CN115294039A - Steel coil end surface defect detection method - Google Patents
Steel coil end surface defect detection method Download PDFInfo
- Publication number
- CN115294039A CN115294039A CN202210886416.2A CN202210886416A CN115294039A CN 115294039 A CN115294039 A CN 115294039A CN 202210886416 A CN202210886416 A CN 202210886416A CN 115294039 A CN115294039 A CN 115294039A
- Authority
- CN
- China
- Prior art keywords
- data
- steel coil
- end surface
- surface defect
- stage
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- 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
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The application discloses a method and a system for detecting end surface defects of a steel coil, computer equipment and a storage medium. The method comprises the following steps: preprocessing the end face data of the steel coil, wherein the travel data comprise a gray level image and a depth image; performing multi-information fusion on the preprocessed data; then data enhancement is carried out to expand data samples; extracting features by using a cross-stage convolution network; and fusing the multi-stage features of the previous step and classifying and positioning the defects. The method solves the problems that a neural network with high efficiency is not introduced in the conventional method, the sample is unbalanced and the like, and provides a reliable method for improving the efficiency of detecting the end surface defects of the steel coil.
Description
Technical Field
The application relates to a steel coil end surface defect detection method, in particular to a steel coil end surface defect detection method based on a neural network, and belongs to the technical field of machine vision detection.
Background
In machine vision technology, deep learning is the mainstream method of image processing. With the proposal of the neural network, various fields such as image classification, target detection, image segmentation and the like are developed in a brand-new way. The convolutional neural network plays an important role in the field of target detection. In the investigation, a plurality of production enterprises still adopt the scheme of artificial visual detection, so that the efficiency is low and the physical and psychological health of workers is influenced to a certain extent. The neural network technology is introduced into the detection of the end surface defects of the steel coil, so that a very efficient solution is provided.
Through investigation and research, the current defect detection technology is mainly divided into two types, the first type is a two-stage detection network, the target detection is completed by dividing the target detection into two stages by using a neural network, and generally, a candidate region is firstly extracted, and then classification regression of the region is performed. Such a solution is relatively cumbersome and has a large number of model parameters, and is not used in industry. The other is a single-stage detection network, which utilizes a neural network to complete the classification and regression of the model at one time. The scheme has the advantages of high detection speed and small model quantity, and is more suitable for the field of steel coil production. However, training of the neural network requires a large amount of data as a support, and in an industrial scene, the quality of a shot image is low due to insufficient data and poor environment and a large number of interference factors such as light sources. The collection of samples becomes a very difficult thing, and the effect is often not good when a small amount of samples are used for direct training. Therefore, the proposal of the sample data set aiming at insufficient quantity and low quality has great significance for solving the problem of steel coil end surface defect detection. On one hand, the operation environment of workers can be improved, on the other hand, the product quality can be improved, and the possibility of damaging downstream production equipment due to defects is reduced.
Disclosure of Invention
The application aims to provide a high-efficiency and robust steel coil end surface defect detection method, so that the problem of insufficient steel coil defect data is solved, and the accuracy of steel coil end surface detection is effectively improved.
The application provides a method for detecting end surface defects of a steel coil, which comprises the following steps:
preprocessing the end surface defect data of the steel coil, wherein the data comprise a gray image shot by a 2D camera and a depth image shot by a 3D camera;
performing channel integration and splicing on the preprocessed data;
processing the integrated data by using a convolutional neural network to obtain characteristic data of each stage;
inputting the feature data of each stage into a neural network of a feature pyramid structure for processing to obtain rich and fused multi-stage and multi-scale features;
and inputting the obtained characteristics into the decoupled detection head to obtain a detection result.
Wherein, carry out the channel integration concatenation with the data after the preliminary treatment, include:
channel integration is carried out by using a single-channel depth map and a single-channel gray map;
the third channel is filled as the sum of the depth map and grayscale image pixel values.
The method for processing the integrated data by using the convolutional neural network to obtain the characteristic data of each stage comprises the following steps:
extracting the processed data after channel integration by using a cross-stage convolutional neural network, and obtaining multi-stage characteristics;
the resulting convolved image feature vectors are used to represent a comprehensive representation of the image.
The method comprises the following steps of inputting feature data of each stage into a neural network of a feature pyramid structure to be processed to obtain rich and fused multi-level and multi-scale features, wherein the method comprises the following steps:
inputting each level of feature data of the multi-level features into the feature pyramid structure;
and fusing the feature data of different stages to obtain the final feature vector rich in multi-scale information.
The application still provides a device that coil of strip terminal surface defect detected, includes:
the preprocessing module is used for integrating the input gray level image and the depth image;
the convolution neural network module is used for performing convolution operation on the integrated data to extract features;
the characteristic pyramid module is used for fusing all levels of characteristics in the convolutional network together to obtain multi-scale fused rich information characteristics;
and the detection module is used for processing the obtained rich characteristic information to obtain a detection result.
Further, the preprocessing module includes: and the multi-information fusion module comprises a channel splicing operation and is used for splicing two single-channel data together and adding the two single-channel data before filling a third channel to form data input by the model.
Further, the convolutional neural network structure comprises a CSPNet model, and features are extracted by carrying out convolution by using a cross-phase connection structure.
Further, the feature pyramid module comprises an FPN model and a PAN model, and the feature pyramid structure is utilized to fuse the multi-level features to obtain a feature vector with multi-scale information.
Further, the detection module includes a decoupled classification branch and a regression branch. And classifying the types of the defects by using the classification branch, and detecting the positions of the defects by using the regression branch.
The present application further provides a computer device, which includes one or more processors and one or more memories, where at least one program code is stored in the one or more memories, and when the program code is loaded and executed by the one or more processors, the program code implements the functions of the steel coil end surface defect detection method as described above.
The application also provides a computer storage medium, wherein at least one program code is stored in the computer storage medium, and when the program code is loaded and executed by a processor, the function of the steel coil end surface defect detection method is realized.
In conclusion, the steel coil end surface defect detection method and the steel coil end surface defect detection system can effectively detect the steel coil end surface defects by using an intelligent technology, reduce the labor cost, improve the product quality and reduce the probability of damaging downstream production equipment by defective products.
Drawings
FIG. 1 is a schematic diagram of a steel coil end surface defect;
FIG. 2 is a schematic diagram of a multi-information fusion model;
fig. 3 is a schematic diagram of a steel coil end surface defect detection system provided in the embodiment of the present application;
fig. 4 is a flowchart of a method for detecting a defect on an end face of a steel coil provided in an embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application more apparent, embodiments of the present application will be described in detail below with reference to the accompanying drawings.
The method aims to design a high-performance and robust deep learning model to detect the defects of the end face of the steel coil. The specific defects are shown in fig. 1, and some defects such as damage, folding or fracture can be generated at the edge of the steel coil. And it has the difficulties of large size difference, low sample data quality, etc.
Based on the difficulties, the embodiment provides a method for detecting the end surface defects of the steel coil, which can effectively detect and identify the end surface defects of the steel coil and improve the product quality.
Fig. 3 is a schematic view of a steel coil end surface defect detection system provided in the embodiment of the present application, and as shown in fig. 3, the steel coil end surface defect detection system of the present embodiment mainly includes: the device comprises a multi-information fusion module, a convolution network module, a characteristic pyramid module and a detection head module.
Fig. 4 is a flowchart of a method for detecting end surface defects of a steel coil according to an embodiment of the present application, and as shown in fig. 4, the method according to the embodiment includes the following steps:
s1, acquiring a gray image of the end face of a steel coil through a 2D line scanning camera, and collecting a depth image corresponding to the end face of the steel coil by using a 3D line scanning camera to preliminarily obtain a sample set { G, D };
s2, carrying out channel fusion on the gray-scale image in the G and the depth image in the D, and constructing a data set M with multiple information by taking the sum of the data of the two images as a third channel;
the specific information fusion formula is M i =Concat(G i ,D i ,G i +D i );
Wherein, G i As gray scale map data, D i For depth map data, concat is a channel data splicing function, M i To obtain multiple information data samples.
And S3, enhancing the image in the M obtained in the S2 through data such as horizontal and vertical overturning, noise disturbance, affine transformation and the like. In addition, a 10-pixel block is randomly erased from the data sample to construct new data. Meanwhile, four different pieces of defect data are spliced together in a random proportion to form a new picture with the same size as the original image. The image processing operations are randomly applied to finally expand the data in M by a factor of 10 to generate a final data set M'.
S4, sequentially inputting the obtained data sets into a convolutional neural network structure CSPNet, performing convolutional extraction on image features, and fusing all levels of features by using an introduced feature pyramid structure to solve the problem of different defect sizes;
and step S5, inputting the multi-stage features into the detection head, classifying and positioning the target, and introducing a Focal loss function into a loss function of the target classification to solve the problem of unbalanced classification of partial defect samples. The formula for Focal loss is as follows:
Loss cls (p t )=-α t (1-p t ) γ log(p t )
and S6, storing the weight of the trained neural network model for subsequent defect detection and positioning of new data.
The method provided by the embodiment of the application solves the problems that a novel high-performance neural network is not effectively considered in the conventional method and the problems that the samples are unbalanced, multi-scale and insufficient in quantity are not solved, and provides a reliable prediction method for accurately detecting and positioning the end surface defects of the steel coil.
The embodiment of the application also provides computer equipment, which comprises one or more processors and one or more memories, wherein at least one program code is stored in the one or more memories, and when the program code is loaded and executed by the one or more processors, the functions of the method for detecting the end surface defect of the steel coil are realized.
The embodiment of the application also provides a computer storage medium, which stores a computer program, and the computer program is executed by a processor to realize the method for detecting the end surface defect of the steel coil.
It will be understood by those skilled in the art that all or part of the steps of the above methods may be implemented by instructing the relevant hardware through a program, and the program may be stored in a computer readable storage medium, such as a read-only memory, a magnetic or optical disk, and the like. Alternatively, all or part of the steps of the above embodiments may be implemented using one or more integrated circuits. Accordingly, each module/unit in the above embodiments may be implemented in the form of hardware, and may also be implemented in the form of a software functional module. The present application is not limited to any specific form of hardware or software combination.
While the foregoing is directed to the preferred embodiment of the present application, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.
Claims (10)
1. A method for detecting the end surface defect of a steel coil is characterized by comprising the following steps:
preprocessing the end surface defect data of the steel coil, wherein the data comprise a gray image shot by a 2D camera and a depth image shot by a 3D camera;
performing channel integration and splicing on the preprocessed data;
processing the integrated data by using a convolutional neural network to obtain characteristic data of each stage;
inputting the feature data of each stage into a neural network of a feature pyramid structure for processing to obtain rich and fused multi-stage and multi-scale features;
and inputting the obtained characteristics into the decoupled detection head to obtain a detection result.
2. The method for detecting the end surface defect of the steel coil as claimed in claim 1, wherein the channel integration and splicing are performed on the preprocessed data, and the method comprises the following steps:
channel integration is carried out by using a single-channel depth map and a single-channel gray map;
the third channel is filled as the sum of the depth map and grayscale image pixel values.
3. The method for detecting the end face defect of the steel coil according to claim 2, wherein the step of processing the integrated data by using a convolutional neural network to obtain the characteristic data of each stage comprises the following steps:
extracting the processed data after channel integration by using a cross-stage convolutional neural network, and obtaining multi-stage characteristics;
the resulting convolved image feature vectors are used to represent a comprehensive representation of the image.
4. The method for detecting the end surface defect of the steel coil as claimed in claim 3, wherein the characteristic data of each stage is input into the neural network of the characteristic pyramid structure to be processed to obtain rich and fused multi-stage and multi-scale characteristics, and the method comprises the following steps:
inputting each level of feature data of the multi-level features into the feature pyramid structure;
and fusing the feature data of different stages to obtain the final feature vector rich in multi-scale information.
5. The utility model provides a coil of strip terminal surface defect detecting system which characterized in that includes:
the preprocessing module is used for integrating the input gray level image and the depth image;
the convolution neural network module is used for performing convolution operation on the integrated data to extract features;
the characteristic pyramid module is used for fusing all levels of characteristics in the convolutional network together to obtain multi-scale fused rich information characteristics;
and the detection module is used for processing the obtained rich characteristic information to obtain a detection result.
6. The steel coil end surface defect detection system of claim 5, wherein the convolutional neural network module comprises: and the cross-stage convolutional neural network module comprises a multilayer convolutional network unit and is used for extracting various features of the image.
7. The steel coil end surface defect detection system of claim 5, wherein the feature pyramid module comprises FPN pyramid and PAN pyramid structures,
wherein the FPN feature pyramid transfers deep convolution features from top to bottom and fuses features at different stages,
and the PAN feature pyramid transmits the top-down multi-stage features obtained by the FPN from top to top, strengthens shallow information again and completes feature fusion.
8. The steel coil end surface defect detection system according to claim 5, wherein the detection module is a decoupling detection head, and comprises a classification module and a positioning module, the input data of the detection module is rich information characteristics of multi-scale fusion, and the classification and regression calculation of the detection module is utilized to obtain a final detection result.
9. A computer device characterized in that it comprises one or more processors and one or more memories, in which at least one program code is stored, which when loaded and executed by the one or more processors, implements the functions of the steel coil end surface defect detection method according to any one of claims 1 to 4.
10. A computer storage medium, characterized in that at least one program code is stored in the computer storage medium, and when being loaded and executed by a processor, the program code realizes the functions of the steel coil end surface defect detection method according to any one of claims 1 to 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210886416.2A CN115294039A (en) | 2022-07-26 | 2022-07-26 | Steel coil end surface defect detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210886416.2A CN115294039A (en) | 2022-07-26 | 2022-07-26 | Steel coil end surface defect detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115294039A true CN115294039A (en) | 2022-11-04 |
Family
ID=83824497
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210886416.2A Pending CN115294039A (en) | 2022-07-26 | 2022-07-26 | Steel coil end surface defect detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115294039A (en) |
-
2022
- 2022-07-26 CN CN202210886416.2A patent/CN115294039A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103279765B (en) | Steel wire rope surface damage detection method based on images match | |
CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
CN111460927B (en) | Method for extracting structured information of house property evidence image | |
CN110827312A (en) | Learning method based on cooperative visual attention neural network | |
CN117252815A (en) | Industrial part defect detection method, system, equipment and storage medium based on 2D-3D multi-mode image | |
CN113989604A (en) | Tire DOT information identification method based on end-to-end deep learning | |
CN113780484B (en) | Industrial product defect detection method and device | |
CN114612418A (en) | Method, device and system for detecting surface defects of mouse shell and electronic equipment | |
CN114419006A (en) | Method and system for removing watermark of gray level video characters changing along with background | |
CN116580026B (en) | Automatic optical detection method, equipment and storage medium for appearance defects of precision parts | |
CN113610024A (en) | Multi-strategy deep learning remote sensing image small target detection method | |
CN113313688A (en) | Energetic material medicine barrel identification method and system, electronic equipment and storage medium | |
CN112967224A (en) | Electronic circuit board detection system, method and medium based on artificial intelligence | |
CN111179278B (en) | Image detection method, device, equipment and storage medium | |
CN115797970B (en) | Dense pedestrian target detection method and system based on YOLOv5 model | |
CN112364687A (en) | Improved Faster R-CNN gas station electrostatic sign identification method and system | |
CN110889418A (en) | Gas contour identification method | |
CN115294039A (en) | Steel coil end surface defect detection method | |
CN113139932B (en) | Deep learning defect image identification method and system based on ensemble learning | |
CN115937492A (en) | Transformer equipment infrared image identification method based on feature identification | |
CN111680691B (en) | Text detection method, text detection device, electronic equipment and computer readable storage medium | |
CN111882545B (en) | Fabric defect detection method based on bidirectional information transmission and feature fusion | |
CN113850167A (en) | Commodity identification method and system based on edge calculation and machine deep learning | |
CN114445767A (en) | Method and system for detecting foreign matters transmitted by transmission belt | |
CN116580277B (en) | Deep learning-based bottom electronic identification tag missing image identification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |