CN115294109A - Real wood board production defect identification system based on artificial intelligence, and electronic equipment - Google Patents
Real wood board production defect identification system based on artificial intelligence, and electronic equipment Download PDFInfo
- Publication number
- CN115294109A CN115294109A CN202211205044.9A CN202211205044A CN115294109A CN 115294109 A CN115294109 A CN 115294109A CN 202211205044 A CN202211205044 A CN 202211205044A CN 115294109 A CN115294109 A CN 115294109A
- Authority
- CN
- China
- Prior art keywords
- defect
- image
- plate
- free
- defective
- 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 114
- 239000002023 wood Substances 0.000 title claims abstract description 110
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 36
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 29
- 239000007787 solid Substances 0.000 claims abstract description 98
- 230000002950 deficient Effects 0.000 claims abstract description 63
- 238000000034 method Methods 0.000 claims abstract description 45
- 230000006870 function Effects 0.000 claims description 40
- 238000004364 calculation method Methods 0.000 claims description 35
- 238000009826 distribution Methods 0.000 claims description 22
- 230000011218 segmentation Effects 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 14
- 239000000203 mixture Substances 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 8
- 239000000126 substance Substances 0.000 claims description 6
- 238000000638 solvent extraction Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 13
- 238000001914 filtration Methods 0.000 abstract description 6
- 238000011156 evaluation Methods 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013145 classification model Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the field of image recognition, in particular to a solid wood board production recognition system and electronic equipment based on artificial intelligence. The method comprises the following steps: the panel image acquisition module is used for acquiring a panel filtering image; the defect plate detection module is used for acquiring the optimal attribution type of each pixel point; constructing a loss function to divide all pixel points in the panel filtering image, taking the panel with the pixel point category number larger than 1 as a suspected defect panel, calculating the defect degree of the panel, and taking the panel as the defect panel when the defect degree is larger than a first threshold value; and the non-defective plate classification module is used for calculating a first index and a second index of the non-defective plate image classification, calculating a classification identification index of every two non-defective plates, and when the classification identification index is greater than a second threshold value, classifying every two corresponding non-defective plates into the same category. The method can not only detect and identify the defective solid wood boards, but also classify the non-defective solid wood boards, and has small calculated amount and high evaluation precision.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a solid wood board production defect recognition system and electronic equipment based on artificial intelligence.
Background
The detection, identification and sorting of solid wood panels after their production is a crucial step for solid wood panels. When the surface of the solid wood board has defects, the requirements cannot be met during subsequent use, the quality of furniture is reduced, and the types of the solid wood boards in each batch are different due to the difference of the purposes and the coloring process. Therefore, the consistency of the manufacturing process and the appearance quality of the solid wood board can be realized for the detection and the identification of the solid wood board.
At present, the solid wood board is generally detected and identified through manual visual identification, and the defect detection, board classification and the like are carried out on the solid wood board through professional operators.
In order to solve the problems, the invention provides an artificial intelligence-based solid wood board production identification system and electronic equipment.
Disclosure of Invention
The invention provides a solid wood board production identification system and electronic equipment based on artificial intelligence, which aim to solve the existing problems and comprise the following steps: the panel image acquisition module is used for acquiring a panel filtering image; the defect plate detection module is used for acquiring the optimal attribution type of each pixel point; constructing a loss function to divide all pixel points in the panel filtering image, taking the panel with the pixel point category number larger than 1 as a suspected defect panel, calculating the defect degree of the suspected defect panel, and taking the suspected defect panel as a defect panel when the defect degree is larger than a first threshold value; and the non-defective plate classification module is used for calculating a first index and a second index of the non-defective plate image classification, calculating a classification identification index of every two non-defective plates, and when the classification identification index is greater than a second threshold value, classifying every two corresponding non-defective plates into the same category.
According to the technical means provided by the invention, the plate is identified based on the solid wood plate image data, the defective solid wood plate can be detected and identified, the non-defective solid wood plate can be classified through the established plate classification model, so that corresponding reference opinions can be provided for related workers, the solid wood plate is processed and sorted in a targeted manner, and the method has the advantages of small calculated amount, high detection speed, high evaluation precision and the like.
The invention adopts the following technical scheme:
a solid wood board production defect identification system based on artificial intelligence comprises an image acquisition module, an image processing module, a clustering module, a first calculation module, a segmentation module, a defect judgment module, a first index calculation module, a second calculation module and a classification module;
an image acquisition module: the method is used for collecting the surface image of the plate in the plate production process.
An image processing module: and denoising the plate surface image acquired by the image acquisition module to obtain a plate denoising image.
A clustering module: the method is used for clustering all pixel points in the plate denoising image to obtain the best category attribution of each pixel point.
A first calculation module: and constructing a loss function by utilizing the Gaussian function corresponding to the optimal category attribution of each pixel point obtained by the clustering module and the Gaussian functions corresponding to the pixel points in other categories, and calculating the loss index of each pixel point by utilizing the constructed loss function.
A segmentation module: and segmenting the loss index of each pixel point by using the loss index of each pixel point obtained by the first calculation module.
And the defect judging module is used for judging whether the plate image acquired by the acquisition module has defects according to the segmentation result of the segmentation module, and if the area segmented by the segmentation module is 1, the acquired plate is a defect-free plate.
The first index calculation module: the method comprises the steps of obtaining a structural distribution matrix of each non-defective plate image, calculating cosine similarity of each column of vectors in the structural distribution matrix of every two non-defective plate images, and taking the mean value of the cosine similarity of all the column vectors as a first classification index of the non-defective plate images.
The second index calculation module: obtaining component difference images of all the channels of the defect-free plate images R, G and B, calculating the color characteristic parameter difference of each component difference image of every two defect-free plate images, and calculating the classification second index of the defect-free plate images according to the color characteristic parameter difference value of each component difference image of every two defect-free plate images.
A second calculation module: and establishing a plate classification identification model according to the first defect-free plate image classification index and the second defect-free plate image classification index, and calculating the classification identification indexes of every two defect-free plate images.
A classification module: and when the classification identification index obtained in the second calculation module is larger than a preset threshold value, dividing every two corresponding five-defect plates into the same category.
Further, the expression of the loss function is as follows:
wherein, the first and the second end of the pipe are connected with each other,expressing the category corresponding to the maximum Gaussian function value of the ith pixel pointThe best attribution category for the ith pixel point,is the Gaussian function value of the ith pixel point in the kth Gaussian model, and K is highThe total number of the Gaussian models, M is the total number of the pixel points,is a self-defined function.
Further, when the area partitioned by the partitioning module is not 1, the collected plates are suspected-defect plates, the defect degree of the suspected-defect plates is calculated according to the number of suspected-defect pixel points in images of the suspected-defect plates, and when the defect degree of the suspected-defect plates is larger than a defect threshold value, the suspected-defect plates are defect plates.
Further, the method for calculating the image defect degree of the suspected-defect plate comprises the following steps of:
taking the area with the largest number of pixel points in the suspected defect plate image as a normal area, taking the area formed by other pixel points in various categories as a suspected defect area, acquiring the number of the suspected defect pixel points in all the suspected defect areas, and calculating the defect degree of the suspected defect plate image, wherein the expression is as follows:
wherein, the number of suspected defect pixel points in the S suspected defect plate image,the number of the types of the pixel points in the suspected defect plate image,the defect degree of the suspected defect plate image is obtained.
Further, a real wood board production defect identification system based on artificial intelligence, the method for obtaining the structure distribution matrix of each non-defective board image comprises the following steps:
and establishing a Gaussian mixture model corresponding to the image of the defect-free plate according to the Gaussian model function values of all pixel points in each image of the defect-free plate, and establishing a structural distribution matrix corresponding to the image of the defect-free plate according to the Gaussian parameters in the Gaussian mixture model of each image of the defect-free plate.
Further, a real wood board production defect identification system based on artificial intelligence, a method for calculating the color characteristic parameter difference of each component difference image of every two non-defective board images comprises the following steps:
the component differential images of the defect-free plate image are respectivelyA corresponding component difference image;
obtaining the corresponding histogram of each component difference imageTaking the color characteristic parameters as the color characteristic parameters of each component differential image, and calculating the expression of the color characteristic parameter difference of each component differential image of every two defect-free plate images as follows:
wherein, the first and the second end of the pipe are connected with each other,representing the ith plate image without defects and the jth plate image without defectsThe color feature difference of the component difference images,representing images of the ith defect-free panelThe color characteristics of the component differential images,representing images of the jth defect-free panelThe color characteristics of the component differential images,representing the weight of each dimension of the histogram, wherein l represents the dimension;
similarly, the ith non-defective plate image and the jth non-defective plate image are calculatedColor feature difference of component difference imageAndcolor feature difference of component differential images。
Further, the method for calculating the second index of the defect-free plate image classification based on the artificial intelligence solid wood plate production defect identification system comprises the following steps:
wherein, the first and the second end of the pipe are connected with each other,a second indicator of image classification representing the ith defect-free sheet image and the jth defect-free sheet image,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature difference of the component difference images,representing the ith plate image without defects and the jth plate image without defectsThe color feature differences of the component differential images,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature differences of the component differential images,respectively, weight parameters.
Further, the real wood board production defect identification system based on artificial intelligence is characterized in that an expression of the board classification identification model is as follows:
wherein the content of the first and second substances,representing classification identification indexes of the ith non-defective plate image and the jth non-defective plate image,representing a first index of image classification representing the ith defect free slab image and the jth defect free slab image,image representing the ith defect-free plate image and the jth defect-free plate imageThe second index is classified.
An electronic device for identifying defects in production of solid wood boards based on artificial intelligence comprises any one of the systems for identifying defects in production of solid wood boards based on artificial intelligence.
The beneficial effects of the invention are: according to the technical means provided by the invention, the identification of the solid wood board is realized based on the image data of the solid wood board, the detection and identification of the defective solid wood board can be realized, the classification of the non-defective solid wood board can be realized through the established board classification model, so that the corresponding reference opinions can be provided for related workers, the solid wood board is processed and sorted in a targeted manner, and the system has the beneficial effects of small calculated amount, high detection speed, high evaluation precision and the like, so that the system is suitable for information system integration services such as the production field, an artificial intelligence system, an intelligent home system and the like, and can be applied to the development of application software such as computer vision and hearing software, biological feature identification software and the like.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic structural diagram of a system for identifying defects in production of solid wood panels based on artificial intelligence according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a system for identifying defects in the production of solid wood panels based on artificial intelligence according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, a schematic structural diagram of a real wood panel production defect identification system based on artificial intelligence according to an embodiment of the present invention is provided, and the system includes an image acquisition module, an image processing module, a clustering module, a first calculation module, a segmentation module, a defect determination module, a first index calculation module, a second calculation module, and a classification module.
An image acquisition module: the method is used for collecting the surface image of the plate in the plate production process.
An image processing module: and denoising the plate surface image acquired by the image acquisition module to obtain a plate denoising image.
A clustering module: the method is used for clustering all the pixel points in the panel denoising image to obtain the optimal category attribution of each pixel point.
A first calculation module: and constructing a loss function by using the Gaussian function corresponding to the best category attribution of each pixel point obtained by the clustering module and the Gaussian functions corresponding to the pixel points in other categories, and calculating the loss index of each pixel point by using the constructed loss function.
A segmentation module: and segmenting the loss index of each pixel point by using the loss index of each pixel point obtained by the first computing module.
And the defect judging module is used for judging whether the plate image acquired by the acquisition module has defects according to the segmentation result of the segmentation module, and if the area segmented by the segmentation module is 1, the acquired plate is a defect-free plate.
The first index calculation module: and acquiring a structural distribution matrix of each non-defective plate image, calculating the cosine similarity of each column of vectors in the structural distribution matrix of every two non-defective plate images, and taking the mean value of the cosine similarities of all the columns of vectors as a first classification index of the non-defective plate images.
The second index calculation module: obtaining component difference images of all channels R, G and B of the defect-free plate images, calculating the color characteristic parameter difference of each component difference image of every two defect-free plate images, and calculating the classification second index of the defect-free plate images according to the color characteristic parameter difference value of each component difference image of every two defect-free plate images.
A second calculation module: and establishing a plate classification identification model according to the first defect-free plate image classification index and the second defect-free plate image classification index, and calculating the classification identification indexes of every two defect-free plate images.
A classification module: and when the classification identification index obtained in the second calculation module is greater than a preset threshold value, dividing every two five corresponding defective plates into the same category.
As shown in fig. 2, a schematic flow chart of a real wood board production defect identification system based on artificial intelligence according to an embodiment of the present invention is provided, including:
101. an image acquisition module: the method is used for collecting the surface image of the plate in the plate production process.
The method is mainly used for realizing detection and identification of the solid wood board based on image data and an artificial intelligence mode, therefore, the method is provided with image acquisition equipment for acquiring the image data of the solid wood board, wherein an implementer of the shooting range and the angle of a camera adjusts the image acquisition equipment according to the actual situation. According to the invention, the camera is arranged right above the solid wood board to collect the front-view image of the board, so that the detection and identification precision of the system is improved.
102. An image processing module: and denoising the plate surface image acquired by the image acquisition module to obtain a plate denoising image.
The method adopts the self-adaptive median filtering algorithm to carry out denoising processing on the solid wood board image, eliminates noise data in the image and improves the image quality. Therefore, the filtered image data of the solid wood board can be obtained according to the method provided by the invention and used for identifying and detecting the solid wood board.
103. A clustering module: the method is used for clustering all pixel points in the plate denoising image to obtain the best category attribution of each pixel point.
For the obtained solid wood board filtering image, the solid wood board pixel points are subjected to clustering analysis based on a clustering algorithm, the solid wood board image data are clustered by adopting a mean clustering algorithm to obtain a plurality of clustering categories, the preliminary classification of the solid wood board pixel points is realized, and the solid wood board pixel points are marked as K category numbers for the subsequent detailed classification of the solid wood board pixel points.
104. A first calculation module: and constructing a loss function by using the Gaussian function corresponding to the best category attribution of each pixel point obtained by the clustering module and the Gaussian functions corresponding to the pixel points in other categories, and calculating the loss index of each pixel point by using the constructed loss function.
Establishing corresponding Gaussian models based on gray values of pixels in the cluster categories through Gaussian distribution functions, wherein each cluster category corresponds to one Gaussian modelAnd obtaining K Gaussian models for classifying the solid wood board pixel points. For solid wood plate pixel points, the invention brings the solid wood plate pixel points into each single Gaussian modelAcquiring corresponding Gaussian function values, and forming each acquired Gaussian function value into a 1-dimensional class vector:
wherein, the first and the second end of the pipe are connected with each other,representing the category vector corresponding to the pixel point i, wherein each element of the category vector is a category,the probability value (0, 1) representing the attribute of the pixel point i to the category k]) The higher the probability value, the higher the probability that the pixel belongs to the category k.
After the solid wood board pixel point i is subjected to category vector acquisition, the methodThe invention relates to a pixel pointClass with maximum probability valueAs a pixel pointBest attribution category of. According to the method, the optimal attribution category of each pixel point can be obtainedAnd M is the total number of the solid wood board pixel points.
In order to ensure that the classification of all the pixel points is more accurate and improve the identification precision of the pixel points, the invention constructs a loss function for supervising the classification process of the pixel points of the solid wood board.
The expression of the loss function is as follows:
wherein, the first and the second end of the pipe are connected with each other,expressing the category corresponding to the maximum Gaussian function value of the ith pixel pointIs the best attribution category of the ith pixel point,is the Gaussian function value of the ith pixel point in the kth Gaussian model, K is the total number of the Gaussian models, M is the total number of the pixel points,is a custom function, wherein。
According to the constructed loss function, the smaller the loss function is, the maximum probability of the pixel point on the best attribution category is achieved, the minimum probability of other categories is guaranteed, and the better category division effect of the pixel point of the solid wood board is achieved.
105. A segmentation module: and segmenting the loss index of each pixel point by using the loss index of each pixel point obtained by the first calculation module.
According to the loss function established by the invention, the optimal algorithm is further combined to search the corresponding pixel point division result when the loss function is minimum, so as to obtain the optimal division result of the solid wood plate pixel points, the optimal algorithm comprises a plurality of genetic algorithms, simulated annealing algorithms, gradient descent methods and the like, and the method is not limited by the invention.
106. And the defect judging module is used for judging whether the plate acquired by the image acquisition module has defects according to the result of the segmentation module, and if the area segmented by the segmentation module is 1, the acquired plate is a defect-free plate.
For the number N of pixel point categories acquired after the optimal division, the method comprises the following steps: when the temperature is higher than the set temperatureWhen the wood board is in use, the surface of the solid wood board is determined to be free of defects, and the corresponding solid wood board is a non-defective board; otherwise, the solid wood board is considered as a suspected defect board, and the defect condition of the solid wood board needs to be analyzed so as to obtain the defect degree of the solid wood board.
For the suspected-defect plate, based on the classification of the pixel points, connected domains corresponding to all the classes of the solid-wood plate can be obtained, and the defects of the solid-wood plate are generally small regions with different defect types.
The method for calculating the image defect degree of the suspected defect plate comprises the following steps:
taking the area with the largest number of pixel points in the suspected defect plate image as a normal area, taking the area formed by other pixel points in various categories as a suspected defect area, acquiring the number of the suspected defect pixel points in all the suspected defect areas, and calculating the defect degree of the suspected defect plate image, wherein the expression is as follows:
wherein the number of suspected defect pixel points in the S suspected defect plate image,the number of the categories of the pixel points in the suspected defect plate image,the defect degree of the suspected defect plate image is shown.
The defect degree value range is [0,1], the larger the function value is, the higher the defect degree is, and in consideration of the fact that inherent material noise points may exist on the surface of the solid wood board in the actual production process, and a certain fault tolerance rate exists before the solid wood board leaves a factory in the actual situation, therefore, the defect degree first threshold value T (0.35) is set, when the defect degree of the suspected defect board is higher than the first threshold value T, the solid wood board is considered to be the defect board, otherwise, the board is considered to be the normal solid wood board meeting the subsequent use conditions, and the board is divided into the defect-free boards.
The first index calculation module: the method comprises the steps of obtaining a structural distribution matrix of each non-defective plate image, calculating cosine similarity of each column of vectors in the structural distribution matrix of every two non-defective plate images, and taking the mean value of the cosine similarity of all the column vectors as a first classification index of the non-defective plate images.
The method for acquiring the structure distribution matrix of each defect-free plate image comprises the following steps:
and establishing a Gaussian mixture model corresponding to the image of the defect-free plate according to the Gaussian model function values of all pixel points in each image of the defect-free plate, and obtaining a structural distribution matrix corresponding to the image of the defect-free plate according to the Gaussian parameters in the Gaussian mixture model of each image of the defect-free plate.
For the image data of the surface of the non-defective solid wood board, a Gaussian mixture model is established for representing the distribution condition of the image data of the surface of the solid wood board, and the Gaussian mixture model specifically comprises the following steps:
in the formula (I), the compound is shown in the specification,and calculating the mixed weight of the Gaussian mixture model by using an EM (effective velocity) algorithm to obtain a more accurate solid wood board Gaussian mixture model.
Each defect-free solid wood board corresponds to one Gaussian mixture model, each single Gaussian model of the Gaussian mixture model has three Gaussian parameters, and each Gaussian mixture model comprisesThe invention constructs a solid wood floor structure distribution matrix based on the Gaussian parametersThe structural distribution matrix may be true to true. The method is characterized in that the distribution condition of the texture structure on the surface of the wood board is used for classifying the defect-free solid wood board.
For any two defect-free solid wood boards, the invention calculates the distance between each column vector in the corresponding two structural distribution matrixesCosine similarity, asThe cosine similarity between corresponding a-th row column vectors in a structural distribution matrix of the non-defective plate i and the non-defective plate j is represented, the cosine similarity of the A-row elements is obtained, and the mean value of the cosine similarity of the A-row elements is used as a first index between the non-defective plate i and the non-defective plate jThe higher the first index is, the more similar the types corresponding to the two defect-free solid wood panels are.
108. The second index calculation module: obtaining component difference images of all the channels of the defect-free plate images R, G and B, calculating the color characteristic parameter difference of each component difference image of every two defect-free plate images, and calculating the classification second index of the defect-free plate images according to the color characteristic parameter difference value of each component difference image of every two defect-free plate images.
Further, in order to improve the accuracy of identifying and dividing the type of the solid wood board, the solid wood board is classified based on the color characteristic parameters of the surface of the board, and the second index of the image classification of the solid wood board is extracted.
For a defect-free solid wood plate, the method acquires component images of R, G and B channels, accurately detects the color distribution condition of the surface of the solid wood plate, extracts more accurate color characteristic parameters, and further acquiresAnd the corresponding component difference image data is used for detecting and identifying the color distribution condition of the surface of the solid wood board.
Obtaining component difference images、、Then, the invention obtains the histogram corresponding to each component difference image based on the component difference value of the pixel point, the dimension of the histogram is 255 and is marked asAnd taking the color parameters as the color characteristic parameters of the solid wood board. It should be noted that the histogram obtaining method and process are well known in the art, and the invention is not described in detail. And acquiring each histogram corresponding to the component difference image of each non-defective solid wood panel according to the method, and classifying each non-defective solid wood panel.
The method for calculating the color characteristic parameter difference of each component difference image of every two defect-free plate images comprises the following steps:
the component differential images of the defect-free plate image are respectivelyA corresponding component difference image;
obtaining the corresponding histogram of each component difference imageThe color characteristic parameters of the difference images of the components are used as the color characteristic parameters of the difference images of the components, and the similarity degree of the color characteristic parameters of any two flawless solid wood boards is obtained, so that the difference of the color characteristic parameters of the two flawless solid wood boards is calculated for the difference images of the components respectively to obtain the color characteristic parameters of the flawless solid wood boardsTaking the component difference image as an example, the expression for calculating the color characteristic parameter difference of the component difference image of every two defect-free plate images is as follows:
wherein the content of the first and second substances,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature difference of the component difference images,representing images of the ith defect-free panelThe color characteristics of the component differential images,representing images of the jth defect-free panelThe color characteristics of the component differential images,representing the weight of each dimension of the histogram, wherein l represents the dimension;
similarly, the ith defect-free plate image and the jth defect-free plate image are calculatedColor feature difference of component difference imageAndcolor feature difference of component difference image。
The method for calculating the second index of the defect-free plate image classification comprises the following steps:
wherein the content of the first and second substances,a second indicator of image classification representing the ith defect-free sheet image and the jth defect-free sheet image,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature differences of the component differential images,representing the ith plate image without defects and the jth plate image without defectsThe color feature difference of the component difference images,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature differences of the component differential images,respectively, the weight parameters are set asThe larger the second index value is, the two defect-free solid wood boards are corresponding toThe higher the degree of similarity of i and j.
109. A second calculation module: and establishing a plate classification identification model according to the first and second defect-free plate image classification indexes, and calculating the classification identification indexes of every two defect-free plate images.
The method carries out classification on the solid wood board based on the extracted board characteristic indexes, and establishes a board type identification model
The expression of the classification recognition model is as follows:
wherein, the first and the second end of the pipe are connected with each other,representing classification identification indexes of the ith non-defective plate image and the jth non-defective plate image,representing a first index of image classification representing the ith defect-free slab image and the jth defect-free slab image,and the second index of the image classification of the ith non-defective plate image and the jth non-defective plate image is represented.
110. A classification module: and when the classification identification index obtained in the second calculation module is larger than a preset threshold value, dividing every two corresponding five-defect plates into the same category.
And normalizing the obtained classification identification indexes to ensure that the values are in [0,1]. For any two solid wood boards, the method can obtain the identification indexes of the corresponding classification identification models, and when the identification indexes are higher than the threshold value (set to be 0.85 by the method), the method classifies the two solid wood boards into one category.
The specific classification process is as follows:
firstly, judging defects of the collected templates, calculating the defect degree of the plate to be detected according to the steps in the defect judging module, and dividing the plate to be detected into defective plates when the defect degree of the plate to be detected is greater than a threshold value;
further, when the plate to be detected is detected to be a non-defective plate, feature extraction is carried out on the non-defective plate, the non-defective plate is taken as a category, the plate to be detected is collected continuously, the next non-defective plate is obtained and feature extraction is carried out, classification indexes of the two non-defective plates are extracted through the first index calculation module and the second index calculation module, classification identification indexes of the two non-defective plates are calculated through the second calculation module, and when the identification indexes are higher than a threshold value, the two non-defective plates are divided into a category;
continuously collecting the plates to be detected, judging whether the plates are defective plates or not, further comparing the plates with any one of the plates in the previous category, and calculating an identification index, wherein if the identification index is not higher than a threshold value, the plates are independently used as one category;
continuously collecting the plates to be detected, comparing the next five-defect plate with any one of the two previous categories respectively, and further classifying the plate into a first category or a second category, or the plate is singly in one category;
the collecting, judging and comparing processes are carried out in sequence, all the plates to be detected in the production process are compared, so that the classification of all the non-defective solid wood plates to be detected can be realized, and the detection and identification of the solid wood plates are realized.
The invention also comprises an electronic device for identifying the production defects of the solid wood boards based on the artificial intelligence, which comprises any one of the systems for identifying the production defects of the solid wood boards based on the artificial intelligence.
According to the technical means provided by the invention, the identification of the solid wood board is realized based on the image data of the solid wood board, the detection and identification of the defective solid wood board can be realized, the classification of the non-defective solid wood board can be realized through the established board classification model, so that the corresponding reference opinions can be provided for related workers, the solid wood board is processed and sorted in a targeted manner, and the system has the beneficial effects of small calculated amount, high detection speed, high evaluation precision and the like, so that the system is suitable for information system integration services such as the production field, an artificial intelligence system, an intelligent home system and the like, and can be applied to the development of application software such as computer vision and hearing software, biological feature identification software and the like.
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, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (9)
1. A solid wood board production defect identification system based on artificial intelligence is characterized by comprising an image acquisition module, an image processing module, a clustering module, a first calculation module, a segmentation module, a defect judgment module, a first index calculation module, a second calculation module and a classification module;
an image acquisition module: the method is used for collecting the surface image of the plate in the plate production process;
an image processing module: denoising the plate surface image acquired by the image acquisition module to obtain a plate denoising image;
a clustering module: the method comprises the steps of clustering all pixel points in a panel denoising image to obtain the optimal category attribution of each pixel point;
a first calculation module: constructing a loss function by utilizing the Gaussian function corresponding to the best category attribution of each pixel point obtained by the clustering module and the Gaussian functions corresponding to the pixel points in other categories, and calculating the loss index of each pixel point by utilizing the constructed loss function;
a segmentation module: dividing the loss index of each pixel point by using the loss index of each pixel point obtained by the first calculation module;
a defect judging module: judging whether the plate image acquired by the acquisition module has defects according to the segmentation result of the segmentation module, if the area segmented by the segmentation module is 1, the acquired plate is a non-defective plate;
the first index calculation module: acquiring a structural distribution matrix of each non-defective plate image, calculating the cosine similarity of each column of vectors in the structural distribution matrix of every two non-defective plate images, and taking the mean value of the cosine similarities of all the columns of vectors as a first classification index of the non-defective plate images;
the second index calculation module: acquiring component difference images of all channels R, G and B of the defect-free plate images, calculating the color characteristic parameter difference of each component difference image of every two defect-free plate images, and calculating a classification second index of the defect-free plate images according to the color characteristic parameter difference value of each component difference image of every two defect-free plate images;
a second calculation module: establishing a plate classification identification model according to the first index and the second index of the defect-free plate image classification, and calculating the classification identification indexes of two defect-free plate images;
a classification module: and when the classification identification index obtained in the second calculation module is larger than a preset threshold value, dividing every two corresponding five-defect plates into the same category.
2. The system of claim 1, wherein the loss function is expressed as follows:
wherein, the first and the second end of the pipe are connected with each other,expressing the category corresponding to the maximum Gaussian function value of the ith pixel pointThe best attribution category for the ith pixel point,for the ith pixel point at the kthThe Gaussian function value in the Gaussian model, K is the total number of the Gaussian models, M is the total number of the pixel points,is a self-defined function.
3. The system of claim 1, wherein when the area partitioned by the partitioning module is not 1, the collected board is a suspected-defective board, the defect degree of the board is calculated according to the number of suspected-defective pixel points in the image of the suspected-defective board, and when the defect degree of the suspected-defective board is greater than the defect threshold, the suspected-defective board is a defective board.
4. The system of claim 3, wherein the method of calculating the degree of the suspected-defect image defect of the solid wood panel comprises:
taking the area with the largest number of pixel points in the suspected defect plate image as a normal area, taking the area formed by other pixel points in various categories as a suspected defect area, acquiring the number of the suspected defect pixel points in all the suspected defect areas, and calculating the defect degree of the suspected defect plate image, wherein the expression is as follows:
5. The artificial intelligence-based solid wood panel production defect identification system of claim 1, wherein the method for obtaining the structural distribution matrix of each defect-free panel image comprises:
and establishing a Gaussian mixture model corresponding to the image of the defect-free plate according to the Gaussian model function values of all pixel points in each image of the defect-free plate, and establishing a structural distribution matrix corresponding to the image of the defect-free plate according to the Gaussian parameters in the Gaussian mixture model of each image of the defect-free plate.
6. The system for identifying defects in production of solid wood panels based on artificial intelligence as claimed in claim 1, wherein the method for calculating the difference of color characteristic parameters of each component differential image of every two defect-free panel images comprises:
the component differential images of the defect-free plate image are respectivelyA corresponding component difference image;
obtaining the corresponding histogram of each component difference imageTaking the color characteristic parameters as the color characteristic parameters of each component differential image, and calculating the expression of the color characteristic parameter difference of each component differential image of every two defect-free plate images as follows:
wherein the content of the first and second substances,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature difference of the component difference images,representing images of the ith defect-free panelThe color characteristics of the component differential images,representing images of the jth defect-free panelThe color characteristics of the component differential images,representing the weight of each dimension of the histogram, wherein l represents the dimension;
7. The artificial intelligence based solid wood panel production defect identification system of claim 6, wherein the method for calculating the second index of defect-free panel image classification comprises:
wherein the content of the first and second substances,a second indicator of image classification representing the ith defect-free sheet image and the jth defect-free sheet image,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature difference of the component difference images,representing the ith defect-free plate image and the jth defect-free plate imageThe color feature difference of the component difference images,representing the ith plate image without defects and the jth plate image without defectsThe color feature difference of the component difference images,respectively, are weight parameters.
8. The system for identifying defects in production of solid wood panels based on artificial intelligence as claimed in claim 1, wherein the expression of the panel classification identification model is:
wherein the content of the first and second substances,representing classification identification indexes of the ith non-defective plate image and the jth non-defective plate image,representing a first index of image classification representing the ith defect-free slab image and the jth defect-free slab image,and the second index of the image classification of the ith non-defective plate image and the jth non-defective plate image is represented.
9. The utility model provides a real wood panel production defect discernment electronic equipment based on artificial intelligence which characterized in that includes: the artificial intelligence based solid wood panel production defect identification system of any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211205044.9A CN115294109A (en) | 2022-09-30 | 2022-09-30 | Real wood board production defect identification system based on artificial intelligence, and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211205044.9A CN115294109A (en) | 2022-09-30 | 2022-09-30 | Real wood board production defect identification system based on artificial intelligence, and electronic equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115294109A true CN115294109A (en) | 2022-11-04 |
Family
ID=83834191
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211205044.9A Pending CN115294109A (en) | 2022-09-30 | 2022-09-30 | Real wood board production defect identification system based on artificial intelligence, and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115294109A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116452587A (en) * | 2023-06-15 | 2023-07-18 | 山东兴华钢结构有限公司 | Environment-friendly building structure steel plate defect identification method |
CN117173444A (en) * | 2023-06-08 | 2023-12-05 | 南京林业大学 | Edge banding board appearance defect detection method and system based on improved YOLOv7 network model |
-
2022
- 2022-09-30 CN CN202211205044.9A patent/CN115294109A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117173444A (en) * | 2023-06-08 | 2023-12-05 | 南京林业大学 | Edge banding board appearance defect detection method and system based on improved YOLOv7 network model |
CN116452587A (en) * | 2023-06-15 | 2023-07-18 | 山东兴华钢结构有限公司 | Environment-friendly building structure steel plate defect identification method |
CN116452587B (en) * | 2023-06-15 | 2023-08-18 | 山东兴华钢结构有限公司 | Environment-friendly building structure steel plate defect identification method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113989279B (en) | Plastic film quality detection method based on artificial intelligence and image processing | |
CN115082683B (en) | Injection molding defect detection method based on image processing | |
CN108537215B (en) | Flame detection method based on image target detection | |
CN109978822B (en) | Banana maturity judging modeling method and judging method based on machine vision | |
CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
CN115294109A (en) | Real wood board production defect identification system based on artificial intelligence, and electronic equipment | |
CN114549522A (en) | Textile quality detection method based on target detection | |
CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
CN116205919B (en) | Hardware part production quality detection method and system based on artificial intelligence | |
CN108181316B (en) | Bamboo strip defect detection method based on machine vision | |
CN114627125B (en) | Stainless steel tablet press surface quality evaluation method based on optical means | |
CN113724231B (en) | Industrial defect detection method based on semantic segmentation and target detection fusion model | |
CN111815564B (en) | Method and device for detecting silk ingots and silk ingot sorting system | |
CN103824059A (en) | Facial expression recognition method based on video image sequence | |
CN111127499A (en) | Security inspection image cutter detection segmentation method based on semantic contour information | |
CN114972356B (en) | Plastic product surface defect detection and identification method and system | |
CN110400293B (en) | No-reference image quality evaluation method based on deep forest classification | |
CN114140669B (en) | Welding defect recognition model training method and device and computer terminal | |
CN113935666B (en) | Building decoration wall tile abnormity evaluation method based on image processing | |
CN111914902B (en) | Traditional Chinese medicine identification and surface defect detection method based on deep neural network | |
CN113838054A (en) | Mechanical part surface damage detection method based on artificial intelligence | |
CN115994907B (en) | Intelligent processing system and method for comprehensive information of food detection mechanism | |
CN111161237A (en) | Fruit and vegetable surface quality detection method, storage medium and sorting device thereof | |
CN114463843A (en) | Multi-feature fusion fish abnormal behavior detection method based on deep learning | |
CN113989196A (en) | Vision-based earphone silica gel gasket appearance defect detection method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20230607 Address after: No. 1805, Building 17, Yihe International City (Phase I), Changjiang Road Economic Development Zone, Kuancheng District, Changchun City, 130000 Jilin Province Applicant after: Jilin Junyu Technology Co.,Ltd. Address before: No. 1866, Building 1, Group 3, Sunliqiao Village, Xingdong Town, Tongzhou District, Nantong City, Jiangsu Province, 226000 Applicant before: Nantong Hengxiang Wood Industry Co.,Ltd. |
|
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221104 |