CN105160654A - Towel label defect detecting method based on feature point extraction - Google Patents
Towel label defect detecting method based on feature point extraction Download PDFInfo
- Publication number
- CN105160654A CN105160654A CN201510405923.XA CN201510405923A CN105160654A CN 105160654 A CN105160654 A CN 105160654A CN 201510405923 A CN201510405923 A CN 201510405923A CN 105160654 A CN105160654 A CN 105160654A
- Authority
- CN
- China
- Prior art keywords
- image
- value
- registration
- detected
- intermediate value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- 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
- G06T7/001—Industrial image inspection using an image reference 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
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30124—Fabrics; Textile; Paper
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a towel label defect detecting method based on ORB feature point extraction. The method comprises the following steps of acquiring images of towel labels to be detected, performing a series of image preprocessing operations on the acquired images to be detected; extracting feature points of template images and images to be detected through an ORB feature extraction algorithm; searching for an optimal matching point according to an LSH (Local Sensitive Hash) search algorithm; adopting RANSAC to reject mistaken matching points to acquire parameters required for affine transformation; establishing a corresponding homography matrix to acquire a registered image; performing image difference operations on the registered image and the template images; respectively counting the proportions of pixel value 0 and pixel value 1 in binary images according to the difference result, comparing the proportions with preset experience threshold values to judge whether defects are present in the images; and finally judging whether the towel label is a qualified one or not.
Description
Technical field
The invention belongs to industrial automation, relate to the towel label defect inspection method that a kind of distinguished point based extracts.
Background technology
In recent years, enterprise is guided to initiate since " machine substitution " strategy along with government takes the lead, the company in increasing design automation field and enterprise, for institute's produced problem of the quality (whether existing defects) of produced product or industrial work piece, develop automatic checkout system on streamline that a set of cover combines based on computer vision and CASE(Computer Aided Software Engineering), greatly reduce the manpower, material resources and financial resources of enterprise, also greatly for enterprise has saved cost.In some enterprise-like corporations, traditional workpiece or the defects detection of product mainly rely on human eye to recognize and identify, efficiency and low, therefore enterprise-like corporation urgently relies on innovative technology to improve the competitive power of enterprise, so the development of automatic production line defect detecting system is very urgent, the innovation for enterprise drives and provides the new motive force of development.
Summary of the invention
The present invention is directed to the deficiency not possessing real-time when traditional SIFT (ScaleInvariantFeatureTransform scale invariability Feature Conversion) algorithm extracts feature, ORB (OrientedFASTandRotatedBRIEF is based on Fast Corner Detection algorithm and rotate BRIEF Feature Descriptor) is adopted to combine based on BRIEF (BinaryRobustIndependentElementaryFeatures two-value robust isolated footing feature interpretation algorithm) feature extraction, registration is carried out to testing image and the template image that makes in advance, finally according to absolute difference algorithm, absolute difference is carried out to Prototype drawing and figure to be checked, compare based on difference result and the threshold value property entered that arranges based on experience value, whether qualifiedly label is passed judgment on by interpretational criteria, whether reach certain standard, thus realize effective detection of defect, this invention real-time Detection results is good.
The technical scheme that technical solution problem of the present invention is taked is:
Step (1). production standard tag template image, industrial camera is adopted to carry out collection image procossing to the label on production line according to tradesman in most cases, from all images collected, select the standard form figure image closely of feeling and will reach.
Step (2). the real-time label image to be detected (online acquisition) of industrial camera collection, technician is controlled by the true-time operation of industrial computer client.
Step (3). pretreatment operation is carried out to the label image to be detected that real-time production line collects.
Step (4). adopt ORB feature extraction algorithm to extract the unique point of tag template image and image to be detected.
Step (5). the unique point of the tag standards template image extracted and image to be detected is carried out Feature Points Matching, adopts LSH (LocalitySensitiveHashing local sensitivity Hash) searching algorithm to search out the feature point pairs of optimized coupling.
Step (6). to utilization RANSAC (RandomSampleConsensus stochastic sampling consistency algorithm) algorithm, error matching points is rejected to the characteristic matching point filtering out optimization.
Step (7). according to RANSAC algorithm reject error matching points to after matching double points set up the homography matrix (HomographyMatrix) of image subject to registration and template image, homography matrix is formed by 8 parameters, the mapping point of image subject to registration to template image is obtained according to homography matrix, thus realize the result of registration, obtain the image after registration.
Step (8). the image after image registration to be detected and standard form image are carried out absolute calculus of differences, draw difference result, this difference result and the threshold value preset are compared, if difference result exceeds this thresholding, then assert that this label product that will detect is unacceptable product, is faulty goods.
Beneficial effect of the present invention:
(1) enhance the real-time of towel label defects detection, effectively can realize on-line checkingi, improve and detect degree of accuracy and detection efficiency.
(2) stability required for system has also been possessed simultaneously, experimentally adopted the present invention defects detection rate can be brought up to more than 96%.
Accompanying drawing explanation
Fig. 1 towel label defects detection process flow diagram.
Concrete embodiment:
Below in conjunction with Figure of description, the present invention is described further.
According to Figure of description (1), implementation step is described in detail:
Step (1). production standard tag template image, industrial camera is adopted to carry out collection image procossing to the label on production line according to tradesman in most cases, from all images collected, select the standard form figure image closely of feeling and will reach.
Realizing a most key step of towel label defects detection is exactly the process making template image, and the quality of a width template image is directly connected to the result of defects detection, embodies very important meaning.In industrial automation, a kind of image of making template of the Corpus--based Method method of average becomes a kind of focus now, but the complicated and efficiency of its manufacturing process is low allows again it have a greatly reduced quality.Suppose to collect N width realtime graphic on a production line, use g
m(x, y) represents, m=1,2,3 ... N, follows the image obtained each time and is newly expressed as f to template image
m(x, y), m=1,2,3 ... N, so adopts general statistical average rule for shown in following formula,
f
1(x,y)＝g
1(x,y),
But on actual production line, the making of standard form image obtains by the experience of technician often, according to randomness, first by the real-time towel label image that industrial camera collection is a large amount of, select it by technician and think that the standard form image that can serve as, the principle of foundation are exactly consistent in the large absolutely degree in the image selected and the aspect designing standard and the requirement reached.
Step (2). the real-time label image to be detected (online acquisition) of industrial camera collection, technician is controlled by the true-time operation of industrial computer client.
Detect on streamline at towel label, industrial camera CCD (Charge-coupledDevice charge coupled cell) video camera or CMOS (ComplementaryMetal-Oxide-Semiconductor complementary metal oxide semiconductor (CMOS)) are firmly fixed on support, to reduce the fine jitter that video camera may produce in the course of the work, thus reduce the noise produced due to mechanical vibration, thus have influence on final testing result.The real-time image signal collected is delivered in the client at industrial computer interface, carries out real-time policer operation.
Step (3). pretreatment operation is carried out to the label image to be detected that real-time production line collects.
Although preliminary work is done more abundant, but inevitably the inside more or less must be doped into some noises, although we are negligible, but because precision of the presently claimed invention is very high, so the impact in order to reduce noise that may be potential further, treat the method filtering noise that detected image uses medium filtering, treat detected image and carry out Gaussian smoothing, to reduce the interference of noise.In view of traditional median filtering algorithm is more consuming time, efficiency comparison is low, the present invention is directed to this problem and proposes fast parallel median filtering algorithm, uses the present invention to improve detection efficiency, and speed aspect also can promote, there is certain actual production meaning.For 3 × 3 medium filterings, suppose that in 3 × 3 windows, pixel distribution is P (i, j), i=0,1,2, j=0,1,2; Concrete steps are as follows:
The first step, calculates maximal value, intermediate value and minimum value respectively by each the row pixel value in window, because there are 3 row, so just obtains 3 groups of data:
Maximal value group: Max0=max [P (0,0), P (1,0), P (2,0)],
Max1＝max[P(0,1),P(1,1),P(2,1)]，
Max2=max [P (0,2), P (1,2), P (2,2)] (wherein max function is got in max representative)
Intermediate value group: Med0=med [P (0,0), P (1,0), P (2,0)],
Med1＝med[P(0,1),P(1,1),P(2,1)]，
Med2=med [P (0,2), P (1,2), P (2,2)] (wherein median is got in med representative)
Minimum value group: Min0=min [P (0,0), P (1,0), P (2,0)],
Min1＝min[P(0,1),P(1,1),P(2,1)]，
Min2=min [P (0,2), P (1,2), P (2,2)] (wherein minimum value function is got in min representative)
Second step: can be drawn by these three groups of data, the minimum value in the maxima and minima group in maximal value group must be maximal value and the minimum value of 9 elements in window, can not be intermediate value, therefore leaves 7 elements and compares; And the maximal value in intermediate value group is greater than 5 elements to I haven't seen you for ages, the minimum value in intermediate value group is at least less than 5 elements, can not be intermediate value, and remaining 5 elements compare thus; Have again the intermediate value in maximal value group to be also at least be greater than 5 elements, and the intermediate value in minimum value group is at least less than 5 elements, also can not be intermediate value, last only remaining 3 elements compare, and are:
Minimum M axmin=min [Max0, Max1, Max2] in maximal value group,
Intermediate value Medmed=med [Med0, Med1, Med2] in intermediate value group,
Maximal value Minmax=max [Min0, Min1, Min2] in minimum value group;
3rd step: finally find out intermediate value and be intermediate value in 9 elements, i.e. med [Maxmin, Medmed, Minmax] in 3 elements that finally will compare.
The suitable method that mathematical morphology can be adopted to combine is used for eliminating those isolated noises, adopts the Erodent Algorithm (structural element) used required in the process of mathematical morphology then manually will determine according to actual state.If the words of the same size of character image contained in towel label, then entire image adopts unified structural element; If the size differences of the character image contained in towel label is larger, then segmentation image can be adopted to adopt different structure element to carry out the method for morphological transformation respectively.
Step (4). adopt ORB feature extraction algorithm to extract the unique point of tag template image and image to be detected.
What local invariant feature detective operators ORB detected that unique point adopts is FAST corner detection operator, and for the skirt response that occurs and do not produce multiple dimensioned shortcoming when detecting unique point of FAST angle point operator, ORB has made improvement to this.ORB algorithm has introduced the Harris angle point that FAST algorithm extracts local feature region, according to the sequence of angle point value size, and N number of unique point before larger by the number reservation angle point value of unique point, thus remove unique point skirt response.ORB algorithm is the same with SIFT algorithm, needs to set up multilayer pyramid, extract minutiae in every layer of pyramid diagram picture, thus obtains dimensional information, for setting up descriptor.ORB, by calculating FAST unique point neighborhood center intensity determination direction parameter, calculates the neighborhood square of image-region, show that circle shaped neighborhood region barycenter and unique point and barycenter angle are defined as the direction of FAST unique point by neighborhood square.After obtaining unique point, need to be described unique point, ORB adopts BRIEF descriptor algorithm to be described unique point, and BRIEF algorithm is expressed by pixel grey scale relatively a small amount of in unique point neighborhood to be described, contrast.But BRIEF is more responsive for noise ratio, so select 5 × 5 pixel windows to each scale-of-two descriptor in 31 × 31 neighborhood of pixels in ORB, noise decrease disturbs.According to the unique point direction parameter that ORB draws, in order to allow the coupling that can realize different characteristic point descriptor, needing Feature Descriptor to give direction, and being consistent with unique point direction.The ORB descriptor pair that finds correlativity lower through greedy algorithm, gets 256 as Feature Descriptor afterwards.
Step (5). the unique point of the tag standards template image extracted and image to be detected is carried out Feature Points Matching, adopts LSH searching algorithm to search out the feature point pairs of optimized coupling.
Adopt local sensitivity hash algorithm to search to the descriptor obtained, use Hamming distance to filter out most suitable matching result pair.Hamming distance is for the calculating distance function based on scale-of-two descriptor, employing be the principle of XOR.Suppose the scale-of-two descriptor operator generating template image and image to be detected, this Hamming distance is therebetween calculated as follows shown in the formula of face:
The binary features descriptor of template image is
Templ(0)templ(1)…templ(i)…templ(255)；
(illustrating: the value that wherein templ (i) represents is not 0 is just 1, total total 256bit value)
The binary features descriptor of image to be detected is
candid(0)candid(1)…candid(j)…candid(255)；
(illustrating: the value that wherein candid (j) represents is not 0 is exactly 1, Zong total 256bit value) Hamming distance HamDistance computing formula so between two descriptors is as follows:
HamDistance=Count ((templ (i) XORcandid (j))==1), i=0,1,2 ... 255, j=0,1,2 ... 255, wherein XOR represents XOR symbol, and Count function representation statistics XOR result is the number of 1, is exactly Hamming distance by the number coming out 1.The Hamming distance come out and the distance threshold preset are compared, if the Hamming distance come out is less than distance threshold, then retain these matching double points, otherwise weed out these matching double points.
Step (6). to utilization RANSAC algorithm, error matching points is rejected to the characteristic matching point filtering out optimization.
By RANSAC algorithm, to the feature after first coupling to screening, weeding out erroneous matching, according to coupling to calculating the optimum anglec of rotation, translational movement, pantograph ratio etc., calculating the homography matrix needing to carry out mapping transformation thus.
Step (7). according to RANSAC algorithm reject error matching points to after matching double points set up the homography matrix of image subject to registration and template image, form homography matrix by 8 parameters, as follows
General h
22=1, and h
20and h
21also negligible, usually calculate by 0, six so left parameters are exactly the matrix needing to carry out the affined transformation mapped in fact.Element h in homography matrix
02with element h
12representative needs the size of carrying out translation; And element h
00and h
11then represent scaling size; Element h
01and h
10represent the parameter of shear transformation.But generally for and can simplify calculating, the point that we get is in specific position, wants to solve this 8 parameters, must set up 8 equations, and such as, for image obj_scene to be checked, four points that we extract are to being:
Obj_scene [0]=(0,0), obj_scene [1]=(cols, 0), obj_scene [2]=(cols, rows), obj_scene [3]=(0, rows), wherein cols and rows represents height value and the width value of got image respectively.By 4 to putting setting up eight parametric equations, solve 8 parameters in homography matrix.
Obtain the mapping point of image subject to registration to template image according to homography matrix, thus realize the result of registration, obtain the image after registration.
And affined transformation be according to template image set up from image subject to registration to template image a series of mapping transformation relations, set up one funtcional relationship one to one, making by image conversion subject to registration in the space coordinates identical with template image, is that the committed step of last defects detection---image difference computing hides the foreshadowing.The success or not of affined transformation is larger for image registration accuracy impact.In six parameters of affined transformation, on actual flow waterline, translation and rotation parameter can neglect substantially because to gather the industrial camera of image be substantially static on a moving belt, there is not skew and the rotation of image, it is envisaged that Image scaling coefficients.
Step (8). the image after image registration to be detected and standard form image are carried out absolute calculus of differences, draw difference result, this difference result and the threshold value preset are compared, if difference result exceeds this thresholding, then assert that this label product that will detect is unacceptable product, is faulty goods.
Maximum variance between clusters Otsu is used to carry out binary conversion treatment to image before carrying out calculus of differences to image.Two kinds of thresholding methods are divided into again: based on the Threshold segmentation of the overall situation and the Threshold segmentation based on local inside Otsu maximum variance between clusters.Generally, the effect that the Threshold segmentation based on the overall situation reaches is not desirable especially, and thus we adopt the dividing method based on local threshold in most cases.In order to prevent template image and detected image from there is small pixel at character edge carrying out xor operation, therefore by template image corrosion primary, thus eliminate those negligible isolated pixel points not affecting testing result.In template image, after corrosion, pixel value is the point of 0, carries out and operation with detected image, after template image corrosion pixel value be 1 point and detected image carry out xor operation computing, avoid the impact because edge little deviation causes pixel to remain.Finally be worth according to statistical pixel 0 value in the result figure of difference image and pixel 1 ratio accounting for the total pixel of bianry image respectively, empirically value presets a threshold value, according to decision rule, if the pixel drawn 0 proportion is greater than set threshold value, then judge that this image is as defect image, thus assert that this towel label is substandard product.
Claims (1)
1. the towel label defect inspection method of distinguished point based extraction, is characterized in that the method comprises the following steps:
Step (1). production standard tag template image, specifically: adopt industrial camera to carry out collection image procossing to the label on production line according to tradesman, from all images collected, the immediate image of standard form figure selected He will reach;
Step (2). the real-time label image to be detected of industrial camera collection;
Step (3). pretreatment operation is carried out to the label image to be detected that real-time production line collects;
Step (4). adopt ORB feature extraction algorithm to extract the unique point of standard label template image and image to be detected;
Step (5). the unique point of the standard label template image extracted and image to be detected is carried out Feature Points Matching, adopts LSH searching algorithm to search out the feature point pairs of optimized coupling;
Step (6). to utilization RANSAC (RandomSampleConsensus stochastic sampling consistency algorithm) algorithm, error matching points is rejected to the characteristic matching point filtering out optimization;
Step (7). according to RANSAC algorithm reject error matching points to after matching double points set up the homography matrix of image subject to registration and template image, homography matrix is formed by eight parameters, the mapping point of image subject to registration to template image is obtained according to homography matrix, thus realize the result of registration, obtain the image after registration;
Step (8). the image after image registration to be detected and standard form image are carried out absolute calculus of differences, draw difference result, this difference result and the threshold value preset are compared, if difference result exceeds this thresholding, then assert that this label product that will detect is unacceptable product, is faulty goods;
In described step (3), pretreatment operation comprises fast parallel median filtering algorithm, for 3 × 3 medium filterings, supposes that in 3 × 3 windows, pixel distribution is P (i, j), i=0,1,2, j=0,1,2;
The first step, calculates maximal value, intermediate value and minimum value respectively by each the row pixel value in window, because there are 3 row, so just obtains 3 groups of data:
Maximal value group: Max0=max [P (0,0), P (1,0), P (2,0)],
Max1＝max[P(0,1),P(1,1),P(2,1)]，
Max2＝max[P(0,2),P(1,2),P(2,2)]
Intermediate value group: Med0=med [P (0,0), P (1,0), P (2,0)],
Med1＝med[P(0,1),P(1,1),P(2,1)]，
Med2＝med[P(0,2),P(1,2),P(2,2)]
Minimum value group: Min0=min [P (0,0), P (1,0), P (2,0)],
Min1＝min[P(0,1),P(1,1),P(2,1)]，
Min2＝min[P(0,2),P(1,2),P(2,2)]
Wherein max function is got in max representative, and median is got in med representative, and minimum value function is got in min representative;
Second step: can be drawn by these three groups of data, the minimum value in the maxima and minima group in maximal value group must be maximal value and the minimum value of 9 elements in window, can not be intermediate value, therefore leaves 7 elements and compares; And the maximal value in intermediate value group is greater than 5 elements to I haven't seen you for ages, the minimum value in intermediate value group is at least less than 5 elements, can not be intermediate value, and remaining 5 elements compare thus; Have again the intermediate value in maximal value group to be also at least be greater than 5 elements, and the intermediate value in minimum value group is at least less than 5 elements, also can not be intermediate value, last only remaining 3 elements compare, and are:
Minimum M axmin=min [Max0, Max1, Max2] in maximal value group,
Intermediate value Medmed=med [Med0, Med1, Med2] in intermediate value group,
Maximal value Minmax=max [Min0, Min1, Min2] in minimum value group;
3rd step: finally find out intermediate value and be intermediate value in 9 elements, i.e. med [Maxmin, Medmed, Minmax] in 3 elements that finally will compare;
Described step (5) is specifically: use Hamming distance to filter out optimized matching result pair; Hamming distance is for the calculating distance function based on scale-of-two descriptor, employing be the principle of XOR;
The binary features descriptor of template image is
Templ(0)templ(1)…templ(i)…templ(255)；
The binary features descriptor of image to be detected is
candid(0)candid(1)…candid(j)…candid(255)；
Hamming distance HamDistance computing formula so between two binary features descriptors is as follows:
HamDistance=Count ((templ (i) XORcandid (j))==1), i=0,1,2 ... 255, j=0,1,2 ... 255, wherein XOR represents XOR symbol, and Count function representation statistics XOR result is the number of 1, is exactly Hamming distance by the number coming out 1; The Hamming distance come out and the distance threshold preset are compared, if the Hamming distance come out is less than distance threshold, then retain these matching double points, otherwise weed out these matching double points;
Described step (7) is specifically: according to RANSAC algorithm reject error matching points to after matching double points set up the homography matrix of image subject to registration and template image, form homography matrix by eight parameters, as follows
Here h
22=1, and h
20and h
2calculate by 0, six so left parameters are exactly the matrix needing to carry out the affined transformation mapped in fact; Element h in homography matrix
02with element h
12representative needs the size of carrying out translation; And element h
00and h
11then represent scaling size; Element h
01and h
10represent the parameter of shear transformation; In order to can calculating be simplified, for image obj_scene to be checked, four points of extraction to for:
Obj_scene [0]=(0,0), obj_scene [1]=(cols, 0), obj_scene [2]=(cols, rows), obj_scene [3]=(0, rows), wherein cols and rows represents height value and the width value of got image respectively; By 4 to putting setting up eight parametric equations, solve 8 parameters in homography matrix; Obtain the mapping point of image subject to registration to template image according to homography matrix, thus realize the result of registration, obtain the image after registration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510405923.XA CN105160654A (en) | 2015-07-09 | 2015-07-09 | Towel label defect detecting method based on feature point extraction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510405923.XA CN105160654A (en) | 2015-07-09 | 2015-07-09 | Towel label defect detecting method based on feature point extraction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105160654A true CN105160654A (en) | 2015-12-16 |
Family
ID=54801496
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510405923.XA Pending CN105160654A (en) | 2015-07-09 | 2015-07-09 | Towel label defect detecting method based on feature point extraction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105160654A (en) |
Cited By (40)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106886707A (en) * | 2017-03-21 | 2017-06-23 | 广东网金控股股份有限公司 | A kind of method and device of image authentication |
| CN107067425A (en) * | 2017-04-25 | 2017-08-18 | 西安理工大学 | Image synchronization alignment schemes are printed in a kind of print quality automatic detection |
| CN107203990A (en) * | 2017-04-02 | 2017-09-26 | 南京汇川图像视觉技术有限公司 | A kind of labeling damage testing method based on template matches and image quality measure |
| CN107492091A (en) * | 2017-07-06 | 2017-12-19 | 东莞理工学院 | Label look detection method and terminal device based on machine vision |
| CN108548820A (en) * | 2018-03-28 | 2018-09-18 | 浙江理工大学 | Cosmetics paper labels defect inspection method |
| CN108763263A (en) * | 2018-04-03 | 2018-11-06 | 南昌奇眸科技有限公司 | A kind of trade-mark searching method |
| CN108804499A (en) * | 2018-04-03 | 2018-11-13 | 南昌奇眸科技有限公司 | A kind of trademark image retrieval method |
| CN108845998A (en) * | 2018-04-03 | 2018-11-20 | 南昌奇眸科技有限公司 | A kind of trademark image retrieval matching process |
| CN108874855A (en) * | 2018-04-03 | 2018-11-23 | 南昌奇眸科技有限公司 | A kind of trade-mark searching method based on similar area detection segmentation |
| CN109085178A (en) * | 2018-08-28 | 2018-12-25 | 武汉科技大学 | A kind of accurate on-line monitoring method of defect fingerprint and feedback strategy for increasing material manufacturing |
| CN109636797A (en) * | 2018-12-21 | 2019-04-16 | 国网福建省电力有限公司 | A kind of method of contactless flexible direct current converter valve module shape changing detection |
| CN109934809A (en) * | 2019-03-08 | 2019-06-25 | 深慧视(深圳)科技有限公司 | A kind of paper labels character defect inspection method |
| CN109934777A (en) * | 2019-01-09 | 2019-06-25 | 深圳市三宝创新智能有限公司 | Image local invariant feature extraction method, apparatus, computer equipment and storage medium |
| CN110070090A (en) * | 2019-04-25 | 2019-07-30 | 上海大学 | A kind of logistic label information detecting method and system based on handwriting identification |
| CN110245674A (en) * | 2018-11-23 | 2019-09-17 | 浙江大华技术股份有限公司 | Template matching method, device, equipment and computer storage medium |
| CN110245615A (en) * | 2019-06-17 | 2019-09-17 | 河海大学常州校区 | A kind of signature false distinguishing method, system and storage medium |
| CN110288547A (en) * | 2019-06-27 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image denoising model |
| CN110443295A (en) * | 2019-07-30 | 2019-11-12 | 上海理工大学 | Improved images match and error hiding reject algorithm |
| CN110472662A (en) * | 2019-07-10 | 2019-11-19 | 上海理工大学 | Image matching algorithm based on improved ORB algorithm |
| CN110503633A (en) * | 2019-07-29 | 2019-11-26 | 西安理工大学 | A kind of applique ceramic disk detection method of surface flaw based on image difference |
| CN110866911A (en) * | 2019-11-14 | 2020-03-06 | 华雁智能科技(集团)股份有限公司 | Dial defect detection method and device, image processing equipment and storage medium |
| CN110991446A (en) * | 2019-11-22 | 2020-04-10 | 上海欧冶物流股份有限公司 | Label identification method, device, equipment and computer readable storage medium |
| CN111145230A (en) * | 2019-12-26 | 2020-05-12 | 北京新唐思创教育科技有限公司 | Image registration quality detection method, device, equipment and storage medium |
| CN111932515A (en) * | 2020-08-10 | 2020-11-13 | 成都数之联科技有限公司 | Short circuit detection method and system for product residual defects and defect classification system |
| CN112258466A (en) * | 2020-10-19 | 2021-01-22 | 燕山大学 | Cold-rolled strip vibration line online detection system and method based on gray level processing |
| CN112288734A (en) * | 2020-11-06 | 2021-01-29 | 西安工程大学 | Printed fabric surface defect detection method based on image processing |
| CN112330597A (en) * | 2020-10-14 | 2021-02-05 | 浙江大华技术股份有限公司 | Image difference detection method and device and computer equipment |
| CN112381765A (en) * | 2020-10-23 | 2021-02-19 | 杭州翔毅科技有限公司 | Equipment detection method, device, equipment and storage medium based on artificial intelligence |
| CN112489042A (en) * | 2020-12-21 | 2021-03-12 | 大连工业大学 | Metal product printing defect and surface damage detection method based on super-resolution reconstruction |
| CN112528055A (en) * | 2020-11-27 | 2021-03-19 | 国家电网有限公司大数据中心 | Method and device for constructing image sample library of power transmission and distribution equipment |
| CN112950623A (en) * | 2021-03-29 | 2021-06-11 | 云印技术(深圳)有限公司 | Mark identification method and system |
| CN113111828A (en) * | 2021-04-23 | 2021-07-13 | 中国科学院宁波材料技术与工程研究所 | Three-dimensional defect detection method and system for bearing |
| CN113706523A (en) * | 2021-09-10 | 2021-11-26 | 华能广东海门港务有限责任公司 | Method for monitoring belt deviation and abnormal operation state based on artificial intelligence technology |
| CN114240939A (en) * | 2022-02-24 | 2022-03-25 | 苏州浪潮智能科技有限公司 | Method, system, equipment and medium for detecting appearance defects of mainboard components |
| CN114348368A (en) * | 2022-01-29 | 2022-04-15 | 红云红河烟草(集团)有限责任公司 | Method and system for evaluating soft package cigarette package forming quality |
| CN114757913A (en) * | 2022-04-15 | 2022-07-15 | 电子科技大学 | Display screen defect detection method |
| WO2022267287A1 (en) * | 2021-06-25 | 2022-12-29 | 浙江商汤科技开发有限公司 | Image registration method and related apparatus, and device and storage medium |
| CN116128935A (en) * | 2022-12-30 | 2023-05-16 | 华中科技大学 | Air cluster infrared sequence registration method, multi-scale target spectrum measurement method and system |
| CN116563289A (en) * | 2023-07-11 | 2023-08-08 | 凯德技术长沙股份有限公司 | Labeling quality detection method and system based on machine vision |
| CN117252840A (en) * | 2023-09-26 | 2023-12-19 | 西安交通大学 | Photovoltaic array defect elimination evaluation method and device and computer equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103546740A (en) * | 2013-09-03 | 2014-01-29 | 小米科技有限责任公司 | Method and device for detecting camera and terminal equipment |
-
2015
- 2015-07-09 CN CN201510405923.XA patent/CN105160654A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103546740A (en) * | 2013-09-03 | 2014-01-29 | 小米科技有限责任公司 | Method and device for detecting camera and terminal equipment |
Non-Patent Citations (1)
| Title |
|---|
| 冯江: ""基于构件的印刷品缺陷检测机器视觉***软件开发研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (58)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106886707B (en) * | 2017-03-21 | 2020-06-09 | 广东网金控股股份有限公司 | Image authentication method and device |
| CN106886707A (en) * | 2017-03-21 | 2017-06-23 | 广东网金控股股份有限公司 | A kind of method and device of image authentication |
| CN107203990B (en) * | 2017-04-02 | 2020-12-18 | 南京汇川图像视觉技术有限公司 | Label breakage detection method based on template matching and image quality evaluation |
| CN107203990A (en) * | 2017-04-02 | 2017-09-26 | 南京汇川图像视觉技术有限公司 | A kind of labeling damage testing method based on template matches and image quality measure |
| CN107067425A (en) * | 2017-04-25 | 2017-08-18 | 西安理工大学 | Image synchronization alignment schemes are printed in a kind of print quality automatic detection |
| CN107067425B (en) * | 2017-04-25 | 2019-09-27 | 西安理工大学 | A kind of print quality prints image synchronization alignment schemes in detecting automatically |
| CN107492091A (en) * | 2017-07-06 | 2017-12-19 | 东莞理工学院 | Label look detection method and terminal device based on machine vision |
| CN108548820A (en) * | 2018-03-28 | 2018-09-18 | 浙江理工大学 | Cosmetics paper labels defect inspection method |
| CN108548820B (en) * | 2018-03-28 | 2023-08-15 | 浙江理工大学 | Cosmetic paper label defect detection method |
| CN108874855A (en) * | 2018-04-03 | 2018-11-23 | 南昌奇眸科技有限公司 | A kind of trade-mark searching method based on similar area detection segmentation |
| CN108845998B (en) * | 2018-04-03 | 2022-04-29 | 南昌奇眸科技有限公司 | Trademark image retrieval and matching method |
| CN108845998A (en) * | 2018-04-03 | 2018-11-20 | 南昌奇眸科技有限公司 | A kind of trademark image retrieval matching process |
| CN108874855B (en) * | 2018-04-03 | 2022-06-21 | 南昌奇眸科技有限公司 | Trademark retrieval method based on similar area detection segmentation |
| CN108804499A (en) * | 2018-04-03 | 2018-11-13 | 南昌奇眸科技有限公司 | A kind of trademark image retrieval method |
| CN108763263A (en) * | 2018-04-03 | 2018-11-06 | 南昌奇眸科技有限公司 | A kind of trade-mark searching method |
| CN109085178A (en) * | 2018-08-28 | 2018-12-25 | 武汉科技大学 | A kind of accurate on-line monitoring method of defect fingerprint and feedback strategy for increasing material manufacturing |
| CN109085178B (en) * | 2018-08-28 | 2021-02-12 | 武汉科技大学 | Defect fingerprint accurate online monitoring and feedback method for additive manufacturing |
| CN110245674B (en) * | 2018-11-23 | 2023-09-15 | 浙江大华技术股份有限公司 | Template matching method, device, equipment and computer storage medium |
| CN110245674A (en) * | 2018-11-23 | 2019-09-17 | 浙江大华技术股份有限公司 | Template matching method, device, equipment and computer storage medium |
| CN109636797B (en) * | 2018-12-21 | 2022-08-05 | 国网福建省电力有限公司 | Method for detecting deformation of non-contact flexible direct current converter valve module |
| CN109636797A (en) * | 2018-12-21 | 2019-04-16 | 国网福建省电力有限公司 | A kind of method of contactless flexible direct current converter valve module shape changing detection |
| CN109934777A (en) * | 2019-01-09 | 2019-06-25 | 深圳市三宝创新智能有限公司 | Image local invariant feature extraction method, apparatus, computer equipment and storage medium |
| CN109934809A (en) * | 2019-03-08 | 2019-06-25 | 深慧视(深圳)科技有限公司 | A kind of paper labels character defect inspection method |
| CN110070090A (en) * | 2019-04-25 | 2019-07-30 | 上海大学 | A kind of logistic label information detecting method and system based on handwriting identification |
| CN110070090B (en) * | 2019-04-25 | 2021-06-25 | 上海大学 | Logistics label information detection method and system based on handwritten character recognition |
| CN110245615B (en) * | 2019-06-17 | 2022-09-13 | 河海大学常州校区 | Signature authentication method, system and storage medium |
| CN110245615A (en) * | 2019-06-17 | 2019-09-17 | 河海大学常州校区 | A kind of signature false distinguishing method, system and storage medium |
| CN110288547A (en) * | 2019-06-27 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image denoising model |
| CN110472662A (en) * | 2019-07-10 | 2019-11-19 | 上海理工大学 | Image matching algorithm based on improved ORB algorithm |
| CN110472662B (en) * | 2019-07-10 | 2023-12-29 | 上海理工大学 | Image matching method based on improved ORB algorithm |
| CN110503633A (en) * | 2019-07-29 | 2019-11-26 | 西安理工大学 | A kind of applique ceramic disk detection method of surface flaw based on image difference |
| CN110443295A (en) * | 2019-07-30 | 2019-11-12 | 上海理工大学 | Improved images match and error hiding reject algorithm |
| CN110866911B (en) * | 2019-11-14 | 2022-08-30 | 华雁智能科技(集团)股份有限公司 | Dial defect detection method and device, image processing equipment and storage medium |
| CN110866911A (en) * | 2019-11-14 | 2020-03-06 | 华雁智能科技(集团)股份有限公司 | Dial defect detection method and device, image processing equipment and storage medium |
| CN110991446A (en) * | 2019-11-22 | 2020-04-10 | 上海欧冶物流股份有限公司 | Label identification method, device, equipment and computer readable storage medium |
| CN111145230A (en) * | 2019-12-26 | 2020-05-12 | 北京新唐思创教育科技有限公司 | Image registration quality detection method, device, equipment and storage medium |
| CN111932515A (en) * | 2020-08-10 | 2020-11-13 | 成都数之联科技有限公司 | Short circuit detection method and system for product residual defects and defect classification system |
| CN112330597A (en) * | 2020-10-14 | 2021-02-05 | 浙江大华技术股份有限公司 | Image difference detection method and device and computer equipment |
| CN112258466A (en) * | 2020-10-19 | 2021-01-22 | 燕山大学 | Cold-rolled strip vibration line online detection system and method based on gray level processing |
| CN112381765A (en) * | 2020-10-23 | 2021-02-19 | 杭州翔毅科技有限公司 | Equipment detection method, device, equipment and storage medium based on artificial intelligence |
| CN112288734A (en) * | 2020-11-06 | 2021-01-29 | 西安工程大学 | Printed fabric surface defect detection method based on image processing |
| CN112528055A (en) * | 2020-11-27 | 2021-03-19 | 国家电网有限公司大数据中心 | Method and device for constructing image sample library of power transmission and distribution equipment |
| CN112489042A (en) * | 2020-12-21 | 2021-03-12 | 大连工业大学 | Metal product printing defect and surface damage detection method based on super-resolution reconstruction |
| CN112950623A (en) * | 2021-03-29 | 2021-06-11 | 云印技术(深圳)有限公司 | Mark identification method and system |
| CN113111828A (en) * | 2021-04-23 | 2021-07-13 | 中国科学院宁波材料技术与工程研究所 | Three-dimensional defect detection method and system for bearing |
| WO2022267287A1 (en) * | 2021-06-25 | 2022-12-29 | 浙江商汤科技开发有限公司 | Image registration method and related apparatus, and device and storage medium |
| CN113706523B (en) * | 2021-09-10 | 2023-08-15 | 华能广东海门港务有限责任公司 | Belt deviation and abnormal operation state monitoring method based on artificial intelligence technology |
| CN113706523A (en) * | 2021-09-10 | 2021-11-26 | 华能广东海门港务有限责任公司 | Method for monitoring belt deviation and abnormal operation state based on artificial intelligence technology |
| CN114348368B (en) * | 2022-01-29 | 2023-08-15 | 红云红河烟草(集团)有限责任公司 | Method and system for evaluating soft package cigarette package forming quality |
| CN114348368A (en) * | 2022-01-29 | 2022-04-15 | 红云红河烟草(集团)有限责任公司 | Method and system for evaluating soft package cigarette package forming quality |
| CN114240939A (en) * | 2022-02-24 | 2022-03-25 | 苏州浪潮智能科技有限公司 | Method, system, equipment and medium for detecting appearance defects of mainboard components |
| CN114757913A (en) * | 2022-04-15 | 2022-07-15 | 电子科技大学 | Display screen defect detection method |
| CN116128935A (en) * | 2022-12-30 | 2023-05-16 | 华中科技大学 | Air cluster infrared sequence registration method, multi-scale target spectrum measurement method and system |
| CN116128935B (en) * | 2022-12-30 | 2023-11-28 | 华中科技大学 | Air cluster infrared sequence registration method, multi-scale target spectrum measurement method and system |
| CN116563289A (en) * | 2023-07-11 | 2023-08-08 | 凯德技术长沙股份有限公司 | Labeling quality detection method and system based on machine vision |
| CN116563289B (en) * | 2023-07-11 | 2023-09-29 | 凯德技术长沙股份有限公司 | Labeling quality detection method and system based on machine vision |
| CN117252840A (en) * | 2023-09-26 | 2023-12-19 | 西安交通大学 | Photovoltaic array defect elimination evaluation method and device and computer equipment |
| CN117252840B (en) * | 2023-09-26 | 2024-04-05 | 西安交通大学 | Photovoltaic array defect elimination evaluation method and device and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105160654A (en) | Towel label defect detecting method based on feature point extraction | |
| WO2022170706A1 (en) | Defect detection method and apparatus for mold monitoring, and device and medium | |
| CN102999886B (en) | Image Edge Detector and scale grating grid precision detection system | |
| CN105894036A (en) | Image feature template matching method being applied to detection of mobile phone screen defects | |
| CN102998316B (en) | Transparent liquid impurity detection system and detection method thereof | |
| CN105913415A (en) | Image sub-pixel edge extraction method having extensive adaptability | |
| Jadhav et al. | A fruit quality management system based on image processing | |
| CN103778624A (en) | Fabric defect detection method based on optical threshold segmentation | |
| CN110335233B (en) | Highway guardrail plate defect detection system and method based on image processing technology | |
| CN107490583A (en) | A kind of intermediate plate defect inspection method based on machine vision | |
| CN102073872A (en) | Image-based method for identifying shape of parasite egg | |
| Fu et al. | Research on image-based detection and recognition technologies for cracks on rail surface | |
| CN104966283A (en) | Imaging layered registering method | |
| Haider et al. | A hybrid method for edge continuity based on Pixel Neighbors Pattern Analysis (PNPA) for remote sensing satellite images | |
| CN103136530A (en) | Method for automatically recognizing target images in video images under complex industrial environment | |
| Al-Mahadeen et al. | Signature region of interest using auto cropping | |
| CN103136518A (en) | Preprocessing algorithm of fingerprint image automatic identification system | |
| Song et al. | Crack detection algorithm for photovoltaic image based on multi-scale pyramid and improved region growing | |
| Xue et al. | Detection of Various Types of Metal Surface Defects Based on Image Processing. | |
| Peng et al. | Automatic Crack Detection by Multi-Seeding Fusion on 1mm Resolution 3D Pavement Images | |
| Machhale et al. | Implementation of number recognition using adaptive template matching and feature extraction method | |
| Kumar et al. | A robust and fast text extraction in images and video frames | |
| CN106996934A (en) | Polyphaser vision detection system and detection method | |
| CN105115987A (en) | Inversion defect detection method for tapered roller based on digital filtering | |
| Lei et al. | Concrete surface crack recognition in complex scenario based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151216 |
|
| WD01 | Invention patent application deemed withdrawn after publication |