CN108447054A - Defects in timber sample acquiring method, device, electronic equipment and storage medium - Google Patents
Defects in timber sample acquiring method, device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN108447054A CN108447054A CN201810240671.3A CN201810240671A CN108447054A CN 108447054 A CN108447054 A CN 108447054A CN 201810240671 A CN201810240671 A CN 201810240671A CN 108447054 A CN108447054 A CN 108447054A
- Authority
- CN
- China
- Prior art keywords
- wood
- sample data
- wood sample
- defect
- qualified
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 235
- 238000000034 method Methods 0.000 title claims abstract description 51
- 239000002023 wood Substances 0.000 claims abstract description 436
- 230000002950 deficient Effects 0.000 claims abstract description 98
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 18
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 13
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 13
- 230000002194 synthesizing effect Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 13
- 238000002372 labelling Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 abstract description 22
- 238000012797 qualification Methods 0.000 abstract 1
- 238000013528 artificial neural network Methods 0.000 description 25
- 238000010586 diagram Methods 0.000 description 13
- 241000894007 species Species 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- 241000238631 Hexapoda Species 0.000 description 6
- 241000607479 Yersinia pestis Species 0.000 description 6
- 238000004891 communication Methods 0.000 description 6
- 238000004590 computer program Methods 0.000 description 6
- DSSYKIVIOFKYAU-XCBNKYQSSA-N (R)-camphor Chemical compound C1C[C@@]2(C)C(=O)C[C@@H]1C2(C)C DSSYKIVIOFKYAU-XCBNKYQSSA-N 0.000 description 5
- 244000082946 Tarchonanthus camphoratus Species 0.000 description 5
- 235000005701 Tarchonanthus camphoratus Nutrition 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 5
- 235000018185 Betula X alpestris Nutrition 0.000 description 4
- 235000018212 Betula X uliginosa Nutrition 0.000 description 4
- 241000196324 Embryophyta Species 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 235000009984 Pterocarpus indicus Nutrition 0.000 description 3
- 244000086363 Pterocarpus indicus Species 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012372 quality testing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
-
- 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
- G06T2207/30161—Wood; Lumber
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present disclosure discloses a kind of defects in timber sample acquiring method, device, electronic equipment and storage medium.Method includes：Obtain qualified wood sample data；Has defect characteristic image according to qualified wood sample data acquisition；The existing defect characteristic image is obtained from defective wood sample database；The existing defect characteristic image and the qualified wood sample Data Synthesis are obtained into defect wood sample data identical with the timber classification of qualification wood sample data.In this way, at the training initial stage of artificial intelligence model, the missing due to defect wood sample is overcome, leads to train the artificial intelligence model None- identified defects in timber come and the low problem of accuracy of identification, and the acquisition modes for simplifying defect wood sample data, save cost.
Description
Technical Field
The disclosure relates to the technical field of artificial intelligence, in particular to a method and a device for acquiring a wood defect sample, electronic equipment and a computer-readable storage medium.
Background
Wood factories require quality testing of wood, including sample quality ratings and defect sample identification. From initial manual visual inspection, to machine learning inspection today, each new stage has an improvement in performance or efficiency over the previous old stage. Machine learning detection requires training of the neural network first, and the trained model is used for detection. Neural network training requires training samples with labeled classification data. The sample data is generally obtained by collecting wood images in a wood factory and performing classification and labeling by a labeling person.
However, in general, in a wood processing plant, a lot of qualified wood samples can be obtained, few defective samples are obtained, the unbalanced distribution of the samples makes the defective samples not easy to obtain, and thus the trained neural network can only identify the qualified wood, but has low capacity or accuracy for identifying the defective wood. Furthermore, different species of wood have different qualifying sample characteristics and cannot be directly transplanted into a new wood species or product with an identification model of another species. Therefore, there is a need for a method for synthesizing wood defect samples, which can synthesize sufficient defect samples of wood species required by factories using existing resources, and further accelerate neural network training for identifying a new wood product or variety.
Disclosure of Invention
The embodiment of the disclosure provides a wood defect sample obtaining method and device, electronic equipment and a computer readable storage medium.
In a first aspect, an embodiment of the present disclosure provides a method for obtaining a wood defect sample, including:
acquiring qualified wood sample data;
acquiring an existing defect characteristic image according to qualified wood sample data; the existing defect characteristic image is obtained from an existing defect wood sample database;
and synthesizing the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data.
Optionally, acquiring an existing defect feature image according to qualified wood sample data, including:
when the number of first existing defective wood sample data of which the wood category is the same as that of the qualified wood sample data in the existing defective wood sample database is less than or equal to a first preset threshold value, obtaining second existing defective wood sample data which is different from that of the qualified wood sample data from the existing defective wood sample database;
and extracting the existing defect characteristic image from the second existing defect wood sample data.
Optionally, the method further comprises:
and in the existing defective wood sample database, when the number of first existing defective wood sample data of which the wood category is the same as that of the qualified wood sample data is greater than a first preset threshold value, taking the first existing defective wood sample data as the defective wood sample data.
Optionally, the qualified wood sample data comprises a qualified wood sample image and labeling data; synthesizing the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data, wherein the method comprises the following steps:
and covering part or all of the existing defect characteristic image on a preset area of the qualified wood sample image to obtain defect wood sample data with the same wood category as that of the qualified wood sample data.
Optionally, the method further comprises:
and adjusting the texture features of the boundary region of the synthesized existing defect feature image, so that the difference between the texture features of the boundary region and the texture features of the existing defect feature image and/or the texture features of the qualified wood sample is smaller than a second preset threshold value.
Optionally, adjusting the texture feature of the boundary region of the synthesized existing defect feature image includes:
and filtering the texture features of the boundary region by using a low-pass filter.
Optionally, adjusting the texture feature of the boundary region of the existing defect feature image on the qualified wood sample image includes:
and carrying out image processing on the boundary area by using a trained artificial intelligence model, so that the image style of the boundary area comes from the existing defect characteristic image, and the image content of the boundary area comes from the qualified wood sample image.
Optionally, synthesizing the existing defect feature image and the qualified wood sample data to obtain defect wood sample data of the same wood category as the qualified wood sample data, further comprising:
generating marking data of the sample data of the defective wood; the marking data comprises defect marks and wood categories in the qualified wood sample data.
Optionally, the method further comprises:
training based on the qualified wood sample data and the defect sample data to obtain a wood defect identification model; or,
training based on the qualified wood sample data and the defect sample data to obtain an identification model of the wood defects, and outputting the identification model of the wood defects to a client.
Optionally, the method further comprises:
and outputting the sample data of the defective wood to a client.
In a second aspect, an embodiment of the present disclosure provides a wood defect sample acquiring device, including:
a first obtaining module configured to obtain qualified wood sample data;
the second acquisition module is configured to acquire an existing defect characteristic image according to qualified wood sample data; the existing defect characteristic image is obtained from an existing defect wood sample database;
and the synthesis module is configured to synthesize the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data.
Optionally, the second obtaining module includes:
a first obtaining sub-module, configured to obtain, in the existing defective wood sample database, a second existing defective wood sample data that is different from the wood category of the qualified wood sample data from the existing defective wood sample database when the number of first existing defective wood sample data that is the same as the wood category of the qualified wood sample data is less than or equal to a first preset threshold;
an extraction submodule configured to extract the existing defect feature image from the second existing defect wood sample data.
Optionally, the apparatus further comprises:
a third obtaining module, configured to, when the number of first existing defective wood sample data in the existing defective wood sample database, which is the same as the wood category of the qualified wood sample data, is greater than a first preset threshold, take the first existing defective wood sample data as the defective wood sample data.
Optionally, the qualified wood sample data comprises a qualified wood sample image and labeling data; the synthesis module comprises:
and the synthesis submodule is configured to cover part or all of the existing defect characteristic image on a preset area of the qualified wood sample image to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data.
Optionally, the apparatus further comprises:
and the adjusting module is configured to adjust the texture features of the boundary area of the synthesized existing defect feature image, so that the difference between the texture features of the boundary area and the texture features of the existing defect feature image and/or the texture features of the qualified wood sample is smaller than a second preset threshold value.
Optionally, the adjusting module includes:
a filtering submodule configured to filter the textural features of the junction region with a low pass filter.
Optionally, the adjusting module includes:
and the image processing submodule is configured to perform image processing on the boundary area by using a trained artificial intelligence model, so that the image style of the boundary area comes from the existing defect characteristic image, and the image content of the boundary area comes from the qualified wood sample image.
Optionally, the synthesis module further includes:
the labeling submodule is configured to generate labeling data of the defective wood sample data; the marking data comprises defect marks and wood categories in the qualified wood sample data.
Optionally, the apparatus further comprises:
a first training module configured to train to obtain a wood defect identification model based on the qualified wood sample data and the defect sample data; or,
and the second training module is configured to train to obtain a wood defect identification model based on the qualified wood sample data and the defect sample data, and output the wood defect identification model to the client.
Optionally, the apparatus further comprises:
an output module configured to output the defective wood sample data to a client.
The functions can be realized by hardware, and the functions can also be realized by executing corresponding software by hardware. The hardware or software includes one or more modules corresponding to the above-described functions.
In one possible design, the structure of the wood defect sample acquiring device includes a memory and a processor, the memory is used for storing one or more computer instructions supporting the wood defect sample acquiring device to execute the wood defect sample acquiring method in the first aspect, and the processor is configured to execute the computer instructions stored in the memory. The wood defect sample acquiring device may further include a communication interface for the wood defect sample acquiring device to communicate with other equipment or a communication network.
In a third aspect, an embodiment of the present disclosure provides an electronic device, including a memory and a processor; wherein the memory is configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method steps of the first aspect.
In a fourth aspect, the disclosed embodiments provide a computer-readable storage medium for storing computer instructions for a wood defect sample acquiring apparatus, which contains computer instructions for executing the wood defect sample acquiring method in the first aspect.
