CN112734690A - Surface defect detection method and device and computer readable storage medium - Google Patents
Surface defect detection method and device and computer readable storage medium Download PDFInfo
- Publication number
- CN112734690A CN112734690A CN202011493629.6A CN202011493629A CN112734690A CN 112734690 A CN112734690 A CN 112734690A CN 202011493629 A CN202011493629 A CN 202011493629A CN 112734690 A CN112734690 A CN 112734690A
- Authority
- CN
- China
- Prior art keywords
- image
- network
- defect detection
- sub
- detected
- 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 133
- 238000001514 detection method Methods 0.000 title claims abstract description 98
- 238000003860 storage Methods 0.000 title claims description 7
- 238000012549 training Methods 0.000 claims abstract description 48
- 230000011218 segmentation Effects 0.000 claims abstract description 46
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 8
- 238000000638 solvent extraction Methods 0.000 claims abstract description 8
- 238000011176 pooling Methods 0.000 claims description 26
- 230000002950 deficient Effects 0.000 claims description 22
- 230000000903 blocking effect Effects 0.000 claims description 16
- 230000008014 freezing Effects 0.000 claims description 6
- 238000007710 freezing Methods 0.000 claims description 6
- 238000005192 partition Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 239000002699 waste material Substances 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30164—Workpiece; Machine component
Abstract
The invention is suitable for the technical field of image recognition and provides a surface defect detection method, which comprises the following steps: collecting surface images of industrial parts; partitioning an original large-size image of the surface image into a plurality of subgraphs to obtain a subgraph training data set; constructing a defect detection network; inputting subgraphs in a subgraph training data set into the defect detection network for training to obtain a trained defect detection network; carrying out segmentation processing on an image to be detected, and segmenting the image to be detected into a plurality of sub-image block images; and inputting the plurality of sub-image block images into a trained defect detection network to obtain the detection results of the plurality of sub-image block images. The method provided by the invention solves the problem of detection difficulty increase caused by poor detection precision, small data volume and small defects of the existing surface defects.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a surface defect method and device and a computer readable storage medium.
Background
In the industrial production and manufacturing process, quality evaluation and detection are particularly critical steps, reducing the delivery of waste products and ensuring the high quality of finished products are always targets sought by factories, and detection, particularly product defect detection, is a remaining problem in the industrial field. The traditional quality control method relies on training quality inspectors to manually screen products, however, with continuous enlargement of the scale of a factory and continuous high efficiency and intellectualization of an industrial product production line, the defects of labor consumption and low efficiency seriously affect the automatic reform and rapid development of the factory. In the early stage, people try to use a traditional method in the field of machine vision to construct an industrial defect detection model, however, when defect features are extracted by using traditional algorithm modeling, manual design still needs to be relied on, and due to the factors of various industrial defects, large uncertainty and the like, the performance of the algorithm is difficult to improve, and actual requirements cannot be met.
In recent years, deep learning is developed rapidly under the promotion of big data, as the deep learning can directly learn features from bottom-layer data, and an automatic learning process has rapid adaptability to new products, a plurality of researchers build a target detection framework based on a deep learning algorithm, and pure supervised learning can achieve good results on a plurality of tasks. However, the deep learning algorithm has disadvantages for industrial detection: firstly, the industrial defect data set does not have the characteristics of diversity, richness and the like, and a good enough model is difficult to train; secondly, the model trained by using the existing data does not have the capability of reasoning and identifying new categories; thirdly, it is impossible to accurately detect and identify targets with too small regions.
Unlike conventional target detection tasks, industrial defect detection has its unique characteristics:
1. the industrial defect image is difficult to collect, and the number of samples is small;
2. in the industrial field, the resolution of the acquired image of the part is high, the size is large, but the defect area is too small;
3. each product manufactured by a factory may have tens or hundreds of defect types;
4. there are various morphologies and varieties of each defect;
5. each defect may exist in any one location of the product.
Especially, the performance of the detection model is very tested by the characteristics of data lack, small defect area and the like.
Disclosure of Invention
The embodiment of the invention aims to provide a surface defect detection method, aiming at solving the problem that the detection method cannot meet the requirements due to the fact that the existing detection method is few in surface defect image data and small in defect.
The embodiment of the invention is realized in such a way that a surface defect detection method comprises the following steps: collecting surface images of industrial parts;
partitioning an original large-size image of the surface image into a plurality of subgraphs to obtain a subgraph training data set;
constructing a defect detection network, wherein the defect detection network comprises a segmentation network and a classification network, the segmentation network carries out pixel-level positioning on the surface defects, and the classification network judges whether the surface images have defects or not;
inputting subgraphs in a subgraph training data set into the defect detection network for training to obtain a trained defect detection network;
the method comprises the steps of carrying out blocking processing on an image to be detected, and blocking the image to be detected into a plurality of sub-image block images;
inputting the plurality of sub-image block images into a trained defect detection network to obtain detection results of the plurality of sub-image block images;
and judging whether the image to be detected has defects according to the detection results of the sub-image blocks, and if any sub-image block in the sub-image blocks is a defective image, judging that the image to be detected has defects.
Further, the image acquired by acquiring the surface image of the industrial part comprises a defective image and a non-defective image.
Further, the constructing the defect detection network includes:
building a segmentation network, wherein the segmentation network consists of a convolution layer, a pooling layer and a full-connection layer;
and building a classification network, wherein the classification network consists of two parts, the first part consists of a convolutional layer and a pooling layer, and the second part is a global pooling network.
Further, the inputting the subgraph in the subgraph training data set into the defect detection network for training, and the obtaining of the trained defect detection network includes:
inputting subgraphs in the subgraph training data set into a defect detection network, training the segmentation network, and obtaining the trained segmentation network;
freezing parameters of the trained segmentation network, inputting subgraphs in the subgraph training data set into a classification network, training the classification network, and obtaining the trained classification network;
and freezing parameters of the trained segmentation network and classification network, inputting subgraphs in the subgraph training data set into the defect detection network for training, and acquiring the trained defect detection network.
Further, the step of partitioning the original large-size image of the surface image into a plurality of sub-images specifically includes: and performing partition cropping on the original large-size image from left to right based on the sliding window, wherein adjacent subgraphs are partially overlapped.
It is also an object of another embodiment of the present invention to provide a surface defect detecting apparatus, the apparatus including:
the acquisition module is used for acquiring an image to be detected on the surface of the industrial part;
the blocking module is used for blocking the original large-size image of the surface image to be detected into a plurality of sub-image block images;
the defect detection network module comprises a segmentation network and a classification network, wherein the segmentation network is used for carrying out pixel-level positioning on the surface defects, and the classification network is used for judging whether the image to be detected on the surface has defects or not;
and the judging module is used for judging whether the image to be detected on the surface has defects according to the detection result of the sub-image block image.
Further, the determining whether the surface image has a defect according to the detection result of the sub-image specifically includes: and if any sub-image block image in the plurality of sub-image block images is a defect image, judging that the image to be detected has defects.
Further, the segmentation network consists of a convolution layer, a pooling layer and a full-link layer; the classification network is composed of two parts, the first part is composed of a convolutional layer and a pooling layer, and the second part is a global pooling network.
Another embodiment of the present invention also provides a computer-readable storage medium, having a computer program stored thereon, which, when being executed by a processor, implements the steps in the surface defect detecting method according to any one of the above.
The invention has the beneficial effects that: by partitioning the original image and respectively carrying out defect detection on all sub-image blocks in the constructed segmentation-classification network, the problem of imbalance proportion between the original image and a defect area is effectively avoided by combining the detection results of the sub-images, the training data volume of the model is increased to a certain extent for processing the sub-images of the original image, and the problem of small-area defect detection in the industrial field can be effectively solved.
Drawings
FIG. 1 is a flow chart of a method for detecting surface defects according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of partitioning an image according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating steps for constructing a defect detection network according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of building a classification network according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of building a classification network according to an embodiment of the present invention;
FIG. 6 is a schematic flow chart of defect detection network training according to an embodiment of the present invention
Fig. 7 is a schematic diagram of a surface defect detecting apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
According to the invention, the original image is partitioned, and all the sub-image blocks are respectively subjected to defect detection in the constructed segmentation-classification network, so that the problem of imbalance proportion between the original image and a defect area is effectively avoided by combining the sub-image detection results, the training data volume of the model is increased to a certain extent for processing the sub-images of the original image, and the problem of small-area defect detection in the industrial field can be effectively solved.
Specific implementations of the present invention are described in detail below with reference to specific examples.
Example one
Referring to fig. 1, a surface defect method according to a first embodiment of the present invention includes steps S01 to S07:
and S001, acquiring an image of the surface of the industrial part.
The surface of each part is shot into a plurality of images which do not overlap with each other. To ensure high quality of the image, the parts should be photographed in a controlled environment, and high resolution should be ensured to avoid missing and catching micro cracks or other defects due to pixel problems. For example, an industrial image acquisition system composed of a computer host, an industrial camera, an LED light source, a photoelectric sensor, and the like. The working process is as follows: firstly, initializing equipment and self-checking the equipment, then driving an industrial camera (a planar array CCD sensor) by a computer host through software, wherein the industrial camera is only in a state of waiting for acquiring an image signal at the moment, and when a photoelectric sensor does not detect an object, the industrial camera continues to wait for acquiring the image signal at the moment; when the photoelectric sensor detects that the product passes by, the LED light source is turned on, the industrial camera is triggered to collect digital image signals of the parts, and then the LED light source is turned off.
And S002, partitioning the original large-size image of the surface image into a plurality of subgraphs to obtain a subgraph training data set.
The industrial surface defect image has the characteristics of overlarge size of the whole image and localized defect characteristics, and the original image is directly subjected to convolution extraction to easily cause characteristic loss, so that the image is considered to be subjected to blocking pretreatment. Taking fig. 2 as an example, the original image is cropped in a partition manner from left to right based on a sliding window, and in order to ensure that each region can be detected and to prevent missing of defective regions of single-digit pixel size, the overlapping ratio between image blocks can be set during cropping. Naming the clipped image blocks according to a certain regular form, and finishing manual labeling of the surface defect image by adopting an image enhancement technology. If the cropped image blocks are named as a1, a2 … … An, the sub-picture training data set formed by the above sub-pictures may be named as subset a. The invention can acquire a plurality of industrial surface images, so a plurality of sub-image training data sets can be formed.
The original large-size image of the surface image is partitioned into a plurality of sub-images, the original large-size image is subjected to partition clipping from left to right based on a sliding window, and adjacent sub-images are partially overlapped.
And S003, constructing a defect detection network, wherein the defect detection network comprises a segmentation network and a classification network, the segmentation network is used for carrying out pixel-level positioning on the surface defects, and the classification network is used for judging whether the surface image has defects or not.
And step S004, inputting the subgraphs in the subgraph training data set into the defect detection network for training, and obtaining the trained defect detection network.
Step S005, the image to be detected is processed in a blocking mode, and the image to be detected is blocked into a plurality of sub-image block images.
Step S006, inputting the plurality of sub-image block images into a trained defect detection network, and obtaining the detection results of the plurality of sub-image block images.
Step S007, determining whether the image to be detected has a defect according to the detection result of the plurality of sub-pattern block images, and if any sub-pattern block image in the plurality of sub-pattern block images is a defective image, determining that the image to be detected has a defect.
Judging the defect result of the image to be detected by combining the detection results of all the sub-block images, and if one of the N sub-block images of the image to be detected is a defect image, determining that the image to be detected has defects; and if the N sub-image block images are non-defective images, judging that the image to be detected is non-defective.
According to the surface defect detection method, the original image is subjected to blocking processing and then the blocked image is subjected to defect detection, the problem that the proportion of the original image and a defect area is not adjusted is effectively avoided by determining whether the original image has defects according to the defect detection result of the blocked image, the training data amount of a model is increased to a certain extent by processing the sub-image of the original image, and the problem of small-area defect detection in the industrial field can be effectively solved.
Example two
In another embodiment of the present invention, the images acquired by acquiring the surface images of the industrial parts include defective and non-defective images.
By acquiring defective images and non-defective images and training the defective images and the non-defective images, the data volume can be increased, and the judgment precision of a defect detection network is improved.
EXAMPLE III
In an embodiment of the present invention, referring to fig. 3, the constructing the defect detection network includes steps S031-S032:
step S031, a segmentation network is built, and the segmentation network consists of a convolution layer, a pooling layer and a full-connection layer.
As shown in fig. 4, the partition network is composed of a convolutional layer, a pooling layer, and a fully-connected layer. The convolution kernel size and number of convolution layers may vary depending on the size of the image block, but to capture smaller feature details in an image, the present segmentation network requires the use of large-sized convolution kernels in the deeper layers, and the step size of all convolution kernels is set to 1. The segmentation network of the invention adopts the pooling layer to complete the down-sampling operation, and the pooling layer is used to replace the convolution layer with the step length, thereby ensuring that small but important details survive in the down-sampling process. The segmentation network established by the invention is mainly used for carrying out pixel-level positioning on surface defects of industrial parts, and after the segmented sub-images are input into the segmentation network, not only can a segmentation mask be obtained, but also a feature map for storing complete features can be obtained. The segmentation network can not only keep the possible detail characteristics in the image blocks, but also has a larger receptive field.
Step S032, building a classification network, wherein the classification network is composed of two parts, the first part is composed of a convolution layer and a pooling layer, and the second part is a global pooling network.
As shown in fig. 5, the classification network is composed of two parts: the first partial structure is shown in the left half of fig. 5 and consists of a convolutional layer and a pooling layer. The input of the first part is a feature map obtained by dividing the previous layer of a full connection layer in the network, the size of a convolution kernel in the network can be changed according to the size of an image block, the number of channels of the network is increased along with the reduction of feature resolution, and the down-sampling operation is finished by adopting a pooling layer, so that the loss of details is avoided; the second part is a global pooling network, as shown in the right part of fig. 5, global maximum and global average pooling is performed on the segmentation mask output result of the segmentation network and the output of the first part of the classification network, each output is a value due to the global pooling, namely, the problem that the dimensionality of the segmentation network output and the additional network output is not matched is eliminated, and finally, whether the sub-image block has defects or not is judged through a full connection layer.
Example four
In an embodiment of the present invention, referring to fig. 6, the inputting the subgraph in the subgraph training dataset into the defect detection network for training, and the obtaining of the trained defect detection network includes the following steps S041-S043:
s041, inputting the subgraphs in the subgraph training data set into a defect detection network, training the segmentation network, and acquiring the trained segmentation network;
when the segmentation network is trained, a loss function of the segmentation network is established and trained by minimizing the loss function of the segmentation network, namely, the network is trained based on a random gradient descent mode.
S042, freezing parameters of the trained segmentation network, inputting subgraphs in the subgraph training data set into a classification network, training the classification network, and obtaining the trained classification network;
the classification network comprises a first CNN classification network and a second CNN classification network, when the classification network is trained, the parameters of the trained segmentation network are frozen, the loss function of the first CNN classification network is proposed, and the loss function of the first CNN classification network is minimized for training; and establishing a loss function of the second CNN classification network, and minimizing the loss function of the second CNN classification network for training.
S043, freezing parameters of the trained segmentation network and classification network, inputting subgraphs in the subgraph training data set into the defect detection network for training, and obtaining the trained defect detection network.
And establishing a loss function of the defect detection network, and training by minimizing the loss function of the defect detection network.
EXAMPLE five
Another embodiment of the present invention provides a defect surface inspection apparatus, referring to fig. 7, the apparatus including:
the acquisition module is used for acquiring an image to be detected on the surface of the industrial part;
the blocking module is used for blocking the original large-size image of the surface image to be detected into a plurality of sub-image block images;
the defect detection network module comprises a segmentation network and a classification network, wherein the segmentation network is used for carrying out pixel-level positioning on the surface defects, and the classification network is used for judging whether the image to be detected on the surface has defects or not;
and the judging module is used for judging whether the image to be detected on the surface has defects according to the detection result of the sub-image block image.
Further, the determining whether the surface image has a defect according to the detection result of the sub-image specifically includes: and if any sub-image block image in the plurality of sub-image block images is a defect image, judging that the image to be detected has defects.
Further, the segmentation network consists of a convolution layer, a pooling layer and a full-link layer; the classification network is composed of two parts, the first part is composed of a convolutional layer and a pooling layer, and the second part is a global pooling network.
According to the defect surface detection device, the acquisition module is used for acquiring an image to be detected of the industrial surface, the blocking module is used for blocking the image to be detected, all sub-block images obtained through blocking are input into the defect detection network module for detection, whether the sub-block images are defective or not is obtained, and finally the judgment module is used for judging whether the image to be detected is defective or not. According to the method, the original large-size image is blocked, the detection result of the blocked image is combined to judge the detection result of the original image, the problem that the proportion of the original image and the defect area is not adjusted is effectively avoided, the training data volume of the model is increased to a certain extent for processing the sub-image of the original image, and the problem of small-area defect detection in the industrial field can be effectively solved.
EXAMPLE six
The present application further provides another embodiment, which is to provide a computer readable storage medium storing a defective surface detection method program, the defective surface detection method program being executable by at least one processor to cause the at least one processor to perform the steps of the defective surface detection method as described above.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present application.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (9)
1. A method of industrial surface defect detection, the method comprising:
collecting surface images of industrial parts;
partitioning an original large-size image of the surface image into a plurality of subgraphs to obtain a subgraph training data set;
constructing a defect detection network, wherein the defect detection network comprises a segmentation network and a classification network, the segmentation network carries out pixel-level positioning on the surface defects, and the classification network judges whether the surface images have defects or not;
inputting subgraphs in a subgraph training data set into the defect detection network for training to obtain a trained defect detection network;
the method comprises the steps of carrying out blocking processing on an image to be detected, and blocking the image to be detected into a plurality of sub-image block images;
inputting the plurality of sub-image block images into a trained defect detection network to obtain detection results of the plurality of sub-image block images;
and judging whether the image to be detected has defects according to the detection results of the plurality of sub-image block images, wherein if any sub-image block image in the plurality of sub-image block images is a defective image, the image to be detected is judged to have defects.
2. The industrial surface defect detection method of claim 1, wherein the images captured of the surface of the industrial component include defective and non-defective images.
3. The industrial surface defect detection method of claim 1, wherein said constructing a defect detection network comprises:
building a segmentation network, wherein the segmentation network consists of a convolution layer, a pooling layer and a full-connection layer;
and building a classification network, wherein the classification network consists of two parts, the first part consists of a convolutional layer and a pooling layer, and the second part is a global pooling network.
4. The industrial surface defect detection method of claim 1, wherein the inputting a subgraph in a subgraph training dataset into the defect detection network for training comprises:
inputting subgraphs in the subgraph training data set into a defect detection network, training the segmentation network, and obtaining the trained segmentation network;
freezing parameters of the trained segmentation network, inputting subgraphs in the subgraph training data set into a classification network, training the classification network, and obtaining the trained classification network;
and freezing parameters of the trained segmentation network and classification network, inputting subgraphs in the subgraph training data set into the defect detection network for training, and acquiring the trained defect detection network.
5. The industrial surface defect detection method of claim 1, wherein the partitioning of the original large-size image of the surface image into a plurality of subgraphs is specifically: and performing partition cropping on the original large-size image from left to right based on the sliding window, wherein adjacent subgraphs are partially overlapped.
6. A defective surface detecting apparatus, characterized in that the apparatus comprises:
the acquisition module is used for acquiring an image to be detected on the surface of the industrial part;
the blocking module is used for blocking the original large-size image of the surface image to be detected into a plurality of sub-image block images;
the defect detection network module comprises a segmentation network and a classification network, wherein the segmentation network is used for carrying out pixel-level positioning on the surface defects, and the classification network is used for judging whether the image to be detected on the surface has defects or not;
and the judging module is used for judging whether the image to be detected on the surface has defects according to the detection result of the sub-image block image.
7. The apparatus of claim 6, wherein the determination module is specifically to: and if any sub-image block image in the plurality of sub-image block images is a defect image, judging that the image to be detected has defects.
8. The apparatus of claim 6, wherein the split network is comprised of a convolutional layer, a pooling layer, and a fully-connected layer; the classification network is composed of two parts, the first part is composed of a convolutional layer and a pooling layer, and the second part is a global pooling network.
9. A computer-readable storage medium, having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for surface defect detection according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011493629.6A CN112734690A (en) | 2020-12-17 | 2020-12-17 | Surface defect detection method and device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011493629.6A CN112734690A (en) | 2020-12-17 | 2020-12-17 | Surface defect detection method and device and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112734690A true CN112734690A (en) | 2021-04-30 |
Family
ID=75602628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011493629.6A Pending CN112734690A (en) | 2020-12-17 | 2020-12-17 | Surface defect detection method and device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112734690A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113327241A (en) * | 2021-06-16 | 2021-08-31 | 南昌航空大学 | Visual detection method and system for surface defects of bearing end face |
| CN113379747A (en) * | 2021-08-16 | 2021-09-10 | 深圳市信润富联数字科技有限公司 | Wood defect detection method, electronic device, device and storage medium |
| CN113469997A (en) * | 2021-07-19 | 2021-10-01 | 京东科技控股股份有限公司 | Method, device, equipment and medium for detecting plane glass |
| CN113538436A (en) * | 2021-09-17 | 2021-10-22 | 深圳市信润富联数字科技有限公司 | Method and device for detecting part defects, terminal equipment and storage medium |
| CN113674203A (en) * | 2021-07-14 | 2021-11-19 | 歌尔股份有限公司 | Defect detection model training method and device and defect detection method and device |
| CN113689431A (en) * | 2021-10-27 | 2021-11-23 | 常州微亿智造科技有限公司 | Industrial product appearance defect detection method and device |
| CN114549393A (en) * | 2021-08-03 | 2022-05-27 | 正泰集团研发中心(上海)有限公司 | Image annotation method, device, equipment and computer storage medium |
| CN114596314A (en) * | 2022-05-09 | 2022-06-07 | 合肥联亚制衣有限公司 | Training method, device, equipment and medium for cloth flaw detection model |
| CN114612444A (en) * | 2022-03-16 | 2022-06-10 | 南京航空航天大学 | Fine defect analysis method based on progressive segmentation network |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392896A (en) * | 2017-07-14 | 2017-11-24 | 佛山市南海区广工大数控装备协同创新研究院 | A kind of Wood Defects Testing method and system based on deep learning |
| CN108985337A (en) * | 2018-06-20 | 2018-12-11 | 中科院广州电子技术有限公司 | A kind of product surface scratch detection method based on picture depth study |
| CN109583489A (en) * | 2018-11-22 | 2019-04-05 | 中国科学院自动化研究所 | Defect classifying identification method, device, computer equipment and storage medium |
| CN110097544A (en) * | 2019-04-25 | 2019-08-06 | 武汉精立电子技术有限公司 | A kind of display panel open defect detection method |
| CN110796637A (en) * | 2019-09-29 | 2020-02-14 | 郑州金惠计算机系统工程有限公司 | Training and testing method and device of image defect detection model and storage medium |
| CN111044525A (en) * | 2019-12-30 | 2020-04-21 | 歌尔股份有限公司 | Product defect detection method, device and system |
| CN111652098A (en) * | 2020-05-25 | 2020-09-11 | 四川长虹电器股份有限公司 | Product surface defect detection method and device |
| CN111667455A (en) * | 2020-04-28 | 2020-09-15 | 广东三三智能科技有限公司 | AI detection method for various defects of brush |
| CN111768381A (en) * | 2020-06-29 | 2020-10-13 | 北京百度网讯科技有限公司 | Part defect detection method and device and electronic equipment |
| CN112037219A (en) * | 2020-09-27 | 2020-12-04 | 上海电力大学 | Metal surface defect detection method based on two-stage convolution neural network |
-
2020
- 2020-12-17 CN CN202011493629.6A patent/CN112734690A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392896A (en) * | 2017-07-14 | 2017-11-24 | 佛山市南海区广工大数控装备协同创新研究院 | A kind of Wood Defects Testing method and system based on deep learning |
| CN108985337A (en) * | 2018-06-20 | 2018-12-11 | 中科院广州电子技术有限公司 | A kind of product surface scratch detection method based on picture depth study |
| CN109583489A (en) * | 2018-11-22 | 2019-04-05 | 中国科学院自动化研究所 | Defect classifying identification method, device, computer equipment and storage medium |
| CN110097544A (en) * | 2019-04-25 | 2019-08-06 | 武汉精立电子技术有限公司 | A kind of display panel open defect detection method |
| CN110796637A (en) * | 2019-09-29 | 2020-02-14 | 郑州金惠计算机系统工程有限公司 | Training and testing method and device of image defect detection model and storage medium |
| CN111044525A (en) * | 2019-12-30 | 2020-04-21 | 歌尔股份有限公司 | Product defect detection method, device and system |
| CN111667455A (en) * | 2020-04-28 | 2020-09-15 | 广东三三智能科技有限公司 | AI detection method for various defects of brush |
| CN111652098A (en) * | 2020-05-25 | 2020-09-11 | 四川长虹电器股份有限公司 | Product surface defect detection method and device |
| CN111768381A (en) * | 2020-06-29 | 2020-10-13 | 北京百度网讯科技有限公司 | Part defect detection method and device and electronic equipment |
| CN112037219A (en) * | 2020-09-27 | 2020-12-04 | 上海电力大学 | Metal surface defect detection method based on two-stage convolution neural network |
Non-Patent Citations (1)
| Title |
|---|
| 刘力哲: "基于深度学习的表面缺陷检测算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113327241A (en) * | 2021-06-16 | 2021-08-31 | 南昌航空大学 | Visual detection method and system for surface defects of bearing end face |
| CN113674203A (en) * | 2021-07-14 | 2021-11-19 | 歌尔股份有限公司 | Defect detection model training method and device and defect detection method and device |
| WO2023284712A1 (en) * | 2021-07-14 | 2023-01-19 | 歌尔股份有限公司 | Defect detection model training method, defect detection method and electronic device |
| CN113469997A (en) * | 2021-07-19 | 2021-10-01 | 京东科技控股股份有限公司 | Method, device, equipment and medium for detecting plane glass |
| CN113469997B (en) * | 2021-07-19 | 2024-02-09 | 京东科技控股股份有限公司 | Method, device, equipment and medium for detecting plane glass |
| CN114549393A (en) * | 2021-08-03 | 2022-05-27 | 正泰集团研发中心(上海)有限公司 | Image annotation method, device, equipment and computer storage medium |
| CN114549393B (en) * | 2021-08-03 | 2024-01-19 | 正泰集团研发中心(上海)有限公司 | Image labeling method, device, equipment and computer storage medium |
| CN113379747A (en) * | 2021-08-16 | 2021-09-10 | 深圳市信润富联数字科技有限公司 | Wood defect detection method, electronic device, device and storage medium |
| CN113538436A (en) * | 2021-09-17 | 2021-10-22 | 深圳市信润富联数字科技有限公司 | Method and device for detecting part defects, terminal equipment and storage medium |
| CN113689431A (en) * | 2021-10-27 | 2021-11-23 | 常州微亿智造科技有限公司 | Industrial product appearance defect detection method and device |
| CN114612444A (en) * | 2022-03-16 | 2022-06-10 | 南京航空航天大学 | Fine defect analysis method based on progressive segmentation network |
| CN114596314A (en) * | 2022-05-09 | 2022-06-07 | 合肥联亚制衣有限公司 | Training method, device, equipment and medium for cloth flaw detection model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112734690A (en) | Surface defect detection method and device and computer readable storage medium | |
| CN110060237B (en) | Fault detection method, device, equipment and system | |
| CN111667455B (en) | AI detection method for brushing multiple defects | |
| CN109613002B (en) | Glass defect detection method and device and storage medium | |
| CN110610475B (en) | Visual defect detection method of deep convolutional neural network | |
| CN110826520B (en) | Port grab bucket detection method based on improved YOLOv3-tiny algorithm | |
| CN112598672A (en) | Pavement disease image segmentation method and system based on deep learning | |
| CN111325713A (en) | Wood defect detection method, system and storage medium based on neural network | |
| CN113096098A (en) | Casting appearance defect detection method based on deep learning | |
| CN111160301A (en) | Tunnel disease target intelligent identification and extraction method based on machine vision | |
| CN111223129A (en) | Detection method, detection device, monitoring equipment and computer readable storage medium | |
| CN114782391A (en) | Method, system and device for constructing defect detection model of few-sample industrial image | |
| CN112200790B (en) | Cloth defect detection method, device and medium | |
| CN113658131A (en) | Tour type ring spinning broken yarn detection method based on machine vision | |
| CN112150460A (en) | Detection method, detection system, device, and medium | |
| CN110929795A (en) | Method for quickly identifying and positioning welding spot of high-speed wire welding machine | |
| CN114781514A (en) | Floater target detection method and system integrating attention mechanism | |
| CN115082451A (en) | Stainless steel soup ladle defect detection method based on image processing | |
| CN115018797A (en) | Screen defect detection method, screen defect detection device and computer-readable storage medium | |
| CN116385430A (en) | Machine vision flaw detection method, device, medium and equipment | |
| CN113177924A (en) | Industrial production line product flaw detection method | |
| CN110827263B (en) | Magnetic shoe surface defect detection system and detection method based on visual identification technology | |
| CN112085017A (en) | Tea tender shoot image segmentation method based on significance detection and Grabcut algorithm | |
| CN114898153A (en) | Two-stage surface defect identification method combining classification and detection | |
| CN111160100A (en) | Lightweight depth model aerial photography vehicle detection method based on sample generation |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210430 |