (en) | Method for detecting irregular defects of industrial products | |
| CN103543394A (en) | Discharge ultraviolet imaging quantization parameter extraction method of high-voltage electric equipment | |
| CN108846831B (en) | Band steel surface defect classification method based on combination of statistical characteristics and image characteristics | |
| CN101169867B (en) | Image dividing method, image processing apparatus and system | |
| CN115345802A (en) | Remote monitoring method for operation state of electromechanical equipment | |
| CN114648511A (en) | Accurate extraction and identification method for escherichia coli contour | |
| CN116843581B (en) | Image enhancement method, system, device and storage medium for multi-scene graph | |
| CN116934787A (en) | Image processing method based on edge detection | |
| CN110427979B (en) | Road water pit identification method based on K-Means clustering algorithm | |
| CN111161276A (en) | Iris normalized image forming method | |
| CN112446417B (en) | Spindle-shaped fruit image segmentation method and system based on multilayer superpixel segmentation | |
| CN110781936A (en) | Construction method of threshold learnable local binary network based on texture description and deep learning and remote sensing image classification method | |
| CN115063603A (en) | Wood annual ring line accurate extraction and restoration method based on edge information | |
| CN112986277B (en) | Detection method for hot-rolled strip steel finish rolling roll mark | |
| CN113496159B (en) | Multi-scale convolution and dynamic weight cost function smoke target segmentation method | |
| CN115035117A (en) | Strip steel overlapping detection analysis method and system based on artificial intelligence | |
| CN117975444B (en) | Food material image recognition method for food crusher | |
| CN118470024B (en) | Food foreign matter rapid detection method and system based on image characteristics | |
| CN115908412B (en) | Bamboo strip defect detection method and terminal based on line laser image | |
| CN118609063B (en) | Electric vehicle household intelligent identification method based on secondary AI algorithm |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220114 |
|
| WW01 | Invention patent application withdrawn after publication |