CN116402753A - Improved YOLOv 5-based steel defect detection method - Google Patents
Improved YOLOv 5-based steel defect detection method Download PDFInfo
- Publication number
- CN116402753A CN116402753A CN202310206730.6A CN202310206730A CN116402753A CN 116402753 A CN116402753 A CN 116402753A CN 202310206730 A CN202310206730 A CN 202310206730A CN 116402753 A CN116402753 A CN 116402753A
- Authority
- CN
- China
- Prior art keywords
- steel
- defect
- steel defect
- improved
- defect detection
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 132
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 128
- 239000010959 steel Substances 0.000 title claims abstract description 128
- 238000001514 detection method Methods 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims description 18
- UQSXHKLRYXJYBZ-UHFFFAOYSA-N Iron oxide Chemical compound [Fe]=O UQSXHKLRYXJYBZ-UHFFFAOYSA-N 0.000 claims description 10
- 230000006870 function Effects 0.000 description 26
- 238000004519 manufacturing process Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000002950 deficient Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 239000000470 constituent Substances 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- NDLPOXTZKUMGOV-UHFFFAOYSA-N oxo(oxoferriooxy)iron hydrate Chemical compound O.O=[Fe]O[Fe]=O NDLPOXTZKUMGOV-UHFFFAOYSA-N 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000519995 Stachys sylvatica Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000009467 reduction Effects 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Geometry (AREA)
- Probability & Statistics with Applications (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a steel defect detection method based on improved YOLOv5, which comprises the steps of obtaining a sample image of a steel defect, and marking different defect types on the sample image so as to form a steel defect data set; the method comprises the steps that a Focal EIoU Loss positioning Loss function is adopted by a YOLOv5 network model, a coordinate attention module is introduced in front of a prediction head of the YOLOv5 network model, an improved YOLOv5 network model is built, and a steel defect data set is trained through the improved YOLOv5 network model so as to obtain a steel defect detection model; inputting the obtained steel defect image to be detected into the steel defect detection model to obtain the defect type and the defect position of the steel image to be detected. The invention can reduce the detection cost and the false judgment rate, and can improve the detection stability, thereby improving the accuracy of steel defect detection.
Description
Technical Field
The invention relates to a steel defect detection method based on improved YOLOv 5.
Background
The strip steel is one of important raw materials of steel, is widely applied to mechanical manufacture, aerospace and transportation and plays an important role in various production and living, but in the production process of the strip steel, due to the limitation of industrial technology and the influence of a production process, various defects such as surface oil spots, almond-shaped defects, white spots, scratches and the like can be caused on the surface of the strip steel. These defects can affect the corrosion resistance and the service life of the strip to a large extent. The existing defect detection means mainly uses manual naked eye detection, has low detection efficiency of workers, high labor intensity and high production cost, and can not meet the requirements of strip steel surface defect detection. The deep learning is performed by automatically extracting and learning the defect characteristics through the convolutional neural network without designing artificial characteristic factors, so that the deep neural network has the characteristics of strong learning capacity and high robustness, and is gradually becoming a mainstream method for detecting the defects of the strip steel.
Disclosure of Invention
The invention aims to solve the technical problems, and provides an improved YOLOv 5-based steel defect detection method and device, which can reduce detection cost and misjudgment rate and improve detection stability so as to improve steel defect detection precision.
The invention adopts the technical scheme that:
an improved YOLOv 5-based steel defect detection method comprises the following steps:
1) Acquiring a sample image of the steel defect, and marking different defect types on the sample image so as to form a steel defect data set;
2) The method comprises the steps that a Focal EIoU Loss positioning Loss function is adopted by a YOLOv5 network model, a coordinate attention module is introduced in front of a prediction head of the YOLOv5 network model, an improved YOLOv5 network model is built, and a steel defect data set is trained through the improved YOLOv5 network model so as to obtain a steel defect detection model;
3) Inputting the obtained steel defect image to be detected into the steel defect detection model to obtain the defect type and the defect position of the steel image to be detected.
Further, the coordinate attention module performs the following operations:
wherein,,
y c an output feature map representing the coordinate attention module;
x c an input feature map of the coordinate attention module;
Further, the Focal EIoU Loss function is:
L FocalEIOU ＝IOU γ L EIOU
wherein,,
L FocalEIOU representing the Focal EIoU Loss function,
γ in order to control the hyper-parameters of the curve radians,
the IOU represents the area cross-correlation of the real and predicted frames,
ρ 2 (b,b gt ) The square of the euclidean distance between the coordinates of the two center points of the target prediction frame and the target real frame,
b is the coordinates of the central point of the prediction frame,
b gt is the center point coordinates of the real frame,
w c and h c The width and height of the smallest rectangular box containing the predicted and real boxes respectively,
w, h are the width and height of the target prediction frame respectively,
w gt ，h gt the width and height of the target real box, respectively.
Further, before training the steel defect data set through the improved YOLOv5 network model, the steel defect data set is clustered through a K-means++ clustering algorithm, so that a priori frame suitable for steel defect detection is obtained.
Further, the defect types of the sample image of the steel defect include: iron oxide, plaque, cracks, pits, inclusions, scratches.
The invention has the following beneficial effects:
according to the invention, the improved YOLOv5 is improved by introducing the coordinate attention module and adopting the Focal EIoU Loss function, the improved YOLOv5 is trained by using the steel defect data set to obtain the steel defect detection model, and the steel image to be detected is input into the steel defect detection model to obtain the defect type of the steel image, so that the detection cost and the false judgment rate can be reduced, the detection stability can be improved, and the steel defect detection precision is improved.
Drawings
FIG. 1 is a flow chart of a method for detecting defects in a steel material based on improved YOLOv5 according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating the operation of the coordinate attention module according to one embodiment of the present invention;
FIG. 3 is a diagram of the algorithm structure of YOLOv5 before modification;
FIG. 4 is a diagram illustrating an algorithm configuration of improved YOLOv5 in accordance with one embodiment of the present invention;
FIG. 5 is a graph showing the identification result of the improved YOLOv5 steel defect detection according to one embodiment of the present invention;
FIG. 6 is a block schematic diagram of a modified Yolov 5-based steel defect detection apparatus according to an embodiment of the present invention;
fig. 7 is a block diagram schematically illustrating an improved YOLOv 5-based steel defect detection apparatus according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
FIG. 1 is a flow chart of a steel defect detection method based on improved YOLOv5 according to an embodiment of the present invention.
As shown in fig. 1, the improved YOLOv 5-based steel defect detection method according to the embodiment of the invention comprises the following steps:
s1, acquiring a sample steel image, and marking the defect type of the sample steel image to form a steel defect data set.
In one embodiment of the invention, some defective steel images may be crawled from the network or taken from the factory, pipeline, etc. production site. And then, marking the defect type of the obtained steel image to form a steel defect data set. Among them, the defect types of the steel image include iron oxide, plaque, crack, pit, inclusion, scratch, and the like. Taking 6 defect types of the steel image including ferric oxide, plaque, crack, pit, inclusion and scratch as examples, the defect types can be used as labels to label the steel image.
S2, training the improved YOLOv5 through a steel defect data set to obtain a steel defect detection model, wherein a coordinate attention module is introduced in front of a prediction head of the improved YOLOv5, and the improved YOLOv5 adopts a Focal EIoU Loss function.
Before training the improved YOLOv5 with the steel defect dataset, further comprising: and clustering the steel defect data set by using a K-means++ clustering algorithm to obtain a priori frame suitable for steel defect detection. The method comprises the following specific steps: randomly selecting one point in the label information set of the steel defect data set as a clustering center point; calculating the shortest distance D (x) between each point of the label information set and the existing cluster center points, calculating the probability that each point is selected as the next cluster center point, then selecting the next cluster center by using a wheel disc method, and repeating the steps to select 8 cluster center points; according to the process, 9 clustering center points can be obtained, the distance from each point in the label information set to the 9 clustering center points is calculated, and 9 clusters can be obtained by distributing each clustering center according to the distance; the average value of each cluster is taken as the center of each cluster, and then the calculation of the distance is repeated until the center of the cluster is no longer changed or little changed.
As shown in fig. 2, to further improve the accuracy of steel defect detection, a coordinate attention module is embedded before the YOLOv5 network pre-measurement head to suppress the interference of irrelevant information and strengthen the attention of defect information. Wherein the coordinate attention module performs the following operations:
specifically, given an input feature map x c First, each channel is encoded along the horizontal and vertical coordinates, respectively, using a pooling layer with pooling kernel sizes (H, 1) and (1, w).
Thus, the output of the c-th channel of height h can be expressed as:
likewise, the output of the c-th channel of width w can be written as:
the 2 transforms aggregate features along two spatial directions, respectively, to obtain a pair of direction-perceived feature maps. The feature map obtained by the transformation is subjected to splicing operation, and then is subjected to transformation operation by using convolution and activation functions:
f＝δ(F([z h ,z w ]))
g h ＝δ(F h (f h ))
g w ＝δ(F w (f w ))
where F represents convolution and delta represents sigmoid activation function.
Compared with the CIoU Loss function, the Focal EIoU Loss function not only splits the Loss term with aspect ratio into the difference value of the predicted width and height and the minimum external frame width and height, improves the convergence speed and regression accuracy of the model, but also solves the problem of sample imbalance in the task of optimizing the boundary frame regression, reduces the contribution of a large number of anchor frames which are less overlapped with the target frame to the boundary frame regression, and enables the whole regression process to be more focused on the high-quality anchor frames. The specific calculation of the Focal EIoU Loss function is as follows:
L FocalEIOU ＝IOU γ L EIOU
wherein L is FocalEIOU Representing Focal EIoU Loss function, IOU represents the area intersection ratio of a real frame and a predicted frame, ρ 2 (b,b gt ) The square of Euclidean distance between two center point coordinates of a target prediction frame and a target real frame is obtained, b is the center point coordinate of the prediction frame, and b gt Is the center point coordinate of the real frame, w c And h c Width and height of the smallest rectangular frame containing the predicted frame and the real frame, respectively, w and h are the orderTarget prediction frame width and height, w gt ，h gt The width and height of the target real box, respectively.
And introducing a coordinate attention module according to the original YOLOv5, and establishing a steel defect detection model based on the YOLOv5 improvement. Specifically, as shown in fig. 3, the YOLOv5 trunk feature extraction network before improvement adopts a CSP network, the CSP network is composed of 4C 3 residual blocks, basic constituent units in each residual block are convolution, batch normalization and activation functions, a feature pyramid network is adopted to fuse a shallow feature map with a deep feature map, and finally the network generates feature maps with the sizes of 20×20, 40×40 and 80×80 through three YOLOv prediction heads to predict results. Fig. 4 is a network structure diagram of the improved YOLOv5 algorithm, in which a coordinate attention mechanism is introduced before three YOLOv pre-measurement heads of the original YOLOv5, so as to inhibit the interference of irrelevant information and strengthen the attention of defect information.
According to the invention, the steel defect data set is trained through the improved YOLOv5, and finally, a model with optimal performance is selected to detect the steel defects. The method comprises the following specific steps: dividing the steel defect data set into a training set, a verification set and a test set according to the proportion of 8:1:1; loading an improved YOLOv5 model, and initializing super parameters, wherein the super parameters comprise: the coefficient values of the loss functions of each part are selected from the dynamic value, the weight attenuation coefficient, the learning rate, the training batch size, the iteration round number and the weight attenuation coefficient; performing a series of data preprocessing operations such as filtering and noise reduction, saturation adjustment and the like on images in a training set, loading a network model, and then performing feature extraction, defect positioning and defect classification on the input training set images; training each round, calculating a loss function by the network, and then carrying out parameter optimization by adopting an SGD optimizer; and calculating mAP (Mean Average Precision, average precision) of the network model on the verification set, judging whether the mAP of the model is optimal, and if so, storing the model.
S3, acquiring a steel image to be detected.
The steel image to be detected can be obtained by shooting at the detection station of the production line.
S4, inputting the steel image to be detected into a steel defect detection model to obtain the defect type and the position of the steel image to be detected.
The defect detection results of some embodiments of the present invention are shown in fig. 5.
According to the improved YOLOv 5-based steel defect detection method provided by the embodiment of the invention, the YOLOv5 is improved by introducing the coordinate attention module and adopting the Focal EIoU Loss function, the modified YOLOv5 is trained by using the steel defect data set to obtain the steel defect detection model, and the steel image to be detected is input into the steel defect detection model to obtain the defect type of the steel image, so that the detection cost and the false judgment rate can be reduced, the detection stability can be improved, and the steel defect detection precision is improved.
In order to realize the improved YOLOv 5-based steel defect detection method of the embodiment, the invention also provides an improved YOLOv 5-based steel defect detection device.
As shown in fig. 6, the improved YOLOv 5-based steel defect detection apparatus according to the embodiment of the present invention includes a first acquisition unit 10, a training unit 20, a second acquisition unit 30, and a detection unit 40.
The first acquiring unit 10 is configured to acquire a sample steel image, and label the defect type of the sample steel image to form a steel defect data set.
The training unit 20 is configured to train the modified YOLOv5 through the steel defect dataset to obtain a steel defect detection model, wherein the modified YOLOv5 incorporates a coordinate attention module in front of a prediction head, and the modified YOLOv5 employs a Focal EIoU Loss function.
The second acquisition unit 30 is used for acquiring an image of the steel material to be detected.
The detecting unit 40 is used for inputting the steel image to be detected into the steel defect detecting model to obtain the defect type and position of the steel image to be detected.
The first acquisition unit 10 may be used to crawl some defective steel images from the network or take defective steel images from a factory, pipeline, or the like. And marking the defect type of the obtained steel image to form a steel defect data set. Among them, the defect types of the steel image include iron oxide, plaque, crack, pit, inclusion, scratch, and the like. The steel image is marked by taking 6 defect types of the steel image, including ferric oxide, plaque, cracks, pits, inclusions and scratches, as labels.
As shown in fig. 7, the improved YOLOv 5-based steel defect detection apparatus according to the embodiment of the present invention may further include: the clustering unit 50, the clustering unit 50 is configured to perform a clustering operation on the steel defect data set using a K-means++ clustering algorithm before training the modified YOLOv5 through the steel defect data set, so as to obtain an a priori frame suitable for steel defect detection.
The method comprises the following specific steps: randomly selecting one point in the label information set of the steel defect data set as a clustering center point; calculating the shortest distance D (x) between each point of the label information set and the existing cluster center point, calculating the probability that each point is selected as the next cluster center point, namely selecting the next cluster center by using a wheel disc method, and repeating the steps to select 8 cluster center points; according to the process, 9 clustering center points can be obtained, the distance from each point in the label information set to the 9 clustering center points is calculated, and 9 clusters can be obtained by distributing each clustering center according to the distance; the average value of each cluster is taken as the center of each cluster, and then the calculation of the distance is repeated until the center of the cluster is no longer changed or little changed.
In order to further improve the accuracy of steel defect detection, a coordinate attention module is embedded in front of a YOLOv5 network pre-measurement head to inhibit interference of irrelevant information and strengthen attention of defect information. Wherein the coordinate attention module performs the following operations:
specifically, given an input feature map x c First, each channel is encoded along the horizontal and vertical coordinates, respectively, using a pooling layer with pooling kernel sizes (H, 1) and (1, w).Thus, the output of the c-th channel of height h can be expressed as:
likewise, the output of the c-th channel of width w can be written as:
the 2 transforms aggregate features along two spatial directions, respectively, to obtain a pair of direction-perceived feature maps. The feature map obtained by the transformation is subjected to splicing operation, and then is subjected to transformation operation by using convolution and activation functions:
f＝δ(F([z h ,z w ]))
g h ＝δ(F h (f h ))
g w ＝δ(F w (f w ))
where F represents convolution and delta represents sigmoid activation function.
Compared with the CIoU Loss function, the Focal EIoU Loss function not only splits the Loss term with aspect ratio into the difference value of the predicted width and height and the minimum external frame width and height, improves the convergence speed and regression accuracy of the model, but also solves the problem of sample imbalance in the task of optimizing the boundary frame regression, reduces the contribution of a large number of anchor frames which are less overlapped with the target frame to the boundary frame regression, and enables the whole regression process to be more focused on the high-quality anchor frames. The specific calculation of the Focal EIoU Loss function is as follows:
wherein L is FocalEIOU Representing Focal EIoU Loss function, IOU represents the area intersection ratio of a real frame and a predicted frame, ρ 2 (b,b gt ) Coordinates of two central points of a target prediction frame and a target real frameB is the center point coordinates of the prediction frame, b gt Is the center point coordinate of the real frame, w c And h c Width and height of the smallest rectangular frame containing the predicted frame and the real frame, w and h are the width and height of the target predicted frame, respectively, w gt ，h gt The width and height of the target real box, respectively.
And introducing a coordinate attention module according to the original YOLOv5, and establishing a steel defect detection model based on the YOLOv5 improvement. Specifically, the YOLOv5 trunk feature extraction network before improvement adopts a CSP network, the CSP network consists of 4C 3 residual blocks, basic constituent units in each residual block are convolution, batch normalization and activation functions, a feature pyramid network is adopted to fuse a shallow feature map with a deep feature map, and finally the network can generate feature maps with the sizes of 20×20, 40×40 and 80×80 through three YOLOv prediction heads to predict results. The improved YOLOv5 introduces a coordinate attention mechanism before the original YOLOv5 three YOLOv pre-measuring heads, so as to inhibit the interference of irrelevant information and strengthen the attention of defect information.
The training unit 20 is configured to train the steel defect dataset through the improved YOLOv5, and finally select a model with optimal performance for detecting the steel defect. The method comprises the following specific steps: dividing the steel defect data set into a training set, a verification set and a test set according to the proportion of 8:1:1; loading an improved YOLOv5 model, and initializing super parameters, wherein the super parameters comprise: the coefficient values of the loss functions of each part are selected from the dynamic value, the weight attenuation coefficient, the learning rate, the training batch size, the iteration round number and the weight attenuation coefficient; and filtering and denoising the images in the training set. Adjusting a series of data preprocessing operations such as saturation, loading a network model, and then carrying out feature extraction, defect positioning and defect classification on an input training set image; training each round, calculating a loss function by the network, and then carrying out parameter optimization by adopting an SGD optimizer; and calculating the mAP of the network model on the verification set, judging whether the mAP of the model is optimal, and if so, storing the model.
According to the improved YOLOv 5-based steel defect detection device, the YOLOv5 is improved by introducing the coordinate attention module and adopting the Focal EIoU Loss function, the improved YOLOv5 is trained by using the steel defect data set to obtain a steel defect detection model, and the steel image to be detected is input into the steel defect detection model to obtain the defect type of the steel image, so that the detection cost and the false judgment rate can be reduced, the detection stability can be improved, and the steel defect detection precision is improved.
The foregoing is merely a preferred embodiment of the invention, and it should be noted that modifications could be made by those skilled in the art without departing from the principles of the invention, which modifications would also be considered to be within the scope of the invention.
Claims (5)
1. A steel defect detection method based on improved YOLOv5 is characterized in that: the method comprises the following steps:
1) Acquiring a sample image of the steel defect, and marking different defect types on the sample image so as to form a steel defect data set;
2) The method comprises the steps that a Focal EIoU Loss positioning Loss function is adopted by a YOLOv5 network model, a coordinate attention module is introduced in front of a prediction head of the YOLOv5 network model, an improved YOLOv5 network model is built, and a steel defect data set is trained through the improved YOLOv5 network model so as to obtain a steel defect detection model;
3) Inputting the obtained steel defect image to be detected into the steel defect detection model to obtain the defect type and the defect position of the steel image to be detected.
2. The improved YOLOv 5-based steel defect detection method of claim 1, wherein: the coordinate attention module performs the following operations:
wherein,,
y c representing the coordinate noteThe output profile of the force module,
x c for the input of the coordinate attention module,
3. The improved YOLOv 5-based steel defect detection method of claim 1, wherein: the Focal EIoU Loss function is:
L FocalEIOU ＝IOU γ L EIOU
wherein,,
L FocalEIOU representing the Focal EIoU Loss function,
gamma is the super parameter of the radian of the control curve,
the IOU represents the area cross-correlation of the real and predicted frames,
ρ 2 (b,b gt ) The square of the euclidean distance between the coordinates of the two center points of the target prediction frame and the target real frame,
b is the coordinates of the central point of the prediction frame,
b gt is the center point coordinates of the real frame,
w c and h c The width and height of the smallest rectangular box containing the predicted and real boxes respectively,
w, h are the width and height of the target prediction frame respectively,
w gt ，h gt the width and height of the target real box, respectively.
4. The improved YOLOv 5-based steel defect detection method of claim 1, wherein: before training the steel defect data set through the improved YOLOv5 network model, clustering the steel defect data set by using a K-means++ clustering algorithm to obtain a priori frame suitable for steel defect detection.
5. The improved YOLOv 5-based steel defect detection method of claim 1, wherein: the defect types of the sample image of the steel defect comprise: iron oxide, plaque, cracks, pits, inclusions, scratches.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310206730.6A CN116402753A (en) | 2023-03-06 | 2023-03-06 | Improved YOLOv 5-based steel defect detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310206730.6A CN116402753A (en) | 2023-03-06 | 2023-03-06 | Improved YOLOv 5-based steel defect detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116402753A true CN116402753A (en) | 2023-07-07 |
Family
ID=87013219
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310206730.6A Pending CN116402753A (en) | 2023-03-06 | 2023-03-06 | Improved YOLOv 5-based steel defect detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116402753A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116664558A (en) * | 2023-07-28 | 2023-08-29 | 广东石油化工学院 | Method, system and computer equipment for detecting surface defects of steel |
-
2023
- 2023-03-06 CN CN202310206730.6A patent/CN116402753A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116664558A (en) * | 2023-07-28 | 2023-08-29 | 广东石油化工学院 | Method, system and computer equipment for detecting surface defects of steel |
| CN116664558B (en) * | 2023-07-28 | 2023-11-21 | 广东石油化工学院 | Method, system and computer equipment for detecting surface defects of steel |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113450307B (en) | Product edge defect detection method | |
| CN109711474B (en) | Aluminum product surface defect detection algorithm based on deep learning | |
| CN111179229B (en) | Industrial CT defect detection method based on deep learning | |
| CN111080620B (en) | Road disease detection method based on deep learning | |
| CN113160139B (en) | Attention-based steel plate surface defect detection method of Faster R-CNN network | |
| CN110009618B (en) | Shaft part surface quality detection method and device | |
| CN109035233B (en) | Visual attention network system and workpiece surface defect detection method | |
| CN114627383B (en) | Small sample defect detection method based on metric learning | |
| CN109215009B (en) | Continuous casting billet surface image defect detection method based on deep convolution neural network | |
| CN111612758A (en) | Tunnel surface defect detection method based on convolutional neural network | |
| CN111681240A (en) | Bridge surface crack detection method based on YOLO v3 and attention mechanism | |
| CN111982910B (en) | Weak supervision machine vision detection method and system based on artificial defect simulation | |
| CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
| CN111079734B (en) | Method for detecting foreign matters in triangular holes of railway wagon | |
| CN111767874B (en) | Pavement disease detection method based on deep learning | |
| CN111353396A (en) | Concrete crack segmentation method based on SCSEOCUnet | |
| CN115995056A (en) | Automatic bridge disease identification method based on deep learning | |
| CN116612106A (en) | Method for detecting surface defects of optical element based on YOLOX algorithm | |
| CN110992314A (en) | Pavement defect detection method and device and storage medium | |
| CN116402753A (en) | Improved YOLOv 5-based steel defect detection method | |
| CN112991271A (en) | Aluminum profile surface defect visual detection method based on improved yolov3 | |
| CN117036243A (en) | Method, device, equipment and storage medium for detecting surface defects of shaving board | |
| CN115457044A (en) | Pavement crack segmentation method based on class activation mapping | |
| CN115456944A (en) | Ceramic substrate appearance defect intelligent identification method based on target detection | |
| CN111524121A (en) | Road and bridge fault automatic detection method based on machine vision technology |
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 |