CN102998316B - Transparent liquid impurity detection system and detection method thereof - Google Patents
Transparent liquid impurity detection system and detection method thereof Download PDFInfo
- Publication number
- CN102998316B CN102998316B CN201210560198.XA CN201210560198A CN102998316B CN 102998316 B CN102998316 B CN 102998316B CN 201210560198 A CN201210560198 A CN 201210560198A CN 102998316 B CN102998316 B CN 102998316B
- Authority
- CN
- China
- Prior art keywords
- image
- detection
- impurity
- product
- identification
- 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.)
- Expired - Fee Related
Links
Abstract
The invention discloses a transparent liquid impurity detection system. The transparent liquid impurity detection system comprises an operation panel and a host computer, a DSP (digital signal processor), a plurality of CCD (charge coupled device) cameras and a photoelectric switch which are sequentially connected, an encoder, an exclusion device and an exclusion confirmation switch which are connected in parallel, and a PLC (programmable logic controller) station. A detection method of the detection system comprises the following steps of: collecting sequence images of a product to be detected by utilizing the plurality of the CCD cameras and storing the sequence images; performing image background suppression by image pretreatment; performing treatment on the images after background suppression to realize detection and tracking of an object; and extracting features of the object, performing impurity identification according to the features and judging whether impurities exist or not. According to the transparent liquid impurity detection system disclosed by the invention, the acquisition speed and treatment speed of the image sequences can be increased to the greatest extent; and an impurity identification algorithm which is adopted can effectively increase the identification speed and further achieve the purposes of completely replacing artificial detection, increasing detection quality and speed, saving production cost and improving product quality and production benefits on the premise of performing good segmentation, tracking and identification on the visible object.
Description
Technical field
The present invention relates to a kind of detection system, be specifically related to a kind ofly can in the high-speed production lines of the industries such as medicine, wine, beverage, in testing transparent liquid, whether have transparency liquid Impurity Detection System and the detection method thereof of impurity.
Background technology
In transparency liquid, the automatic detection of impurity is one of very important test item during product quality detects, and it has broad application prospects at production fields such as medicine, food and drink and chemical industry.Domestic most dependence at present manually realizes the detection of visible foreign matters, it is large that Manual light procuratorial organ formula has labour intensity, efficiency is low, testing result subjectivity is strong, testing staff's fatiguability, loss undulatory property is large and be subject to the shortcomings such as testing staff's physiologic factor impact, can not meet the high-speed and high-precision requirement of automatic production line.Detection mode based on machine vision has adopted the principle of Manual light inspection, replaces people's eyes with video camera, with computing machine, replaces human brain to analyze the character of visible foreign matters.This mode is safe and reliable, to not injury of human body, and is applicable to online detection, and therefore the detection method based on machine vision will replace Manual light procuratorial organ formula gradually.
In transparency liquid impurity automatically detects, the problem that needs to solve is how effectively to improve detection speed and in the little accuracy rate that how to improve detection of prerequisite that guarantees high detection speed.Can be divided into following problem specifically: the quick obtaining of (1) image and processing problem.Liquid impurity identification needs the multidate information of impurity, therefore needs its image sequence of acquisition and processing, and this is proposing higher requirement to system aspect Image Acquisition and processing; (2) image background suppresses problem.For the detection of small foreign matter, the removal that picture noise disturbs is one of the problem that must consider.In order to highlight object, need to select targetedly corresponding background Restrainable algorithms according to the feature of impurity object and ground unrest, for subsequent detection identification lays the foundation.(3) the fast effective identification problem of impurity.Small foreign matter target detection need to adopt high-resolution vision system, and this can increase the data volume of processing undoubtedly, in impurities identification process, need to carry out identifying processing to multiframe low signal-to-noise ratio sequence image simultaneously, and this also can cause algorithm complicated, and operand increases.Therefore must find a kind of suitable algorithm realizes impurity and identifies fast.
Summary of the invention
The object of the invention is, for overcoming above-mentioned the deficiencies in the prior art, provides a kind of transparency liquid Impurity Detection System and detection method thereof.
For achieving the above object, the present invention adopts following technical proposals:
A transparency liquid Impurity Detection System, comprises guidance panel, for Site Detection information read, setting parameter and control; Host computer, is placed in Control Room, for remote monitoring; Also comprise in turn and to connect
Dsp processor, parallel organization, for improving image processing speed, and realizes information extraction by the algorithm of setting;
Several CCD cameras, order is arranged on production line, and is connected to dsp processor respectively, and matches with LED light source, for obtaining product image to be checked;
Optoelectronic switch, for detection of product to be checked position, and triggers LED light source and CCD camera to obtain product image to be checked; And be connected in parallel
Scrambler, follows the tracks of for the position of product to be checked;
Flush trimmer, for rejecting unacceptable product from production line;
Reject and confirm switch, for detection of unacceptable product, whether correctly rejected; And
PLC station, for the treatment of confirm the signal of switch from dsp processor, optoelectronic switch, scrambler and rejecting, and forms decision-making according to the differentiation algorithm of setting, and action command is conveyed to flush trimmer, and while and host computer carry out data communication.
A detection method for transparency liquid Impurity Detection System, step is as follows:
Step1: utilize many CCD collected by camera product sequence images to be checked storage;
Step2: carry out image background inhibition by image pre-service;
Step3: the image after background is suppressed is processed, realize target detects and follows the tracks of;
Step4: target is carried out to feature extraction, carry out impurities identification according to its feature, determine whether impurity.
In described Step1, described product sequence image to be checked is by order, to be arranged on the transparent body image of the same product to be checked that many CCD cameras on production line extract successively.
In described Step2, the additive model of model Noise image, is subordinate to decision rule according to setting, is carried out cluster, then carries out image background inhibition, mainly comprises the following steps:
2a) assignment;
2b) upgrade gain vector;
2c) upgrade filter weights;
2d) upgrade inverse matrix.
In described Step3, the concrete steps of described object detecting and tracking are as follows:
3a) in the first two field picture of image sequence, according to region area size each contour area in tag identifier image;
3b) in sequence image subsequently, adopt the method for preemptive type to carry out contour area association, realize multiobject Data association and tracking.
In described Step4, the concrete steps of described impurities identification are as follows:
4a) according to specified rule, the multiple goal of preliminary election is sorted, realize the pre-detection of impurity, thus the larger impurity of identification;
4b) for the impurity larger with bubble similarity, with gray scale, area, speed and acceleration, as characteristic of division, adopt improved fuzzy least squares support vector machine identification.
The invention has the beneficial effects as follows, the present invention carries out fuzzy clustering to the liquid image obtaining, cluster result and the improved least squares filtering background that contains assorted image that combines is suppressed, by hierarchical identification mode, improve recognition speed and effect, adopt the object detecting and tracking method based on spacetime correlation, realize after the identification of excessive impurity and normal molecule, by utilizing fuzzy least squares support vector machine to realize effective differentiation of impurity and bubble.The present invention effectively reduces production costs, and improves impurities identification rate and recognition speed, thereby improves the quality of products and speed of production, avoids the product that contains impurity to come into the market, and affects corporate image.
The present invention can improve obtaining and processing speed of image sequence to greatest extent; The impurities identification algorithm adopting is carrying out visible target under the good prerequisite of cutting apart, following the tracks of and identifying, can effectively improve its recognition speed, thereby reach, replace manual detection completely, improve to detect quality and speed, save production cost, improve the quality of products and the object of productivity effect.
Accompanying drawing explanation
Fig. 1 is transparency liquid Impurity Detection System structural representation of the present invention;
Fig. 2 is detection method process flow diagram of the present invention;
Wherein, 1. guidance panel, 2. host computer, 3.DSP processor, 4.CCD camera, 5. optoelectronic switch, 6.LED light source, 7.PLC station, 8. scrambler, 9. flush trimmer, 10. rejects and confirms switch.
Embodiment
Below in conjunction with drawings and Examples, the present invention will be further elaborated, should be noted that following explanation is only in order to explain the present invention, does not limit its content.
Fig. 1 is the structural representation of transparency liquid Impurity Detection System of the present invention, and this detection system comprises guidance panel 1, for Site Detection information read, setting parameter and control; Host computer 2, is placed in Control Room, for remote monitoring; Also comprise the dsp processor 3 connecting in turn, parallel organization, for improving image processing speed, and realizes information extraction by the algorithm of setting; Several CCD cameras 4, order is arranged on production line, and is connected to respectively dsp processor 3, and matches with LED light source 6, for obtaining product image to be checked; Optoelectronic switch 5, for detection of product to be checked position, and triggers LED light source 6 and CCD camera 4 to obtain product image to be checked; And the scrambler 8 being connected in parallel, for the position of product to be checked, follow the tracks of; Flush trimmer 9, for rejecting unacceptable product from production line; Reject and confirm switch 10, for detection of unacceptable product, whether correctly rejected; And PLC station 7, for the treatment of confirm the signal of switch 10 from dsp processor 3, optoelectronic switch 5, scrambler 8 and rejecting, and form decision-making according to the differentiation algorithm of setting, and action command is conveyed to flush trimmer 9, while and host computer 2 carry out data communication.
Fig. 2 is the process flow diagram of detection method of the present invention, and step is as follows:
Step1: utilize many CCD collected by camera product sequence images to be checked storage: with the transparent body image that is sequentially arranged on many CCD cameras on production line and extracts successively same product to be checked.
Image acquisition process is as follows:
When optoelectronic switch detects product arrival image capture position to be checked, trigger light source and camera and realize the image acquisition to tested product, the image that gathers reads in after dsp processor the first histogram of computed image, according to image histogram, picture quality is carried out to anticipation, if picture quality meets the requirements, memory image carry out subsequent treatment, realizes information extraction by the algorithm of setting; If picture quality is undesirable, give up image, current location accumulated value adds 1, if the accumulated value of a certain position reaches setting value in a certain setting-up time section, illustrates that the image capturing system of this position has problem, stopping alarm.
Step2: carry out image background inhibition by image pre-service: the additive model of model Noise image, according to setting, be subordinate to decision rule, carried out cluster, then carry out image background inhibition.The process that image background suppresses is as follows:
Known image Point Target model is:
Wherein, the ideal image that D (x, y) is target; X, y is pixel coordinate; G is the maximal value of target gray scale function; E is the truth of a matter of natural logarithm; (x
0, y
0) be the center position of target; (
σx, σ
y) be the divergence of target on spatial domain.The additive model of Noise image is:
f(x,y)=D(x,y)+B(x,y)+v(x,y) (2)
Wherein, B (x, y) representative image background clutter, generally presents universe gray scale non-stationary and the accurate smooth performance of local gray scale; V (x, y), for measuring noise, is assumed to be the white Gaussian process of zero-mean conventionally.
By iterative formula (3), (4), input picture f (x, y) is carried out to fuzzy C-means clustering:
Wherein, N is the number of sample in sample set; C is cluster number; v
jfor cluster centre, u
kfor sample, μ
j(u
k) be the membership function of k sample to j class.
For a width pixel, be the image that M is capable and M is listed as,, its gray matrix f (x, y) is sequentially arranged in to a M by dictionary writing
2dimensional vector, is subordinate to decision rule according to setting, input picture f (x, y) can be carried out to cluster, as shown in Equation (5), and Γ wherein
jfor assigning to all samples of j class, C is classification number.
f(x,y)={Γ
j,j∈[1,C]} (5)
Then in conjunction with the least square recurrence filter, carry out background inhibition, the least square recurrence filter algorithm of take in single subdomain Z is example, supposes f
1(n) one-dimensional data for obtaining by back-shaped scanning, γ is forgetting factor, algorithm is:
2a) assignment
d(n)=f
1(n) (6)
w
T(n)=[w
-l(n),...,w
-1(n-1),w
1(n+1),...,w
l(n)] (8)
Wherein, d (n) is Expected Response output, f
1(n) one-dimensional data for obtaining by back-shaped scanning,
for input data, w
t(n) be weight matrix, n is iterations, the coordinate that l is filter weights.
2b) upgrade gain vector
Adopt following formula to upgrade:
Wherein, g (n) is adaptive gain vector
2c) upgrade filter weights
2d) upgrade inverse matrix
Wherein, P (n) is inverse correlation matrix.Choose original bulk w (0)=0, P (0)=ε
-1i, ε is very little real number, I is unit matrix.When n > 4l, w restrains substantially, then w value is now assigned to w (0), restarts recursion, can obtain background estimating value:
Will
be converted into
can obtain the image after background suppresses:
Wherein, τ (x
n, y
n) be the image after background inhibition, f (x
n, y
n) be original image.
Step3: the image after background is suppressed is processed, realize target detects and follows the tracks of, and process is as follows:
3a) in the first two field picture of image sequence, big or small with each contour area in tag identifier image according to region area:
Image after known Background suppression is I (t), with label L=(L
1..., L
m) m contour area in identification image I (t).In the first two field picture of image sequence, according to region area size, identify these regions;
3b) in sequence image subsequently, adopt the method for preemptive type to carry out contour area association, realize multiobject Data association and tracking:
Utilize following formula can realize tracked target judgement of the matching between frame before and after sequence image:
Wherein, m is shape S
1to S
2mapping, k
1and k
2for corresponding shape curvature.The central point that first of above formula is region, to the distance summation of flex point, is used d
verrepresent; Second is region area, uses s
arerepresent, α is weight coefficient.Therefore region contour feature can be expressed as S
k(d
ver, s
are), utilize the interframe that region contour feature can realize target thing to follow the tracks of.
Step4: target is carried out to feature extraction, carry out impurities identification according to its feature, determine whether impurity:
4a) according to specified rule, the multiple goal of preliminary election is sorted, realize the pre-detection of impurity, thus the larger impurity of identification; And algorithm is simpler, recognition speed is very fast, can effectively identify larger impurity.
In tracing process, according to target property, can to the multiple goal of preliminary election, sort according to specified rule, realize the pre-detection of impurity.Take target gray scale and area sorts as feature, and its ordering rule is:
Wherein, γ
gand γ
sbe respectively the gray scale g of k target
kwith area s
kweight coefficient,
with
average gray and average area for M target.γ
pbe change in location p between the picture frame of k target
kweight coefficient.Utilize above formula by the o of target
ivalue is arranged according to order from big to small.According to priori, available following formula is realized the pre-detection of impurity:
In formula, Th
1and Th
2for impurity judgment threshold, I=1 represents that i target is impurity, and I=0 represents that i target is not impurity; but normal particle or noise in liquid; I=2 represents to judge, this is mainly because the target in this threshold range is likely impurity, is likely also bubble.
4b) for the impurity larger with bubble similarity, with gray scale, area, speed and acceleration, as characteristic of division, adopt improved fuzzy least squares support vector machine identification.Due to the pre-identification in early stage, for larger impurity, can consider, so algorithm only consider the size impurity similar with bubble how with effective differentiation problem of bubble.This hierarchical identification mode can effectively improve accuracy of identification and speed.
The given training set with category label: (x
1, y
1), (x
2, y
2) ..., (x
m, y
m), wherein, training sample input point x
i∈ R
n, y
i∈ 1,1}, and i=1,2,, M.Unknown function regression expression is:
Build fuzzy least squares support vector machine, optimization problem is:
Its equation of constraint is:
Wherein, M is sample number,
be two real number fields
nonlinear Mapping, weight vector
error variance ξ
i∈ R
n, b is deviation variables, C is the compromise of maximum class interval and minimum classification error, p
ifor membership function.Structure Lagrange function:
In formula, a
iit is Lagrange multiplier.According to KTT optimal conditions, can obtain optimum solution:
Wherein, y=(y
1, y
2..., y
m)
t; P=((p
1c)
-1, (p
2c)
-1..., (p
mc)
-1)
t; 1=(1,1 ..., 1)
t; Ω is square formation, and the element of the capable l row of k is
k(x
k, x
l) be kernel function, it meets Mercer condition.Try to achieve optimum solution a
*, b
*after can obtain the decision function of sorter:
Design for degree of membership, is defined as follows membership function:
Wherein,
X
i, x
jfor sample, ρ (x
i) be sample rate, M (X) is the sample size of sample set X, d (x
i, x
j) be measurement functions, T is for controlling the threshold value of sample peripheral extent; The density of the similar sample within threshold value is defined as positive density ρ
+(x
i), and the density of foreign peoples's sample is defined as to negative density ρ
-(x
i), the average density of positive sample is ρ (x
i).
When adopting above-mentioned fuzzy least squares support vector machine to realize the classification of impurity and bubble, select gray scale g
i, area s
i, speed v
iwith acceleration a
ias characteristic of division, because movement velocity and the track of impurity and bubble exists very big-difference, so apply the correct differentiation that above-mentioned feature can realize impurity and bubble.
Although above-mentioned, by reference to the accompanying drawings the specific embodiment of the present invention is described; but be not limiting the scope of the invention; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various modifications that creative work can make or distortion still in protection scope of the present invention.
Claims (3)
1. a detection method for transparency liquid Impurity Detection System, is characterized in that, step is as follows:
Step1: utilize many CCD collected by camera product sequence images to be checked storage;
Step2: carry out image background inhibition by image pre-service;
Step3: the image after background is suppressed is processed, realize target detects and follows the tracks of;
Step4: target is carried out to feature extraction, carry out impurities identification according to its feature, determine whether impurity;
In described Step1, described product sequence image to be checked is by order, to be arranged on the transparent body image of the same product to be checked that many CCD cameras on production line extract successively;
In described Step2, the additive model of model Noise image, is subordinate to decision rule according to setting, is carried out cluster, then carries out image background inhibition, mainly comprises the following steps:
2a) assignment;
2b) upgrade gain vector;
2c) upgrade filter weights;
2d) upgrade inverse matrix;
Described transparency liquid Impurity Detection System, comprises guidance panel, for Site Detection information read, setting parameter and control; Host computer, is placed in Control Room, for remote monitoring; It is characterized in that, also comprise the dsp processor connecting in turn, parallel organization, for improving image processing speed, and realizes information extraction by the algorithm of setting;
Several CCD cameras, order is arranged on production line, and is connected to dsp processor respectively, and matches with LED light source, for obtaining product image to be checked;
Optoelectronic switch, for detection of product to be checked position, and triggers LED light source and CCD camera to obtain product image to be checked; And the scrambler being connected in parallel, for the position of product to be checked, follow the tracks of;
Flush trimmer, for rejecting unacceptable product from production line;
Reject and confirm switch, for detection of unacceptable product, whether correctly rejected;
And PLC station, for the treatment of confirm the signal of switch from dsp processor, optoelectronic switch, scrambler and rejecting, and form decision-making according to the differentiation algorithm of setting, action command is conveyed to flush trimmer, while and host computer carry out data communication.
2. detection method as claimed in claim 1, is characterized in that, in described Step3, the concrete steps of described object detecting and tracking are as follows:
3a) in the first two field picture of image sequence, according to region area size each contour area in tag identifier image;
3b) in sequence image subsequently, adopt the method for preemptive type to carry out contour area association, realize multiobject Data association and tracking.
3. detection method as claimed in claim 1, is characterized in that, in described Step4, the concrete steps of described impurities identification are as follows:
4a) according to specified rule, the multiple goal of preliminary election is sorted, realize the pre-detection of impurity, thus the larger impurity of identification;
4b) for the impurity larger with bubble similarity, with gray scale, area, speed and acceleration, as characteristic of division, adopt improved fuzzy least squares support vector machine identification.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210560198.XA CN102998316B (en) | 2012-12-20 | 2012-12-20 | Transparent liquid impurity detection system and detection method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210560198.XA CN102998316B (en) | 2012-12-20 | 2012-12-20 | Transparent liquid impurity detection system and detection method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102998316A CN102998316A (en) | 2013-03-27 |
CN102998316B true CN102998316B (en) | 2014-11-26 |
Family
ID=47927159
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210560198.XA Expired - Fee Related CN102998316B (en) | 2012-12-20 | 2012-12-20 | Transparent liquid impurity detection system and detection method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102998316B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103913469A (en) * | 2014-04-21 | 2014-07-09 | 南京农业大学 | System for detecting and processing contaminants on surfaces of poultry carcasses online |
CN104535006B (en) * | 2015-01-21 | 2017-10-27 | 杭州电子科技大学 | A kind of bottle cap gap width evaluation method of utilization transmission-type illumination imaging systems |
CN105115986A (en) * | 2015-08-12 | 2015-12-02 | 中南大学 | System and method for detecting visible foreign substances in traditional Chinese medicine oral liquid bottles |
CN105930852B (en) * | 2016-05-09 | 2019-06-21 | 山东省科学院海洋仪器仪表研究所 | A kind of bubble image-recognizing method |
CN106373140B (en) * | 2016-08-31 | 2020-03-27 | 杭州沃朴物联科技有限公司 | Transparent and semitransparent liquid impurity detection method based on monocular vision |
CN107525772A (en) * | 2017-06-28 | 2017-12-29 | 温州大学 | The method and system of one main laminaria quality testing |
CN107316297B (en) * | 2017-07-07 | 2020-09-01 | 山东商务职业学院 | Visual detection method for edible oil impurities |
CN109900716A (en) * | 2019-04-12 | 2019-06-18 | 中民筑友科技投资有限公司 | A kind of system and method detecting PC component presentation quality |
CN111062257A (en) * | 2019-11-21 | 2020-04-24 | 四川极智朗润科技有限公司 | Micro target identification method based on morphological and kinematic characteristics |
CN112525920B (en) * | 2020-11-23 | 2023-03-10 | 山东科技大学 | Liquid perspective imaging interference filtering system |
CN112666047B (en) * | 2021-01-14 | 2022-04-29 | 新疆大学 | Liquid viscosity detection method |
CN115496897B (en) * | 2022-10-10 | 2023-10-31 | 中国核动力研究设计院 | Deionized water solid impurity detection method and system based on digital image processing |
Family Cites Families (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE2260602A1 (en) * | 1972-12-12 | 1974-06-20 | Leitz Ernst Gmbh | DEVICE FOR DETECTION OF CONTAMINATION IN CLEAR LIQUIDS |
DE4408699A1 (en) * | 1994-03-15 | 1995-09-21 | Mbm Kontroll Systeme Gmbh | Test method for the detection of impurities in liquids and device for carrying out the method |
CN2357327Y (en) * | 1998-06-10 | 2000-01-05 | 株洲金兰机电公司 | Device for automatically identifying residue in liquid bottle |
JP2002221498A (en) * | 2000-11-24 | 2002-08-09 | Sukiyan Technol:Kk | System for detecting foreign matter |
CN101241086A (en) * | 2008-03-03 | 2008-08-13 | 中国科学院光电技术研究所 | Detector for measuring glass material air blister and impurity |
CN101354359B (en) * | 2008-09-04 | 2010-11-10 | 湖南大学 | Method for detecting, tracking and recognizing movement visible exogenous impurity in medicine liquid |
CN101413904B (en) * | 2008-11-25 | 2011-09-14 | 湖南大学 | Impurity real time machine vision detection method and apparatus in special-shaped bottle-packaging solution |
CN101819162A (en) * | 2010-05-13 | 2010-09-01 | 山东大学 | Empty bottle wall defect detection method and device |
CN101819164B (en) * | 2010-05-14 | 2011-06-01 | 山东大学 | Device and method for detecting impurities after filling of beverage |
CN102175693B (en) * | 2011-03-08 | 2014-04-09 | 中南大学 | Machine vision detection method of visual foreign matters in medical medicament |
-
2012
- 2012-12-20 CN CN201210560198.XA patent/CN102998316B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN102998316A (en) | 2013-03-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102998316B (en) | Transparent liquid impurity detection system and detection method thereof | |
CN111292305B (en) | Improved YOLO-V3 metal processing surface defect detection method | |
CN105388162B (en) | Raw material silicon chip surface scratch detection method based on machine vision | |
CN107389701A (en) | A kind of PCB visual defects automatic checkout system and method based on image | |
Liming et al. | Automated strawberry grading system based on image processing | |
CN108985169B (en) | Shop cross-door operation detection method based on deep learning target detection and dynamic background modeling | |
CN109859181A (en) | A kind of PCB welding point defect detection method | |
CN104112269B (en) | A kind of solar battery laser groove parameter detection method and system based on machine vision | |
He et al. | Deep regression neural network for industrial surface defect detection | |
CN104637058B (en) | A kind of volume of the flow of passengers identify statistical methods based on image information | |
CN106875381A (en) | A kind of phone housing defect inspection method based on deep learning | |
CN105260749B (en) | Real-time target detection method based on direction gradient binary pattern and soft cascade SVM | |
CN103927534A (en) | Sprayed character online visual detection method based on convolutional neural network | |
CN107486415A (en) | Thin bamboo strip defect on-line detecting system and detection method based on machine vision | |
CN105160654A (en) | Towel label defect detecting method based on feature point extraction | |
CN105044122A (en) | Copper part surface defect visual inspection system and inspection method based on semi-supervised learning model | |
CN104021375A (en) | Model identification method based on machine learning | |
CN107230203A (en) | Casting defect recognition methods based on human eye vision attention mechanism | |
CN114972356B (en) | Plastic product surface defect detection and identification method and system | |
CN105403147A (en) | Embedded bottle pre-form detection system and detection method | |
CN111080691A (en) | Infrared hot spot detection method and device for photovoltaic module | |
CN207238542U (en) | A kind of thin bamboo strip defect on-line detecting system based on machine vision | |
CN110443791A (en) | A kind of workpiece inspection method and its detection device based on deep learning network | |
CN111161237A (en) | Fruit and vegetable surface quality detection method, storage medium and sorting device thereof | |
CN103177266A (en) | Intelligent stock pest identification system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20141126 Termination date: 20161220 |