CN109785290A - Normalized steel plate defect detection method is shone based on local light - Google Patents
Normalized steel plate defect detection method is shone based on local light Download PDFInfo
- Publication number
- CN109785290A CN109785290A CN201811556028.8A CN201811556028A CN109785290A CN 109785290 A CN109785290 A CN 109785290A CN 201811556028 A CN201811556028 A CN 201811556028A CN 109785290 A CN109785290 A CN 109785290A
- Authority
- CN
- China
- Prior art keywords
- steel plate
- image
- data
- detection method
- local light
- 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.)
- Granted
Links
Landscapes
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The present invention relates to one kind to shine normalized steel plate defect detection method based on local light, contains following steps: 1, being taken pictures using colored area array cameras to surface of steel plate, collects the raw image data of a frame Baeyer array；It 2, is the grayscale image of single channel component by image real time transfer；3, processing grayscale image obtains down-sampled gray level image；4, it is handled using cell type mean filter, obtains local photometric data；5, using asking the mode of quotient to acquire the luminance difference data between gray level image texture and local photometric data；6, luminance difference data are done into thresholding processing, obtains bianry image；7, it carries out connected domain and surrounds frame detection；8, it calculates connected domain and surrounds frame area and scored and filtered, defect is sorted out；Effective detection to the non-planar steel plate defect for being difficult to detect under plane steel plate defect or uneven illumination can be achieved in the present invention, and detection sensitivity is high, at low cost, it is easy to accomplish.
Description
(1), technical field
The present invention relates to a kind of steel plate defect detection methods, and in particular to one kind is lacked based on local light according to normalized steel plate
Fall into detection method.
(2), background technique
The defects detection of most metals is primarily upon the features such as scratch, fracture, pit, greasy dirt, for traditional vision
For detection device, detect a kind of planar metal object be it is relatively simple, only need to dispose sufficiently uniform light source, overcome bloom
Reflection can directly apply typical edge enhancement algorithm, then handle to obtain testing result by thresholding.Non-planar metal object
Body is primarily referred to as the metal product with certain curvature or components, the detection for non-planar steel plate, based on traditional defect
Detection system can not dispose accurate enough, ideal light-source system, and therefore, non-planar sample to be tested is difficult to overcome light source uneven
One this characteristic, non-planar bring reflection problems be generalized to traditional threshold detection method can not on curved face product.
General defect detecting system is broadly divided into roller expansion detection system and part is sliced two kinds of detection system, and examines
Measurement equipment is divided into line-scan digital camera and two kinds of area array cameras again, or even also has the detection device based on laser three-D reconstruction technique.One
As defect detecting system there is the problem that volume is big, debugging is difficult, deployment is difficult, volume essentially consist in greatly the body machine of roller expansion compared with
Greatly.Roller is unfolded used in detection system to be line-scan digital camera rather than area array cameras, and line-scan digital camera is mainly used in based on scanning
In the detection device of the image acquisition technology of splicing, therefore, area array cameras is not available, and line-scan digital camera is at high cost, deployment
Required precision is also high, can not flexibly extend in the defects detection of other metalloid objects；Another kind is based on area array cameras
The defect detecting system of array, the limitation of the system is can only could be in splicing by designing light source ideal enough
Do not lead to the problem of hot spot.Therefore, both defect detecting systems are unsuitable for predominantly detecting method as non-planar steel plate.
With manufacturing development, more and more abnormal curved surface part of the steel plate quality control requirements with production technology and
The progress of application scenarios and improve, the deployment complexity and deployment speed of detection device also become in the evaluation of detection scheme
Hold.Therefore, the manufacturer of current manufacturing industry especially speciality sheet metal needs a kind of deployment flexible, is easy to build, is easy to adjust
The novel system of examination, meanwhile, implementation cost also requires lower.
Existing plate defect detection method can not realize the defects detection of non-planar steel plate, main cause well
It is that light source can generate different reflecting effect on the surface of curved surface, and then higher requirement is generated to light source, needs to make
The area source etc. that with annular coaxial light source or more can accurately put.
The non-planar steel plate defect detection of curved surface is difficult to the main reason for flexibly implementing and is that its curved surface can not be overcome to cause
Illuminance it is different, existing detection method does not get rid of the high request problem of light source well.
(3), summary of the invention
The technical problem to be solved by the present invention is providing one kind based on local light according to normalized steel plate defect detection side
Method, this method can realize effective inspection to the non-planar steel plate defect for being difficult to detect under plane steel plate defect or uneven illumination
It surveys, detection sensitivity is high, at low cost, it is easy to accomplish.
Technical solution of the present invention:
One kind shining normalized steel plate defect detection method based on local light, contains following steps:
Step 1: being taken pictures using colored area array cameras to surface of steel plate, the sensor of colored area array cameras is collected
The raw image data of one frame Baeyer array；
Step 2: handling raw image data by reduction formula for the grayscale image of single channel component；
Step 3: being handled to obtain down-sampled gray level image by the grayscale image to single channel component；
Step 4: handling gray level image using cell type mean filter, steel plate Luminance Distribution is calculated, part is obtained
Photometric data；
Step 5: being obtained using asking the mode of quotient to compare the luminance difference between gray level image texture and local photometric data
To luminance difference data；
Step 6: luminance difference data are done thresholding processing, bianry image is obtained；
Step 7: carrying out surrounding frame detection to the connected domain of closure processing, obtaining surrounding frame to the processing of bianry image closure
Center anchor point position；
Step 8: being calculated with scoring formula inside the encirclement frame of positioning, steel plate is significantly lacked according to score value
It is trapped into capable classification.
It is that the method for the present invention can overcome without the use of the reason of gray scale camera using colored area array cameras in step 1
Most of non-idealities, at the same it is not high to the locating depth required precision of camera, and common gray scale camera usually has 10~14bit
The even higher dynamic range such as not detects steel plate defect, the grayscale that general defect can be presented under 8bit precision
Variation is generally higher than 20 grayscale, and therefore, theory, which has 8bit sensors A DC, to be implemented based on this method.
In step 2, tetra- points of R, G, G, B in raw image data array are obtained using the CFA sample window of 2 × 2 sizes
Amount, and it is scaled the grayscale image of single channel component.
General Baeyer pixel array is a kind of 2 × 2 pixel arrangement, and the pixel intensity for extraction is as a result, can be with
Following matrix indicates:
Wherein, CFA is the abbreviation of Color Filter Array, also referred to as color filter array, is used at this group pattern
The expression of CFA title, the position of coordinate when CFA (x, y) refers to CFA array using 2 × 2 size as unit
For gray processing Baeyer array data, shown in a kind of simple and effective following formula of embodiment:
Wherein, the single channel component value that G (x, y) is is obtained by the pixel maximizing to CFA unit.
The ADC precision of general area array sensor is 10bit, while retaining 10bit information by a kind of mapping means
Subsequent information processing is carried out using 8bit.
Define herein it is a kind of with precision compression remap expression formula:
Obviously when input locating depth is 10, and output locating depth is 8:
Further, in order to obtain accurately high optical information, need to round up to floating point result:
So far the image that can switch to 8bit shaping has been obtained.
Wherein, I (x, y) is finally obtained single channel gray level image.
In step 3, for considering for speed up processing and noise reduction problem, adopted using the square of integral multiple width
Sample window carries out the grayscale image of single channel component down-sampled to obtain down-sampled gray level image.
The down-sampled convolution kernel of N × N size is defined, coefficient is expressed as follows:
When sliding convolution kernel x-axis and y-axis step size is N, a kind of down-sampled effect of mean denoising can be realized.
In step 4, down-sampled image is handled using cell type filter, the local normalized Local size of illumination
For filter convolution kernel size, convolution expression formula is as follows:
Wherein, Wf(x, y) is the mean filter sliding window that anchor point is located at convolution kernel origin position (the generally upper left corner),
Size is an integer for being greater than 1.Summation operation is located at window WfInterior operation, finally multiplied by scaling constant C.
Scaling constant C is defined as follows:
Scale constant as the inverse for the included pixel quantity of window of summing.
In step 5, use the mode for asking quotient are as follows: using division calculation make local photometric data and greyscale image data into
Row normalized.
Ask quotient's operation to F (x, y) and I (x, y), normalizes local luminance, the image Z that obtains that treated.
So far, the image Z obtained has as the splicing result uniform with the illumination that line scan camera obtains, can be obvious
Ground obtains defect information, and can direct thresholding.
In step 6, in actual application, if process of the invention is to execute frame by frame and is a kind of scanning workpiece
Form then needs to guarantee that the processing time of present treatment process is constant controllable.
In general, input picture P, then thresholding can indicate are as follows:
In above formula, C is that the maximum of current locating depth indicates range:
C=2depth- 1=28- 1=255
Processing speed caused by cpu cache hit rate being avoided to be lower by tabling look-up due to above-mentioned calculating is inconsistent
Problem constructs thresholding look-up table TLUT herein:
For the image of grayscale i, after defining a threshold value Threshold:
Based on this, thresholding table lookup operations may be implemented, guarantee that the calculating time is constant.
Finally, bianry image B has been obtained:
B (x, y)=TLUT (Z (x, y))
In step 7, general connected domain analysis is carried out, obtains bounding box Recti。
In step 8, by calculate connected domain area and surround frame area quotient score, according to score value differentiation compared with
For significant defect type, highlight area and shadow effect region.
Connected domain area is calculated using integration method and surrounds frame area.
A classifying method is defined herein to determine whether defect is scratch, since remaining defect is that greasy dirt, pit etc. are small
And cost function is sorted out in intensive defect, definition:
Wherein, to ROI (Recti) in connected domain pixel summation, connected domain area is obtained, to RectiInterior length and width are asked
Product, obtains bounding box area.
The final quotient for being scored at the two.Therefore, when S is sufficiently small, then it is assumed that be scratch, the general threshold value that is arranged is
0.08。
Beneficial effects of the present invention:
1, the present invention uses area array cameras and simple area source, can be realized by using local light according to method for normalizing
Effective detection to the non-planar steel plate defect for being difficult to detect under plane steel plate defect or uneven illumination, detection sensitivity are high.
2, it is overcome in the present invention and needs special testing equipment or particular cameras in conventional steel plates defect inspection method, overcome
Light-source brightness unbalanced problem, without using line scan camera or camera array, also without using special super uniform
The high-precision light sources such as LED bar graph light source and area source, there is equipment to integrate, and easy, testing cost is low, integration mode is flexible.
3, the present invention requires low to putting for camera, uses tripod or the simple fixed camera of other methods, energy
It is enough neatly implement under various operating conditions processing workshop, in storehouse scene.
4, the present invention can be integrated in the form of software in any computer equipment or typical Mechatronic Systems in, to place
The performance requirement for managing device is lower, it is easy to accomplish.
(4), Detailed description of the invention
Fig. 1 is that processing local light shines normalized algorithm block diagram in the present invention；
Fig. 2 is the flow diagram that present invention control handling duration establishes thresholding look-up table.
(5), specific embodiment
Contain following steps according to normalized steel plate defect detection method based on local light:
Step 1: being taken pictures using colored area array cameras to surface of steel plate, the sensor of colored area array cameras is collected
The raw image data of one frame Baeyer array；
Step 2: handling raw image data by reduction formula for the grayscale image of single channel component；
Step 3: being handled to obtain down-sampled gray level image by the grayscale image to single channel component；
Step 4: handling gray level image using cell type mean filter, steel plate Luminance Distribution is calculated, part is obtained
Photometric data；
Step 5: being obtained using asking the mode of quotient to compare the luminance difference between gray level image texture and local photometric data
To luminance difference data；
Step 6: luminance difference data are done thresholding processing, bianry image is obtained；
Step 7: carrying out surrounding frame detection to the connected domain of closure processing, obtaining surrounding frame to the processing of bianry image closure
Center anchor point position；
Step 8: being calculated with scoring formula inside the encirclement frame of positioning, steel plate is significantly lacked according to score value
It is trapped into capable classification.
It is that the method for the present invention can overcome without the use of the reason of gray scale camera using colored area array cameras in step 1
Most of non-idealities, at the same it is not high to the locating depth required precision of camera, and common gray scale camera usually has 10~14bit
The even higher dynamic range such as not detects steel plate defect, the grayscale that general defect can be presented under 8bit precision
Variation is generally higher than 20 grayscale, and therefore, theory, which has 8bit sensors A DC, to be implemented based on this method.
In step 2, tetra- points of R, G, G, B in raw image data array are obtained using the CFA sample window of 2 × 2 sizes
Amount, and it is scaled the grayscale image of single channel component.
General Baeyer pixel array is a kind of 2 × 2 pixel arrangement, and the pixel intensity for extraction is as a result, can be with
Following matrix indicates:
Wherein, CFA is the abbreviation of Color Filter Array, also referred to as color filter array, is used at this group pattern
The expression of CFA title, the position of coordinate when CFA (x, y) refers to CFA array using 2 × 2 size as unit
For gray processing Baeyer array data, shown in a kind of simple and effective following formula of embodiment:
Wherein, the single channel component value that G (x, y) is is obtained by the pixel maximizing to CFA unit.
The ADC precision of general area array sensor is 10bit, while retaining 10bit information by a kind of mapping means
Subsequent information processing is carried out using 8bit.
Define herein it is a kind of with precision compression remap expression formula:
Obviously when input locating depth is 10, and output locating depth is 8:
Further, in order to obtain accurately high optical information, need to round up to floating point result:
So far the image that can switch to 8bit shaping has been obtained.
Wherein, I (x, y) is finally obtained single channel gray level image.
In step 3, for considering for speed up processing and noise reduction problem, adopted using the square of integral multiple width
Sample window carries out the grayscale image of single channel component down-sampled to obtain down-sampled gray level image.
The down-sampled convolution kernel of N × N size is defined, coefficient is expressed as follows:
When sliding convolution kernel x-axis and y-axis step size is N, a kind of down-sampled effect of mean denoising can be realized.
In step 4, down-sampled image is handled using cell type filter, the local normalized Local size of illumination
For filter convolution kernel size, convolution expression formula is as follows:
Wherein, Wf(x, y) is the mean filter sliding window that anchor point is located at convolution kernel origin position (the generally upper left corner),
Size is an integer for being greater than 1.Summation operation is located at window WfInterior operation, finally multiplied by scaling constant C.
Scaling constant C is defined as follows:
Scale constant as the inverse for the included pixel quantity of window of summing.
In step 5, use the mode for asking quotient are as follows: using division calculation make local photometric data and greyscale image data into
Row normalized.
Ask quotient's operation to F (x, y) and I (x, y), normalizes local luminance, the image Z that obtains that treated.
So far, the image Z obtained has as the splicing result uniform with the illumination that line scan camera obtains, can be obvious
Ground obtains defect information, and can direct thresholding.
Local light is as shown in Figure 1 according to normalized process flow.
In step 6, in actual application, if process of the invention is to execute frame by frame and is a kind of scanning workpiece
Form then needs to guarantee that the processing time of present treatment process is constant controllable.
In general, input picture P, then thresholding can indicate are as follows:
In above formula, C is that the maximum of current locating depth indicates range:
C=2depth-1=28- 1=255
Processing speed caused by cpu cache hit rate being avoided to be lower by tabling look-up due to above-mentioned calculating is inconsistent
Problem constructs thresholding look-up table TLUT herein:
For the image of grayscale i, after defining a threshold value Threshold:
The flow chart of specific building threshold value look-up table TLUT is as shown in Figure 2.
Based on this, thresholding table lookup operations may be implemented, guarantee that the calculating time is constant.
Finally, bianry image B has been obtained:
B (x, y)=TLUT (Z (x, y))
In step 7, general connected domain analysis is carried out, obtains bounding box Recti。
In step 8, by calculate connected domain area and surround frame area quotient score, according to score value differentiation compared with
For significant defect type, highlight area and shadow effect region.
Connected domain area is calculated using integration method and surrounds frame area.
A classifying method is defined herein to determine whether defect is scratch, since remaining defect is that greasy dirt, pit etc. are small
And cost function is sorted out in intensive defect, definition:
Wherein, to ROI (Recti) in connected domain pixel summation, connected domain area is obtained, to RectiInterior length and width are asked
Product, obtains bounding box area.
The final quotient for being scored at the two.Therefore, when S is sufficiently small, then it is assumed that be scratch, the general threshold value that is arranged is
0.08。
Claims (6)
1. one kind shines normalized steel plate defect detection method based on local light, it is characterized in that: containing following steps:
Step 1: being taken pictures using colored area array cameras to surface of steel plate, the sensor of colored area array cameras collects a frame
The raw image data of Baeyer array；
Step 2: handling raw image data by reduction formula for the grayscale image of single channel component；
Step 3: being handled to obtain down-sampled gray level image by the grayscale image to single channel component；
Step 4: handling gray level image using cell type mean filter, steel plate Luminance Distribution is calculated, obtains local light photograph
Data；
Step 5: being obtained bright using asking the mode of quotient to compare the luminance difference between gray level image texture and local photometric data
Spend variance data；
Step 6: luminance difference data are done thresholding processing, bianry image is obtained；
Step 7: carrying out surrounding frame detection to the connected domain of closure processing, obtaining surrounding frame center to the processing of bianry image closure
Anchor point position；
Step 8: being calculated with scoring formula inside the encirclement frame of positioning, divided according to defect of the score value to steel plate
Class.
2. according to claim 1 shine normalized steel plate defect detection method based on local light, it is characterized in that: the step
In rapid two, tetra- components of R, G, G, B in raw image data array are obtained using the CFA sample window of 2 × 2 sizes, and convert
For the grayscale image of single channel component.
3. according to claim 1 shine normalized steel plate defect detection method based on local light, it is characterized in that: the step
In rapid three, using integral multiple width square sample window to the grayscale image of single channel component carry out it is down-sampled obtain it is down-sampled
Gray level image.
4. according to claim 1 shine normalized steel plate defect detection method based on local light, it is characterized in that: the step
In rapid five, the mode for asking quotient is used are as follows: make local photometric data and greyscale image data that place be normalized using division calculation
Reason.
5. according to claim 1 shine normalized steel plate defect detection method based on local light, it is characterized in that: the step
It in rapid eight, is scored by calculating connected domain area with the quotient for surrounding frame area, defect type, bloom is distinguished according to score value
Region and shadow effect region.
6. according to claim 5 shine normalized steel plate defect detection method based on local light, it is characterized in that: using product
Divide method to calculate connected domain area and surrounds frame area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811556028.8A CN109785290B (en) | 2018-12-19 | 2018-12-19 | Steel plate defect detection method based on local illumination normalization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811556028.8A CN109785290B (en) | 2018-12-19 | 2018-12-19 | Steel plate defect detection method based on local illumination normalization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109785290A true CN109785290A (en) | 2019-05-21 |
| CN109785290B CN109785290B (en) | 2020-01-24 |
Family
ID=66497307
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811556028.8A Active CN109785290B (en) | 2018-12-19 | 2018-12-19 | Steel plate defect detection method based on local illumination normalization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109785290B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375676A (en) * | 2022-10-24 | 2022-11-22 | 微山三利特不锈钢有限公司 | Stainless steel product quality detection method based on image recognition |
| GB2611146A (en) * | 2021-09-22 | 2023-03-29 | Lenovo Beijing Ltd | Processing method and processing device |
| CN117173156A (en) * | 2023-10-23 | 2023-12-05 | 杭州百子尖科技股份有限公司 | Pole piece burr detection method, device, equipment and medium based on machine vision |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103971133A (en) * | 2014-04-13 | 2014-08-06 | 北京工业大学 | Automatic steel plate surface defect recognition method based on case-based reasoning |
| US20140307944A1 (en) * | 2013-04-11 | 2014-10-16 | Samsung Techwin Co., Ltd. | Panel inspection method and apparatus |
| CN104318556A (en) * | 2014-10-16 | 2015-01-28 | 东北大学 | Silicon steel plate surface defect image detection method under oil pollution interference |
| CN104794491A (en) * | 2015-04-28 | 2015-07-22 | 重庆大学 | Fuzzy clustering steel plate surface defect detection method based on pre classification |
| CN108181316A (en) * | 2018-03-13 | 2018-06-19 | 广西师范大学 | A kind of bamboo strip defect detection method based on machine vision |
| CN108876768A (en) * | 2018-05-30 | 2018-11-23 | 杭州舜浩科技有限公司 | Light guide plate shadow defect inspection method |
| CN108982540A (en) * | 2018-09-10 | 2018-12-11 | 五邑大学 | A kind of screw apparent flaws detection method, device and equipment |
-
2018
- 2018-12-19 CN CN201811556028.8A patent/CN109785290B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140307944A1 (en) * | 2013-04-11 | 2014-10-16 | Samsung Techwin Co., Ltd. | Panel inspection method and apparatus |
| CN103971133A (en) * | 2014-04-13 | 2014-08-06 | 北京工业大学 | Automatic steel plate surface defect recognition method based on case-based reasoning |
| CN104318556A (en) * | 2014-10-16 | 2015-01-28 | 东北大学 | Silicon steel plate surface defect image detection method under oil pollution interference |
| CN104794491A (en) * | 2015-04-28 | 2015-07-22 | 重庆大学 | Fuzzy clustering steel plate surface defect detection method based on pre classification |
| CN108181316A (en) * | 2018-03-13 | 2018-06-19 | 广西师范大学 | A kind of bamboo strip defect detection method based on machine vision |
| CN108876768A (en) * | 2018-05-30 | 2018-11-23 | 杭州舜浩科技有限公司 | Light guide plate shadow defect inspection method |
| CN108982540A (en) * | 2018-09-10 | 2018-12-11 | 五邑大学 | A kind of screw apparent flaws detection method, device and equipment |
Non-Patent Citations (6)
| Title |
|---|
| WEN-JING ZHANG ET AL.: "Automatic Printing Plate Defect Detection Based on a Simplified Homocentric Square Filter", 《2016 9TH INTERNATIONAL CONGRESS ON IMAGE AND SIGNAL PROCESSING, BIOMEDICAL ENGINEERING AND INFORMATICS》 * |
| 孙创开: "基于机器视觉的彩钢板缺陷检测和智能分类研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 崔巍 等: "基于彩色CMOS图像传感器的图像压缩预处理***的设计", 《科学技术与工程》 * |
| 王震宇: "基于机器视觉钢板表面缺陷检测技术研究", 《计算机与现代化》 * |
| 许强 等: "钎焊缺陷定量检测分析***设计", 《EQUIPMENT MANUFACTURING TECHNOLOGY》 * |
| 高如新 等: "基于钢板表面缺陷检测的图像增强研究", 《河南理工大学学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2611146A (en) * | 2021-09-22 | 2023-03-29 | Lenovo Beijing Ltd | Processing method and processing device |
| CN115375676A (en) * | 2022-10-24 | 2022-11-22 | 微山三利特不锈钢有限公司 | Stainless steel product quality detection method based on image recognition |
| CN117173156A (en) * | 2023-10-23 | 2023-12-05 | 杭州百子尖科技股份有限公司 | Pole piece burr detection method, device, equipment and medium based on machine vision |
| CN117173156B (en) * | 2023-10-23 | 2024-02-20 | 杭州百子尖科技股份有限公司 | Pole piece burr detection method, device, equipment and medium based on machine vision |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109785290B (en) | 2020-01-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI805857B (en) | System and method for characterizing a specimen | |
| CN107678192B (en) | Mura defect detection method based on machine vision | |
| JP4150390B2 (en) | Appearance inspection method and appearance inspection apparatus | |
| CN108332689A (en) | A kind of optical measuring system and method for detection surface roughness and surface damage | |
| CN111693549A (en) | Method for detecting and classifying defects of mobile phone cover plate glass | |
| CN110389127A (en) | A kind of identification of cermet part and surface defects detection system and method | |
| CN109785290A (en) | Normalized steel plate defect detection method is shone based on local light | |
| CN114719749B (en) | Metal surface crack detection and real size measurement method and system based on machine vision | |
| CN113658155A (en) | Object surface flaw detection and analysis method and device based on photometric stereo | |
| JP5433201B2 (en) | Wafer backside evaluation method | |
| JP2017040600A (en) | Inspection method, inspection device, image processor, program and record medium | |
| WO2001084126A2 (en) | Directional lighting and method to distinguish three dimensional information | |
| CN113237889A (en) | Multi-scale ceramic detection method and system | |
| CN110426395B (en) | Method and device for detecting surface of solar EL battery silicon wafer | |
| CN109622404B (en) | Automatic sorting system and method for micro-workpieces based on machine vision | |
| CN111968100A (en) | Machine vision detection method and system | |
| JP5224288B2 (en) | Surface inspection apparatus and surface inspection method | |
| CN108805854B (en) | Method for rapidly counting tablets and detecting completeness of tablets in complex environment | |
| WO2019176614A1 (en) | Image processing device, image processing method, and computer program | |
| CN116337869A (en) | Machine vision-based detection system for production of double-sided circuit board | |
| EP3974817A1 (en) | Non-lambertian surface inspection system for line scan | |
| Chang et al. | Fast and high-resolution optical inspection system for in-line detection and labeling of surface defects | |
| CN104713887B (en) | A kind of method for detecting planar disfigurement | |
| CN204649642U (en) | A kind of device of detection plane defect | |
| KR20220154345A (en) | Contour detection device using gaussian-weighted least squares and contour detection method thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20191225 Address after: 414000 north of Baoshan Road, BAISHILING Industrial Park, Yueyang Economic and Technological Development Zone, Yueyang City, Hunan Province Applicant after: Hunan Datong Testing Co., Ltd Address before: Room 102, Building 1, Incubation Park, National University Science Park, Zhengzhou High-tech Zone, Henan Province Applicant before: Liu Yongchen |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |