CN112598642A - High-speed high-precision visual detection method - Google Patents
High-speed high-precision visual detection method Download PDFInfo
- Publication number
- CN112598642A CN112598642A CN202011533384.5A CN202011533384A CN112598642A CN 112598642 A CN112598642 A CN 112598642A CN 202011533384 A CN202011533384 A CN 202011533384A CN 112598642 A CN112598642 A CN 112598642A
- Authority
- CN
- China
- Prior art keywords
- product
- different
- detection
- area
- neural network
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 40
- 230000000007 visual effect Effects 0.000 title claims abstract description 9
- 238000000034 method Methods 0.000 claims abstract description 42
- 238000013135 deep learning Methods 0.000 claims abstract description 20
- 238000000691 measurement method Methods 0.000 claims abstract description 12
- 230000008569 process Effects 0.000 claims abstract description 10
- 230000008859 change Effects 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims description 33
- 238000007747 plating Methods 0.000 claims description 18
- 229910052751 metal Inorganic materials 0.000 claims description 16
- 239000002184 metal Substances 0.000 claims description 16
- 230000007547 defect Effects 0.000 claims description 15
- 238000005259 measurement Methods 0.000 claims description 14
- 238000001746 injection moulding Methods 0.000 claims description 10
- 238000011179 visual inspection Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 5
- 238000009713 electroplating Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 239000003292 glue Substances 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- BQCADISMDOOEFD-UHFFFAOYSA-N Silver Chemical compound [Ag] BQCADISMDOOEFD-UHFFFAOYSA-N 0.000 claims description 2
- 238000005286 illumination Methods 0.000 claims description 2
- 238000003825 pressing Methods 0.000 claims description 2
- 238000006748 scratching Methods 0.000 claims description 2
- 230000002393 scratching effect Effects 0.000 claims description 2
- 229910052709 silver Inorganic materials 0.000 claims description 2
- 239000004332 silver Substances 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 22
- 238000012545 processing Methods 0.000 description 13
- 238000004519 manufacturing process Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000003708 edge detection Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Quality & Reliability (AREA)
- Neurology (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a high-speed high-precision visual detection method, and belongs to the technical field of image recognition. The method specifically comprises the following steps: counting the measuring method of a measurer after the product is changed when the product process is changed; sorting the measurement methods and establishing a table document; selecting the difference key points of the measuring method under different conditions from the generalized measuring method table document; establishing a deep learning network, and classifying and dividing products by using the network; classifying all products in the bearable difference change into the products respectively, and setting a set of optimal parameters for each product; when a new product needs to be detected, defining the product into a certain classified class through a deep learning network, and calling the classified detection parameters; and after the optimal parameters are selected, measuring by using an image recognition algorithm. The invention can improve the data accuracy of visual identification in product detection, the adaptability of software to products and the efficiency of software detection.
Description
Technical Field
The invention belongs to the technical field of image recognition, and relates to a high-speed high-precision visual detection method.
Background
The image processing technology is a technology for processing image information by a computer. The method mainly comprises the steps of image digitization, image enhancement and restoration, image data coding, image segmentation, image identification and the like. The image processing technology comprises four methods of point processing, group processing, geometric processing and frame processing.
The image processing technology has wide application fields, is mainly used for precise positioning, precise measurement and the like in factory automation application, and has the precision reaching the micron level, but the technology has the defect that the influence of the characteristic change of a product to be side can be caused, and when the characteristic of the product to be side changes to a certain extent, the image processing technology is difficult to be adaptively adjusted along with the change. Deep learning is an important breakthrough in the field of artificial intelligence in the last decade. It has been used in speech recognition, natural language processing, computer vision, image and video analysis, multimedia and other fields with great success. The existing deep learning model belongs to a neural network. The origin of neural networks dates back to the 40's of the 20 th century, once seen in the eighties and ninety. The neural network tries to solve various machine learning problems by simulating a brain cognitive mechanism, the technology is mainly used for detecting appearance defects and simple logic judgment functions needing human participation in factory automation application, the technology has strong adaptability, but the technology cannot perform accurate measurement, and the detection speed is slow.
Currently, when the above-mentioned technology is used in some product inspection processes, the inspected object may need to go through a stamping process/injection molding process/electroplating process before being inspected. The production processes are influenced by temperature/humidity/production parameters in the production process, the appearance performance of the production processes has certain differences, and the detection by using a single technology has the following problems: 1) the uncertain change of the production process can cause great differences in color, brightness, warping degree and the like of products, and the traditional image recognition technology realizes detection and measurement by setting a brightness fixed threshold in a fixed area. When the product itself is too different, the real validity of the data is questionable when the detection/measurement is performed in a fixed area by a fixed brightness threshold, which results in the need to frequently adjust the detection area and the detection parameters in the actual production process. 2) The detection of defects on the surface of a product by using a single deep learning algorithm can be very complicated due to different detection requirements of different areas. The injection molding area/metal plating area/metal non-plating area will have different requirements. The more complex the neural network system is, the more network layers are, and the more network layers are, the slower the detection efficiency is.
In addition, the conventional image recognition technology cannot accurately process and detect images with large feature changes. When the product has the problems of warping, deformation and the like, the image recognition technology cannot well track and position. In validating the measurement data, the measured reference/measured position has a large impact on the final data result. For example, in a product drawing, the outer contour of the product is a standard rectangle, the length and the width of the product need to be measured, but a plane projection image of an actual product cannot be the standard rectangle, and even the sides of the rectangle are not straight. In order to ensure the dimension of the regular graph when the irregular graph is measured in the actual drawing, in the original measuring mode, the measuring mode can be adjusted in real time by a skilled measuring person according to the knowledge of the skilled measuring person on the product. But this adjustment of random strain is not possible with conventional image recognition techniques. The image recognition technology can replace the measurement process, but cannot replace the adjustment of the qualified measurement personnel according to the change of the product. The deep learning algorithm needs longer operation time, the traditional image recognition technology is utilized to extract important features, the deep learning algorithm is enabled to operate only aiming at the important features, and the visual inspection efficiency is improved. The method is characterized in that a deep learning algorithm is simply used, the defect features can only be established in a complex neural network system, and then a large number of data information pictures are utilized for learning training (in order to avoid the mutual interference of the defect features needing to be detected in different regions). Therefore, the whole network is very large, and the detection efficiency of the deep learning system which is originally low in detection efficiency is low.
Disclosure of Invention
In view of the above, the present invention provides a high-speed and high-precision visual inspection method, which performs deep fusion between an image processing technique and a deep learning technique, so as to improve the data accuracy of visual recognition in product inspection, the adaptability of software to products, and the efficiency of software inspection.
In order to achieve the purpose, the invention provides the following technical scheme:
a high-speed high-precision visual inspection method comprises the following steps: s1: counting and analyzing the measurement method information of a measurer after the product is changed when the product process is changed, and taking the measurement method information as basic data; s2: classifying and recording the basic data, namely sorting the measurement method information and establishing a measurement method table document; s3: selecting the difference key points of the measuring method under different conditions from the generalized measuring method table document; s4: classifying the measuring method and inputting the corresponding pictures into a database, establishing a deep learning network, and classifying and dividing the product by using the network; s5: classifying all products in the bearable difference change into the products respectively, and setting a set of optimal parameters for each product; s6: when a new product needs to be detected, defining the product into a certain classified class through a deep learning network, and calling the classified detection parameters; s7: and after the optimal parameters are selected, measuring by using an image recognition algorithm to obtain a measured value and generate a data report.
Further, in step S3, the differences of the measurement methods under different conditions mainly include: 1) adjusting the illumination of the light source: when the surfaces of different batches of products have differences, different light source illuminances are adjusted to carry out measurement; 2) adjusting the measuring position: and when the shapes of different batches of products are different, adjusting different measuring positions for measuring.
Further, in the visual inspection method, the neural network system is optimized, three neural network systems are established aiming at different types of defect characteristics of three different areas of a product to be inspected, and in each single neural network system, other areas are divided and filled; each neural network system completes an independent loop of detection environment.
Further, the detection by using three independent neural network systems specifically includes:
s1: dividing the defects into three types according to the detection requirement of the appearance defects, namely an injection molding area, a metal plating area and a metal non-plating area;
s2: according to the difference of the characteristics of the three defects, the color/brightness judgment in an image recognition algorithm is utilized to carry out segmentation and extraction: 1) injection molding area: is not reflective and is black; 2) metal non-plating area: reflecting light, silvery metallic color; 3) metal plating area: reflecting light, golden metallic color;
s3: positioning the position of the product by using an image recognition algorithm to obtain the position coordinate of the product;
s4: after obtaining the position coordinates of the product, extracting a color/brightness area at the position corresponding to the product: 1) injection molding area: reserving a black area on the product, and filling other color areas with white; 2) metal non-plating area: reserving a silver area on the product, and filling other areas with black; 3) metal plating area: reserving a golden area on the product, and filling other areas with black;
s5: the pictures which are respectively extracted by using different effect segmentation are stored in different folders;
s6: different deep learning networks are respectively established for pictures in folders in different areas, and each network also has different branches including but not limited to: scratching, pressing damage, glue overflow, glue shortage, poor electroplating, dirt and foreign matters;
s7: the three neural network systems synchronously operate in different GPUs, and the operation result of each neural network is stored in the cache of the industrial personal computer;
s8: and when all the networks are operated, integrating all the results and outputting a comprehensive result.
The invention has the beneficial effects that: the method provided by the invention utilizes the combination of the traditional image recognition algorithm and the deep learning algorithm to mutually make up the defects and solve the problems of short boards which cannot be solved by using a single algorithm originally. The method can better adapt to the production condition of an automatic factory, reduces the times of on-site maintenance personnel on-line debugging parameters to a certain extent, improves the detection efficiency to a certain extent, enables the efficiency of automatic detection equipment to be higher, correspondingly increases the return on investment rate, and has wide application prospect.
Drawings
In order to make the object, technical scheme and beneficial effect of the invention more clear, the invention provides the following drawings for explanation:
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a block diagram of a detection system.
Detailed Description
The invention carries out deep fusion on an image processing technology and a deep learning technology, uses an AI technology to establish a deep learning neural network system for replacing judgment of a qualification depth measurer, firstly selects an optimal measuring mode through the AI technology of deep learning, and then uses an image recognition technology to carry out accurate measurement. Meanwhile, before the neural network is constructed, the image processing technology is utilized to segment/fill the picture, and a complex neural network is decomposed into a plurality of relatively simple neural network systems, so that the operation efficiency of the neural network systems is improved.
Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Fig. 1 is a schematic flow chart of the method of the present invention, and fig. 2 is a block diagram of a detection system, as shown in the figure, the method provided by the present invention is used for product visual detection to improve detection accuracy, and the specific embodiment is as follows:
s1: counting and analyzing the measurement method information of a measurer after the product is changed when the product process is changed;
s2: the information is induced into 4 categories, and product image information of different categories is induced into a certain category by using a deep learning classification tool;
s3: a large number of pictures of 4 categories are used for constructing a neural network system, so that new products can be accurately summarized into the 4 categories when being identified;
class 1: plating darker, unbent products
Class 2: plated brighter, unbent products
Class 3: plating darker, slightly curved products
Class 4: plated brighter, slightly curved products
S4: according to the difference of the categories, different detection parameters are set:
1. for darker plated products, the edge detection threshold was set to 80
2. For products that are more brightly plated, the edge detection threshold is set to 120
3. The spacing of each measurement position was 0.12mm for the unbent product
4. The spacing of each measurement position was 0.115mm for a slightly curved product.
Before the mode is used, if no person follows to change parameters, the measured data of the automatic detection equipment is compared with the measured data of the measuring chamber, and after the data are compared, the difference of the measured data of the products in the same batch is less than 0.005 mm. However, the difference of the measured data of different batches of products can reach 0.02mm at most, and the difference of the measured data of 0.02mm is not acceptable. After the method is used, the difference of the measurement data of different batches of products is less than 0.01mm, and the difference of the measurement data of 0.01mm can well meet the actual requirement.
The technical scheme provided by the invention can achieve a good effect in the aspect of improving the processing speed, and the specific embodiment is as follows:
s1: according to the difference of the characteristics of the three defects, the color/brightness judgment in an image recognition algorithm is utilized to carry out segmentation and extraction:
1) injection molding area: is not reflective and is black; 2) metal non-plating area: reflecting light, silvery metallic color;
3) metal plating area: reflecting light, golden metallic color;
s2: and respectively establishing a neural network system for each area.
S3: during detection, the neural networks of the electroplating area and the non-electroplating area are distributed to the display card A for operation, and the neural network of the injection molding area is distributed to the display card B for operation
S4: and integrating the operation results and outputting the final detection result after all operations are finished.
Before the method is used, all areas are trained into a neural network system, and about 3000 pictures are needed to construct the neural network system with a good detection effect. After the method is used, only 1000 pictures are needed for constructing a neural network system with a good detection effect. When the neural network is trained and constructed, the improved efficiency is about 60 percent.
Before the method is used, only a single display card can be used for operation due to only one neural network system during actual detection. The time required to test a product was 0.8S. By using the method, the detected area is reduced, and the two display cards can be used for synchronous operation. The speed is increased to 0.25S. The speed is greatly increased, the production period of the equipment in the original 1S production period is increased to 0.45S, and the benefit can be obviously improved.
Finally, it is noted that the above-mentioned preferred embodiments illustrate rather than limit the invention, and that, although the invention has been described in detail with reference to the above-mentioned preferred embodiments, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the scope of the invention as defined by the appended claims.
Claims (4)
1. A high-speed high-precision visual detection method is characterized in that: the method comprises the following steps:
s1: counting and analyzing the measurement method information of a measurer after the product is changed when the product process is changed, and taking the measurement method information as basic data;
s2: classifying and recording the basic data, namely sorting the measurement method information and establishing a measurement method table document;
s3: selecting the difference key points of the measuring method under different conditions from the generalized measuring method table document;
s4: classifying the measuring method and inputting the corresponding pictures into a database, establishing a deep learning network, and classifying and dividing the product by using the network;
s5: classifying all products in the bearable difference change into the products respectively, and setting a set of optimal parameters for each product;
s6: when a new product needs to be detected, defining the product into a certain classified class through a deep learning network, and calling the classified detection parameters;
s7: and after the optimal parameters are selected, measuring by using an image recognition algorithm to obtain a measured value and generate a data report.
2. The high-speed high-precision visual inspection method according to claim 1, characterized in that: in step S3, the important differences of the measurement methods under different conditions mainly include:
1) adjusting the illumination of the light source: when the surfaces of different batches of products have differences, different light source illuminances are adjusted to carry out measurement;
2) adjusting the measuring position: and when the shapes of different batches of products are different, adjusting different measuring positions for measuring.
3. The high-speed high-precision visual inspection method according to claim 2, characterized in that: in the visual detection method, a neural network system is optimized, three neural network systems are established aiming at different types of defect characteristics of three different areas of a product to be detected, and in each single neural network system, other areas are divided and filled; each neural network system completes an independent loop of detection environment.
4. A high-speed high-precision visual inspection method according to claim 3, characterized in that: the detection by using three independent neural network systems specifically comprises the following steps:
s1: dividing the defects into three types according to the detection requirement of the appearance defects, namely an injection molding area, a metal plating area and a metal non-plating area;
s2: according to the difference of the characteristics of the three defects, the color/brightness judgment in an image recognition algorithm is utilized to carry out segmentation and extraction: 1) injection molding area: is not reflective and is black; 2) metal non-plating area: reflecting light, silvery metallic color; 3) metal plating area: reflecting light, golden metallic color;
s3: positioning the position of the product by using an image recognition algorithm to obtain the position coordinate of the product;
s4: after obtaining the position coordinates of the product, extracting a color/brightness area at the position corresponding to the product: 1) injection molding area: reserving a black area on the product, and filling other color areas with white; 2) metal non-plating area: reserving a silver area on the product, and filling other areas with black; 3) metal plating area: reserving a golden area on the product, and filling other areas with black;
s5: the pictures which are respectively extracted by using different effect segmentation are stored in different folders;
s6: different deep learning networks are respectively established for pictures in folders in different areas, and each network also has different branches including but not limited to: scratching, pressing damage, glue overflow, glue shortage, poor electroplating, dirt and foreign matters;
s7: the three neural network systems synchronously operate in different GPUs, and the operation result of each neural network is stored in the cache of the industrial personal computer;
s8: and when all the networks are operated, integrating all the results and outputting a comprehensive result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011533384.5A CN112598642B (en) | 2020-12-22 | High-speed high-precision visual detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011533384.5A CN112598642B (en) | 2020-12-22 | High-speed high-precision visual detection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112598642A true CN112598642A (en) | 2021-04-02 |
CN112598642B CN112598642B (en) | 2024-05-10 |
Family
ID=
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5544256A (en) * | 1993-10-22 | 1996-08-06 | International Business Machines Corporation | Automated defect classification system |
CN104778684A (en) * | 2015-03-06 | 2015-07-15 | 江苏大学 | Method and system thereof for automatically measuring, representing and classifying heterogeneous defects on surface of steel |
CN104850858A (en) * | 2015-05-15 | 2015-08-19 | 华中科技大学 | Injection-molded product defect detection and recognition method |
US20180253836A1 (en) * | 2015-06-16 | 2018-09-06 | South China University Of Technology | Method for automated detection of defects in cast wheel products |
KR20190032908A (en) * | 2017-09-20 | 2019-03-28 | 주식회사 에이치엔에스휴먼시스템 | Method for managing steel quality and system |
CN109636772A (en) * | 2018-10-25 | 2019-04-16 | 同济大学 | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning |
CN110766660A (en) * | 2019-09-25 | 2020-02-07 | 上海众壹云计算科技有限公司 | Integrated circuit defect image recognition and classification system based on fusion depth learning model |
CN110930347A (en) * | 2018-09-04 | 2020-03-27 | 京东方科技集团股份有限公司 | Convolutional neural network training method, and method and device for detecting welding spot defects |
CN111696075A (en) * | 2020-04-30 | 2020-09-22 | 航天图景(北京)科技有限公司 | Intelligent fan blade defect detection method based on double-spectrum image |
CN111951237A (en) * | 2020-08-04 | 2020-11-17 | 上海微亿智造科技有限公司 | Visual appearance detection method |
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5544256A (en) * | 1993-10-22 | 1996-08-06 | International Business Machines Corporation | Automated defect classification system |
CN104778684A (en) * | 2015-03-06 | 2015-07-15 | 江苏大学 | Method and system thereof for automatically measuring, representing and classifying heterogeneous defects on surface of steel |
CN104850858A (en) * | 2015-05-15 | 2015-08-19 | 华中科技大学 | Injection-molded product defect detection and recognition method |
US20180253836A1 (en) * | 2015-06-16 | 2018-09-06 | South China University Of Technology | Method for automated detection of defects in cast wheel products |
KR20190032908A (en) * | 2017-09-20 | 2019-03-28 | 주식회사 에이치엔에스휴먼시스템 | Method for managing steel quality and system |
CN110930347A (en) * | 2018-09-04 | 2020-03-27 | 京东方科技集团股份有限公司 | Convolutional neural network training method, and method and device for detecting welding spot defects |
CN109636772A (en) * | 2018-10-25 | 2019-04-16 | 同济大学 | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning |
CN110766660A (en) * | 2019-09-25 | 2020-02-07 | 上海众壹云计算科技有限公司 | Integrated circuit defect image recognition and classification system based on fusion depth learning model |
CN111696075A (en) * | 2020-04-30 | 2020-09-22 | 航天图景(北京)科技有限公司 | Intelligent fan blade defect detection method based on double-spectrum image |
CN111951237A (en) * | 2020-08-04 | 2020-11-17 | 上海微亿智造科技有限公司 | Visual appearance detection method |
Non-Patent Citations (4)
Title |
---|
LI BIJIANG ET AL.: "Deep Learning: Excellent Method at Surface Defect Detection of Industrial Products", 《2019 IEEE 3RD ADVANCED INFORMATION MANAGEMENT, COMMUNICATES, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE (IMCEC)》, 6 February 2020 (2020-02-06), pages 712 - 716 * |
ZHIXUAN ZHAO ET AL.: "A Surface Defect Detection Method Based on Positive Samples", 《PRICAI 2018: TRENDS IN ARTIFICIAL INTELLIGENCE》, vol. 11013, 27 July 2018 (2018-07-27), pages 473 - 481, XP047481215, DOI: 10.1007/978-3-319-97310-4_54 * |
李少波等: "缺陷检测技术的发展与应用研究综述", 《自动化学报》, vol. 46, no. 11, 24 November 2020 (2020-11-24), pages 2319 - 2336 * |
李欣遥: "基于深度学习的工件微小缺陷诊断方法研究", 《中国优秀硕士学位论文全文数据库(信息科技辑)》, 15 July 2019 (2019-07-15), pages 138 - 1310 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110672617B (en) | Method for detecting defects of silk-screen area of glass cover plate of smart phone based on machine vision | |
CN104794491B (en) | Based on the fuzzy clustering Surface Defects in Steel Plate detection method presorted | |
CN106875373B (en) | Mobile phone screen MURA defect detection method based on convolutional neural network pruning algorithm | |
CN111862064B (en) | Silver wire surface flaw identification method based on deep learning | |
CN109636772A (en) | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning | |
CN109900706A (en) | A kind of weld seam and weld defect detection method based on deep learning | |
CN108355987B (en) | A kind of screen printing of battery quality determining method based on piecemeal template matching | |
CN106127779B (en) | The defect inspection method and system of view-based access control model identification | |
CN109993094A (en) | Fault in material intelligent checking system and method based on machine vision | |
CN109934802A (en) | A kind of Fabric Defects Inspection detection method based on Fourier transformation and morphological image | |
CN104700092B (en) | A kind of small characters digit recognition method being combined based on template and characteristic matching | |
CN111402197B (en) | Detection method for colored fabric cut-parts defect area | |
CN108596880A (en) | Weld defect feature extraction based on image procossing and welding quality analysis method | |
CN104992449A (en) | Information identification and surface defect on-line detection method based on machine visual sense | |
CN113221889B (en) | Chip character anti-interference recognition method and device | |
CN109540925B (en) | Complex ceramic tile surface defect detection method based on difference method and local variance measurement operator | |
CN116188475B (en) | Intelligent control method, system and medium for automatic optical detection of appearance defects | |
CN109829895A (en) | A kind of AOI defect inspection method based on GAN | |
CN109145846A (en) | Material microstructure intelligent recognition analysis system and analysis method | |
CN110473184A (en) | A kind of pcb board defect inspection method | |
CN110009019A (en) | Magnetic material crackle intelligent checking system and method | |
CN109085178A (en) | A kind of accurate on-line monitoring method of defect fingerprint and feedback strategy for increasing material manufacturing | |
CN109840899A (en) | A kind of roughness grade number recognition methods based on depth convolutional neural networks | |
CN106340007A (en) | Image processing-based automobile body paint film defect detection and identification method | |
CN109241948A (en) | A kind of NC cutting tool visual identity method and device |
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 | ||
GR01 | Patent grant |