CN113362276A - Visual detection method and system for plate - Google Patents
Visual detection method and system for plate Download PDFInfo
- Publication number
- CN113362276A CN113362276A CN202110454981.7A CN202110454981A CN113362276A CN 113362276 A CN113362276 A CN 113362276A CN 202110454981 A CN202110454981 A CN 202110454981A CN 113362276 A CN113362276 A CN 113362276A
- Authority
- CN
- China
- Prior art keywords
- color image
- defect
- dimensional
- color
- model
- 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
- 230000000007 visual effect Effects 0.000 title claims abstract description 17
- 238000001514 detection method Methods 0.000 title abstract description 31
- 230000007547 defect Effects 0.000 claims abstract description 225
- 238000000034 method Methods 0.000 claims abstract description 43
- 238000013145 classification model Methods 0.000 claims abstract description 35
- 238000011179 visual inspection Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims description 46
- 238000012545 processing Methods 0.000 claims description 34
- 239000000463 material Substances 0.000 claims description 31
- 230000003321 amplification Effects 0.000 claims description 22
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 22
- 238000005452 bending Methods 0.000 claims description 21
- 238000007781 pre-processing Methods 0.000 claims description 15
- 238000001914 filtration Methods 0.000 claims description 13
- 230000035929 gnawing Effects 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 10
- 238000009499 grossing Methods 0.000 claims description 10
- 239000011347 resin Substances 0.000 claims description 10
- 229920005989 resin Polymers 0.000 claims description 10
- 241000238631 Hexapoda Species 0.000 claims description 9
- 238000005520 cutting process Methods 0.000 claims description 8
- 238000009826 distribution Methods 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 8
- 239000003086 colorant Substances 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 6
- 239000010985 leather Substances 0.000 claims description 5
- 239000007787 solid Substances 0.000 claims description 5
- 239000002023 wood Substances 0.000 claims description 5
- 230000007246 mechanism Effects 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 13
- 230000006872 improvement Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000003384 imaging method Methods 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000005259 measurement Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 241000894006 Bacteria Species 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000005498 polishing Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
- 238000012360 testing method Methods 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
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The invention discloses a visual inspection method for a plate, which comprises the following steps: acquiring three-dimensional point cloud information of a target plate acquired by a 3D contourgraph, constructing a three-dimensional model according to the three-dimensional point cloud information, and identifying the size defect of the target plate according to the three-dimensional model and a preset reference size; the method comprises the steps of obtaining two-dimensional color image information of a target plate collected by a linear array camera, inputting the two-dimensional color image information into a defect recognition model trained in advance to recognize natural defects of the target plate, and inputting the two-dimensional color image information into a color classification model trained in advance to recognize color categories of the target plate. The invention also discloses a visual detection system for the plate. By adopting the invention, the defects of the plate can be comprehensively and accurately detected.
Description
Technical Field
The invention relates to the technical field of solid wood detection, in particular to a visual detection method and a visual detection system for a plate.
Background
In the current solid wood internal picking process, a manual detection mode is still adopted, the whole production line needs manual carrying and visual detection of plates, the operation intensity of workers is generally high, and a large amount of human resources need to be consumed in the picking process; as is well known, workers judge the quality standard more subjectively and have the defects of randomness, large error and the like, so that the quality standard of the solid wood picking plate cannot be ensured; in addition, in the whole production process of the solid wood, a plurality of working procedures are required to be set for manual detection, so that the detection efficiency of the whole production quality is low, and the cost is high; in addition, the requirement of the plate detection on the experience of workers is high, people with relevant skilled experience can quickly detect the defect points of the plate to perform adjustment and treatment, and the culture period of the skilled workers is too long. Therefore, manual detection is the most primitive detection mode which is out of being eliminated.
Accordingly, manufacturers are currently striving to automate measurement equipment, image pre-processing, algorithms, and attempt to address floor measurement and quality inspection issues. At present, some enterprises begin to adopt image measurement mode to identify defect detection, but the problems of high time delay, poor performance and the like generally exist.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a visual inspection method and system for a plate, which can perform comprehensive and accurate inspection on the defects of the plate.
In order to solve the technical problem, the invention provides a visual inspection method for a sheet material, which comprises the following steps: acquiring three-dimensional point cloud information of a target plate acquired by a 3D contourgraph, constructing a three-dimensional model according to the three-dimensional point cloud information, and identifying the size defect of the target plate according to the three-dimensional model and a preset reference size; the method comprises the steps of obtaining two-dimensional color image information of a target plate collected by a linear array camera, inputting the two-dimensional color image information into a defect recognition model trained in advance to recognize natural defects of the target plate, and inputting the two-dimensional color image information into a color classification model trained in advance to recognize color categories of the target plate.
As an improvement of the above scheme, the visual inspection method for a sheet material further includes training a defect recognition model, specifically including: acquiring an original color image; intercepting and merging the original color image to form a basic color image; marking defect positions and defect types in the basic color image to form a marked color image, wherein the defect types comprise dead joints, resin bags, leather inclusion and decay; carrying out generalized amplification processing on the labeled color image to form a sample color image; integrating all sample color images into a sample color image set; and training the defect identification model according to the sample color image set.
As an improvement of the scheme, the defect identification model adopts a DNN model to perform multi-scale feature extraction on an input sample color image, combines semantic information and position information to fuse the multi-scale features, and combines an attention mechanism and a space pyramid structure to extract features so as to identify and output defect types, defect positions and confidence information.
As an improvement of the above scheme, the visual inspection method for a sheet further includes training a color classification model, specifically including: acquiring an original color image; extracting a target foreground region of the original color image through a visual algorithm to form an RO I color image; zooming the target foreground region of the RO I color image and cutting the target foreground region into a plurality of local images; marking color categories in the local images to form marked classified images, wherein the color categories comprise blue change, deep color, medium color and light color; carrying out generalized amplification processing on the labeled classified image to form a sample classified image; integrating all sample classification images into a sample classification image set; training the color classification model according to the sample classification image set.
As an improvement of the scheme, the color classification model adopts a DNN model to carry out multi-scale feature extraction on the input sample classification image, classifies high-dimensional feature vectors, and then identifies image semantic classification information to output color classification and confidence information.
As an improvement of the above scheme, before the two-dimensional color image information is input into a defect recognition model and a color classification model trained in advance, the two-dimensional color image information is preprocessed, and the preprocessing step includes: carrying out SVD on the two-dimensional color image information; carrying out Gabor filtering on the two-dimensional color image information subjected to SVD; and performing difference shadow operation on the two-dimensional color image information after Gabor filtering by adopting an iterative difference shadow method.
As an improvement of the scheme, the size defect comprises any one or more of a length, width and height defect, a curvature defect, a torsion defect, a right-angle defect at two ends, a plane-collapse and head-gnawing defect, a tile-shaped defect and a worm defect; when the length, the width and the height defects are identified, the length of the target plate is the length of the three-dimensional model, the width of the target plate is the width of the three-dimensional model, and the thickness of the target plate is the thickness of the three-dimensional model; when the curvature defect is identified, extracting height distribution along the length direction of the three-dimensional model, selecting the maximum height value as the chord height, and dividing the chord height by the length to obtain the curvature in the length direction; when the torsion resistance defect is identified, calculating the distance from any one angle on one surface of the three-dimensional model to the other three angle planes, and taking the maximum distance value as the torsion resistance; when the two-end right-angle defects are identified, calculating an included angle between planes in the three-dimensional model, and taking the included angle as the two-end right-angle; when the plane-collapse gnawing defect is identified, processing the three-dimensional model by adopting a surface smoothing algorithm, subtracting the smoothed data from the original data to obtain a plane-collapse gnawing area, or fitting an upper surface equation through upper surface three-dimensional point cloud information, fitting a lower surface equation through lower surface three-dimensional point cloud information, calculating the distance from each point to the upper surface equation and the lower surface equation, and drawing a gray level map to obtain the plane-collapse gnawing area; when the tile-shaped defects are identified, fitting a circular equation from the inner side and the outer side of the three-dimensional model, calculating the radius of a circle, and determining the tile-shaped defects according to the radius; and when the insect defect is identified, processing the three-dimensional model by adopting a surface smoothing algorithm, and subtracting the smoothed data from the original data to obtain an insect area.
Correspondingly, the invention also provides a visual inspection system for a sheet material, which comprises: the size defect identification module is used for acquiring three-dimensional point cloud information of a target plate acquired by the 3D contourgraph, constructing a three-dimensional model according to the three-dimensional point cloud information, and identifying the size defect of the target plate according to the three-dimensional model and a preset reference size; the natural defect identification module is used for acquiring two-dimensional color image information of a target plate acquired by the linear array camera and inputting the two-dimensional color image information into a defect identification model trained in advance to identify the natural defect of the target plate; and the color classification and identification module is used for acquiring two-dimensional color image information of the target plate acquired by the linear array camera and inputting the two-dimensional color image information into a color classification model trained in advance to identify the color class of the target plate.
As an improvement of the above scheme, the visual inspection system for a sheet material further comprises a defect recognition training module, and the defect recognition training module comprises: a first acquisition unit configured to acquire an original color image; the merging unit is used for intercepting and merging the original color image to form a basic color image; the first marking unit is used for marking defect positions and defect types in the basic color image so as to form a marked color image, wherein the defect types comprise dead joints, resin bags, leather and decay; the first amplification unit is used for carrying out generalized amplification processing on the labeled color image so as to form a sample color image; a first integration unit for integrating all sample color images into a sample color image set; and the first training unit is used for training the defect identification model according to the sample color image set.
As an improvement of the above scheme, the visual inspection system for a sheet material further includes a color classification training module, and the color classification training module includes: a second acquisition unit for acquiring an original color image; the extracting unit is used for extracting a target foreground region of the original color image through a visual algorithm to form an ROI color image; the cutting unit is used for carrying out scaling processing on a target foreground region of the RO I color image and cutting the target foreground region into a plurality of local images; the second labeling unit is used for labeling color categories in the local image to form a labeled classified image, wherein the color categories comprise blue, deep, medium and light colors; the second amplification unit is used for carrying out generalized amplification processing on the labeled classified image to form a sample classified image; the second integration unit is used for integrating all the sample classification images into a sample classification image set; and the second training unit is used for training the color classification model according to the sample classification image set.
The implementation of the invention has the following beneficial effects:
the invention adopts a three-dimensional laser acquisition technology, and the three-dimensional point cloud information of the plate is acquired by using the 3D contourgraph, so that the accuracy is high; meanwhile, the invention also adopts a three-dimensional imaging technology to integrate and display the collected three-dimensional point cloud information to form a three-dimensional model, and the three-dimensional model is analyzed to identify the size defect of the target plate; the invention also adopts a defect detection technology based on deep learning, and continuously optimizes the model to improve the identification accuracy by researching the granularity of defect detection and defect definition and applying the defect detection technology to the training of the identification model;
furthermore, the invention also adopts a visual image acquisition preprocessing technology to preprocess the image and acquire all the plate surface images as clear as possible.
Drawings
FIG. 1 is a flow chart of an embodiment of a visual inspection method for a sheet material according to the present invention;
FIG. 2 is a schematic view of the bending of the sheet of the present invention;
FIG. 3 is a schematic view of the twist of the sheet of the present invention;
FIG. 4 is a schematic representation of the right angle of the ends of the sheet of the present invention;
FIG. 5 is a schematic view of the tile shape of the sheet of the present invention;
FIG. 6 is a flowchart of an embodiment of a training method of a defect recognition model in the present invention;
FIG. 7 is a schematic diagram of a training method of a defect recognition model according to the present invention;
FIG. 8 is a schematic diagram of a defect recognition model according to the present invention;
FIG. 9 is a flowchart of an embodiment of a method for training a color classification model in the present invention;
FIG. 10 is a schematic diagram of a method of training a color classification model according to the present invention;
FIG. 11 is a schematic diagram of the structure of a color classification model according to the present invention;
FIG. 12 is a schematic view of the visual inspection system for sheet material according to the present invention;
FIG. 13 is a schematic diagram of a defect recognition training module according to the present invention;
FIG. 14 is a schematic diagram of a color classification training module according to the present invention;
FIG. 15 is a schematic diagram of the pre-processing module of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings.
Referring to fig. 1, fig. 1 shows a flowchart of an embodiment of the visual inspection method for a sheet material of the present invention, which includes:
s101, three-dimensional point cloud information of the target plate collected by the 3D contourgraph is obtained.
According to the invention, the four 3D contourmeters are adopted to measure the height of the plate in the upper direction, the lower direction, the left direction and the right direction, so that when the plate to be detected enters a detection area consisting of the four 3D contourmeters from a movement device, four laser lines of the four 3D contourmeters respectively scan four planes of the plate to obtain three-dimensional point cloud information of the plate.
And S102, constructing a three-dimensional model according to the three-dimensional point cloud information.
Through the three-dimensional imaging technology, the collected three-dimensional point cloud information can be integrated and displayed to form a three-dimensional model.
And S103, identifying the size defect of the target plate according to the three-dimensional model and the preset reference size.
It should be noted that the actual size of the plate can be obtained through the three-dimensional model, and then the size defect of the plate can be identified by comparing the actual size with the reference size.
The working process of the 3D contourgraph can be known by integrating the steps S101-S103 as follows:
(1) starting collection: when the plate enters the visual field of the 3D contourgraph, opening the 3D contourgraph;
(2) sweeping and taking the plate: the 3D contourgraph scans the rapidly moving plate, and simultaneously sends scanned three-dimensional point cloud information to a computer;
(3) stopping collection: when the plate passes through the visual field of the 3D contourgraph, stopping the 3D contourgraph, and splicing and reconstructing the collected three-dimensional point cloud information of the whole plate;
(4) whether qualified or not: and comparing the calculation structure with the judgment standard according to the detection rule of each parameter, and identifying the specific defects of the related parameters.
Specifically, the size defect includes any one or more of a length, width and height defect, a curvature defect, a torsion defect, a right angle defect at two ends, a plane-breaking and head-biting defect, a tile-shaped defect, a vermin defect, and a bending defect, but not limited thereto.
The following detailed description of various size defects is made with reference to specific identification methods:
(1) when the length, the width and the height defects are identified, the length of the target plate is the length of the three-dimensional model, the width of the target plate is the width of the three-dimensional model, and the thickness of the target plate is the thickness of the three-dimensional model.
The length of the sheet material can be obtained from the length of the three-dimensional model. For example, a test plate length of 910mm and a transport speed of 1m/s can be made accurate to within 0.5mm using a 2kHz 3D profilometer.
The width of the plate is measured by the X-axis direction laser line width of the vertically arranged 3D contourgraph, and the width precision is within 0.051-0.097 mm.
The thickness of the plate is measured by the X-axis direction laser line width of the left and right 3D contourgraph, and the precision is the same as the width precision.
(2) When the curvature defect is identified, height distribution is extracted along the length direction of the three-dimensional model, the maximum height value is selected as the chord height, and the length is divided by the chord height to obtain the curvature in the length direction.
As shown in fig. 2, the shape of the plate is an elongated rectangular parallelepiped, and according to the definition of the degree of curvature: the chord height of the bend in each meter of length is the bending degree in each meter, and the ratio of the total chord height to the total length of the bend in the total length is the total bending degree. Therefore, the height distribution in the length direction in the three-dimensional model can be converted, the chord height of the maximum value is selected, and the curvature in the length direction is obtained by dividing the value by the total length; because the upper surface and the lower surface are arranged, the maximum value is taken as the bending degree.
(3) When the defect of the torsion resistance is identified, the distance from any one angle on one surface to the other three angle planes in the three-dimensional model is calculated, and the maximum distance value is taken as the torsion resistance.
As shown in fig. 3, the shape of the plate is an elongated rectangular parallelepiped, and according to the definition of the twist degree: one corner (angle A) of a plane is arched upwards, and the other three corners (angle B, C, D) are placed on the same plane, wherein the distance between the suspended corner and the plane of the other three points is the torsion degree. Therefore, the distance between any one corner of one surface and the other triangular plane can be converted into calculation, and the maximum distance is taken as the bending degree; the maximum value is taken because of the upper and lower surfaces.
(4) And when the right-angle defects at the two ends are identified, calculating an included angle between the planes in the three-dimensional model, and taking the included angle as the right-angle at the two ends.
As shown in FIG. 4, the right angle at two ends can be regarded as the angle measurement between two ends and the side edge, and can be converted into an included angle (angle E) between a plane and a plane according to a three-dimensional model for solving.
(5) When the defect of plane collapse and head gnawing is identified, the three-dimensional model is processed by adopting a surface smoothing algorithm, the smoothed data and the original data are subtracted to obtain a plane collapse and head gnawing area, or an upper surface equation is fitted through upper surface three-dimensional point cloud information, a lower surface equation is fitted through lower surface three-dimensional point cloud information, the distance from each point to the upper surface equation and the distance from each point to the lower surface equation are calculated, and a gray level graph is drawn to obtain the plane collapse and head gnawing area.
The surface collapse is mainly caused by that a cutter does not fall into the plate or the height of the plate surface is slightly lower, and is characterized in that the surface of the plate is provided with obvious burrs, so that an obvious dark area is formed at the waist part of the plate; therefore, the surface smoothing algorithm can be used for processing, and the smoothed surface is subtracted from the original data to obtain the collapsed surface area, and the lower surface is processed in the same way. For a few regions with unobvious surface burrs, calculating three-dimensional point cloud information of the upper surface and the lower surface, namely, fitting the three-dimensional point cloud information of the upper surface and the lower surface to obtain two square equations, calculating the distance between each point of the surface and a plane equation, drawing a height gray scale map, and obtaining the area of a related region.
(6) And when the tile-shaped defects are identified, fitting a circular equation from the inner side and the outer side of the three-dimensional model, calculating the radius of the circle, and determining the tile-shaped defects according to the radius.
As shown in fig. 5, the tile bending degree is one of the defects of the plate material, and the tile height is required to be 0.5mm or less, and the thickness after tile removal is required to be 1 mm or more, namely, the thickness of the molded green sheet. The tile shape of the plate can be divided into a longitudinal tile shape and a transverse tile shape according to the provided specification. The tile shape is similar to a section of circular arc, taking the tile shape of the longitudinal cross section as an example, the inner side and the outer side of the tile shape can be fitted with a circular equation, and the radius R and the center of the circle can be obtained. The more severe the tile shape is, the smaller the radius of the obtained circle; the more slight the tiling, the larger or infinite the resulting radius of the circle. Therefore, whether the tile shape is formed can be judged by setting a threshold value of the fitted circle radius. The tile-shaped height can be obtained by calculating the extreme value of the three-dimensional point cloud information of the upper surface and the lower surface; or the height of the arc may be calculated by planar geometry.
(7) And when the insect defect is identified, processing the three-dimensional model by adopting a surface smoothing algorithm, and subtracting the smoothed data from the original data to obtain an insect area.
(8) When the bending defect is identified, the lateral bending part is removed, and the width is wide enough; under the condition of enough width, one side is straight and the plate surface is intact, and the other side is not straight.
The plate is deformed under the load or action, the size of each point along the length direction of the member is different, and the deformed plate is in a curve shape; the regular curve has the greatest deflection at the middle and is called lateral bending.
Correspondingly, the error reporting method of the size defect is to open the setting function of the defect threshold, and when the size defect exceeds the threshold, to prompt the defect to be unqualified or send unqualified information.
And S104, acquiring two-dimensional color image information of the target plate acquired by the linear array camera.
The invention adopts four linear array cameras to sweep the upper surface, the lower surface, the left surface and the right surface of the plate simultaneously, and then carries out color imaging on the surfaces of the plate to obtain two-dimensional color image information of the respective surfaces.
And S105, inputting the two-dimensional color image information into a defect recognition model trained in advance to recognize the natural defects of the target plate.
Accordingly, the natural deficiencies include, but are not limited to, dead joints, resin pockets, entrapped skin, and decay.
In addition, when identifying the flaw of the crack defect (surface crack, end crack and fracture line), the three-dimensional model and the defect identification model are combined for processing. Specifically, the surface cracks can be identified by color imaging of the linear array camera in cooperation with single-angle polishing and in combination with a defect identification model, and the depths of the cracks can be analyzed in combination with a three-dimensional model to further determine the cracks.
And S106, inputting the two-dimensional color image information into a color classification model trained in advance to identify the color class of the target plate.
Accordingly, the color categories include, but are not limited to, blue-shift, dark, medium, light, and mottled colors. The plate material is recognized by a color classification model because the plate material is blue-changed when the discoloring bacteria invade the plate material, the distribution range of the blue-changed defect area is large, and the plate material is blue, light blue and the like.
In addition, when the defect of the wormhole is identified, the wormhole is required to be processed by combining a three-dimensional model and a color classification model. The wormhole is judged according to two main points: 1. the color of the wormholes is darker than that of the surrounding plates and is similar to a circle; 2. the wormholes can produce deep sinks in the panel, and a height difference can be formed after the wormholes are swept by the 3D contourgraph. Therefore, wormholes can be identified through the color classification model, and the height of wormholes can be analyzed by combining a 3D contourgraph.
Therefore, the invention can carry out comprehensive and accurate detection on the defects of the plate, wherein: the defect of length, width and height, the defect of curvature, the defect of torsion resistance, the defect of right angle at two ends, the defect of head biting in a collapsing plane, the defect of tile shape, the defect of living insects and the defect of bending measurement can be identified by a three-dimensional model constructed by three-dimensional point cloud information, the defect of dead joints, resin bags, bark inclusion and decay can be identified by a defect identification model, the defect of blue change and color classification can be identified by a color classification model, the defect of cracks can be identified by combining the three-dimensional model and the defect identification model, and the defect of wormholes can be identified by combining the three-dimensional model and the color classification model.
Further, before inputting the two-dimensional color image information into the defect recognition model and the color classification model which are trained in advance, the two-dimensional color image information needs to be preprocessed. The specific pretreatment steps include:
(1) carrying out SVD on the two-dimensional color image information;
(2) carrying out Gabor filtering on the two-dimensional color image information subjected to SVD;
(3) and performing difference shadow operation on the two-dimensional color image information after Gabor filtering by adopting an iterative difference shadow method.
It should be noted that after the iterative difference image method performs median filtering and image enhancement, the image is subjected to pixel template extraction and difference image operation for multiple times, and a defect map with high contrast can be obtained, so that the recognition interference of the background on the defect is effectively inhibited, and the defect characteristics are highlighted, thereby providing effective conditions for the next step of machine learning and recognition.
Therefore, the invention adopts a processing method combining SVD decomposition, Gabor filtering and iterative difference image method, which not only avoids the omission of the processing capability of the SVD in the non-horizontal and vertical direction, but also reduces the calculated amount in the Gabor filtering and gives consideration to the processing effect and the processing efficiency.
Referring to fig. 6 and 7, a flowchart of an embodiment of a training method of a defect recognition model in the present invention includes:
s201, acquiring an original color image;
s202, intercepting and merging the original color image to form a basic color image;
after the images are acquired by the linear array camera, the length-width ratio of the original image size is too large and is not suitable for model learning, so that the images need to be intercepted and merged according to priori knowledge to form image reconstruction.
S203, marking the defect position and the defect type in the basic color image to form a marked color image;
when the frame is marked, the defect position and the defect type can be marked through manual experience. Specifically, defect classes include, but are not limited to, dead joints, resin pockets, entrapped skin, and decay.
S204, performing generalized amplification processing on the labeled color image to form a sample color image;
the number of defect plates actually collected is often limited, and therefore, generalized amplification can be performed by a dedicated algorithm.
S205, integrating all sample color images into a sample color image set;
and S206, training a defect recognition model according to the sample color image set.
As shown in fig. 8, the defect identification model performs multi-scale feature extraction on an input sample color image by using a DNN model, fuses the multi-scale features by combining semantic information and position information, and extracts features by combining an attention mechanism and a spatial pyramid structure to identify and output defect types, defect positions, and confidence information. Accordingly, the sample color image is a 640 × 640 reconstructed image.
Referring to fig. 9 and fig. 10, a flowchart of an embodiment of a method for training a color classification model according to the present invention includes:
s301, acquiring an original color image;
s302, extracting a target foreground region of the original color image through a visual algorithm to form an ROI color image;
s303, carrying out scaling processing on a target foreground region of the ROI color image, and cutting the target foreground region into a plurality of local images;
s304, marking color categories in the local images to form marked classified images;
when the frame is marked, the color category can be marked through manual experience; specifically, the color categories include, but are not limited to, blue-shifted, dark, medium, light, and mottled colors.
S305, performing generalized amplification processing on the labeled classified image to form a sample classified image;
the number of defect plates actually collected is often limited, and therefore, generalized amplification can be performed by a dedicated algorithm.
S306, integrating all the sample classified images into a sample classified image set;
and S307, training a color classification model according to the sample classification image set.
As shown in fig. 11, the color classification model adopts a DNN model to perform multi-scale feature extraction on the input sample classification image, classifies high-dimensional feature vectors, and then identifies image semantic classification information to output color classification and confidence information. Accordingly, the sample classification image is a 320 × 320 tile image.
Therefore, the original color image is obtained through the linear array camera, the defect characteristics are highlighted through the image preprocessing technology, and the defect characteristics are recognized through the trained model algorithm, so that the aim of replacing manual detection is fulfilled, and the linear array camera has the advantages of high efficiency, low cost and high flexibility.
Referring to fig. 12, fig. 12 shows a specific structure of the visual inspection system 100 for a sheet material according to the present invention, which includes a size defect identification module 1, a natural defect identification module 2, and a color classification identification module 3.
The following respectively describes the three modules:
first, size defect identification module 1
The size defect identification module 1 is used for acquiring three-dimensional point cloud information of a target plate acquired by the 3D contourgraph, constructing a three-dimensional model according to the three-dimensional point cloud information, and identifying the size defect of the target plate according to the three-dimensional model and a preset reference size.
In the invention, four 3D contourgraph instruments are adopted to measure the height of the plate in four directions, namely the upper direction, the lower direction, the left direction and the right direction. When the plate to be detected enters a detection area consisting of four 3D contourmeters from the movement device, opening the four 3D contourmeters; then, four laser lines of the four 3D contourgraph respectively scan four planes of the rapidly moving plate to obtain three-dimensional point cloud information of the plate, and meanwhile, the scanned three-dimensional point cloud information is sent to the size defect identification module 1; when the plate passes through the visual field of the 3D contourgraph, stopping the 3D contourgraph, and splicing and reconstructing the collected three-dimensional point cloud information of the whole plate; and finally, comparing the calculation structure with a judgment standard according to the detection rule of each parameter, and identifying the specific size defect of the related parameter.
Accordingly, the size defect includes any one or more of a length, width and height defect, a curvature defect, a torsion defect, a right angle defect at two ends, a head-biting defect, a tile-shaped defect, a worm defect and a bending defect, but not limited thereto. Specifically, the method comprises the following steps:
when the length, the width and the height defects are identified, the length of the target plate is the length of the three-dimensional model, the width of the target plate is the width of the three-dimensional model, and the thickness of the target plate is the thickness of the three-dimensional model. That is, the length of the sheet material can be obtained from the length of the three-dimensional model; the width of the plate is measured by the X-axis direction laser line width of the vertically arranged 3D contourgraph; the thickness of the plate is measured by the X-axis direction laser line width of the left and right 3D contourgraph.
When the curvature defect is identified, height distribution is extracted along the length direction of the three-dimensional model, the maximum height value is selected as the chord height, and the length is divided by the chord height to obtain the curvature in the length direction. As shown in fig. 2, the shape of the plate is an elongated rectangular parallelepiped, and according to the definition of the degree of curvature: the chord height of the bend in each meter of length is the bending degree in each meter, and the ratio of the total chord height to the total length of the bend in the total length is the total bending degree. Therefore, the height distribution in the length direction in the three-dimensional model can be converted, the chord height of the maximum value is selected, and the curvature in the length direction is obtained by dividing the value by the total length; because the upper surface and the lower surface are arranged, the maximum value is taken as the bending degree.
When the defect of the torsion resistance is identified, the distance from any one angle on one surface to the other three angle planes in the three-dimensional model is calculated, and the maximum distance value is taken as the torsion resistance. As shown in fig. 3, the shape of the plate is an elongated rectangular parallelepiped, and according to the definition of the twist degree: one corner of a plane is arched upwards, and the other three corners are placed on the same plane, wherein the distance between the suspended corner and the plane of the other three points is the torsion degree. Therefore, the distance between any one corner of one surface and the other triangular plane can be converted into calculation, and the maximum distance is taken as the bending degree; the maximum value is taken because of the upper and lower surfaces.
And when the right-angle defects at the two ends are identified, calculating an included angle between the planes in the three-dimensional model, and taking the included angle as the right-angle at the two ends. As shown in FIG. 4, the right angle at two ends can be regarded as the angle measurement between two ends and the side edge, and can be converted into the included angle between a plane and a plane according to a three-dimensional model to be solved.
When the defect of plane collapse and head gnawing is identified, the three-dimensional model is processed by adopting a surface smoothing algorithm, the smoothed data and the original data are subtracted to obtain a plane collapse and head gnawing area, or an upper surface equation is fitted through upper surface three-dimensional point cloud information, a lower surface equation is fitted through lower surface three-dimensional point cloud information, the distance from each point to the upper surface equation and the distance from each point to the lower surface equation are calculated, and a gray level graph is drawn to obtain the plane collapse and head gnawing area. The surface collapse is mainly caused by that a cutter does not fall into the plate or the plate surface is slightly low in height, and is characterized in that the surface of the plate is provided with obvious burrs, so that an obvious dark area is formed at the waist part of the plate; therefore, the surface smoothing algorithm can be used for processing, and the smoothed surface is subtracted from the original data to obtain the collapsed surface area, and the lower surface is processed in the same way. For a few regions with unobvious surface burrs, calculating three-dimensional point cloud information of the upper surface and the lower surface, namely, fitting the three-dimensional point cloud information of the upper surface and the lower surface to obtain two square equations, calculating the distance between each point of the surface and a plane equation, drawing a height gray scale map, and obtaining the area of a related region.
And when the tile-shaped defects are identified, fitting a circular equation from the inner side and the outer side of the three-dimensional model, calculating the radius of the circle, and determining the tile-shaped defects according to the radius. As shown in fig. 5, the tile bending degree is one of the defects of the plate material, and the height of the tile shape is required to be 0.5mm or less, and the thickness after the tile is removed is equal to the thickness of the molded green plate +1 mm. The tile shape of the plate can be divided into a longitudinal tile shape and a transverse tile shape according to the provided specification. The tile shape is similar to a section of circular arc, taking the tile shape of the longitudinal cross section as an example, the inner side and the outer side of the tile shape can be fitted with a circular equation, and the radius and the center of the circle can be obtained. The more severe the tile shape is, the smaller the radius of the obtained circle; the more slight the tiling, the larger or infinite the resulting radius of the circle. Therefore, whether the tile shape is formed can be judged by setting a threshold value of the fitted circle radius. The tile-shaped height can be obtained by calculating the extreme value of the three-dimensional point cloud information of the upper surface and the lower surface; or the height of the arc may be calculated by planar geometry.
And when the insect defect is identified, processing the three-dimensional model by adopting a surface smoothing algorithm, and subtracting the smoothed data from the original data to obtain an insect area.
When the bending defect is identified, the lateral bending part is removed, and the width is wide enough; under the condition of enough width, one side is straight and the plate surface is intact, and the other side is not straight. It should be noted that the plate is deformed under the load or action, and the size of each point along the length direction of the member is different, and the deformed plate is in a curve shape; the regular curve has the greatest deflection at the middle and is called lateral bending.
Correspondingly, the error reporting method of the size defect is to open the setting function of the defect threshold, and when the size defect exceeds the threshold, to prompt the defect to be unqualified or send unqualified information.
Second, natural defect recognition module 2
And the natural defect identification module 2 is used for acquiring two-dimensional color image information of the target plate acquired by the linear array camera, and inputting the two-dimensional color image information into a defect identification model trained in advance to identify the natural defect of the target plate. Accordingly, the natural deficiencies include, but are not limited to, dead joints, resin pockets, entrapped skin, and decay.
It should be noted that the invention adopts four line cameras to scan the four surfaces of the upper, lower, left and right of the plate at the same time, and then carries out color imaging on the surfaces to obtain the two-dimensional color image information of the respective surfaces.
Further, when identifying the flaw of the crack (surface crack, end crack and fracture line), the three-dimensional model and the flaw identification model are combined for processing. Specifically, the surface cracks can be identified by color imaging of the linear array camera in cooperation with single-angle polishing and in combination with a defect identification model, and the depths of the cracks can be analyzed in combination with a three-dimensional model to further determine the cracks.
Third, color classification recognition module 3
And the color classification and identification module 3 is used for acquiring two-dimensional color image information of the target plate acquired by the linear array camera and inputting the two-dimensional color image information into a color classification model trained in advance to identify the color category of the target plate.
Accordingly, the color categories include, but are not limited to, blue-shift, dark, medium, light, and mottled colors. The plate material is recognized by a color classification model because the plate material is blue-changed when the discoloring bacteria invade the plate material, the distribution range of the blue-changed defect area is large, and the plate material is blue, light blue and the like.
Further, when the defect of the wormhole is identified, the wormhole is processed by combining a three-dimensional model and a color classification model. The wormhole is judged according to two main points: 1. the color of the wormholes is darker than that of the surrounding plates and is similar to a circle; 2. the wormholes can produce deep sinks in the panel, and a height difference can be formed after the wormholes are swept by the 3D contourgraph. Therefore, wormholes can be identified through the color classification model, and the height of wormholes can be analyzed by combining a 3D contourgraph.
Therefore, the method and the device can be used for comprehensively and accurately detecting the defects of the plate, can solve the problems of fatigue error of current manual detection, substandard detection rate and the like, and have the advantages of good flexibility, strong performance and high detection rate.
As shown in fig. 13, the visual inspection system 100 for sheet material further includes a defect recognition training module 4; specifically, the defect identification training module 4 includes:
a first acquisition unit 41 for acquiring an original color image;
a merging unit 42, configured to perform intercepting and merging processing on the original color image to form a basic color image; after the images are acquired by the linear array camera, the length-width ratio of the original image size is too large and is not suitable for model learning, so that the images need to be intercepted and merged according to priori knowledge to form image reconstruction.
A first labeling unit 43, configured to label defect positions and defect types in the basic color image to form a labeled color image, wherein the defect types include dead joints, resin bags, leather, and decay; when the frame is marked, the defect position and the defect type can be marked through manual experience. Specifically, defect classes include, but are not limited to, dead joints, resin pockets, entrapped skin, and decay.
A first amplification unit 44, configured to perform a generalized amplification process on the labeled color image to form a sample color image; the number of defect plates actually collected is often limited, and therefore, generalized amplification can be performed by a dedicated algorithm.
A first integration unit 45 for integrating all sample color images into a sample color image set;
a first training unit 46 for training a defect recognition model based on the sample color image set.
As shown in fig. 8, the defect identification model performs multi-scale feature extraction on an input sample color image by using a DNN model, fuses the multi-scale features by combining semantic information and position information, and extracts features by combining an attention mechanism and a spatial pyramid structure to identify and output defect types, defect positions, and confidence information. Accordingly, the sample color image is a 640 × 640 reconstructed image.
As shown in fig. 14, the visual inspection system 100 for a sheet material further includes a color classification training module 5; specifically, the color classification training module 5 includes:
a second acquiring unit 51 for acquiring an original color image;
an extracting unit 52, configured to extract a target foreground region of the original color image through a visual algorithm to form an ROI color image;
a cutting unit 53, configured to perform scaling processing on a target foreground region of the ROI color image, and cut the target foreground region into a plurality of local images;
a second labeling unit 54, configured to label color categories in the local image to form a labeled classified image, where the color categories include blue, deep, medium, and light; when the frame is marked, the color category can be marked through manual experience; specifically, the color categories include, but are not limited to, blue-shifted, dark, medium, light, and mottled colors.
A second amplification unit 55, configured to perform generalized amplification processing on the labeled classified image to form a sample classified image; the number of defect plates actually collected is often limited, and therefore, generalized amplification can be performed by a dedicated algorithm.
A second integration unit 56 for integrating all the sample classification images into a sample classification image set;
and a second training unit 57, configured to train the color classification model according to the sample classification image set.
As shown in fig. 11, the color classification model adopts a DNN model to perform multi-scale feature extraction on the input sample classification image, classifies high-dimensional feature vectors, and then identifies image semantic classification information to output color classification and confidence information. Accordingly, the sample classification image is a 320 × 320 tile image.
Therefore, the invention can obtain the original color image through the linear array camera, highlight the defect characteristics by using the image preprocessing technology and recognize the defect characteristics through the trained model algorithm, thereby achieving the aim of replacing manual detection and having the advantages of high efficiency, low cost and strong flexibility.
As shown in fig. 15, the visual inspection system 100 for a sheet material further includes a preprocessing module 6, and the preprocessing module 6 is further required to preprocess the two-dimensional color image information before inputting the two-dimensional color image information into the defect recognition model and the color classification model trained in advance. Specifically, the preprocessing module 6 includes:
an SVD unit 61 configured to perform SVD decomposition on the two-dimensional color image information;
a Gabor unit 62 configured to perform Gabor filtering on the SVD decomposed two-dimensional color image information;
and the difference shadow unit 63 is configured to perform difference shadow operation on the Gabor-filtered two-dimensional color image information by using an iterative difference shadow method. It should be noted that after the iterative difference image method performs median filtering and image enhancement, the image is subjected to pixel template extraction and difference image operation for multiple times, and a defect map with high contrast can be obtained, so that the recognition interference of the background on the defect is effectively inhibited, and the defect characteristics are highlighted, thereby providing effective conditions for the next step of machine learning and recognition.
Therefore, the preprocessing module 6 adopts a processing method combining SVD decomposition, Gabor filtering and iterative difference image method, which not only avoids the lack of processing capability of SVD in the non-horizontal and vertical directions, but also reduces the calculated amount in Gabor filtering, and gives consideration to both processing effect and processing efficiency.
In addition, the invention adopts the edge calculation technology to process the data in the size defect identification module 1, the natural defect identification module 2, the color classification identification module 3, the defect identification training module 4, the color classification training module 5 and the preprocessing module 6.
It should be noted that, the image data acquired by the traditional machine vision detection image acquisition end needs to be uploaded to the cloud, or is uploaded to the image preprocessing center and the discrimination server through the local area network for processing and identification, and the identification result is returned to the client. Due to the large data volume of the photos, if the production lines are large in number, the centralized image preprocessing and distinguishing service center can be challenged by severe performance, the time delay is large, and the detection efficiency is reduced. The invention adopts the edge computing technology, uses the concept of edge cloud cooperation, takes charge of the functions of image storage, model training, tuning, publishing, management, evaluation, labeling and the like by a public cloud or private cloud machine learning platform, and the trained model can be issued to edge equipment, so that two-dimensional color image information acquired by a linear array camera can be directly identified nearby and a recognition result can be returned, thereby greatly reducing time delay and the characteristics of distributed computing, solving the problems of large load and low performance of a centralized discrimination center and an image processing center, effectively adapting to the application scenes with more production lines and greatly improving the detection efficiency.
Correspondingly, the invention also introduces a middle stage concept, abstractly integrates common general abilities such as a machine learning frame, a deep learning frame, model training, model tuning, model evaluation, model release, model life cycle management, artificial standards, semi-automatic labeling, defect statistics and the like to form the field abilities of a model center, a standard center, an intelligent learning center, a statistical center and the like, and the upper application can realize corresponding functions only by calling an interface, thereby avoiding repeated development and repeated construction.
Therefore, the method combines the image preprocessing technology, the edge computing technology and the deep learning technology, manufactures the defect detection equipment with soft-hard combination and edge cloud cooperation, has the advantages of good flexibility, high detectable rate and low time delay, effectively controls the shipment quality, reduces the after-sale cost, and has profound significance for promoting production line automation and industry 4.0.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.
Claims (10)
1. A method of visual inspection of a sheet material, comprising:
acquiring three-dimensional point cloud information of a target plate acquired by a 3D contourgraph, constructing a three-dimensional model according to the three-dimensional point cloud information, and identifying the size defect of the target plate according to the three-dimensional model and a preset reference size;
the method comprises the steps of obtaining two-dimensional color image information of a target plate collected by a linear array camera, inputting the two-dimensional color image information into a defect recognition model trained in advance to recognize natural defects of the target plate, and inputting the two-dimensional color image information into a color classification model trained in advance to recognize color categories of the target plate.
2. The visual inspection method of a sheet material according to claim 1, further comprising training a defect recognition model, specifically comprising:
acquiring an original color image;
intercepting and merging the original color image to form a basic color image;
marking defect positions and defect types in the basic color image to form a marked color image, wherein the defect types comprise dead joints, resin bags, leather inclusion and decay;
carrying out generalized amplification processing on the labeled color image to form a sample color image;
integrating all sample color images into a sample color image set;
and training the defect identification model according to the sample color image set.
3. The visual inspection method for sheet materials according to claim 2, wherein the defect identification model adopts a DNN model to perform multi-scale feature extraction on the input sample color image, combines semantic information and position information to fuse the multi-scale features, and combines an attention mechanism and a spatial pyramid structure to extract features to identify and output defect types, defect positions and confidence information.
4. The visual inspection method of a sheet material according to claim 1, further comprising color classification model training, specifically comprising:
acquiring an original color image;
extracting a target foreground region of the original color image through a visual algorithm to form an ROI color image;
carrying out scaling processing on a target foreground region of the ROI color image, and cutting the target foreground region into a plurality of local images;
marking color categories in the local images to form marked classified images, wherein the color categories comprise blue change, deep color, medium color and light color;
carrying out generalized amplification processing on the labeled classified image to form a sample classified image;
integrating all sample classification images into a sample classification image set;
training the color classification model according to the sample classification image set.
5. The visual inspection method of sheet material according to claim 4, wherein the color classification model uses a DNN model to perform multi-scale feature extraction on the input sample classification image, classifies high-dimensional feature vectors, and then identifies image semantic classification information to output color classification and confidence information.
6. The visual inspection method of solid wood according to claim 1, further comprising: preprocessing the two-dimensional color image information before inputting the two-dimensional color image information into a defect recognition model and a color classification model which are trained in advance, wherein the preprocessing comprises the following steps:
carrying out SVD on the two-dimensional color image information;
carrying out Gabor filtering on the two-dimensional color image information subjected to SVD;
and performing difference shadow operation on the two-dimensional color image information after Gabor filtering by adopting an iterative difference shadow method.
7. The visual inspection method for the plate according to claim 1, wherein the size defects comprise any one or more of length, width and height defects, bending defects, torsion defects, right-angle defects at two ends, head-biting defects, tile-shaped defects and insect-forming defects;
when the length, the width and the height defects are identified, the length of the target plate is the length of the three-dimensional model, the width of the target plate is the width of the three-dimensional model, and the thickness of the target plate is the thickness of the three-dimensional model;
when the curvature defect is identified, extracting height distribution along the length direction of the three-dimensional model, selecting the maximum height value as the chord height, and dividing the chord height by the length to obtain the curvature in the length direction;
when the torsion resistance defect is identified, calculating the distance from any one angle on one surface of the three-dimensional model to the other three angle planes, and taking the maximum distance value as the torsion resistance;
when the two-end right-angle defects are identified, calculating an included angle between planes in the three-dimensional model, and taking the included angle as the two-end right-angle;
when the plane-collapse gnawing defect is identified, processing the three-dimensional model by adopting a surface smoothing algorithm, subtracting the smoothed data from the original data to obtain a plane-collapse gnawing area, or fitting an upper surface equation through upper surface three-dimensional point cloud information, fitting a lower surface equation through lower surface three-dimensional point cloud information, calculating the distance from each point to the upper surface equation and the lower surface equation, and drawing a gray level map to obtain the plane-collapse gnawing area;
when the tile-shaped defects are identified, fitting a circular equation from the inner side and the outer side of the three-dimensional model, calculating the radius of a circle, and determining the tile-shaped defects according to the radius;
and when the insect defect is identified, processing the three-dimensional model by adopting a surface smoothing algorithm, and subtracting the smoothed data from the original data to obtain an insect area.
8. A visual panel inspection system, comprising:
the size defect identification module is used for acquiring three-dimensional point cloud information of a target plate acquired by the 3D contourgraph, constructing a three-dimensional model according to the three-dimensional point cloud information, and identifying the size defect of the target plate according to the three-dimensional model and a preset reference size;
the natural defect identification module is used for acquiring two-dimensional color image information of a target plate acquired by the linear array camera and inputting the two-dimensional color image information into a defect identification model trained in advance to identify the natural defect of the target plate;
and the color classification and identification module is used for acquiring two-dimensional color image information of the target plate acquired by the linear array camera and inputting the two-dimensional color image information into a color classification model trained in advance to identify the color class of the target plate.
9. The visual inspection system of a sheet material of claim 8 further comprising a defect recognition training module, said defect recognition training module comprising:
a first acquisition unit configured to acquire an original color image;
the merging unit is used for intercepting and merging the original color image to form a basic color image;
the first marking unit is used for marking defect positions and defect types in the basic color image so as to form a marked color image, wherein the defect types comprise dead joints, resin bags, leather and decay;
the first amplification unit is used for carrying out generalized amplification processing on the labeled color image so as to form a sample color image;
a first integration unit for integrating all sample color images into a sample color image set;
and the first training unit is used for training the defect identification model according to the sample color image set.
10. The visual inspection system of sheet material of claim 8 further comprising a color classification training module, said color classification training module comprising:
a second acquisition unit for acquiring an original color image;
the extracting unit is used for extracting a target foreground region of the original color image through a visual algorithm to form an ROI color image;
the cutting unit is used for carrying out scaling processing on a target foreground region of the ROI color image and cutting the target foreground region into a plurality of local images;
the second labeling unit is used for labeling color categories in the local image to form a labeled classified image, wherein the color categories comprise blue, deep, medium and light colors;
the second amplification unit is used for carrying out generalized amplification processing on the labeled classified image to form a sample classified image;
the second integration unit is used for integrating all the sample classification images into a sample classification image set;
and the second training unit is used for training the color classification model according to the sample classification image set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110454981.7A CN113362276B (en) | 2021-04-26 | 2021-04-26 | Visual detection method and system for plates |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110454981.7A CN113362276B (en) | 2021-04-26 | 2021-04-26 | Visual detection method and system for plates |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113362276A true CN113362276A (en) | 2021-09-07 |
| CN113362276B CN113362276B (en) | 2024-05-10 |
Family
ID=77525533
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110454981.7A Active CN113362276B (en) | 2021-04-26 | 2021-04-26 | Visual detection method and system for plates |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113362276B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114199892A (en) * | 2021-12-10 | 2022-03-18 | 江苏雷默智能科技有限公司 | Plate measuring method and system based on machine vision |
| CN114820761A (en) * | 2022-05-07 | 2022-07-29 | 北京毅能博科技有限公司 | XY-direction included angle measuring and motion compensating method based on image micro-scanning platform |
| CN114812440A (en) * | 2022-04-22 | 2022-07-29 | 国网湖北省电力有限公司十堰供电公司 | High-rise building fireproof plugging module defect detection device and method based on line structured light |
| CN114935578A (en) * | 2022-04-29 | 2022-08-23 | 山东铝谷产业技术研究院有限公司 | Machine vision online identification device and method for surface defects of alloy plates and strips |
| CN114998194A (en) * | 2022-04-20 | 2022-09-02 | 广州超音速自动化科技股份有限公司 | Product defect detection method, system and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170140539A1 (en) * | 2015-11-16 | 2017-05-18 | Abb Technology Ag | Three-dimensional visual servoing for robot positioning |
| CN107052086A (en) * | 2017-06-01 | 2017-08-18 | 扬州苏星机器人科技有限公司 | Stamping parts surface defect detection apparatus and detection method based on 3D vision |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| CN108010036A (en) * | 2017-11-21 | 2018-05-08 | 江南大学 | A kind of object symmetry axis detection method based on RGB-D cameras |
| WO2018136262A1 (en) * | 2017-01-20 | 2018-07-26 | Aquifi, Inc. | Systems and methods for defect detection |
| EP3373251A1 (en) * | 2017-03-07 | 2018-09-12 | Trimble AB | Scan colorization with an uncalibrated camera |
| US20190096135A1 (en) * | 2017-09-26 | 2019-03-28 | Aquifi, Inc. | Systems and methods for visual inspection based on augmented reality |
| WO2019104767A1 (en) * | 2017-11-28 | 2019-06-06 | 河海大学常州校区 | Fabric defect detection method based on deep convolutional neural network and visual saliency |
| CN110308151A (en) * | 2019-07-22 | 2019-10-08 | 重庆大学 | A kind of surface of workpiece defect identification method and device based on machine vision |
| CN110533642A (en) * | 2019-08-21 | 2019-12-03 | 深圳新视达视讯工程有限公司 | A kind of detection method of insulator damage |
-
2021
- 2021-04-26 CN CN202110454981.7A patent/CN113362276B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170140539A1 (en) * | 2015-11-16 | 2017-05-18 | Abb Technology Ag | Three-dimensional visual servoing for robot positioning |
| WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
| WO2018136262A1 (en) * | 2017-01-20 | 2018-07-26 | Aquifi, Inc. | Systems and methods for defect detection |
| EP3373251A1 (en) * | 2017-03-07 | 2018-09-12 | Trimble AB | Scan colorization with an uncalibrated camera |
| CN107052086A (en) * | 2017-06-01 | 2017-08-18 | 扬州苏星机器人科技有限公司 | Stamping parts surface defect detection apparatus and detection method based on 3D vision |
| US20190096135A1 (en) * | 2017-09-26 | 2019-03-28 | Aquifi, Inc. | Systems and methods for visual inspection based on augmented reality |
| CN108010036A (en) * | 2017-11-21 | 2018-05-08 | 江南大学 | A kind of object symmetry axis detection method based on RGB-D cameras |
| WO2019104767A1 (en) * | 2017-11-28 | 2019-06-06 | 河海大学常州校区 | Fabric defect detection method based on deep convolutional neural network and visual saliency |
| CN110308151A (en) * | 2019-07-22 | 2019-10-08 | 重庆大学 | A kind of surface of workpiece defect identification method and device based on machine vision |
| CN110533642A (en) * | 2019-08-21 | 2019-12-03 | 深圳新视达视讯工程有限公司 | A kind of detection method of insulator damage |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114199892A (en) * | 2021-12-10 | 2022-03-18 | 江苏雷默智能科技有限公司 | Plate measuring method and system based on machine vision |
| CN114199892B (en) * | 2021-12-10 | 2022-11-18 | 江苏雷默智能科技有限公司 | Plate measuring method and system based on machine vision |
| CN114998194A (en) * | 2022-04-20 | 2022-09-02 | 广州超音速自动化科技股份有限公司 | Product defect detection method, system and storage medium |
| CN114812440A (en) * | 2022-04-22 | 2022-07-29 | 国网湖北省电力有限公司十堰供电公司 | High-rise building fireproof plugging module defect detection device and method based on line structured light |
| CN114935578A (en) * | 2022-04-29 | 2022-08-23 | 山东铝谷产业技术研究院有限公司 | Machine vision online identification device and method for surface defects of alloy plates and strips |
| CN114820761A (en) * | 2022-05-07 | 2022-07-29 | 北京毅能博科技有限公司 | XY-direction included angle measuring and motion compensating method based on image micro-scanning platform |
| CN114820761B (en) * | 2022-05-07 | 2024-05-10 | 北京毅能博科技有限公司 | XY direction included angle measurement and motion compensation method based on image microscopic scanning platform |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113362276B (en) | 2024-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113362276A (en) | Visual detection method and system for plate | |
| KR102254773B1 (en) | Automatic decision and classification system for each defects of building components using image information, and method for the same | |
| CN113066079B (en) | Automatic wood defect detection method, system and storage medium thereof | |
| CN107633516B (en) | Method and device for identifying road surface deformation diseases | |
| CN111667455B (en) | AI detection method for brushing multiple defects | |
| CN102288613B (en) | Surface defect detecting method for fusing grey and depth information | |
| Huang et al. | A pavement crack detection method combining 2D with 3D information based on Dempster‐Shafer theory | |
| CN110210448B (en) | Intelligent face skin aging degree identification and evaluation method | |
| CN111915572B (en) | Adaptive gear pitting quantitative detection system and method based on deep learning | |
| CN102654464A (en) | Copper strip surface defect detection system based on multi-feature fuzzy recognition | |
| CN117152161B (en) | Shaving board quality detection method and system based on image recognition | |
| CN114638797A (en) | Method and device for detecting copper surface defects based on linear array camera | |
| CN109580630A (en) | A kind of visible detection method of component of machine defect | |
| CN110473201A (en) | A kind of automatic testing method and device of disc surface defect | |
| CN113962929B (en) | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line | |
| CN111724354B (en) | Image processing-based method for measuring wheat ear length and wheat ear number of multiple wheat plants | |
| CN113781537B (en) | Rail elastic strip fastener defect identification method and device and computer equipment | |
| CN114881987A (en) | Improved YOLOv 5-based hot-pressing light guide plate defect visual detection method | |
| CN116091505B (en) | Automatic defect detection and classification method and system for sapphire substrate | |
| Mustaffar et al. | Automated pavement imaging program (APIP) for pavement cracks classification and quantification-a photogrammetric approach | |
| CN114529536A (en) | Solid wood quality detection method | |
| CN114494845A (en) | Artificial intelligence hidden danger troubleshooting system and method for construction project site | |
| CN118038443A (en) | Track section state detection method and system | |
| CN114820611B (en) | Artificial intelligence-based mechanical part quality evaluation method and system | |
| CN116740449A (en) | Shaving form detection method and system based on AI (advanced technology attachment) computer vision technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |