CN107240086B - A kind of fabric defects detection method based on integral nomography - Google Patents
A kind of fabric defects detection method based on integral nomography Download PDFInfo
- Publication number
- CN107240086B CN107240086B CN201610180226.3A CN201610180226A CN107240086B CN 107240086 B CN107240086 B CN 107240086B CN 201610180226 A CN201610180226 A CN 201610180226A CN 107240086 B CN107240086 B CN 107240086B
- Authority
- CN
- China
- Prior art keywords
- image
- gradient energy
- fabric
- distribution
- integral
- 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.)
- Active
Links
- 239000004744 fabric Substances 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 title claims abstract description 42
- 230000007547 defect Effects 0.000 title claims abstract description 36
- 238000009826 distribution Methods 0.000 claims abstract description 47
- 238000000034 method Methods 0.000 claims abstract description 39
- 238000004422 calculation algorithm Methods 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 20
- 238000010586 diagram Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 8
- 230000002950 deficient Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000002813 epsilometer test Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 abstract description 3
- 239000004753 textile Substances 0.000 abstract description 2
- 230000003044 adaptive effect Effects 0.000 abstract 1
- 230000004069 differentiation Effects 0.000 abstract 1
- 238000011897 real-time detection Methods 0.000 abstract 1
- 230000008859 change Effects 0.000 description 5
- 230000001133 acceleration Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30124—Fabrics; Textile; Paper
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Analysis (AREA)
Abstract
A kind of fabric defects detection method based on integral nomography, Defect Detection is used for using integral nomography rapidly extracting gradient energy statistical nature, first by carrying out image study to indefectible template, count its gradient energy feature distribution, distribution peaks are extracted, then adaptively seek differentiation of the threshold parameter for subsequent flaw；Then, the gradient energy of window where seeking each pixel by integrating nomography to image to be detected, in conjunction with the threshold parameter, determine current pixel point whether fault, determine whether present image is flaw fabric by counting the fault sum of entire image.One aspect of the present invention accelerates the principle of operation based on integrogram, the gradient energy feature distribution of rapidly extracting textile image, it realizes the real-time detection of fabric defects, on the other hand solves distribution peaks and obtain adaptive flaw decision threshold parameter, realize the accurate segmentation of fabric defects.The method of the present invention not only can guarantee real-time but also have higher accuracy.
Description
Technical Field
The invention relates to the technical field of machine vision and video image processing, in particular to a quick fabric flaw detection method based on an integral map algorithm.
Background
The defect detection of the traditional textile industry mostly takes artificial naked eye detection as a main part, however, the visual sense of human eyes is easy to fatigue, so that the detection is missed, the artificial observation efficiency is low, the labor cost is high, the problem is greatly incompatible with large-scale industrial production, and the defect detection of the fabric is automatically carried out by utilizing computer vision and an image processing algorithm, so that the problem can be effectively solved.
The method based on image filtering extracts fabric texture features in a frequency domain to detect flaws, such as Gabor filtering and wavelet transformation, because the flaw scale direction is uncertain, results in multiple scales and multiple directions are often required to be extracted as feature vectors during filtering, and even if a PCA dimension reduction method is adopted, the detection time of a single-frame image still needs tens of seconds; the method is based on signal statistics, the gray distribution characteristics of the fabric are counted in a spatial domain to identify flaws, such as Local Binary Pattern (LBP) characteristics, gray co-occurrence matrix characteristics, Regular Band (RB) characteristics and the like, the statistical characteristics have good robustness, but when the statistical characteristics are extracted, a plurality of pixel data in a neighborhood are generally required to be utilized, and if a proper acceleration strategy is not adopted, the integral operation amount is increased suddenly; the method for dividing the flaws by directly thresholding the fabric image is simple in operation and fast in operation, but is only effective for fabrics with uniform gray levels such as plain weaves and twill weaves and without texture patterns, and is easy to be interfered by noise. For the industrial application of the automatic flaw detection algorithm, both real-time performance and accuracy need to be satisfied, and according to statistics, only a small number of algorithms can satisfy the real-time performance, and less algorithms with the detection accuracy higher than 90% are needed.
The invention provides a quick fabric flaw detection method based on an integral graph. The integral graph is used for simplifying summation operation in image blocks with any size into three-time addition operation, gradient energy statistical characteristics are extracted rapidly, running time expenditure is reduced greatly, a kernel function is used for fitting asymmetric characteristic distribution to obtain a self-adaptive flaw judgment threshold, and accurate segmentation of flaw areas is achieved.
Disclosure of Invention
The invention aims to solve the problems that: the existing fabric flaw detection system relies on human eye observation, and the efficiency is low; the conventional method for detecting the fabric defects with high accuracy by various complex algorithms has large calculation amount and cannot meet the real-time requirement of industrial production; the existing method capable of rapidly detecting the fabric defects only can deal with simple fabric images and has poor effect on complex texture fabrics. In summary, the existing methods are difficult to achieve compatibility of high real-time performance and high accuracy.
The technical scheme of the invention is as follows: a fabric flaw detection method based on an integral graph algorithm is characterized in that a summation operation in an image block with any size is simplified into a cubic addition operation by utilizing the integral graph algorithm, so that gradient energy characteristics are rapidly extracted for flaw detection, and the method specifically comprises the following steps: firstly, image learning is carried out on a flawless template, gradient energy characteristic distribution of the flawless template is counted, the obtained characteristic distribution is asymmetric, kernel function fitting characteristic distribution is adopted, a peak value of distribution is extracted by combining a mean shift method, and then a threshold parameter is obtained by the peak value in a self-adaptive mode, wherein the threshold parameter is used for distinguishing subsequent flaws; then, the gradient energy of a detection window where each pixel point is located is obtained through an integral graph algorithm for an image to be detected, whether the current pixel point is a defect or not is judged by combining the threshold parameter, whether the current image is a defective fabric or not is judged by counting the total number of the defects of the whole image,
the extraction method of the gradient energy features comprises the following steps: firstly, a gradient image G (x, y) of an original image F (x, y) is obtained, then a gradient energy characteristic image E (x, y) of the G (x, y) is obtained by utilizing an integral image algorithm, and for any pixel point (x, y), the energy characteristic is pixel integration in a window area with the point (x, y) as the center and with the size dw dh.
The specific steps of utilizing an integral graph algorithm to obtain a gradient energy characteristic graph E (x, y) of G (x, y) are as follows:
1) obtaining an integral diagram I (x, y) of G (x, y)
I(x,y)＝I(x-1,y)+I(x,y-1)-I(x-1,y-1)+G(x,y) (1)
2) And (3) obtaining a gradient energy characteristic diagram E (x, y) of an arbitrary point (x, y) according to the characteristics of the integral diagram:
the method for detecting the fabric defects comprises the following specific steps:
1) training and learning threshold parameters for flaw detection using flaw-free template images
Establishing a training set by using images of the flawless fabric, and obtaining a gradient energy characteristic diagram E of the images for each imagetrain(x, y) fitting E with a Kernel functiontrainObtaining kernel density probability distribution P (e) by gradient energy distribution in (x, y), iteratively solving an extreme value of the kernel density probability distribution P (e) by using a mean shift method, obtaining the position of a peak value of the gradient energy distribution, marking the position as mu, dividing the energy distribution into a left part and a right part according to the peak value, respectively calculating two variances sigma1、σ2The parameters mu and sigma are obtained for each flawless image of the same type of fabric in the training set1、σ2Then, the average value is obtained As a threshold parameter in the final defect detection process for this type of fabric;
2) obtaining a gradient energy characteristic diagram E of each image to be detectedtest(x, y) if Etest(x, y) is determined as a defect if the following equation is satisfied:
wherein,the threshold parameter obtained in the step 1), α is a control coefficient;
finally, counting the total number of the defects, if the total number of the defects is larger than a set threshold value TdThe map is judged to be defective.
The nuclear density probability distribution P (e) is:
where N is the total number of pixels in the inpainted image, { en1,2, 3.. N is the gradient energy data of each point in the inpaintless image, b is the kernel function bandwidth, c0For the normalization coefficient, k (z) is a Gaussian kernel profile function, i.e., k (z) exp (-z/2), z ≧ 0.
Threshold parameters mu, sigma1、σ2The specific calculation method comprises the following steps:
1) setting an initial value mu0The gradient energy mean value of the current flaw-free template image is obtained;
2) let mu let1＝m(μ0)+μ0Wherein m (. mu.) is0) Is mu0Amount of mean shift of
In the above formula, the function g (z) ═ -k' (z), k (z) is a gaussian contour function;
3) if μ1-μ0|<Epsilon, let parameter mu be mu1Go to the next step, otherwise let μ0＝μ1And returning to the previous step for iteration, wherein epsilon represents infinitesimal quantity;
4) calculating a threshold parameter σ according to equation (6)1And S is the total number of pixel points on the right side of the peak value of the gradient energy distribution graph.
es∈{en1,2.. N } and es≥μ (6)
5) Calculating a threshold parameter σ according to equation (7)2And Z is the total number of pixel points on the left side of the peak value of the gradient energy distribution graph.
ez∈{en1,2.. N } and ez≤μ (7)。
The invention provides a quick fabric flaw detection method based on an integral graph, which not only meets the real-time performance, but also has higher accuracy. The innovation points are as follows: 1) the method has the advantages that the gradient energy characteristics of the fabric image are rapidly extracted by utilizing an integral graph algorithm for flaw judgment for the first time, the summation operation in an image block with any size is simplified into three-time addition operation by utilizing an integral graph, and the operation cost is greatly reduced; meanwhile, when the integral graph algorithm is used, the gradient graph is firstly made and then the integral graph is made instead of directly making the integral graph, so that the interference of illumination change can be removed, and the detection precision is further improved; 2) the method adopts kernel function fitting asymmetric characteristic distribution, extracts a distribution peak value by combining a mean shift method, and adaptively obtains a threshold parameter, thereby realizing automatic and accurate segmentation of the flaw area.
Drawings
FIG. 1 is a block flow diagram of the method of the present invention.
FIG. 2 is a schematic diagram of the method of the present invention for calculating the area integral based on the integral map.
FIG. 3 is an image of a web to be inspected according to an embodiment of the present invention.
FIG. 4 is a gradient energy profile of an embodiment of the present invention.
FIG. 5 is a diagram illustrating a defect detection result according to an embodiment of the present invention.
FIG. 6 is a comparison of the results of the present invention method with other prior art methods.
Detailed Description
The invention provides a novel fabric flaw detection method, which can quickly and automatically realize high-precision fabric flaw detection. The method mainly comprises three parts of gradient energy feature extraction based on an integral graph, threshold parameter learning based on a kernel function and flaw detection.
As shown in fig. 1, the method includes firstly, performing image learning on a flawless template, counting gradient energy feature distribution of the flawless template, obtaining asymmetric feature distribution, fitting the feature distribution by adopting a kernel function, extracting a peak value of the distribution by combining a mean shift method, and then obtaining a threshold parameter by the peak value in a self-adaptive manner, wherein the threshold parameter is used for distinguishing subsequent flaws; then, the gradient energy of a detection window where each pixel point is located is obtained through an integral graph algorithm for an image to be detected, whether the current pixel point is a defect or not is judged by combining the threshold parameter, whether the current image is a defective fabric or not is judged by counting the total number of the defects of the whole image, and the specific implementation mode is as follows:
1. gradient energy feature extraction based on an integral graph:
in general, the image texture of an unblemished fabric is periodically and uniformly distributed, and for any point (x, y) in the fabric, the energy under a window with a fixed size dwdh is calculated by taking the point as the center, ideally, the energy characteristic should not change along with the change of (x, y), and the appearance of the flaw breaks the periodicity and uniformity of the fabric, so that the energy distribution is changed. Accordingly, the window energy value can be used as the feature description to detect the flaw, but considering that the energy value is susceptible to illumination, the invention uses the gradient energy to firstly obtain the gradient map G (x, y) of the original image F (x, y), and then extracts the energy feature E (x, y) for G (x, y). In the process of obtaining the characteristic E (x, y), the sum of the pixels in the neighborhood of each pixel needs to be counted, and if an acceleration algorithm is not adopted, the instantaneity is difficult to guarantee, so that the method adopts the integral graph acceleration. For image G (x, y), the integral map I (x, y) is defined as:
it can be seen that the value of any point (x, y) in the integral graph I is the sum of the pixel values of all points in the rectangular frame formed by the upper left corner to the current point (x, y) in the source image G. In order to accelerate the operation, an integral graph can be quickly obtained by an algorithm shown in an equation (1):
I(x,y)＝I(x-1,y)+I(x,y-1)-I(x-1,y-1)+G(x,y) (1)
after the integral image I (x, y) is obtained, the pixel integral of any rectangular area on the image G (x, y) can be quickly calculated by I (x, y), as shown in FIG. 2, so that I (x, y)1,y1)＝R(A)，I(x2,y2)＝R(A)+R(B)，I(x3,y3)＝R(A)+R(C)，I(x4,y4) R (a) + R (b) + R (c) + R (D), R (·) function represents the area integral, therefore the integral R (D) of area D can be calculated as follows:
R(D)＝I(x4,y4)-I(x2,y2)-I(x3,y3)+I(x1,y1)
according to the formula, no matter how large the area of the rectangular area D is, the sum of the pixel values can be obtained by only three times of operations, and the calculation amount is greatly reduced. Accordingly, the gradient energy feature extraction method based on the integral graph comprises the following specific steps:
1) obtaining a gradient map G (x, y) of the fabric image F (x, y);
2) obtaining an integral diagram I (x, y) of G (x, y) according to the formula (1);
3) for any point (x, y), a gradient energy map E (x, y) is obtained, i.e. a window of size dw dh centered on the point (x, y)
Pixel integration in the mouth region:
2. kernel function based threshold parameter learning:
ideally, for a flawless image, the gradient energy E (x, y) should not change with the change of (x, y), and in practice, the textures of different areas on the fabric may not be completely consistent, and noise may also be introduced during image acquisition, so the gradient energy E (x, y) tends to be distributed, and in order to obtain a suitable threshold parameter to segment the flaw, one of the most direct methods is to fit the distribution of E (x, y) by using a gaussian model, and set the parameter as the mean and variance of the gaussian distribution. However, the distribution of E (x, y) is not always symmetrical, and may exhibit different attenuation characteristics on both sides of the density peak, so that a deviation is easily generated if fitting with a gaussian model having symmetry. For this purpose, the invention proposes to fit the gradient energy distribution using a kernel function to obtain its kernel density probability distribution p (e):
where N is the number of pixels of the flawless image, { en1,2, 3.. N is a gradient energy feature map E extracted from the flawless imagetrainGradient energy data for each point in (x, y), b is kernel bandwidth, c0For normalizing the coefficients, k (z) is the kernel contour function, the invention selects the Gaussian kernel contour function, i.e. k (z) exp (-z/2), z ≧ 0, and substitutes into formula (4) to solve. In order to obtain the position of the peak value of the gradient energy distribution, the extreme value of the kernel density probability distribution P (e) is obtained by using a mean shift method, the energy distribution is divided into a left part and a right part according to the density peak value, and the variance of each part is respectively calculated for calculating the flaw detection threshold value, and the method specifically comprises the following steps:
1) setting an initial value mu0The gradient energy mean value of the current flaw-free template image is obtained;
2) let mu let1＝m(μ0)+μ0Wherein m (. mu.) is0) Is mu0Mean shift amount of (d):
in the above formula, g (z) ═ k' (z), is the derivation of the gaussian kernel profile function.
3) If μ1-μ0|<E, stopping circulation and making threshold parameter mu equal to mu1(ii) a Otherwise let mu0＝μ1And returning to the previous step, wherein epsilon represents infinitesimal quantity;
4) calculating a threshold parameter σ according to equation (6)1And S is the total number of pixel points on the right side of the peak value of the gradient energy distribution graph.
es∈{en1,2.. N } and es≥μ (6)
5) Calculating a threshold parameter σ according to equation (7)2And Z is the total number of pixel points on the left side of the peak value of the gradient energy distribution graph.
ez∈{en1,2.. N } and ez≤μ (7)
In order to improve the algorithm robustness, parameters mu and sigma are obtained for each flawless image in the training set according to the formulas (5), (6) and (7)1、σ2Then, the average value of all the flawless images is takenAs a threshold parameter in the final flaw detection process.
3. Defect detection
In the detection stage, for each image to be detected, firstly, according to the first part of gradient energy characteristic extraction method based on integral graph, obtaining its gradient energy characteristic graph Etest(x, y). If E istest(x, y) is determined as a defect if the following equation is satisfied:
wherein,for the threshold parameter obtained by training the flawless template image, α is a control coefficient, counting the total number of defects, if the total number of defects is larger than a set threshold value TdThen the image is judged to be defective, TdCan be adjusted by the user according to the actual requirements of quality control.
Fig. 3, 4, 5 and 6 are graphs showing the effect of the implementation of the present invention, the fabric image to be detected is derived from a fabric image dataset provided by the laboratory of the electrical and electronic engineering system of hong kong university, wherein the window width dw and the height dh are both set to 25, the control coefficient α is set to 4, fig. 3 shows 3 fabric images to be detected in the dataset, which are respectively (a), (b) and (c), fig. 4 is a graph of gradient energy characteristics extracted correspondingly, fig. 5 is a detection result of the present invention, and (a), (b) and (c) in fig. 3, 4 and 5 respectively correspond to one image, as can be seen from fig. 4, the gradient energy characteristics of a defect region are significantly different from those of a normal texture region, indicating that the constructed gradient energy characteristics can better distinguish the defect region from the normal texture region, as can be seen from fig. 5, the detection result graph of the algorithm provided can accurately locate defects, fig. 6 is a comparison between the method of the present invention and other methods, (a) is a fabric original graph, (b) is a detection result of the method of the present invention, (c) is a detection result based on LBP characteristics, (d) is a detection result of the algorithm provided by an RB, as can be more accurately shown by the present inventiondWhen the number of the images is 50, only 5 images are mistakenly detected, the accuracy reaches 97%, the average processing time of each image under an MATLAB platform is only 56ms, other existing methods such as a method based on LBP characteristics need tens of seconds under the MATLAB platform, and a method based on an RB algorithm with higher speed in the existing research needs 140ms under a C language environment.
Claims (5)
1. A fabric flaw detection method based on an integral graph algorithm is characterized in that the integral graph algorithm is utilized to simplify summation operation in image blocks of any size into three-time addition operation so as to rapidly extract gradient energy characteristics for flaw detection, and the method specifically comprises the following steps: firstly, image learning is carried out on a flawless template, gradient energy characteristic distribution of the flawless template is counted, the obtained characteristic distribution is asymmetric, kernel function fitting characteristic distribution is adopted, a peak value of distribution is extracted by combining a mean shift method, and then a threshold parameter is obtained by the peak value in a self-adaptive mode, wherein the threshold parameter is used for distinguishing subsequent flaws; then, solving gradient energy of a detection window where each pixel point is located through an integral graph algorithm for an image to be detected, judging whether the current pixel point is a defect or not by combining the threshold parameter, and judging whether the current image is a defective fabric or not by counting the total number of the defects of the whole image;
the extraction method of the gradient energy features comprises the following steps: firstly, a gradient image G (x, y) of an original image F (x, y) is obtained, then a gradient energy characteristic image E (x, y) of the G (x, y) is obtained by utilizing an integral image algorithm, and for any pixel point (x, y), the energy characteristic is pixel integration in a window area with the point (x, y) as the center and with the size dw dh.
2. The fabric defect detecting method of claim 1, wherein the step of using the integral graph algorithm to obtain the gradient energy characteristic graph E (x, y) of G (x, y) comprises the following steps:
1) obtaining an integral diagram I (x, y) of G (x, y)
I(x,y)＝I(x-1,y)+I(x,y-1)-I(x-1,y-1)+G(x,y) (1)
2) And (3) obtaining a gradient energy characteristic diagram E (x, y) of an arbitrary point (x, y) according to the characteristics of the integral diagram:
3. the method for detecting fabric defects according to claim 1 or 2, characterized by the specific steps of:
1) training and learning threshold parameters for flaw detection using flaw-free template images
Establishing a training set by using images of the flawless fabric, and obtaining a gradient energy characteristic diagram E of the images for each imagetrain(x, y) fitting E with a Kernel functiontrainObtaining the nuclear density probability distribution P (e) by the gradient energy distribution in (x, y), and iteratively solving the nuclear density probability distribution P (e) by using a mean shift methodExtreme value, obtaining the peak position of gradient energy distribution, marking as mu, dividing the energy distribution into left and right parts according to the peak value, respectively calculating two parts of variance sigma1、σ2The parameters mu and sigma are obtained for each flawless image of the same type of fabric in the training set1、σ2Then, the average value is obtained As a threshold parameter in the final defect detection process for this type of fabric;
2) obtaining a gradient energy characteristic diagram E of each image to be detectedtest(x, y) if Etest(x, y) is determined as a defect if the following equation is satisfied:
wherein,the threshold parameter obtained in the step 1), α is a control coefficient;
finally, counting the total number of the defects, if the total number of the defects is larger than a set threshold value TdThe map is judged to be defective.
4. The fabric defect detection method of claim 3, wherein the kernel density probability distribution P (e) is:
wherein N is the total number of pixels of the inpaintless image, { en1,2, 3 … N is gradient energy data of each point in the inpaintless image, b is kernel bandwidth, c is0For normalizing the coefficients, k (z) is GaussianThe kernel contour function, i.e., k (z) ═ exp (-z/2), z ≧ 0.
5. The fabric defect detection method of claim 3, wherein the threshold parameters μ, σ are1、σ2The specific calculation method comprises the following steps:
1) setting an initial value mu0The gradient energy mean value of the current flaw-free template image is obtained;
2) let mu let1＝m(μ0)+μ0Wherein m (. mu.) is0) Is mu0Mean shift amount of (d):
in the above formula, N is the total number of pixels of the flawless image, { en1,2, 3 … N, b is the kernel function bandwidth, and the function g (z) -k' (z), k (z) is a gaussian kernel contour function;
3) if μ1-μ0|<Epsilon, let parameter mu be mu1Go to the next step, otherwise let μ0＝μ1And returning to the previous step for iteration, wherein epsilon represents infinitesimal quantity;
4) calculating a threshold parameter σ according to equation (6)1：
es∈{en1,2.. N } and es≥μ (6)
Wherein S is the total number of pixel points on the right side of the peak value of the gradient energy distribution graph;
5) calculating a threshold parameter σ according to equation (7)2
ez∈{en1,2.. N } and ez≤μ(7)
Wherein Z is the total number of pixels on the left side of the peak value of the gradient energy distribution graph.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610180226.3A CN107240086B (en) | 2016-03-25 | 2016-03-25 | A kind of fabric defects detection method based on integral nomography |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610180226.3A CN107240086B (en) | 2016-03-25 | 2016-03-25 | A kind of fabric defects detection method based on integral nomography |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107240086A CN107240086A (en) | 2017-10-10 |
| CN107240086B true CN107240086B (en) | 2019-05-14 |
Family
ID=59983575
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610180226.3A Active CN107240086B (en) | 2016-03-25 | 2016-03-25 | A kind of fabric defects detection method based on integral nomography |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107240086B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090894B (en) * | 2017-11-24 | 2020-09-08 | 浙江理工大学 | Fabric defect detection method based on distance matching function and perceptual hash algorithm |
| CN109583306B (en) * | 2018-10-31 | 2021-10-08 | 杭州电子科技大学 | Bobbin residual yarn detection method based on machine vision |
| CN109934287A (en) * | 2019-03-12 | 2019-06-25 | 上海宝尊电子商务有限公司 | A kind of clothing texture method for identifying and classifying based on LBP and GLCM |
| TWI783200B (en) * | 2019-12-30 | 2022-11-11 | 緯創資通股份有限公司 | Method and device for determining whether object has defect |
| CN114226262A (en) * | 2020-09-09 | 2022-03-25 | 宜谷京科技实业有限公司 | Flaw detection method, flaw classification method and flaw detection system |
| CN112308854B (en) * | 2020-11-26 | 2023-12-26 | 武汉工程大学 | Automatic detection method and system for chip surface flaws and electronic equipment |
-
2016
- 2016-03-25 CN CN201610180226.3A patent/CN107240086B/en active Active
Non-Patent Citations (5)
| Title |
|---|
| A fast regularity measure for surface defect detection;Du-Ming Tsai 等;《Machine Vision and Applications》;20120104;全文 * |
| A Shift-Tolerant Dissimilarity Measure for Surface Defect Detection;Du-Ming Tsai 等;《IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS》;20120228;第8卷(第1期);全文 * |
| 图像处理在仪表识别和布匹瑕疵检测中的应用;张北航;《中国优秀硕士学位论文全文数据库 信息科技辑》;20090115;全文 * |
| 瑕疵点检测算法研究及应用;王文杰;《中国优秀硕士学位论文全文数据库 信息科技辑》;20110715;全文 * |
| 钢珠瑕疵检测算法的研究;俞哲 等;《仪器仪表学报》;20091031;第30卷(第10期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107240086A (en) | 2017-10-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107240086B (en) | A kind of fabric defects detection method based on integral nomography | |
| CN108961217B (en) | Surface defect detection method based on regular training | |
| CN109934802B (en) | Cloth defect detection method based on Fourier transform and image morphology | |
| CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
| CN107833220B (en) | Fabric defect detection method based on deep convolutional neural network and visual saliency | |
| CN111179225B (en) | Test paper surface texture defect detection method based on gray gradient clustering | |
| CN108090894B (en) | Fabric defect detection method based on distance matching function and perceptual hash algorithm | |
| CN104574353B (en) | The surface defect decision method of view-based access control model conspicuousness | |
| CN107437243B (en) | Tire impurity detection method and device based on X-ray image | |
| CN103530878B (en) | A kind of edge extracting method based on convergence strategy | |
| CN110930390B (en) | Chip pin missing detection method based on semi-supervised deep learning | |
| CN116664559B (en) | Machine vision-based memory bank damage rapid detection method | |
| CN103729856B (en) | A kind of Fabric Defects Inspection detection method utilizing S-transformation signal extraction | |
| CN107622277A (en) | A kind of complex-curved defect classification method based on Bayes classifier | |
| CN115100206B (en) | Printing defect identification method for textile with periodic pattern | |
| CN111415339B (en) | Image defect detection method for complex texture industrial product | |
| CN117635609B (en) | Visual inspection method for production quality of plastic products | |
| CN116128873A (en) | Bearing retainer detection method, device and medium based on image recognition | |
| CN116051541B (en) | Bearing end face gentle abrasion detection method and device based on stroboscopic light source | |
| CN114202544B (en) | Complex workpiece defect detection method based on self-encoder | |
| CN116524269A (en) | Visual recognition detection system | |
| CN112598646B (en) | Capacitance defect detection method and device, electronic equipment and storage medium | |
| Hsu et al. | Vision-based detection of steel billet surface defects via fusion of multiple image features | |
| CN114529515A (en) | Method for automatically identifying internal defects of solar cell | |
| Aydin et al. | A vision based inspection system using gaussian mixture model based interactive segmentation |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201030 Address after: 226019 No.205, building 6, Nantong University, No.9, Siyuan Road, Nantong City, Jiangsu Province Patentee after: Center for technology transfer, Nantong University Address before: 226019 Jiangsu city of Nantong province sik Road No. 9 Patentee before: NANTONG University |
|
| TR01 | Transfer of patent right | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20171010 Assignee: Nantong Shunteng Textile Co.,Ltd. Assignor: Center for technology transfer, Nantong University Contract record no.: X2021980016785 Denomination of invention: A fabric defect detection method based on integral graph algorithm Granted publication date: 20190514 License type: Common License Record date: 20211229 |
|
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 226001 No.9, Siyuan Road, Chongchuan District, Nantong City, Jiangsu Province Patentee after: Nantong University Technology Transfer Center Co.,Ltd. Address before: 226019 No.205, building 6, Nantong University, No.9, Siyuan Road, Nantong City, Jiangsu Province Patentee before: Center for technology transfer, Nantong University |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20171010 Assignee: Yizhong (Nantong) Network Technology Co.,Ltd. Assignor: Nantong University Technology Transfer Center Co.,Ltd. Contract record no.: X2023980053327 Denomination of invention: A Fabric Defect Detection Method Based on Integral Graph Algorithm Granted publication date: 20190514 License type: Common License Record date: 20231221 |