CN106093066B - A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism - Google Patents
A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism Download PDFInfo
- Publication number
- CN106093066B CN106093066B CN201610479587.8A CN201610479587A CN106093066B CN 106093066 B CN106093066 B CN 106093066B CN 201610479587 A CN201610479587 A CN 201610479587A CN 106093066 B CN106093066 B CN 106093066B
- Authority
- CN
- China
- Prior art keywords
- image block
- value
- image
- defect
- notable
- 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
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
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- 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
- G06T2207/30136—Metal
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Biochemistry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism, step is：One, magnetic shoe image is inputted, the method combined is converted using morphologic top cap and bottom cap, enhances image overall intensity contrast degree；Two, gained image uniform is divided into a*b image block, then distinguishes defect image block and non-defective image block using the gray feature amount of the image block after piecemeal；Three, using the significance for improving Itti vision noticing mechanism model calculating gained defect image block, select primary features to form comprehensive notable figure；Four, it selects Otsu threshold partitioning algorithm to comprehensive notable figure thresholding, extracts defect area.The present invention is by utilizing Morphological scale-space, image block and vision noticing mechanism thought, the problems such as effectively overcoming smaller brightness irregularities, magnetic shoe defect area, magnetic shoe interference of texture itself, can quickly and efficiently extract all kinds of magnetic shoe defects, obtain adaptability with very strong.
Description
Technical field
The present invention relates to technical field of machine vision, are paid attention to more specifically to one kind based on improved machine vision
The magnetic tile surface defect detection method of mechanism.
Background technique
Ferrite magnetic shoe is primarily used for a kind of tiles magnet on magneto, and the height of quality directly affects forever
The overall performance of magneto.In magnetic shoe production process, due to technological problems, magnetic shoe surface is easy to appear crackle, breakage, point
The defects of, directly affect the normal use of magnetic shoe.People is used substantially to the judgement of magnetic tile surface defect in industrial production at present
Work detection, detection accuracy is poor, detection efficiency is low and labor cost is high.
With the continuous development of machine vision, the defect detecting technique based on machine vision is had begun in industrial products table
It is used widely in the quality monitoring of face, being detected automatically using machine vision can be improved the production efficiency of enterprise, reduces labour
Cost increases the competitiveness of enterprise.
Illumination is easily led to due to the features such as itself has gray scale difference unobvious, and there are surface radians for magnetic shoe product
Unevenly, image grayscale contrast is low, and high to exploitation precision, fireballing magnetic shoe detection method brings certain difficulty.Li Xue
Qin etc. exists《CAD and graphics journal》Middle proposition utilizes a kind of non-downsampling Contourlet domain adaptive thresholding
The magnetic shoe defect automatic testing method in value face, this method can guarantee that magnetic tile surface defect detection has compared with high-accuracy, still
It is longer to calculate the time.Yu Yongwei etc. proposes that the segmentation of adaptive line Morphology Algorithm lacks according to magnetic shoe surface grey value profile situation
It falls into, but this method is more sensitive to noise.Darabi etc. propose employment artificial neural networks divide defect image, but the algorithm compared with
Complexity, it is computationally intensive, it is not able to satisfy the requirement of on-line real-time measuremen.Above-mentioned algorithm needs to be changed in the accuracy and speed of detection
Into.
Through retrieving, Chinese Patent Application No. 201310020370.7, the applying date is on January 18th, 2013, innovation and creation name
Referred to as：Magnetic tile surface defect feature extraction and defect classification method based on machine vision；This application constructs first is suitble to magnetic
5 scales that watt surface defects characteristic extracts, 8 direction Gabor filter groups, and original image is filtered, obtain 40 width point
Spirogram；The gray average and Variance feature for extracting component map respectively again, form the feature vector of one 80 dimension；And with PCA it is main at
Divide analytic approach and ICA independent component analysis method to carry out dimensionality reduction to the feature vector of 80 dimension of original, removes correlation and redundancy, obtain
The feature vector of 20 dimensions；Characteristic vector data is normalized and is pre-processed, former data are normalized between [0,1]；Finally use
Gridding method and K-CV cross-validation method realize SVM parameter optimization, with training sample data off-line training SVM model；On-line checking
When, pretreated test sample data are input to support vector machines, so that it may realize the automatic Classification and Identification of defect.
For another example Chinese Patent Application No. 201110061144.4, the applying date are on March 14th, 2011, invention and created name
For：Magnetic tile surface defect automatic testing method and device based on machine vision；This application also discloses that one kind passes through machine
The method that vision technique detects magnetic tile surface defect.Specifically detection process is：(1) detected magnetic shoe is placed in conveyer belt
On；(2) start ccd image acquisition device, acquire magnetic shoe surface image and be sent to image processing unit；(3) image procossing list
Member by the image of acquisition through image filtering, image segmentation, after after the processing such as Morphological scale-space, edge detection, processing result is passed
Transport to defect detection unit；(4) one-dimensional digital signal is converted through feature extraction by processing result image；(5) by acquired results
After the training of conceptual schema recognition unit and test, magnetic shoe surface quality is divided into two class of good magnetic shoe and defect magnetic shoe, to reach
To the purpose of defects detection.
Above-mentioned application case can filter out the interference of magnetic shoe surface texture to a certain extent, and the feature of extraction also can be certain
Degree reflects defect information；But above-mentioned application case or that there are algorithm operation quantities is big, the long problem of detection time, or do not consider magnetic
Watt product gray scale difference is unobvious, there are problems that surface radian, lower to the detection accuracy of magnetic tile surface defect, still needs to further
It improves.
Summary of the invention
1. technical problems to be solved by the inivention
The present invention existing magnetic tile surface defect detection algorithm aiming at the problem that equal Shortcomings in detection accuracy and speed,
A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism is provided, realizes the automatic of magnetic shoe defect
Detection.The present invention uses morphology top cap and bottom cap to convert the contrast of enhancing defect area and background image first, then sharp
It whether there is defect with the gray feature amount contrasting detection magnetic shoe between image block each after image block, effectively reduce algorithm
Operand；It is finally the interference for overcoming the normal grinding texture of magnetic shoe to extract defect, is divided using improved visual attention model
Image realizes defect Segmentation and extraction, the present invention can be dry efficiently against noise by calculating defect image block vision saliency value
It disturbs, algorithm accuracy rate is high, and, high reliablity good to the adaptability of different defects.
2. technical solution
In order to achieve the above objectives, technical solution provided by the invention is：
A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism of the invention, step
For：
Step 1: input magnetic shoe surface original image, converts the method combined using morphologic top cap and bottom cap, increases
Strong image overall intensity contrast degree；
Step 2: magnetic shoe image uniform obtained by step 1 is divided into a*b image block, each image block side length is M*N, so
Defect image block and non-defective image block are distinguished using the gray feature amount of the image block after piecemeal afterwards, judges that image whether there is
Defect simultaneously determines tile location where defect；
Step 3: using the significance for improving defect image block obtained by Itti vision noticing mechanism model calculating step 2,
It selects primary features to form notable figure, fused comprehensive notable figure is normalized, obtaining maximum focus-of-attention is
For defect area, directly comprehensive notable figure is split；
Step 4: selecting Otsu threshold partitioning algorithm to comprehensive notable figure thresholding after obtaining comprehensive notable figure, extract
Defect area.
Further, step 1 carries out top cap and the process of bottom cap transformation is：
Top cap Transformation Graphs That(f) and bottom cap Transformation Graphs Bhat(f) calculation formula is as follows：
Wherein, f indicates that original image, γ (f) indicate that the opening operation of original image f, φ (f) indicate closing for original image f
Operation；
Original image f is subtracted into That(f) it reduces the bright detail of image, then subtracts Bhat(f) enhance image comparison
Degree realizes the purpose for emphasizing magnetic shoe defect area, exports image kTHIt indicates, calculation formula is as follows：
kTH=f- λ That(f)-ψBhat(f)
In formula, select the circular configuration that radius is 17mm as the structural element in morphology, parameter factors λ=0.1, ψ
=0.1.
Further, the gray feature amount of image block described in step 2 includes image block mean value Wm, to four directions ash
Spend the entropy W of co-occurrence matrix superpositioncWith improved image block variance Wd；Wherein：
Image block mean value WmFor n before counting after image block grey scale pixel value sorts from small to large local mean values, count
It is as follows to calculate formula：
In formula, f (xi, yj) indicate coordinate (x in regional areai, yj) grey scale pixel value；
It is as follows to the entropy Wc calculation formula of four direction gray level co-occurrence matrixes superposition：
In above formula, k is the side length of gray level co-occurrence matrixes；P (i, j) indicates the statistical probability in matrix at (i, j), t table
Show that gray level co-occurrence matrixes calculate the direction of entropy；
Improved image block variance WdCalculation formula is as follows：
Wd=Std (SN-min(SN))
SN(x)=[SN(x1) ..., SN(xN)]
In above formula, Std () is to ask mean square deviation formula, SNIt (x) is to be added the gray value of vertical direction in image block to be formed
1 × N matrix, f (xi, yj) indicates coordinate be (xi, yj) point grey scale pixel value.
Further, the standard for judging defect image block and non-defective image block in step 2 is：Defect image block
WmValue is respectively less than the W of locating row image blockmValue, WcAnd WdValue is all larger than the W of locating row image blockc、WdValue, the W of defect image blockd
Value is greater than the locating second largest W of row image blockd3 times of value.
Further, the primary features selected in step 3 include：
A. local luminance feature：Using the mean value and variance of local luminance as reference, with each picture in regional area
The mean value and variance of the image block gray value of the gray value and region of element subtract each other to obtain difference, and carry out index letter to the difference
Number processing, obtains local luminance notable figure Sl(x, y), calculation formula are as follows：
Ilocal(xi, yj)=f (xi, yj)-(mlocal(x, y)+μ dlocal(x, y))
In formula, mlocal(x, y) indicates local gray level average value, dlocal(x, y) indicates local gray level variance, f (xi, yj) table
Show that gray value, μ are variance controlling elements, value range is 0~1；
B. global brightness：Using global average value as reference, subtracted each other therewith with the gray value of each pixel, and
Exponential function processing is carried out to difference, obtains the significant S of global brightnessg(x, y), calculation formula are as follows：
Iglobal(xi, yi)=| f (xi, yi)-mglobal|
In formula, mglobalFor the average gray of whole picture magnetic shoe image, sum (Iglobal(xi, yi)) it is all Iglobal(xi,
yi) value sum；
C. frequecy characteristic：Using DoG filter, frequecy characteristic notable figure S is obtainedG(x, y), calculation formula are as follows：
SG(x, y)=| mglobal-IG(x, y) |
In formula, IG(x, y) is through the filtered image of DoG, σ1And σ2It is the standard deviation of gaussian kernel function, σ1∶σ2=5: 1.
Further, the σ1=0.5, σ2=0.1.
Further, aobvious for obtained local luminance notable figure, global brightness notable figure, frequecy characteristic in step 3
Work figure is merged, and the calculation formula for generating comprehensive notable figure is as follows：
3. beneficial effect
Using technical solution provided by the invention, compared with existing well-known technique, there is following remarkable result：
(1) a kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism of the invention, passes through
Defect area and background image contrast in morphologic top cap and bottom cap transformation enhancing image, inhibit the gray scale of high-brightness region
Value, enhances image overall intensity contrast degree, conveniently identifies defect area；
(2) a kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism of the invention, in conjunction with
Image block thought efficiently differentiates defect map using the gray feature amount of defect image block and the difference of non-defective image block
As block and non-defective image block, image is judged with the presence or absence of defect and determines defect place tile location, improves algorithm
Operational efficiency reduces operation time and reduces partitioning algorithm complexity；
(3) a kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism of the invention is improved
Vision noticing mechanism model realizes defect Segmentation and extraction by calculating defect image block vision saliency value, can be effective gram
The interference of magnetic shoe texture is taken, it is accurate to complete to extract the defect on magnetic shoe surface；
(4) a kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism of the invention, effectively
Ground overcomes the problems such as smaller brightness irregularities, magnetic shoe defect area, magnetic shoe interference of texture itself, can quickly and efficiently mention
All kinds of magnetic shoe defects are taken, there is very strong adaptability.
Detailed description of the invention
Fig. 1 is magnetic tile surface defect detection algorithm flow chart of the invention；
(a)~(h) in Fig. 2 is genetic defects figure of the invention；
(a)~(h) in Fig. 3 is notable figure fusion figure of the invention；
(a)~(h) in Fig. 4 is image segmentation figure of the invention.
Specific embodiment
To further appreciate that the contents of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments.
Embodiment 1
In conjunction with Fig. 1, a kind of magnetic tile surface defect detection side based on improved machine vision attention mechanism of the present embodiment
Method includes the following steps：
The first step：Magnetic shoe surface original image is inputted, during image processing, due to the lower region of some gray values
It is easy to obscure with defect area, interferes processing result.The present embodiment proposes to combine using morphologic top cap and the transformation of bottom cap
Method, improve defect area and background contrast, inhibit the gray value of high-brightness region, make image overall intensity contrast degree
Enhanced, conveniently identifies defect area.
Progress top cap and the process of bottom cap transformation are：
Top cap Transformation Graphs ThatIt (f) is and the bottom cap Transformation Graphs B from the opening operation γ (f) of original image f subtracted imagehat(f)
It is that f is subtracted by the closed operation φ (f) of original image, calculation formula is as follows：
The high gray value region of original image f is shown in highlight regions after top cap transformation, highlights after the transformation of bottom cap
The low ash angle value region of original image f is shown in region.It therefore, can will be former to further decrease the gray value of inclined bright part
Beginning, image f subtracted That(f) it reduces the bright detail of image, then subtracts Bhat(f) enhance picture contrast, realization emphasizes magnetic
Watt defect area.Export image kTHIt indicates, calculation formula is as follows：
kTH=f- λ That(f)-ψBhat(f)
In above formula, select radius be 17mm circular configuration as the structural element in morphology, parameter factors take λ=
0.1, ψ=0.1.
Second step：In order to improve algorithm operational efficiency, reducing operation time and reduce partitioning algorithm complexity, in image
It should quickly be detected before segmentation in piece image with the presence or absence of defect area.The present embodiment combines the thought of image block first,
Magnetic shoe image uniform obtained by step 1 is divided into a*b image block, each image block side length is M*N, in order to make the figure after piecemeal
As moderate dimensions just include defect area, the value of M, N need test of many times to measure.Then the image block after piecemeal is utilized
Gray-scale statistical amount, i.e., the image block mean value after piecemeal, to the entropy and improved image block of the superposition of four direction gray level co-occurrence matrixes
Three kinds of gray feature amounts such as variance, the gray feature amount of the different images block by comparing input picture, efficiently differentiate defect
Image block and non-defective image block judge image with the presence or absence of defect and determine defect place tile location.
Using above-mentioned 3 kinds of gray feature amounts, the detailed process for distinguishing defect image block and non-defective image block is：
Since image is divided into a row b column, so to calculate three kinds of gray feature amounts of a*b image block, specially：
A. mean value Wm：It is smaller to the contribution of image mean value since defect area is smaller relative to whole image block area, therefore
When calculating mean value, the pixel for participating in calculating mean value can be counted and scope of statistics is defined in defect area as far as possible.This reality
Apply n before counting after example sorts grey scale pixel value from small to large local mean value Wm, can use gray scale in practice process
Histogram realizes that n value is 250 appropriate.Calculation formula is as follows：
In formula, f (xi, yj) indicate coordinate (x in regional areai, yj) grey scale pixel value；
B. entropy Wc：Gray level co-occurrence matrixes mainly reflect grayscale distribution information, and element is by separately certain in co-occurrence matrix
What the pixel with same grayscale value of distance constituted number.To reduce operand, the present embodiment is by the gray scale of image block
Grade is compressed to 0~15 from 0~255 and characterizes gray level co-occurrence matrixes, including energy, entropy, the moment of inertia and phase with some scalars again
Guan Xing.Entropy W in gray level co-occurrence matrixesgIt is able to reflect the complexity of image, it, can be four by it for the conspicuousness for increasing entropy
The entropy in direction is added, and is Wc.Calculation formula is as follows：
In above formula, k is the side length of gray level co-occurrence matrixes, when tonal gradation is compressed to 16, k=16；P (i, j) is indicated
Statistical probability in matrix at (i, j), t indicate that gray level co-occurrence matrixes calculate the direction of entropy.
C. variance Wd：For the influence for reducing magnetic shoe texture, the present embodiment is first by the gray value of vertical direction in image block
It is added, forms the matrix S of 1 × NM(y)；Effect of the defect area in calculating process is influenced because gray value is added to reduce,
Matrix items subtract the minimum value in matrix when calculating, the mean square deviation of last calculating matrix, and calculation formula is as follows：
Wd=Std (SN-min(SN))
SN(x)=[SN(x1) ..., SN(xN)]
In above formula, Std () is to ask mean square deviation formula, SNIt (x) is to be added the gray value of vertical direction in image block to be formed
1 × N matrix f (xi, yj) indicates coordinate be (xi, yj) point grey scale pixel value.
Count the mean value of each image block after every a line piecemeal, entropy and variance.The standard for determining whether defect block is：Have scarce
Fall into the W of image blockmValue is respectively less than the image block of locating row, WcAnd WdIt is all larger than locating row image block, it should be noted that defect map
As block WdValue is much larger than non-defective image block, so with WdCertain multiple relationship should be set for criterion, the present embodiment takes multiple
For maximum W in a line image blockdValue is greater than the second largest Wd3 times.If there is the parameter of image block in a line image block statistical form
Meeting conditions above then judges the image block for defect image block.
Third step：Since magnetic shoe surface texture is compared with horn of plenty, directly progress Threshold segmentation difficulty is larger, and the present embodiment uses
Improve the significance that Itti vision noticing mechanism model calculates defect image block obtained by step 2.It is aobvious from global saliency value and part
The consideration of work value angle reselects primary features to form notable figure, normalizes to [0,1] to fused comprehensive notable figure,
Obtain again maximum focus-of-attention (i.e. saliency value be 1 position) be defect area, since defect image block size is smaller, so
It need not consider other focus-of-attentions, therefore can be directly to comprehensive notable figure segmentation.Detailed process is：
(1) primary features select
A. local luminance feature：Using the mean value and variance of local luminance as reference, with each picture in regional area
The mean value and variance of the image block gray value of the gray value and region of element subtract each other to obtain difference, and carry out index letter to the difference
Number processing, obtains local luminance notable figure Sl(x, y), calculation formula are as follows：
Ilocal(xi, yj)=f (xi, yj)-(mlocal(x, y)+μ dlocal(x, y))
M in formulalocal(x, y) indicates local gray level average value, dlocal(x, y) indicates local gray level variance, f (xi, yj) table
Show that gray value, μ are that variance controlling elements value range is 0~1.
B. global brightness lacks enough local contrasts, therefore in larger defect area in the middle part of defect area
Using global average value as reference, subtracted each other therewith with the gray value of each pixel, and numerical value processing is carried out to difference, obtained
Global brightness notable figure.Global brightness notable figure Sg(x, y) calculation formula is as follows：
Iglobal(xi, yj)=| f (xi, yj)-mglobal|
In formula, mglobalFor the average gray of whole picture magnetic shoe image, sum (Iglobal(xi, yj)) it is all Iglobal(xi,
yj) value sum.
C. frequecy characteristic：Since the visible of edge and other details can be enhanced in DoG filter while filtering out interference
Property, so selecting DoG filter, calculation formula is as follows：
In formula, σ1And σ2It is the standard deviation of gaussian kernel function, the bandwidth of their proportionate relationship control filter works as σ1=
1.6σ2When, it is exactly edge detector, works as σ1When infinitely great, then DC component is filtered, i.e., background image filters out.
When handling image data, to filter out texture and other interference informations using a lesser Gaussian kernel window, one
As use [1,4,6,4,1]/16.Using DoG filter, frequecy characteristic notable figure is obtained by following formula：
SG(x, y)=| mglobal-IG(x, y) |
S in formulaG(x, y) is frequecy characteristic notable figure, IG(x, y) is through the filtered image of DoG.σ is taken in the present embodiment1∶
σ2=5: 1, σ1=0.5, σ2=0.1.
(2) Fusion Features：Since local luminance notable figure, global brightness notable figure, the significant graph expression of frequecy characteristic are specific
The characteristic in direction has certain limitation, so carrying out fusion generates a comprehensive notable figure, to image segmentation.It calculates
Formula is as follows：
4th step：Big saliva (Ostu) Threshold Segmentation Algorithm is selected to carry out threshold to comprehensive notable figure after obtaining comprehensive notable figure
Value obtains comprehensive notable figure marking area and non-significant region (foreground and background), can be realized to the significant area of original image
The segmentation in domain (defect area).The marking area being partitioned into not completely corresponds to the defect area of original image, Ke Nengcun
Edge region and isolated significant point, but these can be eliminated carefully by carrying out morphology opening operation to thresholded image
Section.
(a)~(h) in Fig. 2 is genetic defects figure of the invention, and (a)~(h) in Fig. 3 is that notable figure of the invention is melted
Figure is closed, as can be seen from Figure 3 relative to the notable figure of primary features, total notable figure has on inhibiting texture information obviously to be mentioned
It is high.(a)~(h) in Fig. 4 is image segmentation figure of the invention, after dividing as can be seen from Figure 4 to comprehensive specific image,
Defect area is substantially achieved extraction.
Schematically the present invention and embodiments thereof are described above, description is not limiting, institute in attached drawing
What is shown is also one of embodiments of the present invention, and actual structure is not limited to this.So if the common skill of this field
Art personnel are enlightened by it, without departing from the spirit of the invention, are not inventively designed and the technical solution
Similar frame mode and embodiment, are within the scope of protection of the invention.
Claims (6)
1. a kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism, step are：
Step 1: input magnetic shoe surface original image, converts the method combined, enhancing figure using morphologic top cap and bottom cap
As overall intensity contrast degree；
Step 2: magnetic shoe image uniform obtained by step 1 is divided into a*b image block, each image block side length is M*N, then sharp
Defect image block and non-defective image block are distinguished with the gray feature amount of the image block after piecemeal, judges image with the presence or absence of defect
And determine tile location where defect；
Step 3: using the significance for improving defect image block obtained by Itti vision noticing mechanism model calculating step 2, selection
Primary features are normalized fused comprehensive notable figure to form notable figure, and obtaining maximum focus-of-attention is to lack
Fall into region；Wherein, the primary features of selection include：
A. local luminance feature：Using the mean value and variance of local luminance as reference, with each pixel in regional area
The mean value and variance of gray value and the image block gray value in the region subtract each other to obtain difference, and carry out at exponential function to the difference
Reason, obtains local luminance notable figure Sl(x, y), calculation formula are as follows：
Ilocal(xi, yj)=f (xi, yj)-(mlocal(x, y)+μ dlocal(x, y))
In formula, mlocal(x, y) indicates local gray level average value, dlocal(x, y) indicates local gray level variance, f (xi, yj) indicate to sit
It is designated as (xi, yj) point grey scale pixel value, μ be variance controlling elements, value range be 0~1；
B. global brightness：Using global average value as reference, subtracted each other therewith with the gray value of each pixel, and to difference
Value carries out exponential function processing, obtains global brightness notable figure Sg(x, y), calculation formula are as follows：
Iglobal(xi, yj)=| f (xi, yj)-mglobal|
In formula, mglobalFor the average gray of whole picture magnetic shoe image, sum (Iglobal(xi, yj)) it is all Iglobal(xi, yj) value
Sum；
C. frequecy characteristic：Using DoG filter, frequecy characteristic notable figure S is obtainedG(x, y), calculation formula are as follows：
SG(x, y)=| mglobal-IG(x, y) |
In formula, IG(x, y) is through the filtered image of DoG, σ1And σ2It is the standard deviation of gaussian kernel function, σ1∶σ2=5: 1；
Step 4: selecting Otsu threshold partitioning algorithm to comprehensive notable figure thresholding after obtaining comprehensive notable figure, defect is extracted
Region.
2. a kind of magnetic tile surface defect detection side based on improved machine vision attention mechanism according to claim 1
Method, it is characterised in that：Step 1 carries out top cap and the process of bottom cap transformation is：
Top cap Transformation Graphs That(f) and bottom cap Transformation Graphs Bhat(f) calculation formula is as follows：
Wherein, f indicates that original image, γ (f) indicate that the opening operation of original image f, φ (f) indicate the closed operation of original image f；
Original image f is subtracted into That(f) it reduces the bright detail of image, then subtracts Bhat(f) enhance picture contrast, it is real
It now emphasizes the purpose of magnetic shoe defect area, exports image kTHIt indicates, calculation formula is as follows：
kTH=f- λ That(f)-ψBhat(f)
In formula, select the circular configuration that radius is 17mm as the structural element in morphology, parameter factors λ=0.1, ψ=
0.1。
3. a kind of magnetic tile surface defect detection based on improved machine vision attention mechanism according to claim 1 or 2
Method, it is characterised in that：The gray feature amount of image block described in step 2 includes image block mean value Wm, it is total to four direction gray scales
The entropy W of raw matrix superpositioncWith improved image block variance Wd；Wherein：
Image block mean value WmFor n before counting after image block grey scale pixel value sorts from small to large local mean values, calculation formula
It is as follows：
In formula, f (xi, yj) indicate that coordinate is (x in regional areai, yj) point grey scale pixel value；
To the entropy W of four direction gray level co-occurrence matrixes superpositioncCalculation formula is as follows：
In above formula, k is the side length of gray level co-occurrence matrixes；P (i, j) indicates the statistical probability in matrix at (i, j), and t indicates ash
Spend the direction that co-occurrence matrix calculates entropy；
Improved image block variance WdCalculation formula is as follows：
Wd=Std (SN(x)-min(SN(x)))
SN(x)=[SN(x1) ..., SN(xN)]
In above formula, Std () is to ask mean square deviation formula, SN(x) for the gray value of vertical direction in image block is added to be formed 1 ×
The matrix of N, f (xi, yj) indicates coordinate be (xi, yj) point grey scale pixel value.
4. a kind of magnetic tile surface defect detection side based on improved machine vision attention mechanism according to claim 3
Method, it is characterised in that：The standard for judging defect image block and non-defective image block in step 2 is：The W of defect image blockmValue is equal
Less than the W of locating row image blockmValue, WcValue is greater than the W of locating row image blockcValue, the W of defect image blockdValue is greater than locating row figure
As the second largest W of blockd3 times of value.
5. a kind of magnetic tile surface defect detection side based on improved machine vision attention mechanism according to claim 4
Method, it is characterised in that：The σ1=0.5, σ2=0.1.
6. a kind of magnetic tile surface defect detection side based on improved machine vision attention mechanism according to claim 5
Method, it is characterised in that：Local luminance notable figure, global brightness notable figure, frequecy characteristic notable figure in step 3 for obtaining
It is merged, the calculation formula for generating comprehensive notable figure is as follows：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610479587.8A CN106093066B (en) | 2016-06-24 | 2016-06-24 | A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610479587.8A CN106093066B (en) | 2016-06-24 | 2016-06-24 | A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106093066A CN106093066A (en) | 2016-11-09 |
| CN106093066B true CN106093066B (en) | 2018-11-30 |
Family
ID=57253943
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610479587.8A Active CN106093066B (en) | 2016-06-24 | 2016-06-24 | A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106093066B (en) |
Families Citing this family (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106790898B (en) * | 2016-12-08 | 2019-06-18 | 华中科技大学 | A kind of mobile phone screen bad point automatic testing method and system based on significance analysis |
| CN106600593B (en) * | 2016-12-19 | 2019-08-09 | 福州大学 | A kind of middle aluminium porcelain ball surface defect inspection method |
| CN107369163B (en) * | 2017-06-15 | 2020-12-01 | 西安微电子技术研究所 | Rapid SAR image target detection method based on optimal entropy dual-threshold segmentation |
| CN107490583A (en) * | 2017-09-12 | 2017-12-19 | 桂林电子科技大学 | A kind of intermediate plate defect inspection method based on machine vision |
| CN108305243B (en) * | 2017-12-08 | 2021-11-30 | 五邑大学 | Magnetic shoe surface defect detection method based on deep learning |
| CN108447050A (en) * | 2018-03-07 | 2018-08-24 | 湘潭大学 | A kind of Surface Flaw dividing method based on super-pixel |
| CN108520274B (en) * | 2018-03-27 | 2022-03-11 | 天津大学 | High-reflectivity surface defect detection method based on image processing and neural network classification |
| CN109211919B (en) * | 2018-09-03 | 2021-04-30 | 珠海格力智能装备有限公司 | Method and device for identifying magnetic tile defect area |
| CN109701890A (en) * | 2018-12-10 | 2019-05-03 | 湖南航天天麓新材料检测有限责任公司 | Magnetic tile surface defect detection and method for sorting |
| CN109858501A (en) * | 2019-02-20 | 2019-06-07 | 云南农业大学 | A kind of two phase flow pattern feature extracting method |
| CN110033434A (en) * | 2019-03-04 | 2019-07-19 | 南京航空航天大学 | A kind of detection method of surface flaw based on texture conspicuousness |
| CN110009638B (en) * | 2019-04-12 | 2023-01-03 | 重庆交通大学 | Bridge inhaul cable image appearance defect detection method based on local statistical characteristics |
| CN110210608B (en) * | 2019-06-05 | 2021-03-26 | 国家广播电视总局广播电视科学研究院 | Low-illumination image enhancement method based on attention mechanism and multi-level feature fusion |
| CN110751623A (en) * | 2019-09-06 | 2020-02-04 | 深圳新视智科技术有限公司 | Joint feature-based defect detection method, device, equipment and storage medium |
| CN110766675B (en) * | 2019-10-22 | 2020-07-10 | 科士恩科技(上海)有限公司 | Solar cell panel defect detection method |
| CN112950526B (en) * | 2019-11-25 | 2024-03-12 | 合肥欣奕华智能机器股份有限公司 | Display defect detection method and device |
| CN110992336A (en) * | 2019-12-02 | 2020-04-10 | 东莞西尼自动化科技有限公司 | Small sample defect detection method based on image processing and artificial intelligence |
| CN111210419B (en) * | 2020-01-09 | 2023-10-20 | 浙江理工大学 | Micro magnetic shoe surface defect detection method based on human visual characteristics |
| CN111476784A (en) * | 2020-04-14 | 2020-07-31 | 孙洁 | Product surface defect online detection method based on image enhancement recognition technology |
| CN111612787B (en) * | 2020-06-19 | 2021-09-14 | 国网湖南省电力有限公司 | Concrete crack high-resolution image lossless semantic segmentation method and device and storage medium |
| CN112200826B (en) * | 2020-10-15 | 2023-11-28 | 北京科技大学 | Industrial weak defect segmentation method |
| CN112330614B (en) * | 2020-10-27 | 2021-06-18 | 哈尔滨市科佳通用机电股份有限公司 | Bottom plate bolt loss detection method based on image processing |
| CN112435232A (en) * | 2020-11-23 | 2021-03-02 | 南京信息工程大学 | Defect detection method based on haar wavelet combined image variance |
| CN113702257A (en) * | 2021-08-09 | 2021-11-26 | 西南石油大学 | Conglomerate pore structure characterization method based on CT three-dimensional data volume |
| CN114240926B (en) * | 2021-12-28 | 2022-12-13 | 湖南云箭智能科技有限公司 | Board card defect type identification method, device and equipment and readable storage medium |
| CN114004837B (en) * | 2022-01-04 | 2022-03-04 | 深圳精智达技术股份有限公司 | Method for detecting foreign matters on cover glass module and related device |
| CN114897772B (en) * | 2022-03-31 | 2024-05-14 | 河南省开仑化工有限责任公司 | Method for regulating and controlling forward vulcanization of rubber based on machine vision |
| CN114627111B (en) * | 2022-05-12 | 2022-07-29 | 南通英伦家纺有限公司 | Textile defect detection and identification device |
| CN115063400B (en) * | 2022-07-22 | 2022-11-01 | 山东中艺音美器材有限公司 | Musical instrument production defect detection method using visual means |
| CN115082485B (en) * | 2022-08-23 | 2023-08-29 | 广东欧达雅包装制品有限公司 | Method and system for detecting bubble defects on surface of injection molding product |
| CN117237270B (en) * | 2023-02-24 | 2024-03-19 | 靖江仁富机械制造有限公司 | Forming control method and system for producing wear-resistant and corrosion-resistant pipeline |
| CN115861317B (en) * | 2023-02-27 | 2023-04-28 | 深圳市伟利达精密塑胶模具有限公司 | Plastic mold production defect detection method based on machine vision |
| CN116012659B (en) * | 2023-03-23 | 2023-06-30 | 海豚乐智科技(成都)有限责任公司 | Infrared target detection method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567731A (en) * | 2011-12-06 | 2012-07-11 | 北京航空航天大学 | Extraction method for region of interest |
| CN103198322A (en) * | 2013-01-18 | 2013-07-10 | 江南大学 | Magnetic tile surface defect feature extraction and defect classification method based on machine vision |
| CN103729848A (en) * | 2013-12-28 | 2014-04-16 | 北京工业大学 | Hyperspectral remote sensing image small target detection method based on spectrum saliency |
| CN103729842A (en) * | 2013-12-20 | 2014-04-16 | 中原工学院 | Fabric defect detection method based on local statistical characteristics and overall significance analysis |
| CN105354831A (en) * | 2015-09-30 | 2016-02-24 | 广东工业大学 | Multi-defect detection method based on image block variance-weighting eigenvalues |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5198420B2 (en) * | 2009-12-18 | 2013-05-15 | 株式会社日立ハイテクノロジーズ | Image processing apparatus, measurement / inspection system, and program |
-
2016
- 2016-06-24 CN CN201610479587.8A patent/CN106093066B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567731A (en) * | 2011-12-06 | 2012-07-11 | 北京航空航天大学 | Extraction method for region of interest |
| CN103198322A (en) * | 2013-01-18 | 2013-07-10 | 江南大学 | Magnetic tile surface defect feature extraction and defect classification method based on machine vision |
| CN103729842A (en) * | 2013-12-20 | 2014-04-16 | 中原工学院 | Fabric defect detection method based on local statistical characteristics and overall significance analysis |
| CN103729848A (en) * | 2013-12-28 | 2014-04-16 | 北京工业大学 | Hyperspectral remote sensing image small target detection method based on spectrum saliency |
| CN105354831A (en) * | 2015-09-30 | 2016-02-24 | 广东工业大学 | Multi-defect detection method based on image block variance-weighting eigenvalues |
Non-Patent Citations (2)
| Title |
|---|
| Visual Attention in Objective Image Quality Assessment: Based on Eye-Tracking Data;Hantao Liu et al.;《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》;20110731;第21卷(第7期);第971-982页 * |
| 磁瓦表面图像的自适应形态学滤波缺陷提取方法;余永维 等;《计算机辅助设计与图形学学报》;20120331;第24卷(第3期);第352-354页第1-3节以及图1-4 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106093066A (en) | 2016-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106093066B (en) | A kind of magnetic tile surface defect detection method based on improved machine vision attention mechanism | |
| Liming et al. | Automated strawberry grading system based on image processing | |
| CN115294113B (en) | Quality detection method for wood veneer | |
| Song et al. | Defect detection in random colour textures | |
| Mittal et al. | Non-destructive image processing based system for assessment of rice quality and defects for classification according to inferred commercial value | |
| Teimouri et al. | On-line separation and sorting of chicken portions using a robust vision-based intelligent modelling approach | |
| CN103034852B (en) | The detection method of particular color pedestrian under Still Camera scene | |
| CN108985170A (en) | Transmission line of electricity hanger recognition methods based on Three image difference and deep learning | |
| CN111582359B (en) | Image identification method and device, electronic equipment and medium | |
| CN110009618A (en) | A kind of Axle Surface quality determining method and device | |
| Dong et al. | A novel method for extracting information on pores from cast thin-section images | |
| CN108287010A (en) | A kind of crab multi objective grading plant and method | |
| CN108921004A (en) | Safety cap wears recognition methods, electronic equipment, storage medium and system | |
| Galsgaard et al. | Circular hough transform and local circularity measure for weight estimation of a graph-cut based wood stack measurement | |
| Abdellah et al. | Defect detection and identification in textile fabric by SVM method | |
| Yro et al. | Cocoa beans fermentation degree assessment for quality control using machine vision and multiclass svm classifier | |
| Gurubelli et al. | Texture and colour gradient features for grade analysis of pomegranate and mango fruits using kernel-SVM classifiers | |
| Wu et al. | Fast processing of foreign fiber images by image blocking | |
| Pratap et al. | Development of Ann based efficient fruit recognition technique | |
| Yeh et al. | Establishing a demerit count reference standard for the classification and grading of leather hides | |
| Chakraborty et al. | Recent developments in paper currency recognition system | |
| CN109509168A (en) | A kind of details automatic analysis method for picture quality objective evaluating dead leaf figure | |
| CN108154116A (en) | A kind of image-recognizing method and system | |
| CN107545565A (en) | A kind of solar energy half tone detection method | |
| Ilonen et al. | Estimation of bubble size distribution based on power spectrum |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |