CN110211112A - A kind of casting defect inspection method based on filtering selective search - Google Patents
A kind of casting defect inspection method based on filtering selective search Download PDFInfo
- Publication number
- CN110211112A CN110211112A CN201910472271.XA CN201910472271A CN110211112A CN 110211112 A CN110211112 A CN 110211112A CN 201910472271 A CN201910472271 A CN 201910472271A CN 110211112 A CN110211112 A CN 110211112A
- Authority
- CN
- China
- Prior art keywords
- region
- size
- similarity
- image
- casting
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 99
- 238000005266 casting Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001914 filtration Methods 0.000 title claims abstract description 26
- 238000007689 inspection Methods 0.000 title claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000013316 zoning Methods 0.000 claims abstract description 5
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 239000012141 concentrate Substances 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 21
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000010845 search algorithm Methods 0.000 description 7
- 230000002950 deficient Effects 0.000 description 5
- 230000011218 segmentation Effects 0.000 description 5
- 238000003708 edge detection Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000000052 comparative effect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000009659 non-destructive testing Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
-
- 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
- G06T7/0006—Industrial image inspection using a design-rule based approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10116—X-ray image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30116—Casting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to casting defect detection fields, and disclose a kind of casting defect inspection method based on filtering selective search.This method comprises: (a) carries out image preprocessing to the radioscopic image of misrun casting to be detected;(b) radioscopic image obtain after image preprocessing is divided into multiple regions, similarity between zoning, multiple regions are scanned for using the method for global search, determine defect area suspicious in radioscopic image, and with this so as to form suspected defects region collection;(c) acceptable defect area size threshold is set, is concentrated in the suspected defects region, the suspected defects region that size is greater than acceptable defect area size threshold is deleted, remaining suspected defects region is required real defect region.Through the invention, operator's subjective fault because of caused by the difference that examination criteria executes is avoided, mitigates labor workload, promotes detection efficiency.
Description
Technical field
The invention belongs to casting defect detection fields, lack more particularly, to a kind of casting based on filtering selective search
Fall into detection method.
Background technique
Foundry engieering is widely used because having many advantages, such as shorter production cycle and flexible production method
In industrial production.Therefore, ensure that quality and its safety in utilization of casting are most important.In recent years, casting manufacturer is continuous
Reinforce the inspecting force to casting defect.
X-ray non-destructive testing technology has in the defects detection field of casting to be widely applied.But by being examined to X-ray
It surveys and is formed by the work that image carries out casting defect identification, mostly still completed by the staff of profession, by the master of personnel
Sight factor is affected, and detection efficiency is lower.In recent years, with the hair of the technologies such as image processing techniques and computer technology
Exhibition, realizes the automatic inspection of casting defect in conjunction with technologies such as computer vision, image procossing, machine learning by X-ray detection
Survey existing successful case.
Widely applied X-ray automatic defect detection technique in industrial production at present is detected compared to traditional artificial defect
Method detection efficiency with higher, but there are the following problems: defect inspection process is cumbersome, heavy workload and result
Information is not easy to maintain and inquires.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of based on filtering selective search
Defect inspection method is cast, is pre-processed by the image to misrun casting to be detected, is then examined using selective search algorithm
All suspicious regions in image are measured, it is for statistical analysis further according to actual defects size, determine acceptable flaw size
Size, and in this, as filter condition, apparent non-defective region is filtered out, to retain all real defect regions, is realized
To the quantitatively characterizing of defect target signature, this method is simple to operation, calculates simply, high-efficient, greatly reduces in calculating process
Workload, and the result precision obtained is high, and realization quickly and effectively calculates.
To achieve the above object, it is proposed, according to the invention, provide a kind of casting defects detection side based on filtering selective search
Method, this method include the following steps:
(a) image preprocessing is carried out for the radioscopic image of misrun casting to be detected, on the one hand eliminates the X ray picture
On the other hand noise as in enhances the contrast of misrun casting and background colour to be detected in the radioscopic image;
(b) radioscopic image for carrying out obtaining after image preprocessing in step (a) is divided into multiple regions, zoning
Between similarity, calculated by serial similarity value and determine combined region, suspicious in the radioscopic image lack is determined with this
Region is fallen into, and so as to form suspected defects region collection;
(c) acceptable defect area size threshold is set, the suspected defects region collection obtained in step (b)
In, delete the suspected defects region that size is greater than the acceptable defect area size threshold, remaining suspected defects region
For required real defect region.
It is further preferred that in step (b), similarity between the zoning, using the method for global search
Multiple regions are scanned for, determine defect area suspicious in radioscopic image, and with this so as to form suspected defects region
Collection, specifically preferably carries out according to the following steps:
(b1) adjacent region is concentrated for the region, is calculated the similarity in each adjacent region, is obtained with this more
A similarity simultaneously forms similarity set;
(b2) by the corresponding adjacent region r of the maximum value in the similarity setpAnd rqMerge forming region rpq, will
Region rpqIt is incorporated to the region to concentrate, while being concentrated in the middle region and deleting region rpAnd rq, the region is updated with this
Collection;
In the similarity set, region r is deletedpOr rqRegion adjacent thereto calculates the similarity obtained, described in calculating
Region rpqThe similarity in region adjacent thereto, and the result of calculating is added in the similarity set, it is updated with this described
Similarity set;
(b3) return step (b2), until the similarity collection is combined into empty set, corresponding region integrates as suspected defects at this time
Target area collection, wherein p and q is the number in region.
It is further preferred that the calculation formula of the similarity is preferably according to following progress in step (b):
S(ri,rj)=a1Scolour(ri,rj)+a2Stexture(ri,rj)+a3Ssize(ri,rj)+a4Sfill(ri,rj)
Wherein, S (ri,rj) it is two adjacent region riAnd rjSimilarity, a1, a2, a3And a4For the weight of setting, it is
Constant, Scolour(ri,rj) it is two adjacent region riAnd rjColor similarity, Stexture(ri,rj) it is two adjacent areas
Domain riAnd rjTextural characteristics similarity, Ssize(ri,rj) it is two adjacent region riAnd rjSize similarity, Sfill(ri,
rj) it is two adjacent region riAnd rjOverlapping degree similarity.
It is further preferred that the Scolour(ri,rj) preferably carry out as follows:
Wherein, k is the dimension of vector, and n is total dimension that vector is taken,It is the k dimensional vector of the color of ith zone,It is the k dimensional vector of the color in j-th of region.
It is further preferred that the Stexture(ri,rj) preferably carried out according to following expression formula:
Wherein, l is the dimension of vector, and m is total dimension of vector,It is the l dimensional vector of the texture of ith zone,It is
The l dimensional vector of the texture in j-th of region.
It is further preferred that the Ssize(ri,rj) preferably carried out according to following expression formula:
Wherein, Ssize(ri) it is region riSize, Ssize(rj) it is region rjThe size of size, SsizeIt (im) is whole
Open the size of image im.
It is further preferred that the Sfill(ri,rj) preferably carried out according to following expression formula:
Wherein, BBijIt is by region riAnd rjThe minimum rectangle that envelope gets up, Size (BBij) it is minimum rectangle BBijRuler
It is very little, Size (ri) it is by region riThe size of the minimum rectangle of envelope, Size (rj) it is by region rjThe ruler of the minimum rectangle of envelope
Very little, Size (im) is by the size of the minimum rectangle of whole image im envelope.
It is further preferred that the acceptable defect area size of setting is preferably according to the following steps in step (c)
It carries out:
(c1) radioscopic image for obtaining multiple misrun castings identifies the defects of all radioscopic images, using minimum square
Shape envelope method calculates the minimum dimension for obtaining each defect, and the minimum dimension of all defect is obtained with this;
(c1) maximum value in the minimum dimension of all defect is calculated, which is set as acceptable defect area
Size threshold.
It is further preferred that in step (a), described image processing include filtering processing and the brightness to image, coloration,
The adjustment of contrast and acutance.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
1, method provided by the invention can by execute automatically obtain cast(ing) surface defect area, execute the time relative to
Artificial detection method significantly reduces, while not needing printing electronic image, can reduce manual working intensity, promote working efficiency;
2, the present invention carries out global search to radioscopic image by using selective search algorithm, can capture all pictures
The tiny flaw that human eye may be omitted is covered in plain sudden change region, to reduce omission factor, and personnel can be prevented because of visual fatigue
Caused by false retrieval and missing inspection, while can avoid difference caused by each staff's examination criteria difference;
3, method provided by the invention can the relative position of defect and size on accurate recording image, and automatically save letter
Breath, at the same with the auto-associatings such as casting number, provide information based data support for casting data interconnection and process intelligent optimization, it is square
Just indirect labor carries out casting quality and detects work.
Detailed description of the invention
Fig. 1 is the casting defect detection side based on filtering selective search constructed by preferred embodiment according to the invention
The flow chart of method;
Fig. 2 is different condition feelings in the detection radioscopic image of algorithms of different constructed by preferred embodiment according to the invention
The comparison figure of defect area under condition;
Fig. 3 is that the whole radioscopic image that algorithms of different constructed by preferred embodiment according to the invention finally obtains lacks
Fall into the comparative result figure in region.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
The present invention provides a kind of filtering selective search method for casting x-ray image defects detection, can be automatic right
Casting radioscopic image carries out defect real-time detection of overall importance, to mitigate labor workload, improves precision and efficiency of detecting.
S1: original image pretreatment.According to casting radioscopic image target signature, suitable image filtering and enhancing are chosen
Combination, to improve the picture quality of subsequent processing.Noise is the major reason of image interference.Piece image in practical applications may be used
There can be various noises, these noises may generate in the transmission, it is also possible to generate in the processing such as quantization.It is right first
Image is filtered.In view of there are the lesser defects of size for casting when filtering, so in selection filter or window
Pay attention to retaining image detail when size.It is grayscale image for casting radioscopic image, can suitably enhances brightness of image, coloration, comparison
Degree and acutance improve subsequent detection effect to improve the comparison of image object and background.
S2: image overall regioselectivity search.Using selective search algorithm to original image carry out it is of overall importance can
Doubt range searching.Selective search algorithm calculates the similarity in image between adjacent subarea domain, neighbouring by constantly merging
Similar area, the final target area for obtaining negligible amounts, to reduce the search range of target.Algorithm execution uses
Multifarious strategy avoids causing region recommendation effect poor because of pure strategy.Its multifarious strategy is embodied in
Multiple color space, such as RGB, HSV and gray scale etc..For the influence for considering the factors such as scene and illumination condition, pass through color
Original color space is transformed into up to 8 kinds of color space by spatial alternation;The initialization of original area multi-threshold, threshold value can root
It is adjusted flexibly according to image type and specific situation to optimum state, in general, threshold value is bigger, and the region of segmentation is fewer;It is a variety of
Regional Similarity merges standard, such as considers color, textural characteristics, size and the overlapping degree etc. of adjacent area.Calculating phase
When seemingly spending, the similarity value of single dimension is normalized between 0~1, the value is bigger, then similarity between the region compared
It is bigger, several single similarity values are finally added, comprehensive similarity value is obtained.Therefore the total realization step of algorithm are as follows:
1) candidate regional ensemble is generated first with cutting method, divides the image into many fritters;
2) color, textural characteristics, size and the overlapping degree similarity of every two adjacent area are then calculated;
1. color similarity
The histogram for obtaining the 25bins of each Color Channel of image, region each in this way are normalized using L1-norm
The vector of available one 75 dimensionColor similarity is calculated by following formula between region:
Wherein, k is the dimension of vector, and n is total dimension that vector is taken,It is the k dimensional vector of the color of ith zone,It is the k dimensional vector of the color in j-th of region.
New region is carried out to calculate its histogram using needs during region merging technique, calculation method:
2. textural characteristics similarity
Here texture uses SIFT-Like feature.Specific practice is calculated 8 different directions of each Color Channel
The gaussian derivative of variances sigma=1, each each color in channel obtain the histogram (L1-norm normalization) of 10bins, thus may be used
To get the vector of one 240 dimensionTexture similarity calculation and color similarity meter between region
Calculation mode is similar, and the textural characteristics calculation of new region and color characteristic calculate identical after merging:
Wherein, l is the dimension of vector, and m is total dimension of vector,It is the l dimensional vector of the texture of ith zone,It is
The l dimensional vector of the texture in j-th of region.
3. size similarity
Here size refers to the number in region comprising pixel.Using the similarity calculation of size, primarily to
Small region is allowed first to merge as far as possible:
Wherein, Ssize(ri) it is region riSize, Ssize(rj) it is region rjThe size of size, SsizeIt (im) is whole
Open the size of image im.
4. overlapping degree similarity
Here primarily to measuring whether two region overlapping degrees are higher, index is the region after merging
Bounding Box (minimum rectangle (not rotating) that can frame region) is smaller, and overlapping degree is higher.Its calculation:
Wherein, Sfill(ri,rj) it is overlapping degree similarity.
Finally above-mentioned similarity calculation mode is grouped together, can be written as follow, wherein ai∈ { 0,1 }:
S(ri,rj)=a1Scolour(ri,rj)+a2Stexture(ri,rj)+a3Ssize(ri,rj)+a4Sfill(ri,rj)
Wherein, S (ri,rj) it is two adjacent region riAnd rjSimilarity, a1, a2, a3And a4For the weight of setting, it is
Constant, Scolour(ri,rj) it is two adjacent region riAnd rjColor similarity, Stexture(ri,rj) it is two adjacent areas
Domain riAnd rjTextural characteristics similarity, Ssize(ri,rj) it is two adjacent region riAnd rjSize similarity, Sfill(ri,
rj) it is two adjacent region riAnd rjOverlapping degree similarity.
5. merging the highest two pieces of regions of similarity.
3) repetitive operation 2), obtain the suspected defects regional ensemble that we want.
S3: actual defects size statistic analysis.The real defect region of image, counts all and lacks in artificial shearing data set
Sunken length and width relative size simultaneously records, while the size section of category analyzing defect, calculates defect proportion in each section
Situation.
S4: it determines filter condition and filters out non-defective region.According to the casting real defect size distribution situation counted,
The size critical value and two-dimensional ratio threshold of real defect are set, the screening conditions of filtered search are determined as.In this base
On plinth, all areas that detected by original selection searching algorithm are subjected to limitation screening, filter out obvious non-defective region,
Retain all real defects, to achieve the effect that precise search image object.
Now it is compared for the detectability of proposed method.These selected images include the defect under various states
Situations such as target, such as target are located in structure, and target imaging is fuzzy and target is overlapping.And with Canny side edge detection
Method and threshold segmentation method OTSU are compared, specific comparative effectiveness, as shown in Fig. 2, (a) is to be detected lack in Fig. 2
The X-ray original image of casting is fallen into, (b) is the defect area found using selective search algorithm in Fig. 2, and (c) is to adopt in Fig. 2
The defect area found with filtering selective search of the invention, the defect that (d) uses Prewitt edge detection to find in Fig. 2
Region, (e) is the defect area found using OTSU Threshold segmentation in Fig. 2, as can be seen from the results, with original selectivity
Searching algorithm is compared, and filtered selective search eliminates some targets of structure class, and the visual effect of defects detection is obvious,
Result is more accurate simultaneously.Although the marginal information of defect can be presented in Prewitt edge detection method, can also detect too
Mostly useless edge interferes to generate huge vision to detection effect.Finally, OTSU threshold segmentation method is above-mentioned several
In the case of it is almost invalid, and there are no detect defect completely in some cases.
As shown in figure 3, (a) is the X-ray original image of misrun casting to be detected in Fig. 3, (b) is using selection in Fig. 3
Property the defect area that finds of searching algorithm, (c) is the defect area found using filtering selective search of the invention in Fig. 3,
(d) uses the defect area that finds of Prewitt edge detection in Fig. 2, and (e) is lacking of being found using OTSU Threshold segmentation in Fig. 3
Region is fallen into, (f) is the comparison of filtering front and back detection accuracy in Fig. 3, it can be seen that original selective search algorithm detects
Many non-defective targets, or even actual defects target is covered, to reduce the quantity of actual defects.Filtered selectivity
All suspicious objects are more specifically retrieved in search, almost without omission.And Prewitt detection method detects many in image
Non-defective feature, and poor visibility, so that target identification is difficult.OTSU thresholding dividing method detects some defects, but still
It cannot so can be clearly seen that the true shape of target, and this method falls flat for the target in background.Finally, comparing
False detection rate and omission factor of the selective search algorithm before and after filtering.It can be seen from the figure that filtered false detection rate drops to
1/3 or so of original value, omission factor are also dropped to practically zero from 4%.This is because reducing after filtering to non-defective target
Selection, leave behind a small amount of other similar target in defect.Meanwhile it reduces the meaningless covering of some defect targets
Rate, to reduce missing inspection number.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (9)
1. a kind of casting defect inspection method based on filtering selective search, which is characterized in that this method includes the following steps:
(a) image preprocessing is carried out for the radioscopic image of misrun casting to be detected, on the one hand eliminated in the radioscopic image
Noise, on the other hand enhance the contrast of misrun casting and background colour to be detected in the radioscopic image;
(b) radioscopic image for carrying out in step (a) obtaining after image preprocessing is divided into multiple regions, between zoning
Similarity, by serial similarity value calculate determine combined region, defect area suspicious in the radioscopic image is determined with this
Domain, and so as to form suspected defects region collection;
(c) acceptable defect area size threshold is set, the suspected defects region obtained in step (b) is concentrated, and is deleted
Except size is greater than the suspected defects region of the acceptable defect area size threshold, remaining suspected defects region is required
Real defect region.
2. a kind of casting defect inspection method based on filtering selective search as described in claim 1, which is characterized in that in step
Suddenly in (b), the similarity between the zoning determines that X is penetrated by the determining combined region of serial similarity value calculating with this
Suspicious defect area in line image, and so as to form suspected defects region collection, specifically preferably carried out according to the following steps:
(b1) adjacent region is concentrated for the region, calculates the similarity in each adjacent region, multiple phases are obtained with this
Like spending and form similarity set;
(b2) by the corresponding adjacent region r of the maximum value in the similarity setpAnd rqMerge forming region rpq, by the area
Domain rpqIt is incorporated to the region to concentrate, while being concentrated in the middle region and deleting region rpAnd rq, the region collection is updated with this;
In the similarity set, region r is deletedpOr rqRegion adjacent thereto calculates the similarity obtained, calculates the region
rpqThe similarity in region adjacent thereto, and the result of calculating is added in the similarity set, it is updated with this described similar
Degree set;
(b3) return step (b2), until the similarity collection is combined into empty set, corresponding region integrates as suspected defects target at this time
Region collection, wherein p and q is the number in region.
3. a kind of casting defect inspection method based on filtering selective search as described in claim 1, which is characterized in that in step
Suddenly in (b), the calculation formula of the similarity is preferably according to following progress:
S(ri,rj)=a1Scolour(ri,rj)+a2Stexture(ri,rj)+a3Ssize(ri,rj)+a4Sfill(ri,rj)
Wherein, S (ri,rj) it is two adjacent region riAnd rjSimilarity, a1, a2, a3And a4It is constant for the weight of setting,
Scolour(ri,rj) it is two adjacent region riAnd rjColor similarity, Stexture(ri,rj) it is two adjacent region riWith
rjTextural characteristics similarity, Ssize(ri,rj) it is two adjacent region riAnd rjSize similarity, Sfill(ri,rj) it is two
A adjacent region riAnd rjOverlapping degree similarity.
4. a kind of casting defect inspection method based on filtering selective search as claimed in claim 3, which is characterized in that described
Scolour(ri,rj) preferably carry out as follows:
Wherein, k is the dimension of vector, and n is total dimension that vector is taken,It is the k dimensional vector of the color of ith zone,It is
The k dimensional vector of the color in j-th of region.
5. a kind of casting defect inspection method based on filtering selective search as claimed in claim 3, which is characterized in that described
Stexture(ri,rj) preferably carried out according to following expression formula:
Wherein, l is the dimension of vector, and m is total dimension of vector,It is the l dimensional vector of the texture of ith zone,It is j-th
The l dimensional vector of the texture in region.
6. a kind of casting defect inspection method based on filtering selective search as claimed in claim 3, which is characterized in that described
Ssize(ri,rj) preferably carried out according to following expression formula:
Wherein, Ssize(ri) it is region riSize, Ssize(rj) it is region rjThe size of size, SsizeIt (im) is whole figure
As the size of im.
7. a kind of casting defect inspection method based on filtering selective search as claimed in claim 3, which is characterized in that described
Sfill(ri,rj) preferably carried out according to following expression formula:
Wherein, BBijIt is by region riAnd rjThe minimum rectangle that envelope gets up, Size (BBij) it is minimum rectangle BBijSize,
Size(ri) it is by region riThe size of the minimum rectangle of envelope, Size (rj) it is by region rjThe size of the minimum rectangle of envelope,
Size (im) is by the size of the minimum rectangle of whole image im envelope.
8. a kind of casting defect inspection method based on filtering selective search as described in claim 1, which is characterized in that in step
Suddenly in (c), the acceptable defect area size of setting is preferably carried out according to the following steps:
(c1) radioscopic image for obtaining multiple misrun castings identifies the defects of all radioscopic images, using minimum rectangle packet
Network method calculates the minimum dimension for obtaining each defect, and the minimum dimension of all defect is obtained with this;
(c1) maximum value in the minimum dimension of all defect is calculated, which is set as acceptable defect area size
Threshold value.
9. such as a kind of described in any item casting defect inspection methods based on filtering selective search of claim 1-8, feature
It is, in step (a), described image processing includes filtering processing and the brightness to image, coloration, contrast and acutance
Adjustment.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910472271.XA CN110211112A (en) | 2019-05-31 | 2019-05-31 | A kind of casting defect inspection method based on filtering selective search |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910472271.XA CN110211112A (en) | 2019-05-31 | 2019-05-31 | A kind of casting defect inspection method based on filtering selective search |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110211112A true CN110211112A (en) | 2019-09-06 |
Family
ID=67790107
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910472271.XA Pending CN110211112A (en) | 2019-05-31 | 2019-05-31 | A kind of casting defect inspection method based on filtering selective search |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110211112A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111008952A (en) * | 2019-10-22 | 2020-04-14 | 浙江蓝卓工业互联网信息技术有限公司 | Object internal defect detection method applied to Fabry-Perot interference imaging |
CN112147160A (en) * | 2020-08-13 | 2020-12-29 | 妙微(永康)科技有限公司 | Ultra-small flaw detection method and system |
CN113220922A (en) * | 2021-06-04 | 2021-08-06 | 北京有竹居网络技术有限公司 | Image searching method and device and electronic equipment |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105891220A (en) * | 2016-06-02 | 2016-08-24 | 昆山市交通工程试验检测中心有限公司 | Pavement marker line defect detecting device and detecting method thereof |
CN107808378A (en) * | 2017-11-20 | 2018-03-16 | 浙江大学 | Complicated structure casting latent defect detection method based on vertical co-ordination contour feature |
CN108596875A (en) * | 2018-03-21 | 2018-09-28 | 浙江大学 | A kind of semiconductor chip flash rapid detection method based on image segmentation algorithm |
US20180328860A1 (en) * | 2017-05-15 | 2018-11-15 | Kla-Tencor Corporation | High accuracy of relative defect locations for repeater analysis |
CN109142533A (en) * | 2018-10-22 | 2019-01-04 | 广东工业大学 | A kind of rapid detection method and equipment of internal defect in cast |
US10199985B2 (en) * | 2016-02-16 | 2019-02-05 | Airbus Ds Gmbh | Technique for detecting a defect in a multi-junction solar cell |
CN109492940A (en) * | 2018-12-06 | 2019-03-19 | 华中科技大学 | A kind of casting furnace lodge follow-up of quality method for moulding pouring product line |
CN109658396A (en) * | 2018-12-06 | 2019-04-19 | 华中科技大学 | A kind of casting x-ray image defect identification method based on selective search and CNN |
CN109684854A (en) * | 2018-11-20 | 2019-04-26 | 华中科技大学 | A kind of bottom data encryption method suitable for management information system in enterprise |
-
2019
- 2019-05-31 CN CN201910472271.XA patent/CN110211112A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10199985B2 (en) * | 2016-02-16 | 2019-02-05 | Airbus Ds Gmbh | Technique for detecting a defect in a multi-junction solar cell |
CN105891220A (en) * | 2016-06-02 | 2016-08-24 | 昆山市交通工程试验检测中心有限公司 | Pavement marker line defect detecting device and detecting method thereof |
US20180328860A1 (en) * | 2017-05-15 | 2018-11-15 | Kla-Tencor Corporation | High accuracy of relative defect locations for repeater analysis |
CN107808378A (en) * | 2017-11-20 | 2018-03-16 | 浙江大学 | Complicated structure casting latent defect detection method based on vertical co-ordination contour feature |
CN108596875A (en) * | 2018-03-21 | 2018-09-28 | 浙江大学 | A kind of semiconductor chip flash rapid detection method based on image segmentation algorithm |
CN109142533A (en) * | 2018-10-22 | 2019-01-04 | 广东工业大学 | A kind of rapid detection method and equipment of internal defect in cast |
CN109684854A (en) * | 2018-11-20 | 2019-04-26 | 华中科技大学 | A kind of bottom data encryption method suitable for management information system in enterprise |
CN109492940A (en) * | 2018-12-06 | 2019-03-19 | 华中科技大学 | A kind of casting furnace lodge follow-up of quality method for moulding pouring product line |
CN109658396A (en) * | 2018-12-06 | 2019-04-19 | 华中科技大学 | A kind of casting x-ray image defect identification method based on selective search and CNN |
Non-Patent Citations (3)
Title |
---|
余永维等: "基于深度学习特征的铸件缺陷射线图像动态检测方法", 《农业机械学报》 * |
周建新等: "基于LABVIEW的焊缝表面缺陷智能检测系统研究", 《焊接技术》 * |
张园: "一种基于纹理合成的疏松缺陷图像生成方法", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111008952A (en) * | 2019-10-22 | 2020-04-14 | 浙江蓝卓工业互联网信息技术有限公司 | Object internal defect detection method applied to Fabry-Perot interference imaging |
CN112147160A (en) * | 2020-08-13 | 2020-12-29 | 妙微(永康)科技有限公司 | Ultra-small flaw detection method and system |
CN112147160B (en) * | 2020-08-13 | 2024-05-03 | 妙微(嘉兴)科技有限公司 | Ultra-small flaw detection method and system |
CN113220922A (en) * | 2021-06-04 | 2021-08-06 | 北京有竹居网络技术有限公司 | Image searching method and device and electronic equipment |
CN113220922B (en) * | 2021-06-04 | 2024-02-02 | 北京有竹居网络技术有限公司 | Image searching method and device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109743879B (en) | Underground pipe gallery leakage detection method based on dynamic infrared thermography processing | |
US11669971B2 (en) | Colony contrast gathering | |
JP6660313B2 (en) | Detection of nuclear edges using image analysis | |
Miller et al. | Classification of breast tissue by texture analysis | |
US8340372B2 (en) | Image analysis | |
CA1161271A (en) | Method and apparatus for measuring mean cell volume of red blood cells | |
CN110211112A (en) | A kind of casting defect inspection method based on filtering selective search | |
CN110210448B (en) | Intelligent face skin aging degree identification and evaluation method | |
JP2001522488A (en) | Method and apparatus for automatic muscle segmentation of digital mammograms | |
EP1699016A2 (en) | Method and apparatus for automated analysis of biological specimen | |
CN109685766A (en) | A kind of Fabric Defect detection method based on region fusion feature | |
CN109145708A (en) | A kind of people flow rate statistical method based on the fusion of RGB and D information | |
Wen et al. | Verifying edges for visual inspection purposes | |
CN110189375A (en) | A kind of images steganalysis method based on monocular vision measurement | |
Ghazal et al. | Automated framework for accurate segmentation of leaf images for plant health assessment | |
CN113252568A (en) | Lens surface defect detection method, system, product and terminal based on machine vision | |
Li et al. | An improved water surface images segmentation algorithm based on the Otsu method | |
CN114882400A (en) | Aggregate detection and classification method based on AI intelligent machine vision technology | |
Lin et al. | Surface defect detection of machined parts based on machining texture direction | |
KR101078756B1 (en) | Method for Detection of Clustered Microcalcifications on Digital Mammogram | |
Ben Chaabane et al. | Color edges extraction using statistical features and automatic threshold technique: application to the breast cancer cells | |
CN108648409B (en) | Smoke detection method and device | |
CN112396618B (en) | Grain boundary extraction and grain size measurement method based on image processing | |
CN116934752B (en) | Glass detection method and system based on artificial intelligence | |
CN109815784A (en) | A kind of intelligent method for classifying based on thermal infrared imager, system 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 |
Application publication date: 20190906 |
|
RJ01 | Rejection of invention patent application after publication |