The technical scheme provided by the embodiment of the disclosure can have the following beneficial effects:
after the qualified wood sample data is obtained, the existing defect characteristic image is extracted from the existing defect sample data, and the existing defect characteristic image is synthesized to the qualified wood sample image, so that the defective wood sample data with the same wood type as the qualified wood sample data is obtained. By the method, the problems that the wood defects cannot be identified and the identification precision is low due to the defect of the defective wood sample in the initial training stage of the artificial intelligence model are solved, the acquisition mode of the defective wood sample data is simplified, and the cost is saved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
Other features, objects, and advantages of the present disclosure will become more apparent from the following detailed description of non-limiting embodiments when taken in conjunction with the accompanying drawings. In the drawings:
fig. 1 shows a flow chart of a method of wood defect sample acquisition according to an embodiment of the present disclosure;
FIG. 2 shows a flow chart of step S102 according to the embodiment shown in FIG. 1;
FIG. 3 shows a schematic diagram of an extracted pre-existing defect feature image overlaid entirely on a qualified wood sample image according to an embodiment of the present disclosure;
FIG. 4 illustrates a schematic view of an existing defect feature image partially overlaid on a qualified wood sample image according to an embodiment of the present disclosure;
FIG. 5 illustrates yet another schematic view of an existing defect feature image partially overlaid on a qualified wood sample image according to an embodiment of the present disclosure;
fig. 6 shows a schematic diagram of a VGG network according to an embodiment of the present disclosure;
fig. 7 illustrates a block diagram of a structure of a wood defect sample acquiring device according to an embodiment of the present disclosure;
fig. 8 is a schematic structural diagram of an electronic device suitable for implementing a wood defect sample acquisition method according to an embodiment of the present disclosure.
Detailed Description
Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. Also, for the sake of clarity, parts not relevant to the description of the exemplary embodiments are omitted in the drawings.
In the present disclosure, it is to be understood that terms such as "including" or "having," etc., are intended to indicate the presence of the disclosed features, numbers, steps, behaviors, components, parts, or combinations thereof, and are not intended to preclude the possibility that one or more other features, numbers, steps, behaviors, components, parts, or combinations thereof may be present or added.
It should be further noted that the embodiments and features of the embodiments in the present disclosure may be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
In a wood processing plant, a lot of qualified wood samples can be obtained, but few defect samples are obtained, the defect samples are not easy to obtain due to unbalanced distribution of the samples, and thus, the trained artificial intelligence model can only identify the qualified wood, but the identification capability or the accuracy of the defective wood is not high enough. Moreover, through experiments, the qualified sample characteristics of different types of wood are different, and an identification model of other types cannot be directly transplanted into a new wood type or product, but the defect sample characteristics of some different types of wood are similar. Therefore, the embodiment of the present disclosure provides a method for obtaining a wood defect sample, which can synthesize a qualified sample of a factory and a defect sample of a cloud to obtain a defect sample of a wood type required by the factory, and is used to accelerate training of an artificial intelligence model for identifying a new wood product or variety.
Fig. 1 shows a flowchart of a wood defect sample acquisition method according to an embodiment of the present disclosure. As shown in fig. 1, the wood defect sample acquiring method includes the following steps S101 to S103:
in step S101, sample data of qualified wood is acquired;
in step S102, acquiring an existing defect characteristic image according to qualified wood sample data; the existing defect characteristic image is obtained from an existing defect wood sample database;
in step S103, synthesizing the existing defect feature image and the qualified wood sample data to obtain defect wood sample data of the same wood category as the qualified wood sample data.
In this embodiment, the qualified wood sample data may be data of a wood sample considered as having no defects, which is provided by a factory, and includes an image of the wood sample, a wood species, and the like. The existing defective wood sample database can store some known defective wood sample data with defects, including images of the defective wood sample data, defect information, wood categories and the like; the defect information may include a location of the defect, a kind of the defect, and the like. The defect feature image may be an image of a defect site extracted from the non-qualified wood sample data. In one embodiment, the existing defect feature image can be obtained according to the qualified wood sample data in a predefined manner, and the existing defect feature image is synthesized to the qualified wood sample image to obtain the defective wood sample data with the same wood type as the qualified wood sample, so as to expand the unqualified wood sample data corresponding to the qualified wood sample and provide a basis for further training the artificial intelligence model.
In an optional implementation manner of this embodiment, as shown in fig. 2, the step S102, namely the step of obtaining the existing defect feature image according to the qualified wood sample data, further includes the following steps S201 to S202:
in step S201, when the number of first existing defective wood sample data in the existing defective wood sample database, which is the same as the wood category of the qualified wood sample data, is less than or equal to a first preset threshold, obtaining second existing defective wood sample data, which is different from the wood category of the qualified wood sample data, from the existing defective wood sample database;
in step S202, the existing defect feature image is extracted from the second existing defect wood sample data.
In this optional implementation manner, the acquired qualified wood sample data may include the wood type of the qualified wood sample, and when the existing defect feature image is acquired, it may be checked first whether the number of the defect wood samples in the existing defect wood sample database, which are the same as the wood type, is not enough, that is, whether the number of the defect wood samples is less than or equal to a first preset threshold value, if so, a defect wood sample different from the wood type is acquired from the existing defect wood sample database, and the defect feature image is extracted from the different defect wood samples, so that the extracted defect feature image is synthesized onto the image of the qualified wood sample. The wood defect types can be various, for example, the wood defect types can be classified into joints, dead knots, decay, cracks, insect pests and the like, so that a second existing defect wood sample data with various wood defect types can be obtained from the existing defect wood sample database, and existing defect feature images of various types can be obtained, so that various defect wood sample data with the same wood type as that of the qualified wood sample but different defect types can be obtained.
In an optional implementation manner of this embodiment, the method for acquiring a wood defect sample further includes:
and in the existing defective wood sample database, when the number of first existing defective wood sample data of which the wood category is the same as that of the qualified wood sample data is greater than a first preset threshold value, taking the first existing defective wood sample data as the defective wood sample data.
In this optional implementation manner, when there are more first existing defective wood sample data of the same kind as the qualified wood sample in the existing defective wood sample database, that is, greater than the first preset threshold, existing defective wood sample data of the same kind as the qualified wood sample may be directly obtained from the existing defective wood sample database, without synthesizing the defective feature images of other kinds of wood.
In an optional implementation manner of this embodiment, the qualified wood sample data includes a qualified wood sample image and label data; synthesizing the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data, wherein the method comprises the following steps:
and covering part or all of the existing defect characteristic image on a preset area of the qualified wood sample image to obtain defect wood sample data with the same wood category as that of the qualified wood sample data.
In this alternative implementation, when synthesizing the image, the image synthesis may be performed by covering part or all of the existing defect feature image in the preset area of the qualified wood sample image. The preset area may be at any position of the qualified wood sample image, such as a middle position, an edge position, and the like. FIG. 3 shows a schematic diagram of an extracted existing defect feature image overlaid on a qualified wood sample image; fig. 4 and 5 are schematic diagrams respectively showing the existing defect feature image partially overlaid on the qualified wood sample image.
In an optional implementation manner of this embodiment, the method for acquiring a wood defect sample further includes: and adjusting the texture features of the boundary region of the synthesized existing defect feature image, so that the difference between the texture features of the boundary region and the texture features of the existing defect feature image and/or the texture features of the qualified wood sample is smaller than a second preset threshold value.
In the optional implementation mode, after the existing defect feature image is synthesized on the qualified wood sample image, the texture feature at the junction of the existing defect feature image and the qualified wood sample image can be adjusted to enable the texture feature at the junction of the existing defect feature image and the qualified wood sample image to be consistent, namely the difference of the texture feature is smaller than a preset threshold value, so that the texture trend on the qualified wood sample image can be naturally transited to the texture trend of the existing defect feature image, and the textures of the existing defect feature image and the qualified wood sample image are organically connected. In the adjusting process, the existing defect characteristic image can be not sacrificed as much as possible, but some small adjustment can be made on the partial image at the boundary of the existing defect characteristic image. The interface region may be a region where the existing defect feature image and the qualified wood sample image interface after the existing defect feature image is partially or completely overlaid on the qualified wood sample image, including but not limited to a portion of the qualified wood sample image, a portion of the existing defect feature image, or a combination thereof. In the operation process, the existing defect feature image can be kept unchanged, the texture features of the part of qualified wood sample images outside the boundary of the existing defect feature image can be adjusted, the qualified wood sample images can also be kept unchanged, the texture features of the part of images inside the boundary of the existing defect feature image can be adjusted, the texture features of the part of images inside the boundary of the existing defect feature image and the texture features of the part of qualified wood sample images outside the boundary of the existing defect feature image can also be adjusted together, the operation can be specifically carried out according to the actual situation, and the operation is not limited herein.
In an optional implementation manner of this embodiment, the step of adjusting the texture feature of the boundary region of the synthesized existing defect feature image further includes the following steps:
and filtering the texture features of the boundary region by using a low-pass filter.
In this alternative implementation, the texture feature may be adjusted by a low-pass filter. For example, the texture features of the qualified wood sample image in the boundary area are greatly different from those of the existing defect feature image, and the difference can generate high-frequency components, so that after passing through a low-pass filter, the high-frequency components can be weakened, and the texture trend differences of the qualified wood sample image and the existing defect feature image tend to be consistent.
In an optional implementation manner of this embodiment, the step of adjusting the texture feature of the boundary region of the synthesized existing defect feature image further includes the following steps:
and carrying out image processing on the boundary area by using a trained artificial intelligence model, so that the image style of the boundary area comes from the existing defect characteristic image, and the image content of the boundary area comes from the qualified wood sample image.
In this optional implementation, an artificial intelligence model, such as a convolutional neural network, may be trained, and the artificial intelligence model may be used to perform style migration on the image of the boundary region, so as to migrate the existing defect feature image to the qualified wood sample image. After training, content representation and style representation in the convolutional neural network can be well separated, the feature response result of a high layer in the convolutional neural network can be used as content representation, and the relation between different features of different layers in the convolutional neural network can be used as style representation. For example, an image is constructed by matching the style representation of the existing defect feature image through two VGG networks, and the average square distance of the gram matrix of the qualified wood sample image and the gram matrix of the generated image is minimized on a white noise image by using a gradient descent algorithm.
FIG. 6 is a schematic diagram of a VGG convolutional neural network in accordance with an embodiment of the present disclosure. As shown in fig. 6, the VGG network has 19 layers, the feature space is provided by 16 convolutional layers and 5 pooling layers,in order to input the defect image,in order to provide a circular envelope at the interface region,an image is generated for the ring envelope that is continuously obtained through gradient descent.
The image information can be more obvious through the artificial intelligence model, and the content of the image and the natural style of the image are separated and then combined through the artificial intelligence model, so that the synthesized image has higher quality and is closer to a real image.
In an optional implementation manner of this embodiment, the step of synthesizing the existing defect feature image and the qualified wood sample data to obtain the defective wood sample data of the same wood category as the qualified wood sample data further includes the following steps:
generating marking data of the sample data of the defective wood; the marking data comprises defect marks and wood categories in the qualified wood sample data.
In this optional implementation, after the existing defect feature image is synthesized onto the qualified wood sample image, the obtained defective wood sample image may be labeled, for example, whether the defective wood sample is defective, the defect type, and the like, and the defect type of the defective wood sample may be labeled as the wood type of the qualified wood sample. By the method, the same kind of defective wood sample data can be obtained from the qualified wood sample data, the defective wood sample is expanded, and the problem that the defective wood sample is difficult to obtain under the actual condition is solved.
In an optional implementation manner of this embodiment, the method further includes the following steps:
training based on the qualified wood sample data and the defect sample data to obtain a wood defect identification model; or,
training based on the qualified wood sample data and the defect sample data to obtain an identification model of the wood defects, and outputting the identification model of the wood defects to a client.
In the optional implementation mode, after the defective wood sample data is obtained, the qualified wood sample data and the defective wood sample data can be respectively used as a positive sample and a negative sample to train the artificial intelligent model, and after the number of the positive sample and the number of the negative sample are accumulated to a certain degree, the wood defect identification model capable of identifying the wood defects can be obtained. If the wood defect identification model is obtained through cloud training, the wood defect identification model can be output to a client, so that the client can use the model to identify wood.
In an optional implementation manner of this embodiment, the method further includes the following steps: and outputting the sample data of the defective wood to a client.
In the optional implementation mode, after the defective wood sample data is obtained, the defective wood sample data can be output to the client, and the client trains the artificial intelligent model by using the defective wood sample data, so that a wood defect identification model with high identification capability is finally obtained.
The method for acquiring the wood defect sample is suitable for the condition that the wood processing plant does not have enough or can not collect enough sample data of the defective wood, and the wood defect sample can be acquired by requesting the cloud system or directly acquiring a wood defect identification model from the cloud system.
The cloud system can collect qualified wood sample data and defective wood sample data from a plurality of different wood processing plant clients, and when a request is made by the client, the client synthesizes the defective wood sample data or directly provides a trained wood defect identification model for the client according to the request of the client. Along with the accumulation of the defective wood samples of the cloud system, the cloud system can balance the number of different wood species in the defective wood sample data, and the situation that the performance of the artificial intelligence model is affected due to the unbalanced distribution of the defect types of the synthesized new defective wood samples is avoided. For example, the wood defect types can be classified into joints, dead knots, decay, cracks, insect pests and the like, and if the sample number of the joints, dead knots and decay types in the cloud database is high and the sample number of the crack and insect pest types is low, the cloud database can directionally accumulate the defective wood samples of the crack and insect pest types.
In the specific application process, the wood factory carries out category marking on the collected qualified wood samples. The wood grain type of the qualified wood sample can be various wood grains such as cross grains, vertical grains, twills, envelope grains and the like, and the wood grain type can also be various wood grains such as camphor wood, birch wood, rosewood, oak and the like, so that the wood grain type label of the qualified wood sample can also include the wood grain type, such as { camphor wood, cross grains }, { camphor wood, vertical grains }, { birch wood, twills }, { birch wood, envelope grains }, { rosewood, vertical grains }, { rosewood, envelope grains } and the like.
In the cloud system, when a newly synthesized defective wood sample is labeled, the defect type in the labeling data can be directly copied and synthesized into the defect type of the used existing defect characteristic image, and the wood type in the labeling data can be changed into the wood type of the qualified wood sample used in the synthesis. For example, after the qualified wood sample of camphor wood and the crack defect wood sample of birch wood are synthesized, the labeled data can be { camphor wood, crack }.
The artificial intelligence model may employ a neural network. There are two ways in which the neural network can be trained. First, the cloud trains a neural network based on all labeled samples. Second, the factory trains a local neural network directly based on all labeled samples.
For the first training mode, the cloud end can also transmit the trained neural network to a wood factory through the network, or only provides wood classification service for the wood factory at the cloud end.
For the second training mode, the factory can also continue to train existing neural networks to extend wood recognition capabilities to new wood species.
In the early stage of neural network training, the neural network cannot work under the conditions that wood samples are few and defective wood samples are very few, and if the neural network waits for a certain number of samples to be accumulated, the neural network cannot perform wood classification work for a long time. Or, even if the neural network is trained to a certain stage, the synthesis of the defective wood samples at the cloud has an obvious effect on improving the performance of the neural network, because the distribution of the defective wood samples in the total samples is unbalanced, the number of the defective wood samples is far smaller than that of the qualified wood samples, and the types of the defective wood samples are single, so that the trained neural network has no sufficient capacity of identifying the defective wood. And through the synthesis of the cloud defective wood sample, the recognition capability of the trained neural network on the defective wood is greatly improved, and the recognition capability is equivalent to that of the qualified wood. For example, if the number of qualified wood samples is 100, wherein the number of horizontal grain 30, vertical grain 25, diagonal grain 25, envelope grain 20, and the number of defective wood samples is 0, the recognition capability of the neural network is: transverse lines 0.9, vertical lines 0.85, diagonal lines 0.85, enveloping lines 0.8 and defects 0, wherein after the defective wood samples are synthesized by a cloud, the number of the defective wood samples is 100, and the number of the defective wood samples is equal to 25, 20 dead knots, 20 decayed stones, 20 cracks and 15 insect pests, so that the recognition capability of the neural network is as follows: 0.9 transverse striation, 0.85 vertical striation, 0.85 diagonal striation, 0.8 enveloping striation, 0.8 movable joint, 0.75 dead joint, 0.75 rotten, 0.75 crack and 0.7 insect pest.
Or, for a neural network with good wood defect recognition capability, when the neural network faces a new wood species, the performance of the neural network is similar to that of the neural network at the initial training stage, and at the moment, the cloud defect wood sample synthesis service has important significance.
With the increasing acquisition quantity of new wood samples, the increasing quantity of the wood-lacking samples, the increasing proportion of the number of the wood-lacking samples to the total samples, the increasing balance of the category distribution of the wood-lacking samples, and the enough number of the locally acquired wood-lacking samples to support the neural network to identify various types of wood defects, the wood factory can not rely on the synthetic service of the cloud-side wood-lacking samples any more.
The following are embodiments of the disclosed apparatus that may be used to perform embodiments of the disclosed methods.
Fig. 7 shows a block diagram of a structure of a wood defect sample acquiring device according to an embodiment of the present disclosure, which may be implemented as part or all of an electronic device by software, hardware, or a combination of both. As shown in fig. 7, the wood defect sample acquiring apparatus includes a first acquiring module 701, a second acquiring module 702, and a synthesizing module 703:
a first obtaining module 701 configured to obtain qualified wood sample data;
a second obtaining module 702, configured to obtain an existing defect feature image according to qualified wood sample data; the existing defect characteristic image is obtained from an existing defect wood sample database;
a synthesizing module 703 configured to synthesize the existing defect feature image and the qualified wood sample data to obtain defect wood sample data that is the same as the wood category of the qualified wood sample data.
In an optional implementation manner of this embodiment, the second obtaining module includes:
a first obtaining sub-module, configured to obtain, in the existing defective wood sample database, a second existing defective wood sample data that is different from the wood category of the qualified wood sample data from the existing defective wood sample database when the number of first existing defective wood sample data that is the same as the wood category of the qualified wood sample data is less than or equal to a first preset threshold;
an extraction submodule configured to extract the existing defect feature image from the second existing defect wood sample data.
In an optional implementation manner of this embodiment, the apparatus further includes:
a third obtaining module, configured to, when the number of first existing defective wood sample data in the existing defective wood sample database, which is the same as the wood category of the qualified wood sample data, is greater than a first preset threshold, take the first existing defective wood sample data as the defective wood sample data.
In an optional implementation manner of this embodiment, the qualified wood sample data includes a qualified wood sample image and label data; the synthesis module comprises:
and the synthesis submodule is configured to cover part or all of the existing defect characteristic image on a preset area of the qualified wood sample image to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data.
In an optional implementation manner of this embodiment, the apparatus further includes:
and the adjusting module is configured to adjust the texture features of the boundary area of the synthesized existing defect feature image, so that the difference between the texture features of the boundary area and the texture features of the existing defect feature image and/or the texture features of the qualified wood sample is smaller than a second preset threshold value.
In an optional implementation manner of this embodiment, the adjusting module includes:
a filtering submodule configured to filter the textural features of the junction region with a low pass filter.
In an optional implementation manner of this embodiment, the adjusting module includes:
and the image processing submodule is configured to perform image processing on the boundary area by using a trained artificial intelligence model, so that the image style of the boundary area comes from the existing defect characteristic image, and the image content of the boundary area comes from the qualified wood sample image.
In an optional implementation manner of this embodiment, the synthesizing module further includes:
the labeling submodule is configured to generate labeling data of the defective wood sample data; the marking data comprises defect marks and wood categories in the qualified wood sample data.
In an optional implementation manner of this embodiment, the apparatus further includes:
a first training module configured to train to obtain a wood defect identification model based on the qualified wood sample data and the defect sample data; or,
and the second training module is configured to train to obtain a wood defect identification model based on the qualified wood sample data and the defect sample data, and output the wood defect identification model to the client.
In an optional implementation manner of this embodiment, the apparatus further includes:
an output module configured to output the defective wood sample data to a client.
The above-mentioned wood defect sample obtaining apparatus corresponds to the wood defect sample obtaining method described in the embodiment shown in fig. 1 and the related parts, and specific details can be referred to the description of the wood defect sample obtaining method, which is not described herein again.
Fig. 8 is a schematic structural diagram of an electronic device suitable for implementing a wood defect sample acquisition method according to an embodiment of the present disclosure.
As shown in fig. 8, the electronic apparatus 800 includes a Central Processing Unit (CPU)801 that can execute various processes in the embodiment shown in fig. 1 described above according to a program stored in a Read Only Memory (ROM)802 or a program loaded from a storage section 808 into a Random Access Memory (RAM) 803. In the RAM803, various programs and data necessary for the operation of the electronic apparatus 800 are also stored. The CPU801, ROM802, and RAM803 are connected to each other via a bus 804. An input/output (I/O) interface 805 is also connected to bus 804.
The following components are connected to the I/O interface 805: an input portion 806 including a keyboard, a mouse, and the like; an output section 807 including a signal such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage portion 808 including a hard disk and the like; and a communication section 809 including a network interface card such as a LAN card, a modem, or the like. The communication section 809 performs communication processing via a network such as the internet. A drive 810 is also connected to the I/O interface 805 as necessary. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 810 as necessary, so that a computer program read out therefrom is mounted on the storage section 808 as necessary.
In particular, according to embodiments of the present disclosure, the method described above with reference to fig. 1 may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program tangibly embodied on a medium readable thereby, the computer program comprising program code for performing the method of fig. 1. In such an embodiment, the computer program may be downloaded and installed from a network via the communication section 809 and/or installed from the removable medium 811.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowcharts or block diagrams may represent a module, a program segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units or modules described in the embodiments of the present disclosure may be implemented by software or hardware. The units or modules described may also be provided in a processor, and the names of the units or modules do not in some cases constitute a limitation of the units or modules themselves.
As another aspect, the present disclosure also provides a computer-readable storage medium, which may be the computer-readable storage medium included in the apparatus in the above-described embodiment; or it may be a separate computer readable storage medium not incorporated into the device. The computer readable storage medium stores one or more programs for use by one or more processors in performing the methods described in the present disclosure.
The foregoing description is only exemplary of the preferred embodiments of the disclosure and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention in the present disclosure is not limited to the specific combination of the above-mentioned features, but also encompasses other embodiments in which any combination of the above-mentioned features or their equivalents is possible without departing from the inventive concept. For example, the above features and (but not limited to) the features disclosed in this disclosure having similar functions are replaced with each other to form the technical solution.
Claims (10)
1. A method for obtaining a wood defect sample is characterized by comprising the following steps:
acquiring qualified wood sample data;
acquiring an existing defect characteristic image according to qualified wood sample data; the existing defect characteristic image is obtained from an existing defect wood sample database;
and synthesizing the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data.
2. The method for obtaining the wood defect sample according to claim 1, wherein obtaining the existing defect feature image according to the qualified wood sample data comprises:
when the number of first existing defective wood sample data of which the wood category is the same as that of the qualified wood sample data in the existing defective wood sample database is less than or equal to a first preset threshold value, obtaining second existing defective wood sample data which is different from that of the qualified wood sample data from the existing defective wood sample database;
and extracting the existing defect characteristic image from the second existing defect wood sample data.
3. The method of claim 1, further comprising:
and in the existing defective wood sample database, when the number of first existing defective wood sample data of which the wood category is the same as that of the qualified wood sample data is greater than a first preset threshold value, taking the first existing defective wood sample data as the defective wood sample data.
4. The method according to claim 1, wherein the qualified wood sample data includes a qualified wood sample image and labeling data; synthesizing the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data, wherein the method comprises the following steps:
and covering part or all of the existing defect characteristic image on a preset area of the qualified wood sample image to obtain defect wood sample data with the same wood category as that of the qualified wood sample data.
5. The method of claim 4, further comprising:
and adjusting the texture features of the boundary region of the synthesized existing defect feature image, so that the difference between the texture features of the boundary region and the texture features of the existing defect feature image and/or the texture features of the qualified wood sample is smaller than a second preset threshold value.
6. The method for obtaining the wood defect sample according to claim 5, wherein the adjusting the texture characteristics of the boundary area of the synthesized existing defect characteristic image comprises:
and filtering the texture features of the boundary region by using a low-pass filter.
7. The method for obtaining a wood defect sample according to claim 5, wherein the adjusting the texture feature of the boundary area of the existing defect feature image on the qualified wood sample image comprises:
and carrying out image processing on the boundary area by using a trained artificial intelligence model, so that the image style of the boundary area comes from the existing defect characteristic image, and the image content of the boundary area comes from the qualified wood sample image.
8. A wood defect sample acquisition device, comprising:
a first obtaining module configured to obtain qualified wood sample data;
the second acquisition module is configured to acquire an existing defect characteristic image according to qualified wood sample data; the existing defect characteristic image is obtained from an existing defect wood sample database;
and the synthesis module is configured to synthesize the existing defect characteristic image and the qualified wood sample data to obtain defect wood sample data which is the same as the wood category of the qualified wood sample data.
9. An electronic device comprising a memory and a processor; wherein,
the memory is configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method steps of any of claims 1-7.
10. A computer-readable storage medium having stored thereon computer instructions, characterized in that the computer instructions, when executed by a processor, carry out the method steps of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810240671.3A CN108447054B (en) | 2018-03-22 | 2018-03-22 | Wood defect sample obtaining method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810240671.3A CN108447054B (en) | 2018-03-22 | 2018-03-22 | Wood defect sample obtaining method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108447054A true CN108447054A (en) | 2018-08-24 |
| CN108447054B CN108447054B (en) | 2021-06-29 |
Family
ID=63196126
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810240671.3A Active CN108447054B (en) | 2018-03-22 | 2018-03-22 | Wood defect sample obtaining method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108447054B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109409414A (en) * | 2018-09-28 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Sample image determines method and apparatus, electronic equipment and storage medium |
| CN110059549A (en) * | 2019-03-11 | 2019-07-26 | 齐鲁工业大学 | A kind of thin wood plate categorizing system and algorithm based on deep learning |
| CN110223277A (en) * | 2019-05-28 | 2019-09-10 | 深圳新视智科技术有限公司 | Method, apparatus, terminal device and the storage medium that image generates |
| CN110675359A (en) * | 2019-06-29 | 2020-01-10 | 创新奇智(南京)科技有限公司 | Defect sample generation method and system for steel coil surface and electronic equipment |
| CN111144244A (en) * | 2019-12-13 | 2020-05-12 | 唐美(北京)文化科技发展有限公司 | Furniture wood type identification method and device |
| CN111681162A (en) * | 2020-06-09 | 2020-09-18 | 创新奇智(成都)科技有限公司 | Defect sample generation method and device, electronic equipment and storage medium |
| CN113674249A (en) * | 2021-08-24 | 2021-11-19 | 重庆忽米网络科技有限公司 | PCB printing quality detection method based on industrial internet |
| CN113807378A (en) * | 2020-06-16 | 2021-12-17 | 纬创资通股份有限公司 | Training data increment method, electronic device and computer readable recording 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 |
| CN115240031A (en) * | 2022-07-21 | 2022-10-25 | 无锡锤头鲨智能科技有限公司 | Method and system for generating plate surface defects based on generation countermeasure network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103258218A (en) * | 2013-05-28 | 2013-08-21 | 清华大学 | Matte detection frame generation method and device and defect detection method and device |
| CN106770321A (en) * | 2016-12-08 | 2017-05-31 | 广东工业大学 | A kind of plastic part defect inspection method interval based on multi thresholds |
| CN107358636A (en) * | 2017-06-16 | 2017-11-17 | 华南理工大学 | A kind of rarefaction defect image generating method based on textures synthesis |
-
2018
- 2018-03-22 CN CN201810240671.3A patent/CN108447054B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103258218A (en) * | 2013-05-28 | 2013-08-21 | 清华大学 | Matte detection frame generation method and device and defect detection method and device |
| CN106770321A (en) * | 2016-12-08 | 2017-05-31 | 广东工业大学 | A kind of plastic part defect inspection method interval based on multi thresholds |
| CN107358636A (en) * | 2017-06-16 | 2017-11-17 | 华南理工大学 | A kind of rarefaction defect image generating method based on textures synthesis |
Non-Patent Citations (1)
| Title |
|---|
| 白雪冰等: "基于Markov随机场的木材表面缺陷图像分割方法", 《林业科技》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109409414A (en) * | 2018-09-28 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Sample image determines method and apparatus, electronic equipment and storage medium |
| CN110059549A (en) * | 2019-03-11 | 2019-07-26 | 齐鲁工业大学 | A kind of thin wood plate categorizing system and algorithm based on deep learning |
| CN110223277A (en) * | 2019-05-28 | 2019-09-10 | 深圳新视智科技术有限公司 | Method, apparatus, terminal device and the storage medium that image generates |
| CN110675359A (en) * | 2019-06-29 | 2020-01-10 | 创新奇智(南京)科技有限公司 | Defect sample generation method and system for steel coil surface and electronic equipment |
| CN111144244A (en) * | 2019-12-13 | 2020-05-12 | 唐美(北京)文化科技发展有限公司 | Furniture wood type identification method and device |
| CN111681162B (en) * | 2020-06-09 | 2023-09-01 | 创新奇智(成都)科技有限公司 | Defect sample generation method and device, electronic equipment and storage medium |
| CN111681162A (en) * | 2020-06-09 | 2020-09-18 | 创新奇智(成都)科技有限公司 | Defect sample generation method and device, electronic equipment and storage medium |
| CN113807378A (en) * | 2020-06-16 | 2021-12-17 | 纬创资通股份有限公司 | Training data increment method, electronic device and computer readable recording medium |
| CN113807378B (en) * | 2020-06-16 | 2024-05-31 | 纬创资通股份有限公司 | Training data increment method, electronic device and computer readable recording medium |
| CN113674249A (en) * | 2021-08-24 | 2021-11-19 | 重庆忽米网络科技有限公司 | PCB printing quality detection method based on industrial internet |
| 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 |
| CN115240031A (en) * | 2022-07-21 | 2022-10-25 | 无锡锤头鲨智能科技有限公司 | Method and system for generating plate surface defects based on generation countermeasure network |
| CN115240031B (en) * | 2022-07-21 | 2023-10-27 | 无锡锤头鲨智能科技有限公司 | Board surface defect generation method and system based on generation countermeasure network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108447054B (en) | 2021-06-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108447054B (en) | Wood defect sample obtaining method and device, electronic equipment and storage medium | |
| US11276162B2 (en) | Surface defect identification method and apparatus | |
| CN112070135B (en) | Power equipment image detection method and device, power equipment and storage medium | |
| EP3843036A1 (en) | Sample labeling method and device, and damage category identification method and device | |
| CN110598784A (en) | Machine learning-based construction waste classification method and device | |
| CN112766110A (en) | Training method of object defect recognition model, object defect recognition method and device | |
| CN111626947A (en) | Map vectorization sample enhancement method and system based on generation of countermeasure network | |
| CN112001901A (en) | Apple defect detection method and system based on convolutional neural network | |
| US10726535B2 (en) | Automatically generating image datasets for use in image recognition and detection | |
| CN109241867B (en) | Method and device for recognizing digital rock core image by adopting artificial intelligence algorithm | |
| CN113256608B (en) | Workpiece defect detection method and device | |
| CN117392042A (en) | Defect detection method, defect detection apparatus, and storage medium | |
| CN108805884A (en) | A kind of mosaic area's detection method, device and equipment | |
| CN113158969A (en) | Apple appearance defect identification system and method | |
| CN114359235A (en) | Wood surface defect detection method based on improved YOLOv5l network | |
| CN117252842A (en) | Aircraft skin defect detection and network model training method | |
| CN113763384B (en) | Defect detection method and defect detection device in industrial quality inspection | |
| CN116309612B (en) | Semiconductor silicon wafer detection method, device and medium based on frequency decoupling supervision | |
| CN109145955A (en) | A kind of Wood Identification Method and system | |
| CN112836724A (en) | Object defect recognition model training method and device, electronic equipment and storage medium | |
| CN116824274A (en) | Small sample fine granularity image classification method and system | |
| CN111507945A (en) | Method for training deep learning defect detection model by using defect-free map | |
| CN116645351A (en) | Online defect detection method and system for complex scene | |
| CN113947567B (en) | Defect detection method based on multitask learning | |
| CN116626059A (en) | Bridge damage detection method and system integrating calculation and report |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |