CN113393426A - Method for detecting surface defects of rolled steel plate - Google Patents
Method for detecting surface defects of rolled steel plate Download PDFInfo
- Publication number
- CN113393426A CN113393426A CN202110564536.6A CN202110564536A CN113393426A CN 113393426 A CN113393426 A CN 113393426A CN 202110564536 A CN202110564536 A CN 202110564536A CN 113393426 A CN113393426 A CN 113393426A
- Authority
- CN
- China
- Prior art keywords
- defect
- mask
- rolled steel
- point cloud
- steel plate
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 112
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 63
- 239000010959 steel Substances 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000001514 detection method Methods 0.000 claims abstract description 58
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 52
- 238000005096 rolling process Methods 0.000 claims abstract description 26
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 12
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 4
- 230000002146 bilateral effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; 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
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a method for detecting surface defects of a rolled steel plate, and relates to the technical field of rolled steel plate defect detection. The detection method comprises the steps that a high-precision linear array scanning camera collects two-dimensional images of the surface of a steel rolling plate and inputs the two-dimensional images into a Mask R-CNN convolutional neural network detection model for identification, and defect types and defect areas on the images are obtained; identifying defect points of the contour point cloud data of the surface of the steel rolling plate by a principal component analysis method and a RANSAC (random sample consensus) plane fitting algorithm, and performing Euclidean clustering on the defect points to obtain defect positioning of the surface of the steel rolling plate; and finally, taking the intersection of the identified defect area of the two-dimensional image and the defect location as the final defect position. The method for detecting the surface defects of the steel rolling plate reduces the labor cost and improves the precision of defect detection.
Description
Technical Field
The invention relates to the technical field of rolled steel plate defect detection, in particular to a method for detecting the surface defects of a rolled steel plate.
Background
Aiming at the problems that the prior steel plate surface quality inspection completely depends on manual experience to detect, the manual labor intensity is high, the working environment is poor, the timeliness and the accuracy are poor, and the like, a detection means or equipment is urgently needed to replace manual detection, the accuracy and the precision of the steel plate blank surface quality detection are improved, the labor intensity and the cost of personnel are greatly reduced, the production quality is improved, the automatic and intelligent development of the steel plate blank production detection is promoted, and the improvement of the production quality and the efficiency is promoted.
In recent years, visual detection technology and equipment gradually become an effective way for defect detection in industrial production, are widely applied to the fields of automobile manufacturing, aerospace, machine tool machining and the like, and the method for replacing the existing artificial steel plate defect detection by using the visual detection technology is an economical and feasible method.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method for detecting the surface defects of a steel rolling plate. The method finally obtains the defect position of the surface of the steel rolling plate by analyzing the two-dimensional image and the contour point cloud data of the surface of the steel rolling plate, and the method improves the detection efficiency and the detection precision of the surface defect of the steel rolling plate.
In order to achieve the purpose, the invention adopts the following technical scheme: a method for detecting surface defects of a rolled steel plate specifically comprises the following steps:
(1) acquiring a two-dimensional image of the surface of a rolled steel plate by a high-precision linear array scanning camera, and scanning the surface of the rolled steel plate by high-precision linear laser scanning equipment to obtain contour point cloud data;
(2) carrying out image denoising and histogram equalization processing on the two-dimensional image acquired in the step (1) to obtain preprocessed data;
(3) constructing a Mask R-CNN convolutional neural network detection model, training the Mask R-CNN convolutional neural network detection model, inputting the preprocessed data obtained in the step (2) into the trained Mask R-CNN convolutional neural network detection model, and identifying defect types and defect areas on the two-dimensional image;
(4) identifying defect points of the contour point cloud data obtained in the step (1) through a principal component analysis method and a RANSAC (random sample consensus) plane fitting algorithm, and performing Euclidean clustering on the defect points to obtain defect positioning of the surface of a steel rolling plate;
(5) and (4) taking the intersection of the defect area of the two-dimensional image identified in the step (3) and the defect location obtained in the step (4) as a final defect position.
Further, the Mask R-CNN convolutional neural network detection model comprises a ResNet-FPN main network, a regional suggestion network and a prediction module which are sequentially connected; the ResNet-FPN main network is used for extracting multi-scale feature images in the preprocessed data; the area suggestion network is used for extracting an interested area in the multi-scale characteristic image, performing pooling operation on the interested area through a RoIAlign layer, and converting the interested area into an output vector with a fixed size; the prediction module performs target classification, bounding box prediction and mask segmentation on the output vector by using the full-link layer and the full convolution layer.
Further, the training process of the Mask R-CNN convolutional neural network detection model specifically comprises the following steps:
(a) collecting images on the surface of a steel rolling plate, carrying out image denoising and histogram equalization on the images, then carrying out manual marking on a defect area, and forming an image data set by the manually marked images;
(b) and inputting the image data set into a Mask R-CNN convolutional neural network detection model for training, and finishing the training of the Mask R-CNN convolutional neural network detection model when a loss function L is converged.
Further, the manually marked content comprises a defect type mark and a defect area mark.
Further, the loss function L is specifically:
L＝Lclass+Lbox+Lmask
wherein L isclassRepresenting the target classification loss function, LboxRepresenting the boundary localization loss function, LmaskRepresenting a mask segmentation penalty function.
Further, the step (4) specifically includes the following sub-steps:
(4.1) scanning the surface of the rolled steel plate by adopting a high-precision line laser scanner to obtain contour point cloud data;
(4.2) calculating the normal line characteristic n of each contour point by principal component analysispComparing the error with the reference normal Z which is (0,0,1), and marking the contour point with the error E exceeding a threshold value as a suspected defect point;
(4.3) carrying out plane fitting on the contour point cloud data by using a RANSAC plane fitting algorithm, removing point clouds in the fitted plane model, and reserving residual point clouds;
and (4.4) taking the intersection of the suspected defect point and the residual point cloud as a defect point set, and performing Euclidean clustering on the defect point set to obtain the defect location of the surface of the steel rolling plate.
Further, the process of comparing errors in step (4.2) is as follows:
E＝arccos(np·Z)
further, the process of step (4.3) is specifically:
(A) randomly selecting three points in the contour point cloud data, and fitting the three points into a plane model;
(B) calculating the distances from all the contour point cloud data to the fitted plane model, and counting the contour point cloud data with the distances smaller than a threshold value;
(C) and (C) repeating the steps (A) - (B) until all the point cloud contour data are fitted with the plane model, taking the plane model with the maximum number of contour point clouds with the distance smaller than the threshold value as the final plane model, removing the point clouds in the fitted plane model, and keeping the residual point clouds.
Compared with the prior art, the invention has the following beneficial effects: the method carries out defect detection on the surface of the rolled steel plate through a high-precision linear array scanning camera and a high-precision line laser scanning device, on one hand, the defects in a two-dimensional image acquired by the camera are automatically identified and positioned by utilizing a Mask R-CNN convolution neural network detection model, so that the efficiency and the precision of the defect detection on the surface of the rolled steel plate are effectively improved, on the other hand, the defect position judgment is carried out on contour point cloud data acquired by the laser scanning device by combining a principal component analysis method and an RANSAC plane fitting algorithm, the robustness and the real-time performance of the defect detection on the surface of the rolled steel plate are ensured, finally, the defect area identified in the two-dimensional image cloud data is synthesized, the intersection of the two defect areas and the point cloud data is used as the final defect position, and the positioning accuracy of the surface defects of the rolled steel plate is further improved. The method for detecting the surface defects of the rolled steel plate can realize real-time online detection, and has the advantages of good timeliness, high detection precision and high detection efficiency.
Drawings
FIG. 1 is a flow chart of the method for detecting the surface defects of a rolled steel plate according to the present invention;
FIG. 2 is a schematic diagram of the defect detection of the two-dimensional image of the surface of the steel rolling plate collected in the present invention;
FIG. 3 is a schematic structural diagram of a Mask R-CNN convolutional neural network detection model in the present invention;
FIG. 4 is a schematic diagram of the defect detection of the profile point cloud data scanned on the surface of a steel rolling plate in the invention.
Detailed Description
The technical scheme of the invention is explained in detail in the following with the accompanying drawings.
FIG. 1 is a flow chart of the method for detecting the surface defects of the rolled steel plate, which specifically comprises the following steps:
(1) acquiring a two-dimensional image of the surface of a rolled steel plate by a high-precision linear array scanning camera, and scanning the surface of the rolled steel plate by high-precision linear laser scanning equipment to obtain contour point cloud data; the defect detection analysis of the surface of the rolled steel plate can be carried out through the two-dimensional image and the contour point cloud data of the surface of the rolled steel plate, and the defect detection accuracy is improved.
(2) Carrying out image denoising and histogram equalization processing on the two-dimensional image acquired in the step (1) to obtain preprocessed data; the image denoising method adopted in the invention is bilateral filtering processing, and bilateral filtering not only considers the spatial position relation of image pixel points, but also considers the pixel value similarity among different pixel points, so that the edge characteristics of the image can be well reserved while removing noise.
(3) The method comprises the steps of constructing a Mask R-CNN convolutional neural network detection model, training the Mask R-CNN convolutional neural network detection model, inputting the preprocessed data obtained in the step (2) into the trained Mask R-CNN convolutional neural network detection model, and identifying defect types and defect areas on a two-dimensional image. As shown in fig. 2, which is a schematic diagram of defect detection of a two-dimensional image of a steel rolling plate surface acquired in the present invention, the Mask R-CNN model not only can realize defect classification and positioning, but also can perform pixel-level segmentation on a defect region.
As shown in fig. 3, the Mask R-CNN convolutional neural network detection model adopted in the present invention includes a ResNet-FPN backbone network, a regional proposal network RPN, and a prediction module, which are connected in sequence; the ResNet-FPN main network is used for extracting multi-scale feature images in the preprocessed data and is formed by connecting a depth residual error network ResNet, a feature pyramid network FPN and a reverse side edge, and a multi-scale feature image C is extracted from a convolution layer at the ith stage of the depth residual error network ResNeti(i is more than or equal to 1 and less than or equal to 5), and then, the feature graph C is processed in the feature pyramid network FPN5Is replicated as P5I.e. P5＝C5From multi-scale feature images Ci(i is more than or equal to 1 and less than or equal to 4) and a feature map P generated in a feature pyramid network FPNj+1(j ═ i) feature map P of i-th stage generated through side connectioni(2. ltoreq. i.ltoreq.5) in which P5＝C5，P6From P5Obtained by convolution with 3 × 3 with step size 2. In the reverse side connection added in the present invention, P is added2Is replicated to N2I.e. N2＝P2Connecting the reverse side edges to generate a feature map Nl(l is more than or equal to 2 and less than or equal to 5) obtaining the product P by convolution with 3 multiplied by 3 with the step length of 2j+1(j is more than or equal to 2 and less than or equal to 5) feature map with the same size as Pj+1(j is more than or equal to 2 and less than or equal to 5) are added to obtain a fusion characteristic diagram Nl+1(l is more than or equal to 2 and less than or equal to 5), and the number of convolution kernels of all convolution operations is 256. The region suggestion network RPN is used for extracting an interested region RoI in the multi-scale characteristic image, performing pooling operation on the interested region RoI through a RoIAlign layer, and converting the interested region RoI into an output vector with a fixed size; the prediction module utilizes the full connection layer FC and the full convolution layer FCN to perform target classification, bounding box positioning and mask segmentation on the output vector, and target detection and segmentation efficiency is effectively improved.
The training process of the Mask R-CNN convolutional neural network detection model specifically comprises the following steps:
(a) collecting images on the surface of a steel rolling plate, carrying out image denoising and histogram equalization on the images, then carrying out manual marking on a defect area, and forming an image data set by the manually marked images; the content of the manual labeling comprises defect type labeling and defect area labeling.
(b) And inputting the image data set into a Mask R-CNN convolutional neural network detection model for training, and finishing the training of the Mask R-CNN convolutional neural network detection model when a loss function L is converged.
The loss function adopted in the invention is specifically as follows:
L＝Lclass+Lbox+Lmask
wherein L isclassRepresenting the target classification loss function, LboxRepresenting a bounding box prediction penalty function, LmaskRepresenting a mask segmentation penalty function, NclsFor normalizing LclassTerm, piRepresenting the probability that the ith anchor box is predicted as a foreground object,
representing the true classification value of the ith anchor, λ represents the equilibrium parameter, NregFor normalizing LboxItem, LregRepresenting regression loss terms，tiThe predicted coordinates representing the target bounding box,
representing the real coordinates of the object, smooth, in relation to the anchor frame of the positive sampleL1() Denotes the smooth L1 loss function defined in Fast R-CNN, m denotes the size of the region of interest RoI, yijA prediction mask value representing a pixel point (i, j), k represents the belonging classification of the region of interest RoI,
and (3) representing the real mask value of the pixel point (i, j) on the k category. The target classification loss, the boundary frame prediction loss and the Mask segmentation loss jointly form the multitask loss of a Mask R-CNN convolutional neural network detection model, the supervision model automatically and efficiently learns effective information in image data, and the defect detection precision is effectively improved.
(4) Identifying defect points of the contour point cloud data obtained in the step (1) through a principal component analysis method and a RANSAC (random sample consensus) plane fitting algorithm, and performing Euclidean clustering on the defect points to obtain defect positioning of the surface of a steel rolling plate; the method specifically comprises the following substeps:
(4.1) scanning the surface of the rolled steel plate by adopting a high-precision line laser scanner to obtain contour point cloud data;
(4.2) calculating the normal line characteristic n of each contour point by principal component analysispComparing the error with the reference normal Z which is (0,0,1), and marking the contour point with the error E exceeding a threshold value as a suspected defect point;
E＝arccos(np·Z)
(4.3) carrying out plane fitting on the contour point cloud data by using a RANSAC plane fitting algorithm, removing point clouds in the fitted plane model, and reserving residual point clouds; the specific process is as follows:
(A) randomly selecting three points in the contour point cloud data, and fitting the three points into a plane model;
(B) calculating the distances from all the contour point cloud data to the fitted plane model, and counting the contour point cloud data with the distances smaller than a threshold value;
(C) and (C) repeating the steps (A) - (B) until all the point cloud contour data are fitted with the plane model, taking the plane model with the maximum number of contour point clouds with the distance smaller than the threshold value as the final plane model, removing the point clouds in the fitted plane model, and keeping the residual point clouds.
And (4.4) taking the intersection of the suspected defect point and the residual point cloud as a defect point set, and performing Euclidean clustering on the defect point set to obtain the defect location of the surface of the steel rolling plate.
Fig. 4 is a schematic diagram of defect detection performed on contour point cloud data scanned on the surface of a steel rolling plate in the invention, a suspected defect point in the point cloud data is extracted by a principal component analysis method, meanwhile, a RANSAC plane fitting algorithm is adopted to extract residual point clouds in a non-plane, and then, an intersection of the suspected defect point and the residual point clouds is selected as a final defect point set, so that the false detection rate of the algorithm is effectively reduced, and the algorithm robustness is improved.
(5) And (4) taking the intersection of the defect area of the two-dimensional image identified in the step (3) and the defect location obtained in the step (4) as a final defect position. On one hand, defects in a two-dimensional image acquired by a camera are automatically identified and positioned by using a convolutional neural network detection model, so that the efficiency and the precision of the detection of the defects on the surface of a steel rolling plate are effectively improved, on the other hand, the defect position judgment is performed on contour point cloud data acquired by laser scanning equipment by combining a principal component analysis method and a RANSAC (random sample consensus) plane fitting algorithm, the robustness and the real-time performance of the detection of the defects on the surface of the steel rolling plate are ensured, finally, the defect regions identified in the two-dimensional image and the point cloud data are integrated, the intersection of the two is used as the final defect position, and the positioning accuracy of the defects on the surface of the steel rolling plate is further improved.
The method for detecting the surface defects of the steel rolling plate is used for detecting the surface defects of 100 rolled steel plates, and the defect detection accuracy reaches over 88.5 percent, so the method can realize the accurate positioning of the surface defects of the steel rolling plate and improve the working efficiency.
It should be noted that the present invention is not limited by the above-mentioned embodiments, and it is obvious to those skilled in the art that similar modifications, substitutions and the like can be made on the above-mentioned technical solutions, and these should also be included in the protection scope of the present invention.
Claims (8)
1. A method for detecting surface defects of a rolled steel plate is characterized by comprising the following steps:
(1) acquiring a two-dimensional image of the surface of a rolled steel plate by a high-precision linear array scanning camera, and scanning the surface of the rolled steel plate by high-precision linear laser scanning equipment to obtain contour point cloud data;
(2) carrying out image denoising and histogram equalization processing on the two-dimensional image acquired in the step (1) to obtain preprocessed data;
(3) constructing a Mask R-CNN convolutional neural network detection model, training the Mask R-CNN convolutional neural network detection model, inputting the preprocessed data obtained in the step (2) into the trained Mask R-CNN convolutional neural network detection model, and identifying defect types and defect areas on the two-dimensional image;
(4) identifying defect points of the contour point cloud data obtained in the step (1) through a principal component analysis method and a RANSAC (random sample consensus) plane fitting algorithm, and performing Euclidean clustering on the defect points to obtain defect positioning of the surface of a steel rolling plate;
(5) and (4) taking the intersection of the defect area of the two-dimensional image identified in the step (3) and the defect location obtained in the step (4) as a final defect position.
2. The method for detecting the surface defects of the rolled steel plate according to claim 1, wherein the Mask R-CNN convolutional neural network detection model comprises a ResNet-FPN trunk network, an area suggestion network and a prediction module which are sequentially connected; the ResNet-FPN main network is used for extracting multi-scale feature images in the preprocessed data; the area suggestion network is used for extracting an interested area in the multi-scale characteristic image, performing pooling operation on the interested area through a RoIAlign layer, and converting the interested area into an output vector with a fixed size; the prediction module performs target classification, bounding box prediction and mask segmentation on the output vector by using the full-link layer and the full convolution layer.
3. The method for detecting the surface defects of the rolled steel plate as claimed in claim 2, wherein the training process of the Mask R-CNN convolutional neural network detection model is as follows:
(a) collecting images on the surface of a steel rolling plate, carrying out image denoising and histogram equalization on the images, then carrying out manual marking on a defect area, and forming an image data set by the manually marked images;
(b) and inputting the image data set into a Mask R-CNN convolutional neural network detection model for training, and finishing the training of the Mask R-CNN convolutional neural network detection model when a loss function L is converged.
4. The method as claimed in claim 3, wherein the contents of the manual labeling include a defect type label and a defect area label.
5. The method for detecting surface defects of rolled steel sheets according to claim 3, wherein said loss function L is specifically:
L＝Lclass+Lbox+Lmask
wherein L isclassRepresenting the target classification loss function, LboxRepresenting the boundary localization loss function, LmaskRepresenting a mask segmentation penalty function.
6. The method for detecting the surface defects of a rolled steel sheet as claimed in claim 1, wherein the step (4) comprises the following steps:
(4.1) scanning the surface of the rolled steel plate by adopting a high-precision line laser scanner to obtain contour point cloud data;
(4.2) calculating the normal line characteristic n of each contour point by principal component analysispComparing the error with the reference normal Z which is (0,0,1), and marking the contour point with the error E exceeding a threshold value as a suspected defect point;
(4.3) carrying out plane fitting on the contour point cloud data by using a RANSAC plane fitting algorithm, removing point clouds in the fitted plane model, and reserving residual point clouds;
and (4.4) taking the intersection of the suspected defect point and the residual point cloud as a defect point set, and performing Euclidean clustering on the defect point set to obtain the defect location of the surface of the steel rolling plate.
7. The method for detecting surface defects of rolled steel sheets according to claim 6, wherein the error comparison in step (4.2) is performed by:
E＝arccos(np·Z)
8. the method for detecting the surface defects of the rolled steel sheet as claimed in claim 6, wherein the step (4.3) is carried out by:
(A) randomly selecting three points in the contour point cloud data, and fitting the three points into a plane model;
(B) calculating the distances from all the contour point cloud data to the fitted plane model, and counting the contour point cloud data with the distances smaller than a threshold value;
(C) and (C) repeating the steps (A) - (B) until all the point cloud contour data are fitted with the plane model, taking the plane model with the maximum number of contour point clouds with the distance smaller than the threshold value as the final plane model, removing the point clouds in the fitted plane model, and keeping the residual point clouds.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110564536.6A CN113393426B (en) | 2021-05-24 | 2021-05-24 | Steel rolling plate surface defect detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110564536.6A CN113393426B (en) | 2021-05-24 | 2021-05-24 | Steel rolling plate surface defect detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113393426A true CN113393426A (en) | 2021-09-14 |
| CN113393426B CN113393426B (en) | 2024-09-17 |
Family
ID=77619030
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110564536.6A Active CN113393426B (en) | 2021-05-24 | 2021-05-24 | Steel rolling plate surface defect detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113393426B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113592862A (en) * | 2021-09-27 | 2021-11-02 | 武汉科技大学 | Point cloud data segmentation method, system, device and medium for steel plate surface defects |
| CN115953453A (en) * | 2023-03-03 | 2023-04-11 | 国网吉林省电力有限公司信息通信公司 | Transformer substation geological deformation monitoring method based on image dislocation analysis and Beidou satellite |
| CN117237360A (en) * | 2023-11-15 | 2023-12-15 | 宁德思客琦智能装备有限公司 | Casting burr detection method and system based on 3D vision |
| CN117237361A (en) * | 2023-11-15 | 2023-12-15 | 苏州拓坤光电科技有限公司 | Grinding control method and system based on residence time algorithm |
| CN117853482A (en) * | 2024-03-05 | 2024-04-09 | 武汉软件工程职业学院(武汉开放大学) | Multi-scale-based composite defect detection method and equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104020177A (en) * | 2014-06-26 | 2014-09-03 | 重庆大学 | Dual-CCD (Charge Coupled Device) scanning imaging detection method for continuous casting slab surface defects |
| CN110232675A (en) * | 2019-03-28 | 2019-09-13 | 昆明理工大学 | Grain surface defects detection and segmenting device and method under a kind of industrial environment |
| CN110310262A (en) * | 2019-06-19 | 2019-10-08 | 上海理工大学 | A kind of method, apparatus and system for detection wheel tyre defect |
| CN111080627A (en) * | 2019-12-20 | 2020-04-28 | 南京航空航天大学 | 2D +3D large airplane appearance defect detection and analysis method based on deep learning |
| JP2021060291A (en) * | 2019-10-07 | 2021-04-15 | 株式会社神戸製鋼所 | Substrate surface defect inspection method |
-
2021
- 2021-05-24 CN CN202110564536.6A patent/CN113393426B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104020177A (en) * | 2014-06-26 | 2014-09-03 | 重庆大学 | Dual-CCD (Charge Coupled Device) scanning imaging detection method for continuous casting slab surface defects |
| CN110232675A (en) * | 2019-03-28 | 2019-09-13 | 昆明理工大学 | Grain surface defects detection and segmenting device and method under a kind of industrial environment |
| CN110310262A (en) * | 2019-06-19 | 2019-10-08 | 上海理工大学 | A kind of method, apparatus and system for detection wheel tyre defect |
| JP2021060291A (en) * | 2019-10-07 | 2021-04-15 | 株式会社神戸製鋼所 | Substrate surface defect inspection method |
| CN111080627A (en) * | 2019-12-20 | 2020-04-28 | 南京航空航天大学 | 2D +3D large airplane appearance defect detection and analysis method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 赵畅: "基于点云数据的钢板表面缺陷三维检测方法研究", 中国优秀硕士学位论文全文数据库（电子期刊）信息科技辑, no. 2021, 15 February 2021 (2021-02-15), pages 32 - 34 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113592862A (en) * | 2021-09-27 | 2021-11-02 | 武汉科技大学 | Point cloud data segmentation method, system, device and medium for steel plate surface defects |
| CN115953453A (en) * | 2023-03-03 | 2023-04-11 | 国网吉林省电力有限公司信息通信公司 | Transformer substation geological deformation monitoring method based on image dislocation analysis and Beidou satellite |
| CN115953453B (en) * | 2023-03-03 | 2023-08-15 | 国网吉林省电力有限公司信息通信公司 | Substation geological deformation monitoring method based on image dislocation analysis and Beidou satellite |
| CN117237360A (en) * | 2023-11-15 | 2023-12-15 | 宁德思客琦智能装备有限公司 | Casting burr detection method and system based on 3D vision |
| CN117237361A (en) * | 2023-11-15 | 2023-12-15 | 苏州拓坤光电科技有限公司 | Grinding control method and system based on residence time algorithm |
| CN117237361B (en) * | 2023-11-15 | 2024-02-02 | 苏州拓坤光电科技有限公司 | Grinding control method and system based on residence time algorithm |
| CN117237360B (en) * | 2023-11-15 | 2024-03-01 | 宁德思客琦智能装备有限公司 | Casting burr detection method and system based on 3D vision |
| CN117853482A (en) * | 2024-03-05 | 2024-04-09 | 武汉软件工程职业学院(武汉开放大学) | Multi-scale-based composite defect detection method and equipment |
| CN117853482B (en) * | 2024-03-05 | 2024-05-07 | 武汉软件工程职业学院(武汉开放大学) | Multi-scale-based composite defect detection method and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113393426B (en) | 2024-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110175982B (en) | Defect detection method based on target detection | |
| CN111292305B (en) | Improved YOLO-V3 metal processing surface defect detection method | |
| CN111223088B (en) | Casting surface defect identification method based on deep convolutional neural network | |
| CN113393426B (en) | Steel rolling plate surface defect detection method | |
| CN111062915B (en) | Real-time steel pipe defect detection method based on improved YOLOv3 model | |
| CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
| CN107230203B (en) | Casting defect identification method based on human eye visual attention mechanism | |
| CN110033431B (en) | Non-contact detection device and detection method for detecting corrosion area on surface of steel bridge | |
| CN111582294A (en) | Method for constructing convolutional neural network model for surface defect detection and application thereof | |
| CN111598856A (en) | Chip surface defect automatic detection method and system based on defect-oriented multi-point positioning neural network | |
| CN112907519A (en) | Metal curved surface defect analysis system and method based on deep learning | |
| CN113538503B (en) | Solar panel defect detection method based on infrared image | |
| CN115439458A (en) | Industrial image defect target detection algorithm based on depth map attention | |
| CN108509950B (en) | Railway contact net support number plate detection and identification method based on probability feature weighted fusion | |
| CN111523540A (en) | Metal surface defect detection method based on deep learning | |
| CN113177924A (en) | Industrial production line product flaw detection method | |
| CN115272204A (en) | Bearing surface scratch detection method based on machine vision | |
| CN115830359A (en) | Workpiece identification and counting method based on target detection and template matching in complex scene | |
| CN113870202A (en) | Far-end chip defect detection system based on deep learning technology | |
| CN111612747A (en) | Method and system for rapidly detecting surface cracks of product | |
| CN111950556A (en) | License plate printing quality detection method based on deep learning | |
| CN111753572A (en) | Complex background low-quality two-dimensional bar code detection method based on deep learning | |
| CN113705564B (en) | Pointer type instrument identification reading method | |
| CN111178405A (en) | Similar object identification method fusing multiple neural networks | |
| CN114331961A (en) | Method for defect detection of an object |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |