CN110517265A - A kind of detection method of surface defects of products, device and storage medium - Google Patents
A kind of detection method of surface defects of products, device and storage medium Download PDFInfo
- Publication number
- CN110517265A CN110517265A CN201910831247.0A CN201910831247A CN110517265A CN 110517265 A CN110517265 A CN 110517265A CN 201910831247 A CN201910831247 A CN 201910831247A CN 110517265 A CN110517265 A CN 110517265A
- Authority
- CN
- China
- Prior art keywords
- image
- detected
- product
- connected region
- target
- 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
Classifications
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
- G06T7/001—Industrial image inspection using an image reference approach
-
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
Abstract
The application provides detection method, device and the storage medium of a kind of surface defects of products, this method comprises: acquisition includes the target image of product to be detected, wherein the product surface to be detected has multiple objects to be detected;The target image is handled, multiple image to be detected are obtained, wherein includes the single object to be detected of product to be detected in every image to be detected;The not bending moment of every image to be detected corresponding d is calculated separately, and according to the d of every image to be detected not bending moment d standard corresponding with the standardized product image being obtained ahead of time not bending moments, to the product progress surface defects detection to be detected.The embodiment of the present application for surface there is the product to be processed of multiple objects to be detected can be realized quick, easy surface defects detection, not need manually to be recognized one by one, detection efficiency is higher.
Description
Technical field
This application involves technical field of image processing, detection method, dress in particular to a kind of surface defects of products
It sets and storage medium.
Background technique
The defects detection on existing product to be processed surface takes the method manually examined one by one mostly, when discovery produces
When product surface has defect, faulty goods is taken out in time, is classified to processing failure go-to field.The product smooth for surface,
Under the reflection of light, there is the part of defect and non-defective part difference is larger, be manually easy to differentiate, but has certain
The product of texture variations, such as surface have multiple objects that need to be observed, artificial to recognize that difficulty increases, and intensive object meeting
Accelerate eye fatigue, adds heavily to the difficulty of the work.
Summary of the invention
The detection method for being designed to provide a kind of surface defects of products, device and storage medium of the embodiment of the present application,
The quick detection that can be realized surface defect by the image that shooting needs testing product, to solve above-mentioned technical problem.
In a first aspect, the embodiment of the present application provides a kind of detection method of surface defects of products, which comprises obtain
It include the target image of product to be detected, wherein the product surface to be detected has multiple objects to be detected;To the mesh
Logo image is handled, multiple image to be detected are obtained, wherein in every image to be detected including product to be detected it is single to
Test object;A not bending moment of the corresponding d of every image to be detected is calculated separately, and constant according to the d of every image to be detected
Square d standard corresponding with the standardized product image being obtained ahead of time not bending moment carries out surface defect inspection to the product to be detected
It surveys, wherein d is positive integer.
In above scheme, for surface have multiple objects to be detected product to be detected, obtain first include individually to
Then image to be detected of test object is treated in detection image using the not bending moment of characterization image geometrical characteristic with the presence or absence of scarce
It is trapped into capable differentiation.This programme, can be direct by defective product according to image using the structure change of image to be detected as foundation
It picks out, more quickly and easy compared to the mode of artificial detection, detection efficiency is considerably higher, meanwhile, this method can be with
Without pre-information such as the types, shape, size of knowing the reason of causing defect and defect, can directly testing product lack
It falls into, it is versatile.
In a kind of possible embodiment, the shape of the multiple object to be detected is identical, according to every mapping to be checked
The d of picture not bending moment d standard corresponding with the standardized product image being obtained ahead of time not bending moments, to the product progress to be detected
Surface defects detection, comprising: calculate separately the d of every image to be detected not bending moment relative to the d standard not bending moment
Variance;If the variance of any image to be detected in multiple described image to be detected be greater than first threshold, it is determined that it is described to
There are surface defects for testing product.
Variance indicates the departure degree of the not bending moment of image to be detected relative to standard not bending moment, when any mapping to be checked
The variance of picture is greater than first threshold, shows that wherein object to be detected and object free of surface defects have significant difference, can be true
The fixed product existing defects to be detected.
In a kind of possible embodiment, the shape of the multiple object to be detected is identical, described to the target figure
As being handled, multiple image to be detected are obtained, comprising: Threshold segmentation is carried out to the target image, obtains K1 connected region
Domain, and determine from the K1 connected region K2 connected region for meeting the first preset requirement, wherein described first in advance
If K1, K2 are positive integer, and K1 > K2 it is required that related to the form parameter of the connected region;It determines every in K2 connected region
The mass center of one connected region, and the position of the target image is mapped to according to each mass center, obtain K2 Target Segmentation images;
K3 image to be detected for meeting the second preset requirement are determined from the K2 Target Segmentation images, wherein second is default
It is required that related to the texture information of the Target Segmentation image, K3 is positive integer, and K2 > K3.
In above scheme, by the way that the first preset requirement is arranged, the connected region shape formed with object to be detected can be filtered out
The biggish connected region of shape difference has the function that tentatively to screen out, by the way that the second preset requirement is arranged, can filter out with it is to be detected
The biggish image of object texture difference, have the function that it is secondary screen out, finally gradually isolate corresponding to object to be detected to
Detection image.
It is described that Threshold segmentation is carried out to the target image in a kind of possible embodiment, obtain K1 connected region
Domain, comprising: the target image is filtered, filtered gray level image is obtained;According to picture each in the gray level image
The gray value of vegetarian refreshments and the size relation of second threshold, obtain the bianry image of the target image, wherein the second threshold
For the product of the maximum gradation value in predetermined coefficient and the gray level image, the single object to be detected on product to be detected is in institute
It states and forms connected region in bianry image;Obtain K1 connected region in the bianry image.
By the filtering to target image, binary conversion treatment, it can obtain what object to be detected in product to be detected was formed
Multiple connected regions.
It is described to determine to meet the first preset requirement from the K1 connected region in a kind of possible embodiment
K2 connected region, comprising: calculate the area of each connected region in the K1 connected region, and with the connected region
Domain have identical standard second-order central away from elliptical eccentricity;It determines to meet following requirement from the K1 connected region
K2 connected region: the area of the connected region is located in first threshold range and the corresponding eccentricity of the connected region
Not less than third threshold value.
Feature Descriptor of the area of connected region with corresponding eccentricity as preliminary screening connected region, passes through setting
The threshold value of area and eccentricity can exclude and shape visibly different connected region excessive with object area gap to be detected,
Two parameters are complimentary to one another, codetermine the screening conditions of connected region.
In a kind of possible embodiment, the position that the target image is mapped to according to each mass center is obtained
K2 Target Segmentation images, comprising: the target image is split centered on the mass center of K2 connected region, obtains K2
Open original segmented image;The gradient orientation histogram of each original segmented image is calculated, the gradient orientation histogram indicates
The statistics of pixel number of the original segmented image on different gradient directions;Determine picture in the gradient orientation histogram
The largest number of gradient directions of vegetarian refreshments, and the object to be detected is corrected in institute according to the gradient direction and preset rotation direction
State the angle in original segmented image, the segmented image after being corrected;Divide centered on the mass center of segmented image after correcting
The other segmented image to after each correction is cut, and obtains K2 Target Segmentation images.
In above scheme, segmentation obtains K2 larger-size original segmented images first, and according to original segmented image
Gradient direction distribution to original segmented image carry out angle correction so that angle, the size of object to be detected in the picture
It reaches unanimity, image is cut using size lesser rectangle frame after angle correction, since object to be detected exists with shape
It is located at the center of image after angle correction, therefore can be cut out object to be detected minimally to draw with the smallest rectangle frame
Enter the ingredient of non-object to be detected, reduces influence of noise.
It is described to determine that meeting second presets from the K2 Target Segmentation images in a kind of possible embodiment
It is required that K3 image to be detected, comprising: calculate the gradient orientation histogram of the Target Segmentation image, the gradient direction is straight
Side's figure indicates the statistics of pixel number of the Target Segmentation image on different gradient directions;Calculate every Target Segmentation figure
Variance of the gradient orientation histogram of picture relative to the corresponding normal gradients direction histogram of standardized product image being obtained ahead of time;
It determines to meet K3 image to be detected required as follows: the Target Segmentation image pair from the K2 Target Segmentation images
The variance answered is less than the 4th threshold value.
Above scheme using gradient orientation histogram can by multiple Target Segmentation images include object to be detected figure
As with include non-object to be detected image be distinguished, when Target Segmentation image obtain variance be not less than the 4th threshold value, then
The texture of the texture information and practical object to be detected that show the Target Segmentation image has significant difference, therefore the Target Segmentation
Image will be removed, and finally leave the K3 Target Segmentation images met the requirements.
It is described to calculate separately a not bending moment of the corresponding d of every image to be detected, packet in a kind of possible embodiment
Include: calculate separately every image to be detected in the horizontal direction and along the vertical direction on second order gradient, and utilize hu not bending moment meter
Calculating multiple not bending moments of the second order gradient, every image to be detected obtains a not bending moment of total D, wherein D is positive integer, and D >
d;The principal component that a not bending moment of the D is determined using Principal Component Analysis Algorithm obtains d not bending moments.
The effect of second order gradient is that smooth region zero, while prominent image to be detected will be changed in image to be detected
The part of middle mutation, it is to be detected if product surface has defect since the surface defect of product is typically all lofty
Image by two gradiometers calculation after defect part will be projected so that not the value of bending moment will with it is free of surface defects
Standard not bending moment value difference away from larger, and then be easy to differentiate in the image to be detected with the presence or absence of defect, in addition, utilizing PCA
Algorithm, by the D of script, bending moment dimensionality reduction is not to d, and one can play the role of dimensionality reduction, reduce characteristic parameter, so that finally
Bending moment can not accurately reflect the structure change of image to be detected more to d of acquisition, secondly can reduce calculation amount, improve and calculate
Method speed.
In a kind of possible embodiment, the multiple object to be detected is multiple shapes of the product surface to be detected
The identical protrusion of shape.For example, product to be detected is bellows, the single ripple on bellows is raised in product surface.
Second aspect, the embodiment of the present application provide a kind of detection device of surface defects of products, and described device includes: to obtain
Module, for obtain include product to be detected target image, wherein the product surface to be detected have it is multiple to be detected
Object;Image detection module obtains multiple image to be detected for handling the target image, wherein every to be checked
It include the single object to be detected of product to be detected in altimetric image;Defects detection module, for calculating separately every mapping to be checked
As a not bending moment of corresponding d, and it is corresponding with the standardized product image being obtained ahead of time according to a not bending moment of the d of every image to be detected
D standard not bending moment, surface defects detection is carried out to the product to be detected.
The third aspect, the embodiment of the present application provide a kind of storage medium, are stored with computer program on the storage medium,
It executes when the computer program is run by processor such as possible embodiment institute any in first aspect or first aspect
The method stated.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application will make below to required in the embodiment of the present application
Attached drawing is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore should not be seen
Work is the restriction to range, for those of ordinary skill in the art, without creative efforts, can be with
Other relevant attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow chart of the detection method of surface defects of products provided by the embodiments of the present application;
Fig. 2 is the specific flow chart of step 102 in the detection method of the application;
Fig. 3 is the specific flow chart of step 103 in the detection method of the application;
Fig. 4 is the specific flow chart of step 1021-1023 in the detection method of the application;
Fig. 5 is the schematic diagram of the detection system of surface defects of products provided by the embodiments of the present application;
Fig. 6 is the schematic diagram of the detection device of surface defects of products provided by the embodiments of the present application;
Fig. 7 is the schematic diagram of electronic equipment provided by the embodiments of the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application is described.
The embodiment of the present application introduces a kind of detection method of surface defects of products, please refers to Fig. 1, and this method includes following step
It is rapid:
Step 101: acquisition includes the target image of product to be detected.
In the present embodiment, product to be detected refers to that surface has the product to be processed of multiple objects to be detected, for example, to be checked
Surveying object can be, multiple patterns of product surface spray printing, alternatively, the periodic undulations due to product surface changes, on surface
The multiple protrusions or sunk area formed.The possible product to be detected of one kind is bellows, and bellows surface has multiple shapes
The identical ripple of shape, wherein single ripple can be used as an object to be detected, therefore this implementation can be used in corrugated pipe products
The method that example provides detects its surface defect.
Step 102: the target image being handled, multiple image to be detected are obtained.
It include the single object to be detected of product to be detected, such as bellows in the present embodiment, in every image to be detected
On single ripple image.
In one embodiment, referring to Fig. 2, step 102 be may include steps of:
Step 1021: Threshold segmentation being carried out to target image, obtains K1 connected region, and from K1 connected region really
Make K2 connected region for meeting the first preset requirement.
The present embodiment obtains corresponding bianry image, each object to be detected after carrying out Threshold segmentation to target image
The connected region of specific shape will be will form in the bianry image, extract the connected region in bianry image, be obtained K1
Connected region.
Multiple object shapes to be detected on product to be detected are identical, therefore the connected region tool formed in bianry image
There is similar shape.It also include non-object to be detected in K1 connected region in addition to the connected region of object to be detected
Therefore interference region by the way that the first preset requirement relevant to the form parameter of connected region is arranged, can be distinguished to be detected
Object and non-object to be detected, the biggish interference region of connected region difference with object to be detected is filtered out, K2 is finally obtained
A connected region has the function that tentatively to screen out.
Step 1022: determining the mass center of each connected region in K2 connected region, and mesh is mapped to according to each mass center
The position of logo image obtains K2 Target Segmentation images.
After step 1021, connected region is K2 by K1 initial variation.It determines in K2 connected region respectively
The mass center of each connected region, and center of mass point is mapped in target image, then respectively to mesh centered on each center of mass point
Logo image is split, and obtains K2 Target Segmentation images;It wherein, include on product to be detected in partial target segmented image
Single object to be detected includes the non-object to be detected on product to be detected in another part Target Segmentation image, that is to say, that
There is the interference image of non-object to be detected to be mixed into K2 Target Segmentation images, step 1023 is used for from K2 Target Segmentation figures
The image including object to be detected is filtered out as in.
Step 1023: K3 image to be detected for meeting the second preset requirement are determined from K2 Target Segmentation images.
For K2 Target Segmentation images, since object to be detected and the texture of non-object to be detected in the picture are clearly
It is different, by the way that relevant to the texture information of Target Segmentation image the second preset requirement is arranged, can will with it is to be detected right
The biggish interference image of the texture difference of elephant is filtered out, have the function that it is secondary screen out, finally obtain K3 really for lacking
Fall into image to be detected of detection.
In above-mentioned steps 1021-1023, K1, K2, K3 are positive integer, and K1 > K2 > K3.
After step 102, it executes step 103: calculating separately the corresponding d not bending moment, and root of every image to be detected
According to d not bending moment d standard corresponding with the standardized product image not bending moment of every image to be detected, treats testing product and carry out
Surface defects detection.
The calculating process of d not bending moment are as follows: calculate separately every image to be detected in the horizontal direction and along the vertical direction on
Second order gradient, and bending moment does not calculate multiple not bending moments of the second order gradient using hu, i.e., in the horizontal direction with vertical direction point
B not bending moments are not obtained, and every image to be detected obtains total d not bending moment, d=2b.Alternatively, another calculating process are as follows: In
Horizontal direction and vertical direction calculate separately to obtain C not bending moments, and every image to be detected obtains a not bending moment of total D, D=2C,
Wherein, b, d, C, D are positive integer, and D > d;It utilizes principal component analysis (Principal Component Analysis, PCA)
Algorithm determines the principal component of a not bending moment of the D, obtains d not bending moments.The portion in a not bending moment of D obtained due to product to be detected
Point not bending moment parameter may practical bring effect it is lower, or even be also possible to interfere subsequent calculate, using PCA algorithm,
By the D of script, to d, one can not play the role of dimensionality reduction, reduce characteristic parameter bending moment dimensionality reduction, so that finally obtain
A not bending moment of d can more accurately reflect the structure change of image to be detected, secondly can reduce calculation amount, improve algorithm speed
Degree.
The effect of second order gradient is that smooth region zero will be changed in image to be detected in the present embodiment, prominent simultaneously
The part being mutated in image to be detected, since the surface defect of product is typically all lofty, if product surface has defect,
So image to be detected by two gradiometers calculation after defect part will be projected, image can be characterized without bending moment
Geometrical characteristic, so that the value of bending moment away from larger, and then will not be easy to sentence with the value difference of standard free of surface defects not bending moment
It whether there is defect not in the image to be detected.
In one embodiment, referring to Fig. 3, step 103 includes:
Step 1031: calculating separately the d of every image to be detected not variance of the bending moment relative to d standard not bending moment.
Bending moment is obtained according to the standardized product image being obtained ahead of time to d standard, and standardized product refers to and production to be detected
The product to be processed free of surface defects of the same batch of product, same model, standardized product image and target image are in same inspection
It surveys to shoot in environment and obtain.Abovementioned steps identical with target image are executed to standardized product image, it is to be checked to obtain K3 standards
Altimetric image, every standard image to be detected are obtained d not bending moments, then calculate the average value of each not bending moment, obtain d mark
Quasi- not bending moment.
D not variance of the bending moment relative to d standard not bending moment are as follows:
Wherein, mpIndicate corresponding p-th of the not bending moment of the image to be detected, MpIndicate that standardized product image is p-th corresponding
Standard not bending moment.
Step 1032: if the variance of any image to be detected in multiple image to be detected is greater than first threshold, really
There are surface defects for fixed product to be detected.
K3 variances sigma is obtained in K3 image to be detected1, threshold value t1 is set, when any one variance in K3 variance is big
In threshold value t1, then show that the product surface to be detected has defect.Optionally, it is lacked when surface occurs in a certain product to be detected of discovery
When falling into, the terminal that result is sent to corresponding staff in time can be will test, staff is reminded to verify.For convenient for
Staff fast and accurately observes defect area, while obtaining image to be detected, will record each to be detected
Object to be detected corresponds to the position in target image in image, will when the variance of a certain image to be detected is greater than threshold value t1
The position for marking out the object to be detected there are surface defect in the target image, saves the observing time of staff, thus
Improve working efficiency.
Both it should be noted that this programme calculates the variance of not bending moment and standard not bending moment, be substantially intended to determine
Extent of deviation, therefore, actually step 1031 not necessarily calculates variance, other are any can to characterize this extent of deviation
Index can be applied in the above-mentioned calculating of the present embodiment.
It should be understood that method provided in this embodiment can also be used for the detection of variform multiple objects to be detected, for example,
While obtaining K3 image to be detected, record the number of each image to be detected, the number can characterize wherein to
Test object corresponds to the position in target image, is calculating a not bending moment of the corresponding d of image to be detected and d standard not bending moment
Variance when, d standard used by different image to be detected not bending moment be not be identical, but according to mapping to be checked
The number of picture finds standard image to be detected of reference numeral in standardized product image, and according to the standard image to be detected
A not bending moment of d obtains d standard not bending moment.
Optionally, referring to Fig. 4, a kind of specific embodiment of step 1021-1023 introduced below, to realize to be detected
The acquisition of image, includes the following steps:
Step 201: target image being filtered, filtered gray level image is obtained.
For example, being filtered using Gabor filter to target image, the parameter configuration of Gabor filter is as follows: wavelength
λ is set as the pixel size of an object to be detected horizontal length in the picture, for example, a section ripple of bellows is in level side
To being 32 pixels, then wavelength is set as 32;Remaining parameter (such as direction, coefficient of variation, length-width ratio, phase offset etc.) can be with
It is configured according to actual needs, the present embodiment does not limit.
Step 202: according to the size relation of the gray value of pixel each in gray level image and second threshold, obtaining target
The bianry image of image.
To filtered gray level image Imggb(i, j) carries out Threshold segmentation processing, obtains a bianry image Img (i, j),
It is as follows that bianry image obtains formula:
Wherein, Imggb(i, j) indicates the gray value of the pixel of the i-th row jth column in image, threshold value t2=q*max
(Imggb(i, j)), max (Imggb(i, j)) it is maximum gradation value in the gray level image, the value range of q is 0 to 1, one
In kind embodiment, any number in 0.3-0.5 is taken.
Step 203: obtaining K1 connected region in bianry image.
Step 204: calculating the area of each connected region in K1 connected region, and with the connected region with identical
Standard second order center away from elliptical eccentricity.
Step 205: from determining that the area of connected region is located in first threshold range and is connected in K1 connected region
The corresponding eccentricity in region is not less than K2 connected region of third threshold value.
Feature Descriptor of the area of connected region with corresponding eccentricity as preliminary screening connected region.Same shape
Object to be detected in filtering, after binaryzation, the area of connected region and have in identical standard second order with the connected region
The elliptical eccentricity of heart square can fluctuate in a certain range, but should totally be not much different.Area and eccentricity are two mutual
The form parameter of benefit can measure the size and shape of connected region respectively, by the threshold range of setting area, can exclude
The excessive connected region with object area gap to be detected, meanwhile, the eccentricity of connected region need to be not less than the third threshold of setting
Value, for example the single ripple of bellows is a partially flat-shaped shape, then third threshold value can be set to it is slightly bigger
Numerical value, finally, while the connected region for meeting area and eccentricity requirement is screened out, obtains the K2 connections met the requirements
Region.
First threshold range and third threshold value are determined according to the concrete condition after the filtering of target image, binary conversion treatment
It is fixed, area and eccentricity can be determined according to the processing result of current target image or the processing result of standardized product image
Threshold value.In one embodiment, first threshold range is set as 50-1000, and the third threshold value of eccentricity is set as 0.95, i.e.,
Area is greater than 1000 or is identified as the region of non-object to be detected less than 50 connected region, and identical, eccentricity is less than
0.95 connected region is also identified as the region of non-object to be detected.
Step 206: target image being split centered on the mass center of K2 connected region, obtains K2 original segmentations
Image.
In the embodiment, before obtaining Target Segmentation image, target image is divided into the identical K2 of size first
Image is opened, that is, the rectangle frame of first size is used to select image-region, the ruler of rectangle frame by center frame of the mass center of connected region
Very little size should can frame select at least two objects to be detected of product to be detected to get to K2 original segmented images;But
It may be noted that the size of rectangle frame can not be excessive, it, will packet in adjacent original segmented image if too many object to be detected of frame choosing
Containing more repeating part, meanwhile, the corresponding rectangle frame of object to be detected positioned at edge will the too many non-object to be detected of frame choosing
Region, be unfavorable for the application of subsequent gradients direction histogram.
Step 207: calculating the gradient orientation histogram of each original segmented image, and determine in gradient orientation histogram
The largest number of gradient directions of pixel.
The gradient direction of each pixel in each original segmented image is calculated, and according to the gradient direction divided in advance
Section the pixel number in each section is counted, obtain gradient orientation histogram, for example, gradient direction from-
180 ° to 180 ° are divided, and the section of a gradient direction is marked off every 10 °, and the division numbers of gradient direction are more, point
Distinguish that ability is higher.Then, it is determined that any gradient direction in the largest number of sections of pixel is the gradient orientation histogram
Main gradient direction.
Step 208: object to be detected is corrected in original segmented image according to the gradient direction and preset rotation direction
Angle, the segmented image after being corrected.
Angle correction is carried out to original segmented image according to determining main gradient direction and preset rotation direction, that is, will
Angle " just " of the object to be detected in original segmented image comes, and finally makes object to be detected angle in the picture, big
Small and shape reaches unanimity, and is convenient for subsequent calculating.
It is noted that the present embodiment is during above-mentioned angle correction, be not whole picture target image is rotated, and
It is that image rotation is carried out using discrete way.It is single, only that discrete way refers to that the target image by product to be detected is divided into
Vertical original segmented image, and multiple discrete original segmented images are rotated respectively, the purpose handled in this way is, right
Entire image is handled the calculating and rotation speed that can undoubtedly slow down image, and discrete multiple original segmented images are due to region
It is small, so that the calculating of histogram and the rotation of image also can faster, so as to promote the speed of angle correction.Further
, test object is treated due to the present embodiment and has carried out angle rotation, so that object to be detected is located at the segmented image after correction
Center, therefore object to be detected can be cut out in step 209 minimally to introduce non-to be checked with the smallest rectangle frame
The ingredient of object is surveyed, influence of noise is reduced.
Step 209: by correct after segmented image mass center centered on respectively to each correction after segmented image into
Row is cut, and obtains K2 Target Segmentation images.
Postrotational image is cut, that is, uses the rectangle frame of the second size using the mass center of segmented image as center frame
Select image-region, the size of rectangle frame should can frame select the single object to be detected of product to be detected, it is final
To K2 Target Segmentation images.
It should be understood that above-mentioned steps 206-209 is only a kind of possible embodiment for obtaining Target Segmentation image, actually answer
In, the object to be detected of product to be detected can be kept direct in imaging by adjusting the camera of photographic subjects image
With the horizontal direction parallel of target image or vertical, then can be omitted above-mentioned angle correction step, that is, is directlyed adopt
The rectangle frame of two sizes, segmentation obtains K2 Target Segmentation images centered on the mass center of K2 connected region.
Step 210: calculating the gradient orientation histogram of Target Segmentation image, and calculate the gradient of every Target Segmentation image
Variance of the direction histogram relative to normal gradients direction histogram.
The step of calculating gradient orientation histogram is identical as above-mentioned steps 207, does not repeat to repeat herein, and final statistics obtains
Pixel number of the Target Segmentation image on different gradient directions.- 180 ° to 180 ° are divided into N number of by gradient orientation histogram
The section of gradient direction.
Variance of the gradient orientation histogram of Target Segmentation image relative to normal gradients direction histogram are as follows:
Wherein, nqFor the corresponding pixel in q-th of gradient direction section in the gradient orientation histogram of Target Segmentation image
Number, NqFor q-th of gradient direction section number of corresponding pixels in the normal gradients direction histogram of standardized product image.
Normal gradients direction histogram is to execute abovementioned steps identical with target image to standardized product image, obtains K3 standard mesh
Segmented image is marked, every standard target segmented image obtains a gradient orientation histogram, then calculates each gradient direction area
Between pixel number mean value after obtain.
Step 211: K3 Target Segmentation images of the variance less than the 4th threshold value are determined from K2 Target Segmentation images,
Obtain K3 image to be detected.
The step for purpose be by obtain multiple Target Segmentation images in include object to be detected image with include
The image of non-object to be detected is distinguished, and the Feature Descriptor for being used to distinguish is the distribution of gradient direction in image, works as mesh
It marks the variance that segmented image obtains and is not less than the 4th threshold value, then show that the texture information of the Target Segmentation image and reality are to be detected
The texture of object has significant difference, therefore the Target Segmentation image will be removed, and finally leaves the K3 targets met the requirements
Segmented image opens image to be detected to get to the K3 for final defects detection.
In the present embodiment, gradient direction can describe the texture information in Target Segmentation image well, simultaneously for light
The variation of line can have good robustness.In the environment of actual photographed product to be detected, not can guarantee has well
Illumination condition, therefore multiple objects to be detected are most likely in the state that part is bright, part is gloomy in obtained target image,
But due to the shape of object to be detected, texture be it is identical, the distribution of the gradient direction of multiple objects to be detected be variation
Little, so as to guarantee the reliability and accuracy of the above method.
It should be understood that above-mentioned steps 201-211 provides only a kind of possible embodiment, other than above scheme,
It is not excluded for realize the acquisition to image to be detected by other means.For example, the side based on template matching can be used
Method, i.e., using an image to be detected template to target image carry out relevant calculation, from target image search with it is to be detected
The similar target of image, and threshold process is carried out to the value after relevant calculation and obtains corresponding bianry image, it then can be again to it
The screening of connected region and the segmentation of image are carried out, to obtain image to be detected.
The detection method of surface defects of products provided in this embodiment can be used in treating on testing product production line
Line defect detection, i.e., after completing the process a certain product, the image of the captured in real-time product and the inspection for obtaining the product in real time
It surveys as a result, furthermore, it can also be used to the offline inspection of a certain product, i.e., after the target image that product to be detected is opened in acquisition one,
The surface defect of the product is detected.
Detection method provided in this embodiment has a characteristic that in actual conditions that the surface defect of product may have more
The different defect kind of kind, shape difference, different sizes, but the texture information of product script can be destroyed, it is lacked so that existing
Sunken object to be detected and other flawless objects has significant difference, and the present embodiment is with the structure change of image to be detected
Foundation can directly pick out defective product, therefore not need to know the reason of causing defect and defect
The pre-information such as type, shape, size, can directly select defective product, and versatility is stronger.
Referring to Fig. 5, the present embodiment also provides a kind of detection system 300 of surface defects of products, for realizing aforementioned implementation
The detection method of example.Detection system 300 includes processing terminal 301 and multiple cameras 302, and each camera 302 is separately positioned on
Around product to be detected, for shooting the target image of product different surfaces to be detected, multiple target images enable
Completely reflect the surface information of product to be detected.Each camera 302 is communicated to connect with processing terminal 301 respectively, processing terminal 301
Multiple target images that multiple cameras 302 are shot are received, and each target image is executed in above method embodiment respectively
Detecting step, if any target image show the product to be detected there are surface defect, i.e., exportable product tool
Defective testing result.Processing terminal 301, which can be laptop, tablet computer, desktop computer, server etc., to be had
Any one in the calculating equipment of image-capable.
Based on the same inventive concept, referring to Fig. 6, a kind of detection dress of surface defects of products is also provided in the embodiment of the present application
400 are set, which includes:
Obtain module 401, for obtain include product to be detected target image, wherein the product table to be detected
Face has multiple objects to be detected;
Image detection module 402 obtains multiple image to be detected for handling the target image, wherein every
Open the single object to be detected in image to be detected including product to be detected;
Defects detection module 403, for calculating separately a not bending moment of the corresponding d of every image to be detected, and according to every
The d of image to be detected not bending moment d standard corresponding with the standardized product image being obtained ahead of time not bending moments, to described to be detected
Product carries out surface defects detection.
Optionally, the shape of the multiple object to be detected is identical, and defects detection module 403 is specifically used for: calculating separately
The d of every image to be detected not variances of the bending moment relative to the d standard not bending moment;If in multiple described image to be detected
Any image to be detected variance be greater than first threshold, it is determined that there are surface defects for the product to be detected.
Optionally, the shape of the multiple object to be detected is identical, and image detection module 402 is specifically used for: to the mesh
Logo image carries out Threshold segmentation, obtains K1 connected region, and determines that meeting first presets from the K1 connected region
It is required that K2 connected region, wherein first preset requirement is related to the form parameter of the connected region, and K1, K2 are
Positive integer, and K1 > K2;It determines the mass center of each connected region in K2 connected region, and is mapped to according to each mass center described
The position of target image obtains K2 Target Segmentation images;Determine to meet second from the K2 Target Segmentation images in advance
If it is required that K3 image to be detected, wherein the second preset requirement is related to the texture information of the Target Segmentation image, and K3 is
Positive integer, and K2 > K3.
Optionally, image detection module 402 is specifically used for: being filtered to the target image, obtains filtered ash
Spend image;According to the size relation of the gray value of pixel each in the gray level image and second threshold, the target is obtained
The bianry image of image, wherein the second threshold is the product of the maximum gradation value in predetermined coefficient and the gray level image,
Single object to be detected on product to be detected forms connected region in the bianry image;It obtains in the bianry image
K1 connected region.
Optionally, image detection module 402 is specifically used for: calculating the face of each connected region in the K1 connected region
Product, and with the connected region have identical standard second-order central away from elliptical eccentricity;From the K1 connected region
In determine to meet the K2 connected region required as follows: the area of the connected region is located in first threshold range and described
The corresponding eccentricity of connected region is not less than third threshold value.
Optionally, image detection module 402 is specifically used for: to the target figure centered on the mass center of K2 connected region
As being split, K2 original segmented images are obtained;Calculate the gradient orientation histogram of each original segmented image, the ladder
Degree direction histogram indicates the statistics of pixel number of the original segmented image on different gradient directions;Determine the ladder
The largest number of gradient directions of pixel in direction histogram are spent, and according to the gradient direction and preset rotation direction house of correction
State angle of the object to be detected in the original segmented image, the segmented image after being corrected;With the segmentation figure after correction
The segmented image after each correction is cut respectively centered on the mass center of picture, obtains K2 Target Segmentation images.
Optionally, image detection module 402 is specifically used for: the gradient orientation histogram of the Target Segmentation image is calculated,
The gradient orientation histogram indicates the statistics of pixel number of the Target Segmentation image on different gradient directions;It calculates
The gradient orientation histogram of every Target Segmentation image is relative to the corresponding normal gradients side of standardized product image being obtained ahead of time
To the variance of histogram;It determines to meet K3 image to be detected required as follows: institute from the K2 Target Segmentation images
The corresponding variance of Target Segmentation image is stated less than the 4th threshold value.
Optionally, defects detection module 403 is specifically used for: calculating separately every image to be detected in the horizontal direction and along perpendicular
The upward second order gradient of histogram, and bending moment does not calculate multiple not bending moments of the second order gradient, every image to be detected using hu
Obtain a not bending moment of total D, wherein D is positive integer, and D > d;The master of a not bending moment of the D is determined using Principal Component Analysis Algorithm
Ingredient obtains d not bending moments.
Optionally, the multiple object to be detected is the identical protrusion of multiple shapes of the product surface to be detected, than
Such as, product to be detected is bellows, and the single ripple on bellows is raised in product surface, the single ripple conduct on bellows
One object to be detected.
The detection device of above-mentioned offer is identical as the technical effect of the basic principle of former approach embodiment and generation, for letter
It describes, the present embodiment part does not refer to place, can refer to the corresponding contents in above-mentioned embodiment of the method, does not do herein superfluous
It states.
The embodiment of the present application also provides a kind of storage medium, is stored with program on the storage medium, when the program is processed
The step of detection method of the surface defects of products provided such as the above embodiments of the present application is provided when device is run.
Referring to Fig. 7, the present embodiment provides a kind of electronic equipment 500, including processor 501 and memory 502, memory
At least one instruction, at least a Duan Chengxu, code set or instruction set, at least one instruction, an at least Duan Cheng are stored in 502
Sequence, code set or instruction set are loaded and are executed by processor 501, to realize the inspection of surface defects of products provided by the above embodiment
Survey method.Electronic equipment 500 can also include communication bus 503, wherein processor 501 and memory 502 pass through communication bus
503 complete mutual communication.Memory 502 may include high-speed random access memory (as caching), can also include
Nonvolatile memory, for example, at least a disk memory, flush memory device or other volatile solid-state parts.It is logical
Letter bus 503 is to connect the circuit of described element and realize transmission between these elements.For example, processor 501 is logical
It crosses communication bus 503 and receives order from other elements, decode the order received, executed according to decoded order and calculate or count
According to processing.
Electronic equipment 500 can correspond to the processing terminal in the detection system of the said goods surface defect, for obtaining
The target image taken carries out image procossing, to realize the detection for treating testing product surface defect.
In embodiment provided herein, it should be understood that disclosed device and method, it can be by others side
Formula is realized.The apparatus embodiments described above are merely exemplary, for example, the division of the module, only one kind are patrolled
Function division is collected, there may be another division manner in actual implementation, in another example, multiple units or components can combine or can
To be integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual
Coupling, direct-coupling or communication connection can be INDIRECT COUPLING or communication link by some communication interfaces, device or unit
It connects, can be electrical property, mechanical or other forms.Each functional module in the embodiment of the present application can integrate to be formed together
One independent part, is also possible to modules individualism, can also be integrated to form one with two or more modules
Independent part.
If it should be noted that function is realized in the form of software function module and sells or make as independent product
Used time can store in a computer readable storage medium.Based on this understanding, the technical solution essence of the application
On in other words the part of the part that contributes to existing technology or the technical solution can embody in the form of software products
Out, which is stored in a storage medium, including some instructions are used so that a computer equipment
(can be personal computer, server or the network equipment etc.) execute each embodiment the method for the application whole or
Part steps.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (Read-Only Memory, ROM) with
Machine accesses various Jie that can store program code such as memory (Random Access Memory, RAM), magnetic or disk
Matter.
Herein, relational terms such as first and second and the like be used merely to by an entity or operation with it is another
One entity or operation distinguish, and without necessarily requiring or implying between these entities or operation, there are any this reality
Relationship or sequence.
The above description is only an example of the present application, the protection scope being not intended to limit this application, for ability
For the technical staff in domain, various changes and changes are possible in this application.Within the spirit and principles of this application, made
Any modification, equivalent substitution, improvement and etc. should be included within the scope of protection of this application.
Claims (11)
1. a kind of detection method of surface defects of products, which is characterized in that the described method includes:
Acquisition includes the target image of product to be detected, wherein the product surface to be detected has multiple objects to be detected;
The target image is handled, multiple image to be detected are obtained, wherein includes to be detected in every image to be detected
The single object to be detected of product;
A not bending moment of the corresponding d of every image to be detected is calculated separately, and according to a not bending moment of the d of every image to be detected and in advance
The corresponding d standard of the standardized product image first obtained not bending moment carries out surface defects detection to the product to be detected,
In, d is positive integer.
2. the method according to claim 1, wherein the shape of the multiple object to be detected is identical, according to every
Not bending moment d standard corresponding with the standardized product image the being obtained ahead of time not bending moment of d of image to be detected, to described to be checked
It surveys product and carries out surface defects detection, comprising:
Calculate separately the d of every image to be detected not variance of the bending moment relative to the d standard not bending moment;
If the variance of any image to be detected in multiple described image to be detected is greater than first threshold, it is determined that described to be checked
Surveying product, there are surface defects.
3. described right the method according to claim 1, wherein the shape of the multiple object to be detected is identical
The target image is handled, multiple image to be detected are obtained, comprising:
Threshold segmentation is carried out to the target image, obtains K1 connected region, and determine from the K1 connected region
Meet K2 connected region of the first preset requirement, wherein the form parameter of first preset requirement and the connected region
Correlation, K1, K2 are positive integer, and K1 > K2;
It determines the mass center of each connected region in K2 connected region, and maps to the position of the target image according to each mass center
It sets, obtains K2 Target Segmentation images;
K3 image to be detected for meeting the second preset requirement are determined from the K2 Target Segmentation images, wherein second
Preset requirement is related to the texture information of the Target Segmentation image, and K3 is positive integer, and K2 > K3.
4. according to the method described in claim 3, it is characterized in that, described carry out Threshold segmentation, acquisition to the target image
K1 connected region, comprising:
The target image is filtered, filtered gray level image is obtained;
According to the size relation of the gray value of pixel each in the gray level image and second threshold, the target image is obtained
Bianry image, wherein the second threshold is the product of the maximum gradation value in predetermined coefficient and the gray level image, to be checked
The single object to be detected surveyed on product forms connected region in the bianry image;
Obtain K1 connected region in the bianry image.
5. according to the method described in claim 3, it is characterized in that, described determine satisfaction from the K1 connected region
K2 connected region of one preset requirement, comprising:
The area of each connected region in the K1 connected region is calculated, and there is identical standard two with the connected region
Rank center away from elliptical eccentricity;
It determines to meet the K2 connected region required as follows: the area position of the connected region from the K1 connected region
In in first threshold range and the corresponding eccentricity of the connected region be not less than third threshold value.
6. according to the method described in claim 3, it is characterized in that, described map to the target image according to each mass center
Position obtains K2 Target Segmentation images, comprising:
The target image is split centered on the mass center of K2 connected region, obtains K2 original segmented images;
The gradient orientation histogram of each original segmented image is calculated, the gradient orientation histogram indicates the original segmentation
The statistics of pixel number of the image on different gradient directions;
It determines the largest number of gradient directions of pixel in the gradient orientation histogram, and according to the gradient direction and presets
Correct angle of the object to be detected in the original segmented image, the segmented image after being corrected in direction of rotation;
The segmented image after each correction is cut respectively centered on the mass center of segmented image after correcting, obtains K2
Open Target Segmentation image.
7. according to the method described in claim 3, it is characterized in that, described determine completely from the K2 Target Segmentation images
K3 image to be detected of the second preset requirement of foot, comprising:
The gradient orientation histogram of the Target Segmentation image is calculated, the gradient orientation histogram indicates the Target Segmentation figure
As the statistics of the pixel number on different gradient directions;
The gradient orientation histogram of every Target Segmentation image is calculated relative to the corresponding mark of standardized product image being obtained ahead of time
The variance of quasi- gradient orientation histogram;
It determines to meet K3 image to be detected required as follows: the Target Segmentation figure from the K2 Target Segmentation images
As corresponding variance is less than the 4th threshold value.
8. the method according to claim 1, wherein described calculate separately every image to be detected corresponding d
Not bending moment, comprising:
Calculate separately every image to be detected in the horizontal direction and along the vertical direction on second order gradient, and utilize hu not bending moment meter
Calculating multiple not bending moments of the second order gradient, every image to be detected obtains a not bending moment of total D, wherein D is positive integer, and D >
d;
The principal component that a not bending moment of the D is determined using Principal Component Analysis Algorithm obtains d not bending moments.
9. method according to claim 1-8, which is characterized in that the multiple object to be detected is described to be checked
Survey the identical protrusion of multiple shapes of product surface.
10. a kind of detection device of surface defects of products, which is characterized in that described device includes:
Obtain module, for obtains include product to be detected target image, wherein the product surface to be detected is with more
A object to be detected;
Image detection module obtains multiple image to be detected for handling the target image, wherein every to be checked
It include the single object to be detected of product to be detected in altimetric image;
Defects detection module, for calculating separately a not bending moment of the corresponding d of every image to be detected, and according to every mapping to be checked
The d of picture not bending moment d standard corresponding with the standardized product image being obtained ahead of time not bending moments, to the product progress to be detected
Surface defects detection.
11. a kind of storage medium, which is characterized in that be stored with computer program, the computer program on the storage medium
Such as claim 1-9 described in any item methods are executed when being run by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910831247.0A CN110517265B (en) | 2019-09-04 | 2019-09-04 | Method and device for detecting surface defects of product and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910831247.0A CN110517265B (en) | 2019-09-04 | 2019-09-04 | Method and device for detecting surface defects of product and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110517265A true CN110517265A (en) | 2019-11-29 |
CN110517265B CN110517265B (en) | 2022-03-01 |
Family
ID=68629640
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910831247.0A Active CN110517265B (en) | 2019-09-04 | 2019-09-04 | Method and device for detecting surface defects of product and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110517265B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111127571A (en) * | 2019-12-03 | 2020-05-08 | 歌尔股份有限公司 | Small sample defect classification method and device |
CN111598832A (en) * | 2020-04-01 | 2020-08-28 | 江门荣信电路板有限公司 | Slot defect marking method and device and storage medium |
CN111693533A (en) * | 2020-06-11 | 2020-09-22 | 南通通富微电子有限公司 | Workpiece surface quality detection method and device and appearance machine |
CN112001841A (en) * | 2020-07-14 | 2020-11-27 | 歌尔股份有限公司 | Image to-be-detected region extraction method and device and product defect detection system |
CN112001902A (en) * | 2020-08-19 | 2020-11-27 | 上海商汤智能科技有限公司 | Defect detection method and related device, equipment and storage medium |
CN113807378A (en) * | 2020-06-16 | 2021-12-17 | 纬创资通股份有限公司 | Training data increment method, electronic device and computer readable recording medium |
CN114419039A (en) * | 2022-03-28 | 2022-04-29 | 武汉市融科优品装饰材料有限公司 | Decorative wallpaper defect detection method based on template matching |
CN115880288A (en) * | 2023-02-21 | 2023-03-31 | 深圳市兆兴博拓科技股份有限公司 | Detection method and system for electronic component welding and computer equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508110A (en) * | 2011-10-10 | 2012-06-20 | 上海大学 | Texture-based insulator fault diagnostic method |
KR101361922B1 (en) * | 2012-10-19 | 2014-02-21 | 서울여자대학교 산학협력단 | Automatic detection system of porosity defect using density and shape information in industrial ct and controlling method therefor |
CN106600600A (en) * | 2016-12-26 | 2017-04-26 | 华南理工大学 | Wafer defect detection method based on characteristic matching |
CN106778734A (en) * | 2016-11-10 | 2017-05-31 | 华北电力大学(保定) | A kind of insulator based on rarefaction representation falls to go here and there defect inspection method |
CN107301422A (en) * | 2017-05-24 | 2017-10-27 | 广西师范大学 | A kind of bamboo cane method for sorting colors and system based on color characteristic and Hu squares |
CN109493339A (en) * | 2018-11-20 | 2019-03-19 | 北京嘉恒中自图像技术有限公司 | A kind of casting inner surface gas hole defect detection method based on endoscopic imaging |
CN109785285A (en) * | 2018-12-11 | 2019-05-21 | 西安工程大学 | A kind of insulator damage testing method based on oval feature fitting |
-
2019
- 2019-09-04 CN CN201910831247.0A patent/CN110517265B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508110A (en) * | 2011-10-10 | 2012-06-20 | 上海大学 | Texture-based insulator fault diagnostic method |
KR101361922B1 (en) * | 2012-10-19 | 2014-02-21 | 서울여자대학교 산학협력단 | Automatic detection system of porosity defect using density and shape information in industrial ct and controlling method therefor |
CN106778734A (en) * | 2016-11-10 | 2017-05-31 | 华北电力大学(保定) | A kind of insulator based on rarefaction representation falls to go here and there defect inspection method |
CN106600600A (en) * | 2016-12-26 | 2017-04-26 | 华南理工大学 | Wafer defect detection method based on characteristic matching |
CN107301422A (en) * | 2017-05-24 | 2017-10-27 | 广西师范大学 | A kind of bamboo cane method for sorting colors and system based on color characteristic and Hu squares |
CN109493339A (en) * | 2018-11-20 | 2019-03-19 | 北京嘉恒中自图像技术有限公司 | A kind of casting inner surface gas hole defect detection method based on endoscopic imaging |
CN109785285A (en) * | 2018-12-11 | 2019-05-21 | 西安工程大学 | A kind of insulator damage testing method based on oval feature fitting |
Non-Patent Citations (1)
Title |
---|
孔月萍,王亚安,王快社: "基于不变矩的带钢数字图像的缺陷检测算法", 《无损检测》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111127571A (en) * | 2019-12-03 | 2020-05-08 | 歌尔股份有限公司 | Small sample defect classification method and device |
CN111598832A (en) * | 2020-04-01 | 2020-08-28 | 江门荣信电路板有限公司 | Slot defect marking method and device and storage medium |
CN111693533A (en) * | 2020-06-11 | 2020-09-22 | 南通通富微电子有限公司 | Workpiece surface quality detection method and device and appearance machine |
CN111693533B (en) * | 2020-06-11 | 2023-01-20 | 南通通富微电子有限公司 | Workpiece surface quality detection method and device and appearance machine |
CN113807378A (en) * | 2020-06-16 | 2021-12-17 | 纬创资通股份有限公司 | Training data increment method, electronic device and computer readable recording medium |
CN112001841A (en) * | 2020-07-14 | 2020-11-27 | 歌尔股份有限公司 | Image to-be-detected region extraction method and device and product defect detection system |
CN112001902A (en) * | 2020-08-19 | 2020-11-27 | 上海商汤智能科技有限公司 | Defect detection method and related device, equipment and storage medium |
CN114419039A (en) * | 2022-03-28 | 2022-04-29 | 武汉市融科优品装饰材料有限公司 | Decorative wallpaper defect detection method based on template matching |
CN115880288A (en) * | 2023-02-21 | 2023-03-31 | 深圳市兆兴博拓科技股份有限公司 | Detection method and system for electronic component welding and computer equipment |
CN115880288B (en) * | 2023-02-21 | 2023-11-14 | 深圳市兆兴博拓科技股份有限公司 | Detection method, system and computer equipment for electronic element welding |
Also Published As
Publication number | Publication date |
---|---|
CN110517265B (en) | 2022-03-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110517265A (en) | A kind of detection method of surface defects of products, device and storage medium | |
CN105957082A (en) | Printing quality on-line monitoring method based on area-array camera | |
CN113570605B (en) | Defect detection method and system based on liquid crystal display panel | |
CN108520514B (en) | Consistency detection method for electronic elements of printed circuit board based on computer vision | |
CN106370671A (en) | PCB (printed circuit board) component detection system and method based on machine vision | |
CN112651968B (en) | Wood board deformation and pit detection method based on depth information | |
Mahalingam et al. | Pcb-metal: A pcb image dataset for advanced computer vision machine learning component analysis | |
CN106501272B (en) | Machine vision soldering tin positioning detection system | |
US8315457B2 (en) | System and method for performing multi-image training for pattern recognition and registration | |
CN113608378B (en) | Full-automatic defect detection method and system based on LCD (liquid crystal display) process | |
Wu et al. | Automated visual inspection of surface mounted chip components | |
CN110781913B (en) | Zipper cloth belt defect detection method | |
TWI765442B (en) | Method for defect level determination and computer readable storage medium thereof | |
CN110503638A (en) | Spiral colloid amount online test method | |
US6714670B1 (en) | Methods and apparatuses to determine the state of elements | |
CN115775246A (en) | Method for detecting defects of PCB (printed circuit board) components | |
CN116309518A (en) | PCB (printed circuit board) detection method and system based on computer vision | |
CN114820439A (en) | PCB bare board defect detection system and method based on AOI | |
CN116228780B (en) | Silicon wafer defect detection method and system based on computer vision | |
CN112541478A (en) | Insulator string stain detection method and system based on binocular camera | |
CN108898584B (en) | Image analysis-based full-automatic veneered capacitor welding polarity discrimination method | |
CN113822836A (en) | Method of marking an image | |
Chen et al. | Sub-pixel edge detection of led probes based on canny edge detection and iterative curve fitting | |
CN116543247A (en) | Data set manufacturing method and verification system based on photometric stereo surface reconstruction | |
CN114758158A (en) | Quality inspection method and system for PCB |
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 |