CN109507193A - A kind of fabric defects detection method based on local contrast enhancing and binary pattern - Google Patents
A kind of fabric defects detection method based on local contrast enhancing and binary pattern Download PDFInfo
- Publication number
- CN109507193A CN109507193A CN201811310641.1A CN201811310641A CN109507193A CN 109507193 A CN109507193 A CN 109507193A CN 201811310641 A CN201811310641 A CN 201811310641A CN 109507193 A CN109507193 A CN 109507193A
- Authority
- CN
- China
- Prior art keywords
- feature
- flaw
- histogram
- value
- local contrast
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/888—Marking defects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Abstract
The invention discloses a kind of fabric defects detection methods based on local contrast enhancing and binary pattern, including training part and part of detecting;The training part obtains a threshold value by the processing to indefectible image, and the part of detecting carries out detection and the label of flaw by using the threshold value that training obtains.Beneficial effects of the present invention: the present invention solve the problems, such as to a certain extent traditional complete local binary patterns in fabric defects detection there are histogram dimension is excessively high and feature redundancy and in zonule image change amplitude it is violent or there are limitations when amplitude of variation is gentle.
Description
Technical field
The present invention relates to the technical field of textile Defect Detection, more particularly to it is a kind of based on local contrast enhancing with
Improve the fabric defects detection method of complete local binary patterns.
Background technique
It is beginning to carry out very early about the research of textile Defect Detection in recent years, in order to solve textile flaw
The automation of detection realizes, foreign study personnel propose substantially three kinds of methods: statistical method, spectrum and based on model
Method.Local binary patterns are initially to be proposed by T.Ojala of Oulu, Finland university et al. in 1996, at 2002,
Timo Ojala et al. has delivered an article about local binary patterns again on PAMI, and what this article was perfectly clear explains
The improved local binary patterns feature of constant multiresolution, grey scale and invariable rotary, equivalent formulations is stated.Use office
For portion's binary pattern as feature extracting method, foreign countries are applied in recognition of face, are being applied to the textile flaw
Defect detection field, F.Tajeripour in 2008 have delivered the textile flaw detection method based on local binary patterns
Article, article is described in detail about the application method of local binary patterns and in textile Defect Detection, office
Portion's binary pattern has excellent experimental result, this is also in next step about using local binary patterns to mention as a kind of feature
Application of the method in textile Defect Detection is taken to have laid a good foundation, facts proved that, local binary patterns are being weaved
Product Defect Detection field is that have more outstanding effect, has certain research significance and value, fast as a kind of speed,
Feature extracting method at low cost also has certain value in practical applications.
In recent years, local binary patterns cause the concern of image procossing and area of pattern recognition scholar.Local binary mould
Formula operator is initially applied to the description of textural characteristics, and since its principle is relatively easy, computation complexity is low while having merged line
The mechanism characteristics and statistical nature of reason, and it is not illuminated by the light the influence of the factors such as variation.Local binary patterns are in recent years due to its expansion
Exhibition method emerges one after another, application range also further expands to recognition of face by texture analysis field, target following detects,
The multiple fields such as medical image analysis.But local binary patterns method is to noise-sensitive, and only considered center pixel and adjacent
The difference symbolic feature of domain pixel, does not account for difference amplitude, is lost a part of data information.In order to make local binary mould
Formula feature extraction is more abundant, and Guo et al. proposes complete local binary patterns method, what complete local binary patterns extracted
Feature relatively comprehensively and has stronger distinguishing ability, is applied in Texture classification, achieves higher discrimination.But
It is that there are no good effects for Defect Detection under Varying Illumination.Experiment shows that care problem is to influence the textile flaw
An important factor for defect detects, illumination is not artificially controlled in the detection process as external factor, in different illumination
Under, different two dimensional images can be obtained, and it is extremely difficult that the two dimensional image based on these under any illumination, which carries out identification,
, therefore illumination pretreatment is a crucial step in textile Defect Detection.
Light irradiation preprocess method is broadly divided into following a few classes at present: preprocess method based on wavelet transformation, from quotient graph
Image space method, Retinex method, the heterogeneous smooth treatment of anisotropy, contrast enhancing etc..In these methods, it is based on histogram
For the Enhancement Method of equalization because its validity and property easy to use are by favor, basic thought is the ash according to input picture
Probability-distribution function is spent to determine its corresponding output gray level value, and promotion figure is reached by the gray scale of expanded images, dynamic range
The purpose of image contrast.Histogram equalization is divided into global and local two kinds, compared with global approach, local histogram equalization
Change the local message that can preferably enhance image.
Summary of the invention
The purpose of this section is to summarize some aspects of the embodiment of the present invention and briefly introduce some preferable implementations
Example.It may do a little simplified or be omitted to avoid this is made in this section and the description of the application and the title of the invention
Partially, the purpose of abstract of description and denomination of invention is fuzzy, and this simplification or omission cannot be used for limiting model of the invention
It encloses.
In view of above-mentioned existing fabric defects detection method there are the problem of, propose the present invention.
Therefore, the fabric defects detection based on local contrast enhancing with binary pattern that it is an object of the present invention to provide a kind of
Method, the problem of being able to solve the textile images Defect Detection for being illuminated by the light influence.
In order to solve the above technical problems, the invention provides the following technical scheme: a kind of based on local contrast enhancing and two
The fabric defects detection method of value mode, including training part and part of detecting;The training part is by indefectible image
Processing obtain a threshold value, the part of detecting carries out detection and the label of flaw by using the threshold value that training obtains.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: the trained part is further comprising the steps of, using the flawless image for not being illuminated by the light influence;To each
Indefectible figure does lattice dividing processing, and the feature histogram of each grid is calculated;Finally ask the average spy of all grid
Levy histogram;The feature histogram of each grid and average histogram are done into relative divergence calculating again and find out divergence value, finally
To threshold value T.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: the part of detecting is further comprising the steps of, does illumination pretreatment to flaw figure;Then by every flaw
In figure the feature histogram of each grid and relative divergence calculating is done with average characteristics histogram;The each value and threshold that will be obtained
Value compares, and the part greater than the threshold value T is labeled as flaw.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: the training department point includes the following steps, obtains all no light shadows with morphology component analyzing method
The cartoon layer and texture layer of indefectible figure are rung, and only retains cartoon layer part;Lattice is done for the cartoon layer to handle and calculate
The feature histogram of each lattice and genitive average characteristics histogram is sought after lattice segmentation;Ask the feature of each grid straight
The relative divergence of side's figure and average histogram simultaneously obtains threshold value T.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: the part of detecting includes the following steps have to suffered illumination effect with local contrast Enhancement Method
Flaw figure pre-processes;The cartoon layer and texture layer of image after obtaining the pretreatment with morphology component analyzing method, and only
Retain cartoon layer part;It calculates all flaw graph cards and leads to each lattice feature histogram and genitive average characteristics histogram in layer
Figure, and relative divergence is sought with the average histogram;By the relative divergence value that obtains compared with threshold value T, greater than threshold value T's
Grid is labeled as flaw.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: it is described there is flaw figure to pre-process suffered illumination effect with local contrast Enhancement Method after, also wrap
It includes and feature extraction is carried out using complete local binary patterns method is improved to each grid.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: image enhancement processing is carried out with local contrast Enhancement Method on the flaw figure for being illuminated by the light influence, it is described
Local contrast Enhancement Method are as follows:
Wherein,
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: it is described to improve complete local binary patterns method, by improving complete local binary pattern operator, improve
Operator describes operator for gradient window difference in the complete local binary patterns of tradition and makes improvement, by comparing two pixels
The mean value size of gray difference amplitude shade of gray difference characteristic in local window is described, calculation method is as follows:
Wherein,
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: between carrying out characteristic value with relative divergence value between the characteristic value extracted in training part and detection part
Quantization compare and carry out threshold value selection, wherein relative divergence is the method for measuring two probability distribution distances, calculate
Method is as follows:
In formula, P (xi) and Q (xi) respectively represent two kinds of probability distribution;N is characterized vector dimension, PiBeing characterized value is i's
The frequency that feature vector occurs in the feature histogram of mapping to be checked;QiBeing characterized value is the feature vector of i in average characteristics
The frequency occurred in histogram.
As one kind of the present invention based on local contrast enhancing and the fabric defects detection method of binary pattern
Preferred embodiment, in which: the threshold value is chosen for the feature histogram in training part by each grid of the flawless figure of calculating
Some relative divergence values are obtained with the relative divergence of average characteristics histogram, the method for selected threshold is in these divergence values
In formula, TkThe divergence value being calculated for k-th of grid in all samples.
Beneficial effects of the present invention: the present invention solves traditional complete local binary patterns to a certain extent and is knitting
When object defect detection there are histogram dimension is excessively high and feature redundancy and zonule image change amplitude acutely or variation width
There are problems that limitation when spending gentle.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, making required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, right
For those of ordinary skill in the art, without any creative labor, it can also be obtained according to these attached drawings
Other attached drawings.Wherein:
Fig. 1 is that of the present invention enhanced based on local contrast is examined with the fabric defects for improving complete local binary patterns
The principle flow chart of survey method;
Fig. 2 is that of the present invention enhanced based on local contrast is examined with the fabric defects for improving complete local binary patterns
The actual experiment process flow diagram flow chart of survey method;
Fig. 3 is that of the present invention enhanced based on local contrast is examined with the fabric defects for improving complete local binary patterns
Feature extraction flow chart during the actual experiment of survey method;
Fig. 4 is schematic diagram of the different light irradiation preprocess methods to star flaw type flaw figure processing result;
Fig. 5 is schematic diagram of the different light irradiation preprocess methods to box-shaped flaw type flaw figure processing result;
Fig. 6 is different light irradiation preprocess methods to star broken ends of fractured bone flaw type flaw figure testing result;
Fig. 7 is different light irradiation preprocess methods to star reticulate pattern flaw type flaw figure testing result;
Fig. 8 is different light irradiation preprocess methods to star stria flaw type flaw figure testing result;
Fig. 9 is different light irradiation preprocess methods to box-shaped broken ends of fractured bone flaw type flaw figure testing result;
Figure 10 is different light irradiation preprocess methods to box-shaped reticulate pattern flaw type flaw figure testing result;
Figure 11 is different light irradiation preprocess methods to box-shaped stria flaw type flaw figure testing result.
Specific embodiment
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, right with reference to the accompanying drawings of the specification
A specific embodiment of the invention is described in detail, it is clear that and described embodiment is a part of the embodiments of the present invention, and
It is not all of embodiment.Based on the embodiments of the present invention, ordinary people in the field is without making creative work
Every other embodiment obtained, all should belong to the range of protection of the invention.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, but the present invention may be used also
To be implemented using other than the one described here other way, those skilled in the art can be without prejudice in the present invention
Similar popularization is done in the case where culvert, therefore the present invention is not limited by the specific embodiments disclosed below.
Secondly, " one embodiment " or " embodiment " referred to herein, which refers to, may be included at least one realization of the present invention
A particular feature, structure, or characteristic in mode." in one embodiment " that different places occur in the present specification is not equal
Refer to the same embodiment, nor the individual or selective embodiment mutually exclusive with other embodiments.
Thirdly, combination schematic diagram of the present invention is described in detail, when describing the embodiments of the present invention, for purposes of illustration only,
Indicate that the sectional view of device architecture can disobey general proportion and make partial enlargement, and the schematic diagram is example, herein not
The scope of protection of the invention should be limited.In addition, the three-dimensional space of length, width and depth should be included in actual fabrication.
Simultaneously in the description of the present invention, it should be noted that the orientation of the instructions such as " upper and lower, inner and outer " in term
Or positional relationship is to be based on the orientation or positional relationship shown in the drawings, and is merely for convenience of description of the present invention and simplification of the description, and
It is not that the device of indication or suggestion meaning or element must have a particular orientation, be constructed and operated in a specific orientation, therefore
It is not considered as limiting the invention.In addition, term " first, second or third " is used for description purposes only, and cannot understand
For indication or suggestion relative importance.
In the present invention unless otherwise clearly defined and limited, term " installation is connected, connection " shall be understood in a broad sense, example
Such as: may be a fixed connection, be detachably connected or integral type connection;It equally can be mechanical connection, electrical connection or directly connect
It connects, can also indirectly connected through an intermediary, the connection being also possible to inside two elements.For the common skill of this field
For art personnel, the concrete meaning of above-mentioned term in the present invention can be understood with concrete condition.
Embodiment 1
Signal referring to Fig.1 illustrates provided by the invention a kind of complete based on local contrast enhancing and improvement
The principle flow chart of the fabric defects detection method of local binary patterns, in order to solve the textile images flaw for being illuminated by the light influence
The problem of defect detects, 100 method of fabric defects detection method are divided into training and test two parts, wherein training part is to nothing
The complete local binary patterns characteristic value of each grid computed improved after the segmentation of flaw image pane simultaneously calculates the characteristic values of all grid
Mean value, the relative divergence by calculating each grid characteristic value and mean value acquire threshold value;And part of detecting by relative divergence with
Threshold value comparison, the grid greater than threshold value are labeled as flaw.Wherein lattice are divided into using the lattice based on adaptive K-Means method
Dividing method carries out piecemeal to image, and cartoon tomographic image at this time is divided into some fritters.It should be noted that in training
It requires to carry out lattice dividing processing to image with the part of detecting first step.
More specifically, flaw detection method 100 includes training and test two parts, and wherein training department point includes following
Step:
Step 102, indefectible image is inputted, image is first done into form and studies point analysis pretreatment and treated image
Only retain cartoon layer part, lattice dividing processing is carried out to all cartoon layers, is marked by image block, and to all pieces;
Step 104, characteristic value calculating is carried out using complete local binary pattern operator is improved to all tag blocks, will counted
All characteristic values after calculation re-start label;
Step 106, the average value M of all characteristic values marked in above-mentioned steps is sought;
Step 108, the relative divergence value in step 104 between all marker characteristic values and average value M is sought, and is dissipated all
Angle value is marked;
Step 110, the maximum value of above-mentioned steps label divergence value is filtered out as threshold value T and is retained.
Part of detecting the following steps are included:
Step 112, input it is to be detected have flaw image, using local contrast Enhancement Method on being illuminated by the light influence flaw
Figure pre-processes, and the method in above-mentioned steps 102 that reuses carries out pretreatment dative segmentation to image, is marked after piecemeal;
Step 114, characteristic value calculating, meter are carried out using complete local binary pattern operator is improved by all pieces after label
All characteristic values re-start label after calculation;
Step 116, the relative divergence value of the characteristic value and average value M marked in above-mentioned steps is sought, and to all divergence values
It is marked;
Step 118, the divergence value after above-mentioned label is compared with threshold value T, if label divergence value is greater than threshold value T,
The tag block is labeled as defect areas, on the contrary it is flawless.
The recall ratio of last testing result is defined as follows with precision ratio:
Wherein, tp indicates that correctly label is positive, and fp indicates that the label of mistake is positive and (is originally that negative flag is positive),
Fn indicates that the label of mistake is negative and (is originally just, label is negative).
By the fabric defects inspection proposed by the invention for being enhanced based on local contrast and improving complete local binary patterns
Survey method, experiment are classified into three parts, are tested in star and box standard database.First part is by comparing not
Next step experiment is carried out to the pretreating effect of flaw image with light irradiation preprocess method.Second part is different by comparison
Influence of the light irradiation preprocess method to textile Defect Detection.Part III is filtered with based on Garbor by means of the present invention
The Comparison of experiment results analysis of wave dative dividing method, gold image subtraction method, cloth forest belt index method and Regular Band method, says
The bright present invention has a distinct increment on recall ratio and precision ratio.
It should be understood that the method for above-mentioned Defect Detection, only carries out EXPERIMENTAL EXEMPLIFICATIONThe explanation with above-mentioned a few class images, it is real
Do not limit to above-described embodiment in the application of border, can according to need and the above method is applied into different places and is tested
Analysis.
Embodiment 2
Referring to the signal of Fig. 2~3, being illustrated as a kind of local contrast that is based on provided by the invention enhances and improves complete
Flow chart of the fabric defects detection method of local binary patterns during actual experiment.Specifically, referring to first implementation
The principle process of example is totally divided into two parts in the present embodiment: training and test.Signal referring to Fig.1, training department point include:
Step 1: obtaining all no lights with MCA method influences the cartoon layer and texture layer of indefectible figure, and only retains card
Logical layer part;
Step 2: doing lattice segmentation (piecemeal) processing for above-mentioned cartoon layer and calculate after above-mentioned lattice are divided each piece (grid)
Feature histogram and seek all pieces of average characteristics histogram;
Step 3: asking the feature histogram of each grid and the KLD divergence value of average histogram and obtain threshold value T.
Part of detecting includes:
Step 4: thering is flaw figure to pre-process suffered illumination effect with LCE method;
Step 5: the cartoon layer and texture layer of image after obtaining above-mentioned pretreatment with MCA, and only retain cartoon layer part;
Step 6: calculating all flaw graph cards and lead to each block feature histogram in layer and seek KLD with above-mentioned average histogram
Divergence value;
Step 7: by the above-mentioned KLD divergence value that obtains compared with threshold value T, the grid greater than threshold value T is labeled as flaw.
Further, it after then thering is flaw figure to pre-process suffered illumination effect with local contrast Enhancement Method, also wraps
It includes and feature extraction is carried out using complete local binary patterns method is improved to each grid, in the complete local binary of above-mentioned improvement
In mode method, M operator only accounts for the grey value difference feature of local window, this allow for when due to uneven illumination or
When the image grayscale that the reasons such as shooting angle variation obtain is unevenly distributed, M operator can omit the texture information of smooth part.
Therefore, herein for this problem, improvement is made to M operator, by the gray difference amplitude and window that compare two pixels
The mean value size of gray difference amplitude describes shade of gray difference characteristic in local window, and improved M operator can indicate are as follows:
Wherein,
Other two operator is identical as traditional complete local binary patterns, and lattice segmentation is shown with feature extraction process such as Fig. 3's
Meaning.
Embodiment 3
Feasibility and correctness of the invention are verified in the present embodiment, and Multi simulation running experiment is carried out in PC machine,
Middle processor is i7-4710HQ, and dominant frequency 2.5GHz inside saves as 8GB, carries out experiment simulation with MATLAB R2016b software.
The 81 width pixels that the present embodiment has used Hong Kong University electrics and electronics engineering system industrial automation laboratory to provide
24 color textile product images of size 256 × 256, these images are converted into 8 gray level images in an experiment.81 width
Image includes two kinds of patterns: star-shaped image and box-shaped image, includes wherein that 26 indefectible and 15 width have flaw in box-shaped image
Image;It include that 25 indefectible and 15 width have flaw image in star-shaped image.The flaw image that has of two kinds of patterns includes 3 kinds of flaws
Defect type: the broken ends of fractured bone, reticulate pattern and stria, all flaw images have the flaw reference map of same size, and flaw reference map is 2
It is worth image.Wherein 1 indicate flaw, 0 indicates background.
There are four evaluation indexes used in the experimental section of the present embodiment: TPR, FPR, PPV, NPV.TPR table shows flaw
Indicate that the pixel of flaw is correctly demarcated as the ratio of flaw by method in reference map;FPR indicates to indicate background in flaw reference map
Pixel be demarcated as the ratio of flaw by method fault;Flaw institute accounting in the flaw reference map of PPV representation method output
Example;Background proportion in the flaw reference map of NPV representation method output.
Further, the present embodiment is to the pretreating effect of textile images for different light irradiation preprocess methods
Different problems, therefore carried out comparative experiments for this problem and verified, and optimal preprocess method is selected to carry out down
The Defect Detection of one step, the signal of experimental result such as Fig. 4, Fig. 5.
By the experimental result of analysis chart 4, Fig. 5, it is not difficult to find that in four kinds of light irradiation preprocess methods, local contrast enhancing
Method is outstanding compared with other methods for the pretreating effect of textile images;LCE can effectively improve minutia
Visualization;Although Retinex and homographic filtering method can also restore the essential characteristic of original image, still there is partial region mistake
Secretly, minutia can not effectively be extracted;From quotient images, treated that brightness of image is excessively high, affects the extraction of local feature, drops
The low contrast of defect areas and flawless region.
To all methods treated textile images be carried out with Defect Detection below, the present embodiment uses above-mentioned four kinds
Light irradiation preprocess method is concentrated in star graph and box-shaped diagram data and carries out Defect Detection comparative experiments, experimental result such as table 1, table
The signal of 2 and Fig. 6-11.
Table 1 is that different pretreatments method influences the detection effect of mulle.
1 test experience of analytical table is as a result, the defect detection effect based on LCE preprocess method is outstanding, for three kinds
The TPR value of flaw type reaches 0.99, and erroneous detection situation is not present.Retinex method is directed to the broken ends of fractured bone and stria flaw class
Although the testing result TPR value of type reaches 0.99 still there are large area erroneous detections for its testing result, for reticulate pattern flaw type
Testing result TPR value only has 0.11, and also has large-scale erroneous detection.Homographic filtering method is directed to reticulate pattern, stria and the broken ends of fractured bone flaw
The testing result TPR value of defect type be respectively 0.91,0.86 and 0.56 and testing result there are the erroneous detections of large area.From quotient graph
Image space method is below 0.3 and still remains the erroneous detection of large area for the testing result TPR value of three kinds of flaw types.
Table 2 is that different pretreatments method influences the detection effect of box-shaped pattern.
2 test experience of analytical table is as a result, the defect detection effect based on LCE preprocess method is outstanding, for reticulate pattern
Reach 0.99 with the TPR value of stria flaw type, there are small range erroneous detection situations for reticulate pattern type detection result.
Retinex method can not almost detect defect areas for the TPR value of the broken ends of fractured bone for 0.01, for reticulate pattern and stria flaw
The testing result TPR value of type is 0.5 or so, but there are large area erroneous detections for its testing result.Homographic filtering method is directed to net
The testing result TPR value of line, stria and broken ends of fractured bone flaw type is respectively 0.60,0.56 and 0.45 and testing result exists greatly
The erroneous detection of area.0.3 is below for the testing result TPR value of three kinds of flaw types from quotient images method and is still remained big
The erroneous detection of area.
In conclusion Retinex method connects only for the experimental result recall ratio TPR and LCE method of star data set
Closely, but the former testing result erroneous detection area is larger, other both of which and LCE method gap are larger, therefore selects the side LCE
Method combines feature extracting method proposed by the present invention to carry out flaw inspection as optimal light irradiation preprocess method.
It is concentrated in star graph and box-shaped diagram data, context of methods is in conjunction with other classical ways LSG, WGIS, BB and RB
LCE light irradiation preprocess method compares experiment.Experimental result is as shown in table 3, table 4.
Table 3 is that distinct methods influence the detection effect of mulle.
Table3The effect of different algorithm on star pattern
3 test experience of analytical table is as a result, context of methods is better than other three kinds of sides in the testing result of broken ends of fractured bone flaw type
Method, context of methods is better than other three kinds of testing results in the testing result of reticulate pattern flaw type, examines in stria flaw type
Survey the testing result that context of methods in result is better than other methods.And this method TPR value in the flaw of all three types is equal
It is optimal, FPR value keeps stablizing.
Table 4 is that distinct methods influence the detection effect of box-shaped pattern.
4 test experience of analytical table is as a result, context of methods is superior to other three kinds in the testing result of three kinds of flaw types
Method and this method the TPR value in the flaw of all three types are optimal, and FPR value keeps stablizing.It is provided by the invention
Method is preferable to mulle and box-shaped pattern textile inspection effect, and a part of recall ratio can achieve 0.99, major part
The recall ratio of testing result is 0.90 or more.
It should be noted that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although referring to preferable
Embodiment describes the invention in detail, those skilled in the art should understand that, it can be to technology of the invention
Scheme is modified or replaced equivalently, and without departing from the spirit and scope of the technical solution of the present invention, should all be covered in this hair
In bright scope of the claims.
Claims (10)
1. a kind of fabric defects detection method based on local contrast enhancing and binary pattern, it is characterised in that: including training
Part and part of detecting;The training part obtains a threshold value by the processing to indefectible image, and the part of detecting is logical
Cross detection and the label that flaw is carried out using the threshold value that training obtains.
2. the fabric defects detection method as described in claim 1 based on local contrast enhancing and binary pattern, feature
Be: the trained part is further comprising the steps of,
Using the flawless image for not being illuminated by the light influence;Lattice dividing processing is done to each indefectible figure, and each is calculated
The feature histogram of grid;Finally seek the average characteristics histogram of all grid;Again by the feature histogram of each grid and flat
Equal histogram does relative divergence calculating and finds out divergence value, finally obtains threshold value T.
3. the fabric defects detection method as claimed in claim 2 based on local contrast enhancing and binary pattern, feature
Be: the part of detecting is further comprising the steps of,
Illumination pretreatment is done to flaw figure;Then by the feature histogram of each grid in every flaw figure and with average characteristics it is straight
Square figure does relative divergence calculating;Obtained each value is compared with threshold value, the part greater than the threshold value T is labeled as flaw.
4. the fabric defects detection method based on local contrast enhancing and binary pattern as described in claims 1 to 3 is any,
It is characterized by: the training department point includes the following steps,
Obtaining all no lights with morphology component analyzing method influences the cartoon layer and texture layer of indefectible figure, and only retains card
Logical layer part;
Lattice is done for the cartoon layer to handle and calculate the feature histogram of each lattice after lattice segmentation and ask genitive
Average characteristics histogram;
It asks the feature histogram of each grid and the relative divergence of average histogram and obtains threshold value T.
5. the fabric defects detection method as claimed in claim 4 based on local contrast enhancing and binary pattern, feature
Be: the part of detecting includes the following steps,
There is flaw figure to pre-process suffered illumination effect with local contrast Enhancement Method;
The cartoon layer and texture layer of image after obtaining the pretreatment with morphology component analyzing method, and only retain cartoon layer portion
Point;
It calculates all flaw graph cards and leads to each lattice feature histogram and genitive average characteristics histogram in layer, and put down with described
Equal histogram seeks relative divergence;
By the relative divergence value that obtains compared with threshold value T, the grid greater than threshold value T is labeled as flaw.
6. the fabric defects detection method as claimed in claim 5 based on local contrast enhancing and binary pattern, feature
Be: it is described there is flaw figure to pre-process suffered illumination effect with local contrast Enhancement Method after, further include to each lattice
Son carries out feature extraction using complete local binary patterns method is improved.
7. special such as the fabric defects detection method described in claim 5 or 6 based on local contrast enhancing and binary pattern
Sign is: carrying out image enhancement processing, the local contrast with local contrast Enhancement Method to the flaw figure for being illuminated by the light influence
Spend Enhancement Method are as follows:
Wherein,
8. the fabric defects detection method as claimed in claim 7 based on local contrast enhancing and binary pattern, feature
It is: it is described to improve complete local binary patterns method, by improving complete local binary pattern operator, operator is improved for biography
Gradient window difference in complete local binary patterns of uniting describes operator and makes improvement, passes through the gray difference width of two pixels of comparison
The mean value size of value describes shade of gray difference characteristic in local window, and calculation method is as follows:
Wherein,
9. the fabric defects detection method as claimed in claim 8 based on local contrast enhancing and binary pattern, feature
It is: is compared between the characteristic value extracted in training part and detection part with the quantization between relative divergence value carries out characteristic value
And threshold value selection is carried out, wherein relative divergence is the method for measuring two probability distribution distances, and calculation method is as follows:
In formula, P (xi) and Q (xi) respectively represent two kinds of probability distribution;N is characterized vector dimension, PiIt is characterized the feature that value is i
The frequency that vector occurs in the feature histogram of mapping to be checked;QiBeing characterized value is the feature vector of i in average characteristics histogram
The frequency occurred in figure.
10. the fabric defects detection method as claimed in claim 9 based on local contrast enhancing and binary pattern, feature
Be: the threshold value is chosen for feature histogram and average characteristics in training part by each grid of the flawless figure of calculating
The relative divergence of histogram obtains some relative divergence values, and the method for selected threshold is in these divergence values
In formula, TkThe divergence value being calculated for k-th of grid in all samples.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811310641.1A CN109507193A (en) | 2018-11-06 | 2018-11-06 | A kind of fabric defects detection method based on local contrast enhancing and binary pattern |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811310641.1A CN109507193A (en) | 2018-11-06 | 2018-11-06 | A kind of fabric defects detection method based on local contrast enhancing and binary pattern |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109507193A true CN109507193A (en) | 2019-03-22 |
Family
ID=65747635
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811310641.1A Pending CN109507193A (en) | 2018-11-06 | 2018-11-06 | A kind of fabric defects detection method based on local contrast enhancing and binary pattern |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109507193A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112802074A (en) * | 2021-01-06 | 2021-05-14 | 江南大学 | Textile flaw detection method based on illumination correction and visual saliency characteristics |
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 |
CN114677376A (en) * | 2022-05-30 | 2022-06-28 | 南通摩瑞纺织有限公司 | Method and system for testing defects in production of woven belts based on illumination pretreatment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101866427A (en) * | 2010-07-06 | 2010-10-20 | 西安电子科技大学 | Method for detecting and classifying fabric defects |
CN105277567A (en) * | 2015-05-21 | 2016-01-27 | 南通大学 | Fabric flaw detection method |
-
2018
- 2018-11-06 CN CN201811310641.1A patent/CN109507193A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101866427A (en) * | 2010-07-06 | 2010-10-20 | 西安电子科技大学 | Method for detecting and classifying fabric defects |
CN105277567A (en) * | 2015-05-21 | 2016-01-27 | 南通大学 | Fabric flaw detection method |
Non-Patent Citations (2)
Title |
---|
王美等: "自适应特征提取的光照鲁棒性人脸识别", 《计算机工程与应用》 * |
赵树志等: "基于改进判别性完整局部二值模式与格分割的织物瑕疵检测方法", 《纺织学报》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN112802074A (en) * | 2021-01-06 | 2021-05-14 | 江南大学 | Textile flaw detection method based on illumination correction and visual saliency characteristics |
CN114677376A (en) * | 2022-05-30 | 2022-06-28 | 南通摩瑞纺织有限公司 | Method and system for testing defects in production of woven belts based on illumination pretreatment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Hanbay et al. | Fabric defect detection systems and methods—A systematic literature review | |
Rasheed et al. | Fabric defect detection using computer vision techniques: a comprehensive review | |
Hu | Automated defect detection in textured surfaces using optimal elliptical Gabor filters | |
Zhang et al. | Fabric defect detection using salience metric for color dissimilarity and positional aggregation | |
CN108090894B (en) | Fabric defect detection method based on distance matching function and perceptual hash algorithm | |
Jin et al. | Automatic fabric defect detection based on an improved YOLOv5 | |
CN107220649A (en) | A kind of plain color cloth defects detection and sorting technique | |
CN110060237A (en) | A kind of fault detection method, device, equipment and system | |
CN108154519A (en) | Dividing method, device and the storage medium of eye fundus image medium vessels | |
CN106770323B (en) | The textile flaw detection method filtered based on hierarchical clustering and Gabor | |
CN109507193A (en) | A kind of fabric defects detection method based on local contrast enhancing and binary pattern | |
CN108027970A (en) | Methods, devices and systems for the medical image for analyzing blood vessel | |
Tajeripour et al. | Developing a novel approach for stone porosity computing using modified local binary patterns and single scale retinex | |
CN109087286A (en) | A kind of detection method and application based on Computer Image Processing and pattern-recognition | |
Zhang et al. | A review of fabric identification based on image analysis technology | |
CN111724372A (en) | Method, terminal and storage medium for detecting cloth defects based on antagonistic neural network | |
Hoseini et al. | Fabric defect detection using auto-correlation function | |
CN111091134A (en) | Method for identifying tissue structure of colored woven fabric based on multi-feature fusion | |
Wen et al. | Verifying edges for visual inspection purposes | |
Li et al. | A weak supervision machine vision detection method based on artificial defect simulation | |
Baboiu et al. | Vascular bifurcation detection in scale-space | |
Barisin et al. | Methods for segmenting cracks in 3d images of concrete: A comparison based on semi-synthetic images | |
Aldemir et al. | An improved gray line profile method to inspect the warp–weft density of fabrics | |
CN102024149A (en) | Method of object detection and training method of classifier in hierarchical object detector | |
CN111275705A (en) | Intelligent cloth inspecting method and device, electronic equipment and storage medium |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190322 |