CN103871053B - Vision conspicuousness-based cloth flaw detection method - Google Patents
Vision conspicuousness-based cloth flaw detection method Download PDFInfo
- Publication number
- CN103871053B CN103871053B CN201410063927.XA CN201410063927A CN103871053B CN 103871053 B CN103871053 B CN 103871053B CN 201410063927 A CN201410063927 A CN 201410063927A CN 103871053 B CN103871053 B CN 103871053B
- Authority
- CN
- China
- Prior art keywords
- notable
- gray
- pixel
- target area
- detection method
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The invention discloses a vision conspicuousness-based cloth flaw detection method which comprises the following steps: (1) collecting an image; (2) processing brightness features; (3) processing direction features; (4) performing multichannel superposition normalizing processing; (5) processing a grey-scale map; (6) performing binaryzation processing; (7) judging a flaw area. Compared with the traditional cloth flaw detection method, the vision conspicuousness-based cloth flaw detection method has the advantages that the operation complexity is reduced, the recognition rate is increased, accurate positioning can be realized, false detection easily caused under the condition that a gray value of a conspicuousness image of a perfection image of detected cloth is higher than a gray value of a perfection part in a flaw image is avoided, the interference of a background during detection is effectively reduced, and the condition that a target area obtained by performing adaptive threshold segmentation on an image of perfect cloth is mistakenly judged to be the flaw area is reduced.
Description
Technical field
The present invention relates to a kind of Fabric Defect detection method is and in particular to a kind of Fabric Defect of view-based access control model significance is examined
Survey method.
Background technology
In modern textile produces, the control of quality is extremely important with detection, and the Defect Detection of cloth is wherein particularly to close
The ingredient of key, at present, more than the detection mode of domestic textile enterprise based on artificial, human eye detection finite rate, detection knot
Fruit is easily affected by subjective factorss, and flase drop, missing inspection etc. easily occur.Artificial Fabric Defect inspection is replaced with advanced Automatic Measurement Technique
Survey, be to improve detection efficiency, reduce labour force, the important measures reducing hand labor intensity and ensureing cloth quality.Both at home and abroad
Scholar achieves much remarkable achievement in the research of automatic testing method.
Fabric Defect detection algorithm is mainly and carries out flaw judgement according to the textural shape of the cloth detecting, and substantially has 3
Class: statistic law, Zymography and be based on modelling.Utilization Fourier transformation, wavelet transformation and gabor that most common of which is
The Spectral Analysis Method of analysis.Fourier transformation is to carry out global change to figure, thus can not be accurately positioned flaw；Gabor analyzes
Change detection better performances, but need to carry out two-dimensional filtering and fusion to multichannel direction, greatly increase calculating complicated
Degree；Wavelet transformation has the advantages that good local time frequency analysis, calculating speed are fast, but set direction difference is so as to can not be very
Describe well X-Y scheme characteristic, lead to Detection results undesirable.
The mankind can fast and effeciently identify defect, and no matter reflective power, the change of defect shape, various processes are made
Difference becoming etc..When cloth is presented on before human eye, its flaw part is more notable, can cause vision attention.That is,
In figure information to be found often focuses on some crucial regions, normally referred to as marking area.If can carry exactly
Take these marking areas, i.e. the region of visually arresting, we just can effectively find the figure information of needs.In order to from
In figure extracts marking area, there has been proposed the computation model that view-based access control model notes.In these models, represented with notable figure and regard
Feel the significance in region.Notable figure is a width X-Y scheme, and the value of its each pixel represents the notable of artwork corresponding pixel points
Property size.The present invention is view-based access control model significance principle, there is provided a kind of method carrying out Defect Detection to cloth figure.
Content of the invention
It is an object of the invention to provide a kind of Fabric Defect detection method of view-based access control model significance, the vision note of simulation people
Meaning mechanism, by gbvs(graph-based visual saliency) based on model, by extract cloth brightness and
Direction character, generates visual saliency map, by analyzing the intensity profile relation of visual saliency map, judges that visual saliency map whether can
In quantity set, and then judge that cloth whether there is flaw.
To achieve the above object of the invention, the technical solution used in the present invention is: a kind of cloth flaw of view-based access control model significance
Defect detection method, comprises the steps:
(1), gather image, figure collection is carried out to cloth, obtains initial gray figure；
(2), gbvs model generates specific image, generates the comprehensive notable figure of described initial gray figure according to gbvs model；
(3), gray-scale maps are processed, and reduce the gray value of each pixel of described comprehensive notable figure, thus obtain weakening background showing
Final notable figure after work property；
(4), binary conversion treatment, according to maximum variance between clusters, carries out adaptive threshold fuzziness to described final notable figure,
Obtain the binary map with prominent significant target part；
(5), defect areas judge, by gray scale quadratic sum and the institute of corresponding target area in relatively described final notable figure
State the gray scale quadratic sum of comprehensive notable figure whole region, to judge target area whether as flaw.
Further technical scheme, step (2) includes as follows step by step:
A (), brightness are processed, carry out gaussian pyramid filtering to described initial gray figure, obtain one group of different scale
Under brightness filter result, to brightness filter result each described, according to the difference between its pixel and Euclidean distance
Set up respective Markov Chain, then ask its Markov to balance
Distribution, the brightness filter result under the described one group of different scale after Markov balanced distribution is returned
One change is processed, and obtains a brightness notable figure；
B (), direction character are processed, described initial gray figure is carried out on four direction with gabor pyramid filtering, corresponding
Obtain the direction character filter result under four groups of different scales, to direction character filter result each described, according between its pixel
Difference and Euclidean distance set up respective Markov Chain, then seek its Markov balanced distribution, Markov balanced
The direction character filter result under described one group of different scale after distribution is normalized, and obtains a direction character and shows
Write figure；
C (), multichannel are superimposed normalization, by the institute in the described brightness notable figure and step (b) in step (a)
State direction character notable figure linear, additive and be normalized, obtain comprehensive notable figure；
Wherein, described step (a) and step (b) no sequencing.
Further technical scheme, step (3) is adopted with the following method: calculates each pixel of described comprehensive notable figure
Average gray, judges whether the gray value of each pixel described is more than described average gray, if described gray value is little
In described average gray, then return-to-zero is carried out to described gray value, otherwise described gray value is equal to it and is put down with described gray scale
The difference of average, obtains the final notable figure weakening background significance.
Further technical scheme, step (5) is adopted with the following method: demarcates described final notable figure and corresponds to described target
Partial region is target area, calculates the pixel gray level quadratic sum of the target area of described final notable figure and described respectively
The pixel gray level quadratic sum of the whole region of comprehensive notable figure, and then it is whole with described to calculate described target area gray scale quadratic sum
The ratio of individual area grayscale quadratic sum, if described ratio is less than judgment threshold, described target area is defect areas, no
Then described target area is not defect areas.
Further technical scheme, the span of described judgment threshold is 0.15～0.4.
Because technique scheme is used, the present invention compared with prior art has the advantage that
1st, by carrying out vision significance analysis to the cloth figure inputting, judge whether to obtain that there is prominent significant target
The binary map in region, and then judge that cloth whether there is flaw, compared with traditional Fabric Defect detection method, present invention reduces
The complexity of computing, improves discrimination, and can be accurately positioned；
2nd, reduced higher than the gray value of the point of meansigma methodss by making grey scale pixel value in comprehensive notable figure, make comprehensive notable figure
The gray value zero of the sub-average point of middle grey scale pixel value, to weaken background (regarding the intact region of cloth as background) significantly
Property so that the significance in defect areas and intact region has obvious gap, it is to avoid due to the notable map generalization of described synthesis
The reason of principle, the gray value in the intact region of the cloth the being detected feelings more taller than the intact part gray value in defect areas
The flase drop being easily caused under condition, significantly reduces interference in detection process for the background；
3rd, because human eye always has the region of relative attention concentration when observation texture is uniformly schemed, therefore view-based access control model is notable
Property the detection method of Fabric Defect easily intact for cloth figure is mistaken for defect areas through the target part that over-segmentation obtains, lead to
The figure characteristic information crossing extraction target area carries out energy judgement, judges whether this target area is area of energy concentration domain, finally
Determine defect areas, calculate the pixel gray level square of the target area corresponding to described target part in described final notable figure
With and described comprehensive notable figure whole region pixel gray level quadratic sum, and then calculate both ratio described, by than
Value size judges this target area whether energy Relatively centralized, and ratio is more than judgment threshold, i.e. energy Relatively centralized, then described mesh
Mark region is flaw, and otherwise described target area is not flaw, by above technical scheme, decreases the generation of erroneous judgement.
Brief description
Fig. 1 is the implementing procedure figure of the Fabric Defect detection method of view-based access control model significance disclosed by the invention；
Fig. 2 is the progressively testing result figure of the intact figure of cloth disclosed by the invention；
Fig. 3 is the progressively testing result figure of cloth typical case's flaw figure disclosed by the invention；
Fig. 4, Fig. 5, Fig. 6, Fig. 7 are the testing result figures of other flaw figures of cloth disclosed by the invention.
Specific embodiment
Below in conjunction with the accompanying drawings and embodiment further describes to the present invention.
Embodiment: referring to Fig. 1, as shown in legend therein, above-mentioned Fabric Defect detection method comprises the steps:
(1), gather image, gather the figure of cloth by industrial camera, obtain initial gray figure, as Fig. 2 (a) is
The initial graph of the intact figure of cloth, such as Fig. 3 (a) are the initial gray figure of cloth typical case's flaw figure.
(2), brightness is processed:
A, by 2-d gaussian filterses device to input above-mentioned initial gray figureCarry out gaussian pyramid filtering, gold
The filtering of word tower refers to initial gray figure is constantly carried out with 1/2 down-sampled and filtering, and the scale factor of filtering subtracts with the reduction of figure
Little, obtain the filter result under one group of different scale, in this example, pyramid grade is 2, that is, the difference obtaining under two yardsticks is bright
Degree characteristic filtering result, i.e. two brightness figures；
Wherein,It is initial gray figure,It is in scale factor to brightnessUnder use Gaussian function
Filter the gaussian pyramid obtaining,For the Gaussian function of 2-d gaussian filterses device,For corresponding in figure pixel
Position,Scale factor for corresponding figure；
B, to brightness figure every describedCarry out Markov balanced distribution, and then normalized, obtain
To brightness notable figure；
If each brightness figureFor, the distance between any two pixels are, then any two pixelsWithDistance:
By characteristic patternEach two pixel all connects two-by-two, and each edge represents the weight of point-to-point transmission, pointTo pointWeightFor:
WhereinFor Gauss weighting parameters；
That is the gray difference between the weight between any two pixels is by two pixels and dot spacing are determined from common
Fixed, rightabout weight is identical, said process be ask every brightness figure respective weights matrix eigenvalue of maximum corresponding
The process of characteristic vector, that is, seek the equilibrium state process of every brightness figure；
By two brightness balanced distribution figuresIt is normalized, obtain a brightness notable figure.
(3), direction character is processed:
A, by gabor wave filter to input initial gray figureGabor pyramid is carried out on four direction
Filtering, pyramid filtering refers to initial gray figure is constantly carried out with 1/2 down-sampled and filtering, scale factor the subtracting with figure of filtering
Little and reduce, obtain the filter result under four groups of different scales, in this example, pyramid grade is 2, that is, obtain four groups of two yardsticks
Under different directions characteristic filtering result, i.e. eight direction character figures；
Wherein,It is the gabor gold word that the gabor function filtering in yardstick direction obtains to direction character
Tower,For the gabor function of gabor wave filter,For the position of corresponding in figure pixel,For yardstick because
Son,(take for direction factor),For the sine wave freuqency of gabor wave filter,；
B, to every direction character figureCarry out Markov balanced distribution, and then normalized, obtain
Brightness notable figure,
If each direction character figureFor, the distance between any two pixels are, then any two pixels
PointWithDistance:
By characteristic patternEach two pixel all connects two-by-two, and each edge represents the weight of point-to-point transmission, pointTo pointWeightFor:
WhereinFor Gauss weighting parameters；
That is the gray difference between the weight between any two pixels is by two pixels and dot spacing are determined from common
Fixed, rightabout weight is identical, said process be ask every direction character figure respective weights matrix eigenvalue of maximum corresponding
The process of characteristic vector, that is, seek the equilibrium state process of every direction character figure；
By eight direction character balanced distribution figuresIt is normalized, obtain direction character notable figure；
(4), it is superimposed normalization, by above-mentioned brightness notable figureWith direction character notable figure
Linear, additive is simultaneously normalized, and obtains and above-mentioned initial gray figureComprehensive notable figure of the same size
, such as Fig. 2 (b) is the comprehensive notable figure of the intact figure of cloth, and such as Fig. 3 (b) is the comprehensive notable figure of cloth typical case's flaw figure.
(5), gray-scale maps are processed, and calculate above-mentioned comprehensive notable figureAverage gray, relatively above-mentioned synthesis is notable
FigureEach pixel gray value and average gray, if above-mentioned pixel gray value is less than average gray, to upper
State pixel gray value and carry out return-to-zero, otherwise above-mentioned pixel gray value is equal to its difference with average gray, obtains
Weaken the final notable figure of background gray scale, such as Fig. 2 (c) is the final notable figure of the intact figure of cloth, and such as Fig. 3 (c) is cloth
The final notable figure of typical flaw figure.
(6), binary conversion treatment, according to maximum variance between clusters, to above-mentioned final notable figureCarry out adaptive thresholding
Value segmentation, obtains the binary map with prominent significant target part, such as Fig. 2 (d) is the two-value of the intact figure of cloth
Figure, such as Fig. 3 (d) are the binary map of cloth typical case's flaw figure.
(7), defect areas judge, demarcate above-mentioned final notable figureIn corresponding to the region of described target part be
Target area, calculates above-mentioned final notable figureThe pixel gray level quadratic sum of target area and described comprehensive notable figureWhole region pixel gray level quadratic sum, and then calculate both ratio.Judgment threshold in the methods of the invention
Span is 0.15～0.4, and this example judgment threshold is taken as 0.2, if above-mentioned ratio is less than judgment threshold 0.2, above-mentioned
Target area is defect areas, and otherwise above-mentioned target area is not defect areas, and such as Fig. 2 (e) is the judgement knot of the intact figure of cloth
Fruit is schemed, and such as Fig. 3 (e) is the judged result figure of cloth typical case's flaw figure.
Calculate the ratio in divided region and the gray scale quadratic sum of whole region, this target area is judged by ratio size
Whether energy Relatively centralized.Ratio is more than judgment threshold, i.e. energy Relatively centralized, is flaw；It is not otherwise flaw.
Referring to Fig. 4-Fig. 7, it is the testing result figure of other several typical case's flaw figures of cloth, wherein (a) is initial graph, (b)
For comprehensive notable figure, (c) is judged result figure.
Claims (4)
1. a kind of Fabric Defect detection method of view-based access control model significance is it is characterised in that comprise the steps:
(1), gather image, figure collection is carried out to cloth, obtains initial gray figure；
(2), gbvs model generates specific image, generates the comprehensive notable figure of described initial gray figure according to gbvs model；Including
As follows step by step:
A (), brightness are processed, carry out gaussian pyramid filtering to described initial gray figure, obtain under one group of different scale
Brightness filter result, to brightness filter result each described, sets up according to the difference between its pixel and Euclidean distance
Respective Markov Chain, then asks its Markov to balance
Distribution, the brightness filter result under the described one group of different scale after Markov balanced distribution is normalized
Process, obtain a brightness notable figure；
B (), direction character are processed, described initial gray figure is carried out on four direction with gabor pyramid filtering, and correspondence obtains
Direction character filter result under four groups of different scales, to direction character filter result each described, according to the difference between its pixel
Different and Euclidean distance sets up respective Markov Chain, then seeks its Markov balanced distribution, by Markov balanced distribution
The direction character filter result under described one group of different scale afterwards is normalized, and obtains a direction character notable
Figure；
C (), multichannel are superimposed normalization, by the described side in the described brightness notable figure and step (b) in step (a)
To characteristic remarkable picture linear, additive and be normalized, obtain comprehensive notable figure；
Wherein, described step (a) and step (b) no sequencing；
(3), gray-scale maps are processed, and reduce the gray value of each pixel of described comprehensive notable figure, thus obtaining weakening background significance
Final notable figure afterwards；
(4), binary conversion treatment, according to maximum variance between clusters, carries out adaptive threshold fuzziness to described final notable figure, obtains
There is the binary map of prominent significant target part；
(5), defect areas judge, the gray scale quadratic sum by corresponding target area in relatively described final notable figure is comprehensive with described
Close the gray scale quadratic sum of notable figure whole region, to judge target area whether as flaw.
2. according to view-based access control model significance described in claim 1 Fabric Defect detection method it is characterised in that step (3) adopt
Following method: calculate the average gray of each pixel of described comprehensive notable figure, judge the gray scale of each pixel described
Whether value is more than described average gray, if described gray value is less than described average gray, described gray value is carried out
Return-to-zero, otherwise described gray value be equal to its difference with described average gray, obtain weakening the final of background significance
Notable figure.
3. according to view-based access control model significance described in claim 1 Fabric Defect detection method it is characterised in that step (5) adopt
Following method: the region demarcating described final notable figure corresponding to described target part is target area, respectively described in calculating
The pixel gray level of the whole region of the pixel gray level quadratic sum of the target area of whole notable figure and described comprehensive notable figure is put down
Fang He, and then calculate the ratio of described target area gray scale quadratic sum and described whole region gray scale quadratic sum, if described ratio
When value is less than judgment threshold, then described target area is defect areas, and otherwise described target area is not defect areas.
4. according to the Fabric Defect detection method described in claim 3 it is characterised in that the span of described judgment threshold is
0.15～0.4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410063927.XA CN103871053B (en) | 2014-02-25 | 2014-02-25 | Vision conspicuousness-based cloth flaw detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410063927.XA CN103871053B (en) | 2014-02-25 | 2014-02-25 | Vision conspicuousness-based cloth flaw detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103871053A CN103871053A (en) | 2014-06-18 |
| CN103871053B true CN103871053B (en) | 2017-02-01 |
Family
ID=50909558
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410063927.XA Active CN103871053B (en) | 2014-02-25 | 2014-02-25 | Vision conspicuousness-based cloth flaw detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103871053B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109300127A (en) * | 2018-09-21 | 2019-02-01 | 中兴新通讯有限公司 | Defect inspection method, device, computer equipment and storage medium |
Families Citing this family (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104198497B (en) * | 2014-09-12 | 2017-01-25 | 苏州大学 | Surface defect detection method based on visual saliency map and support vector machine |
| CN104574353B (en) * | 2014-09-12 | 2017-08-01 | 苏州大学 | The surface defect decision method of view-based access control model conspicuousness |
| CN104700412B (en) * | 2015-03-17 | 2019-01-22 | 苏州大学 | A kind of calculation method of visual saliency map |
| CN105021628B (en) * | 2015-07-19 | 2019-12-17 | 中北大学 | Method for detecting surface defects of optical fiber image inverter |
| CN106157314B (en) * | 2015-12-23 | 2018-12-14 | 河海大学 | A kind of real-time fault diagnosis method of paper making equipment |
| CN105678767B (en) * | 2016-01-07 | 2019-02-19 | 无锡信捷电气股份有限公司 | A kind of cloth surface flaw detection method based on SoC Hardware/Software Collaborative Design |
| CN108268866B (en) * | 2016-12-30 | 2020-07-10 | 纵目科技（上海）股份有限公司 | Vehicle detection method and system |
| CN108629225B (en) * | 2017-03-15 | 2022-02-25 | 纵目科技（上海）股份有限公司 | Vehicle detection method based on multiple sub-images and image significance analysis |
| CN108629778B (en) * | 2017-03-15 | 2022-02-25 | 纵目科技（上海）股份有限公司 | Image salient object detection method and system based on local feature weighting enhancement |
| CN108629806B (en) * | 2017-03-15 | 2022-02-25 | 纵目科技（上海）股份有限公司 | Image salient object detection method and system based on image contrast |
| CN107145824B (en) * | 2017-03-29 | 2020-10-23 | 纵目科技（上海）股份有限公司 | Lane line segmentation method and system based on significance analysis and vehicle-mounted terminal |
| CN107220962B (en) * | 2017-04-07 | 2020-04-21 | 北京工业大学 | Image detection method and device for tunnel cracks |
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107895363A (en) * | 2017-10-31 | 2018-04-10 | 常州大学 | Textile flaw detection method based on template characteristic |
| CN109960984A (en) * | 2017-12-25 | 2019-07-02 | 大连楼兰科技股份有限公司 | Vehicle checking method based on contrast and significance analysis |
| CN108960255A (en) * | 2018-06-28 | 2018-12-07 | 西安工程大学 | Conspicuousness fabric defect detection method based on color similarity and position aggregation |
| CN109816631A (en) * | 2018-12-25 | 2019-05-28 | 河海大学 | A kind of image partition method based on new cost function |
| CN110031461B (en) * | 2019-02-14 | 2022-03-18 | 江苏恒力化纤股份有限公司 | Polyester filament yarn dyeing uniformity test method |
| CN109961437B (en) * | 2019-04-04 | 2021-06-25 | 江南大学 | Method for detecting significant fabric defects based on machine teaching mode |
| CN110298835B (en) * | 2019-07-02 | 2023-07-18 | 广东工业大学 | Leather surface damage detection method, system and related device |
| CN110706205B (en) * | 2019-09-07 | 2021-05-14 | 创新奇智(重庆)科技有限公司 | Method for detecting cloth hole-breaking defect by using computer vision technology |
| CN110751135A (en) * | 2019-12-24 | 2020-02-04 | 广东博智林机器人有限公司 | Drawing checking method and device, electronic equipment and storage medium |
| CN111402225B (en) * | 2020-03-13 | 2023-05-26 | 浙江工业大学 | Cloth folding false-detection defect discriminating method |
| CN111667476B (en) * | 2020-06-09 | 2022-12-06 | 创新奇智(广州)科技有限公司 | Cloth flaw detection method and device, electronic equipment and readable storage medium |
| CN111861996B (en) * | 2020-06-23 | 2023-11-03 | 西安工程大学 | Printed fabric defect detection method |
| CN112669306A (en) * | 2021-01-06 | 2021-04-16 | 北京信息科技大学 | Solar cell defect detection method and system based on saliency map |
| CN113638104B (en) * | 2021-10-14 | 2022-02-08 | 海门市恒昌织带有限公司 | Intelligent yarn cleaning control method and system for bobbin winder |
| CN114511851B (en) * | 2022-01-30 | 2023-04-04 | 中国南水北调集团中线有限公司 | Hairspring algae cell statistical method based on microscope image |
| CN115598138B (en) * | 2022-11-23 | 2023-04-07 | 惠州威尔高电子有限公司 | Power control circuit board flaw detection method and system based on significance detection |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214298A (en) * | 2011-06-20 | 2011-10-12 | 复旦大学 | Method for detecting and identifying airport target by using remote sensing image based on selective visual attention mechanism |
| CN102799882A (en) * | 2012-07-09 | 2012-11-28 | 武汉市科迅智能交通设备有限公司 | License plate positioning method based on visual saliency |
| CN103247059A (en) * | 2013-05-27 | 2013-08-14 | 北京师范大学 | Remote sensing image region of interest detection method based on integer wavelets and visual features |
| CN103309982A (en) * | 2013-06-17 | 2013-09-18 | 武汉大学 | Remote sensing image retrieval method based on vision saliency point characteristics |
| CN103544709A (en) * | 2013-11-07 | 2014-01-29 | 东北大学 | Method for detecting micro defects on surface of silicon steel plate based on significant active contour model |
-
2014
- 2014-02-25 CN CN201410063927.XA patent/CN103871053B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214298A (en) * | 2011-06-20 | 2011-10-12 | 复旦大学 | Method for detecting and identifying airport target by using remote sensing image based on selective visual attention mechanism |
| CN102799882A (en) * | 2012-07-09 | 2012-11-28 | 武汉市科迅智能交通设备有限公司 | License plate positioning method based on visual saliency |
| CN103247059A (en) * | 2013-05-27 | 2013-08-14 | 北京师范大学 | Remote sensing image region of interest detection method based on integer wavelets and visual features |
| CN103309982A (en) * | 2013-06-17 | 2013-09-18 | 武汉大学 | Remote sensing image retrieval method based on vision saliency point characteristics |
| CN103544709A (en) * | 2013-11-07 | 2014-01-29 | 东北大学 | Method for detecting micro defects on surface of silicon steel plate based on significant active contour model |
Non-Patent Citations (3)
| Title |
|---|
| Graph-based visual saliency;Harel J 等;《Advances in neural information processing systems》;20060228;第545-552页 * |
| 一种改进的多通道Gabor滤波器布匹瑕疵检测方法;屈博 等;《纺织学报》;20091231;第30卷(第12期);第37-40页 * |
| 基于图像显著性特征的铸坯表面缺陷检测;吴家伟;《智能系统学报》;20120228;第7卷(第1期);第75-80页 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109300127A (en) * | 2018-09-21 | 2019-02-01 | 中兴新通讯有限公司 | Defect inspection method, device, computer equipment and storage medium |
| CN109300127B (en) * | 2018-09-21 | 2021-02-05 | 深圳新视智科技术有限公司 | Defect detection method, defect detection device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103871053A (en) | 2014-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103871053B (en) | Vision conspicuousness-based cloth flaw detection method | |
| CN110378196B (en) | Road visual detection method combining laser point cloud data | |
| CN109636772A (en) | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning | |
| CN113592845A (en) | Defect detection method and device for battery coating and storage medium | |
| CN104392463B (en) | Image salient region detection method based on joint sparse multi-scale fusion | |
| CN104008399B (en) | The recognition methodss of the gauge pointer shake based on support vector machine in a kind of instrument detection | |
| CN103604809B (en) | A kind of online visible detection method of pattern cloth flaw | |
| CN103996018B (en) | Face identification method based on 4DLBP | |
| CN108009509A (en) | Vehicle target detection method | |
| CN108898610A (en) | A kind of object contour extraction method based on mask-RCNN | |
| CN110310262A (en) | A kind of method, apparatus and system for detection wheel tyre defect | |
| Liu et al. | A rail surface defect detection method based on pyramid feature and lightweight convolutional neural network | |
| CN106056118A (en) | Recognition and counting method for cells | |
| CN107767387B (en) | Contour detection method based on variable receptive field scale global modulation | |
| CN104732536A (en) | Sub-pixel edge detection method based on improved morphology | |
| CN110827260B (en) | Cloth defect classification method based on LBP characteristics and convolutional neural network | |
| CN105046197A (en) | Multi-template pedestrian detection method based on cluster | |
| CN109242848A (en) | Based on OTSU and GA-BP neural network wallpaper defects detection and recognition methods | |
| CN107292310A (en) | A kind of circular pointer dial plate vision positioning and automatic reading method | |
| CN103048329A (en) | Pavement crack detecting method based on active contour model | |
| CN110334594A (en) | A kind of object detection method based on batch again YOLO algorithm of standardization processing | |
| CN104103033A (en) | Image real-time processing method | |
| CN103778627A (en) | Sea oil spill detection method based on SAR image | |
| Ji et al. | Apple grading method based on features of color and defect | |
| CN105225216A (en) | Based on the Iris preprocessing algorithm of space apart from circle mark rim detection |
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 |