WO2016194210A1 - 欠陥画像分類装置および欠陥画像分類方法 - Google Patents
欠陥画像分類装置および欠陥画像分類方法 Download PDFInfo
- Publication number
- WO2016194210A1 WO2016194210A1 PCT/JP2015/066244 JP2015066244W WO2016194210A1 WO 2016194210 A1 WO2016194210 A1 WO 2016194210A1 JP 2015066244 W JP2015066244 W JP 2015066244W WO 2016194210 A1 WO2016194210 A1 WO 2016194210A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- defect
- image
- classification
- unit
- class
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
- G01N21/9501—Semiconductor wafers
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N23/00—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00
- G01N23/22—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by measuring secondary emission from the material
- G01N23/225—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by measuring secondary emission from the material using electron or ion
- G01N23/2251—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by measuring secondary emission from the material using electron or ion using incident electron beams, e.g. scanning electron microscopy [SEM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/98—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
- G01N21/956—Inspecting patterns on the surface of objects
- G01N21/95607—Inspecting patterns on the surface of objects using a comparative method
- G01N2021/95615—Inspecting patterns on the surface of objects using a comparative method with stored comparision signal
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2223/00—Investigating materials by wave or particle radiation
- G01N2223/60—Specific applications or type of materials
- G01N2223/611—Specific applications or type of materials patterned objects; electronic devices
- G01N2223/6116—Specific applications or type of materials patterned objects; electronic devices semiconductor wafer
Definitions
- the present invention relates to a defect image classification apparatus and a defect image classification method for automatically classifying an image obtained by imaging defects generated during the manufacture of a semiconductor wafer.
- An appearance inspection apparatus is an apparatus that inspects a wafer using optical means or an electron beam and outputs position coordinates of detected defects. Since the appearance inspection device needs to inspect the entire wafer surface at high speed, the pixel resolution of the detected image is lowered to the extent that defect detection is possible, and the amount of image data per unit area is reduced to shorten the inspection time. . For this reason, it is difficult to observe the defect in detail using the detection image of the appearance inspection apparatus.
- the defect observation apparatus is an apparatus that images a position with high resolution based on defect coordinates obtained from an appearance inspection apparatus and outputs a captured image. Due to the miniaturization of the semiconductor manufacturing process, a defect observation apparatus SEM using a scanning electron microscope (SEM) is widely used for this observation.
- the defect observation SEM has an ADR (Automatic Defect Review) function that automatically captures and collects defect images on a wafer based on defect coordinates obtained from an appearance inspection apparatus, and an ADC (Automatic Review) that automatically classifies the collected images.
- ADR Automatic Defect Review
- ADC Automatic Review
- Patent Document 1 discloses a method for monitoring ADC learning data during production application.
- Defective classes have various classes even if they are limited to one classification process, and there are cases in which variations in shape and brightness are included in one class. For this reason, it is practically difficult to apply ADC to all processes. For this reason, the classification process is divided into an ADC application process and a visual classification process, or a defect class that uses an ADC result in one classification process and a defect class that is not used, and only a defect class that does not use an ADC result.
- a combination of ADC and visual classification, such as visual classification has also been proposed.
- Patent Document 2 discloses an ADC recipe setting method based on the premise that ADC and visual classification are used together.
- Patent Document 1 in order to stably maintain the ADC classification performance at the time of mass production application, the feature amount of the defect image to be classified by the ADC is compared with the feature amount registered in the learning data, In other words, a technique is disclosed that can stably operate without degrading the ADC classification performance even when mass production is applied by detecting a statistical change between the image and the learning data and instructing learning data update.
- Patent Document 1 discloses a method of using an ADC classification result or a visual classification result as a defect class of an image to be subjected to ADC.
- using the classification result of the ADC subject to performance evaluation cannot eliminate the erroneous classification of the ADC in the first place, and the visual classification also has a classification accuracy rate of 60 at most as described in Non-Patent Document 1. In any case, it is difficult to evaluate the performance of an appropriate ADC.
- Patent Document 2 is based on the premise that the defect class suitable for ADC and the non-suitable defect class are separated, and the defect class unsuitable for ADC is confirmed by visual classification.
- Visual classification is performed from a plurality of ADC recipes prepared in advance.
- a technique for optimizing ADC parameters by selecting a recipe that can reduce the number of defect images to be rotated is disclosed.
- the defect classification work in the combined use of the ADC and the visual classification both stable operation of the ADC and reduction of the visual classification man-hour are achieved.
- the visual classification result is used at the time of ADC parameter optimization, and the influence of the visual classification having a low classification accuracy rate cannot be avoided.
- the explanation was advanced with the words “learning data”, “ADC recipe”, and “ADC parameter” used in each document. These may be interpreted as data necessary for operating the ADC in a broad sense, but more accurately, the ADC parameter is a parameter related to image processing necessary for calculating the image feature amount, learning.
- Data can be defined as a parameter set used by an ADC classification algorithm derived from a feature amount of a teaching image, and an ADC recipe can be defined as a data set for operating an ADC including both ADC parameters and learning data.
- the object of the present invention is to solve the problems of conventional visual classification in the defect classification work in the combined use of ADC and visual classification, and then use ADC and visual classification or ADC and other classification devices in combination for reliability. It is an object of the present invention to provide a defect image classification device and a defect image classification method that enable high ADC performance evaluation and update of ADC learning data.
- an apparatus for classifying defect images includes a storage unit for storing defect images obtained by imaging with other imaging means and a plurality of other defect classification means.
- An image selection unit that selects an image from defect images stored in the storage unit using information on the defect class that classifies the defect, and an image classification unit that classifies the image selected by the image selection unit based on the classification recipe
- a classification performance evaluation unit that evaluates the classification performance of the image classification unit based on the result of classifying the image, and a selection performed by the image selection unit when the result evaluated by the classification performance evaluation unit does not reach a preset standard.
- a learning update unit that updates the classification recipe of the image classification unit using an image.
- an apparatus for classifying defect images includes a storage unit for storing defect images obtained by imaging with other imaging means, and a plurality of other defect classifications.
- An image selection unit that selects an image from defect images stored in the storage unit using information on the defect class classified by the means, and an image classification unit that classifies the image stored in the storage unit based on the classification recipe
- a learning update unit that updates the classification recipe of the image classification unit using the image selected by the image selection unit.
- a defect image obtained by imaging with another imaging means is stored in a storage unit, and a plurality of other defect classifications are performed.
- the image selection unit selects an image from the defect images stored in the storage unit using the defect class information obtained by classifying the defect by means, and the image selection unit selects the image selected based on the classification recipe.
- the classification performance evaluation unit evaluates the classification performance of the image classification unit based on the result of classification and image classification, When the result evaluated by the classification performance evaluation unit does not reach a preset standard, the learning update unit updates the classification recipe of the image classification unit using the image selected by the image selection unit.
- a defect image obtained by imaging with another imaging means is stored in a storage unit, and a plurality of other defect classifications are performed.
- the image selection unit selects an image from the defect images stored in the storage unit, and the image classification unit classifies the image stored in the storage unit based on the classification recipe.
- the learning update unit updates the classification recipe of the image classification unit using the image selected by the image selection unit.
- the present invention it is possible to evaluate the performance of a highly reliable ADC with data having few defect class errors by comparing the results of a plurality of visual classification or classification devices other than the ADC device. Further, by performing ADC learning or learning update with this data, the ADC classification performance can be maintained and improved.
- the present invention classifies selected images using defect class information obtained by classifying defects by a plurality of other defect classification means, and performs classification performance based on the results.
- the image classification recipe is updated when the evaluated result does not reach a preset standard.
- the present invention also relates to a method and apparatus for classifying a defect image, and the defect image stored in the storage unit by the image selection unit using the defect class information classified by a plurality of other defect classification means.
- An image is selected from among the images, and the classification update of the image classification unit is updated by the learning update unit using the selected image.
- FIG. 1 shows an automatic defect classification (ADC) apparatus 100 according to the present invention.
- FIG. 1 also shows a defect image capturing device 102, a yield management system 103, and a visual classification (MDC) device 104 that exchange information with the ADC device 100 via the network 101.
- a plurality of MDC devices 104 are connected to the network.
- the ADC device 100, the defect image capturing device 102, the yield management system 103, and the MDC device 104 are connected to a network.
- other means such as a portable memory device can be used. It doesn't matter.
- the defect image capturing apparatus 102 captures an image of a defect position detected by an appearance inspection apparatus (not shown) at a high magnification and captures the appearance of the defect.
- the defect image capturing apparatus 102 is an optical or SEM (Scanning Electron Microscope). Microscope) type.
- An SEM type is used for a fine device, has a function of automatically capturing an image of a defect position detected by an appearance inspection apparatus, and is called a defect observation SEM or the like.
- the yield management system 103 includes defect coordinates output from an appearance inspection apparatus (not shown), defect images output from the defect image capturing apparatus 102, defect classes (defect types) output from the ADC apparatus 100 and the MDC apparatus 104.
- a defect coordinate is transmitted in response to a request from the defect image capturing apparatus 102, and a defect image is transmitted in response to a request from the ADC apparatus 100 and the MDC apparatus 104.
- the MDC device 104 is a device in which an operator classifies defect images and assigns defect class information to the defect images.
- a defect image is received from the ADC device 100, the defect image capturing device 102, or the yield management system 103, a defect class is assigned by the operator, and the defect class information is transmitted to the ADC device 100 or the yield management system 103. It is assumed that a plurality of MDC devices 104 are connected from MDC 1 to N.
- the MDC apparatus 104 as an example of a defect classification apparatus different from the ADC apparatus 100.
- any apparatus other than the MDC apparatus can be used as long as it can assign a defect class to a defect captured in a defect image. It may be a defect classification device or a defect analysis device.
- Defect image data, defect class information, and the like are transmitted / received to / from the defect image capturing apparatus 102, the yield management system 103, and the MDC apparatus 104 via the data transmitting / receiving unit 110.
- the ADC device 100 includes a data transmission / reception unit 110, a storage unit 111, a defect class comparison unit 112, an image selection unit 113, a classification performance evaluation unit 114, an image classification unit 115, a learning update unit 116, an input / display terminal 117, and a bus 118. I have.
- the storage unit 111 stores defect image data, defect class information, and the like.
- the defect class comparison unit 112 compares a plurality of defect classes with respect to the same defect image obtained by other than the ADC device 100, and compares the defect class obtained as a result of the comparison with the defect class assigned by the image classification unit 115.
- the image selection unit 113 selects a defect image from the defect images stored in the storage unit 111 based on the comparison defect of the defect class comparison unit 112.
- the classification performance evaluation unit 114 evaluates the classification performance of the image classification unit 115 based on the comparison between the defect class obtained as a result of the comparison by the defect class comparison unit 112 and the defect class given by the image classification unit 115. It is.
- the learning update unit 116 updates the ADC processing recipe executed by the image classification unit 115 based on the evaluation result of the classification performance of the image classification unit 115 evaluated by the classification performance evaluation unit 114.
- the input / display terminal 117 displays processing contents and accepts an operator's set value input and the like.
- the bus 118 includes a data transmission / reception unit 110, a storage unit 111, a defect class comparison unit 112, an image selection unit 113, a classification performance evaluation unit 114, an image classification unit 115, a learning update unit 116, and a display terminal 117. Information transmission / reception is performed between them.
- the ADC device 100 may be mounted on any of the defect image capturing device 102, the yield management system 103, or the visual classification device 104.
- a defect image is determined (S200).
- the defect image here is one or more images to be evaluated through this flow.
- This defect image may be an image captured by the defect image capturing apparatus 102 or an image registered in the yield management system 103.
- a defect image is an image obtained from one or a plurality of wafers.
- the defect image is transmitted to a plurality of MDC devices 104 (S201).
- MDC 1 to N the defect images are classified by an operator, and a defect class is assigned.
- the assigned defect class information is received from the MDC device 104 via the data transmission / reception unit 110 and stored in the storage unit 111 (S202).
- the defect classes assigned by the operator in each MDC device 104 (MDCM1 to N) are compared (S203). A comparison method will be described with reference to FIG.
- FIG. 3 shows information displayed on the screen 300 of the display terminal 116.
- the MDC devices 1, 2, 3 displayed in the MDC device column 302 are shown.
- , 4 and 5 are shown in the classification result column 310.
- A, B, and C in the table represent defect classes.
- the defect ID 1 of the defect ID column 301 the information of the defect class B is received only from the MDC device 2 displayed in the MDC device column 302, the other MDC devices are the defect class A, and the MDC devices 1, 2, The defect classes of 3, 4, and 5 do not match.
- the defect ID 2 in the defect ID column 301 has received information of defect class A from all the MDC devices displayed in the MDC device column 302, and the defect classes of the MDC devices 1, 2, 3, 4, and 5 are Match.
- ⁇ is marked in the “match” column 303 of each defect ID in the classification result column 310, and in the “mismatch” column 304 in the case of mismatch.
- a defect ID with a matching defect class has a higher reliability of the defect class by MDC than a defect ID with a mismatched defect class.
- FIG. 3 shows five MDC devices, this is for convenience of explanation, and any number may be used as long as there are a plurality of MDC devices.
- the defect class described in the item of the majority decision 305 in the classification result column 310 is the defect class that has obtained the largest number of votes in the MDC for the defect image.
- the screen 300 may be displayed with at least one selected from the selection buttons of match 321, mismatch 322, and majority 323. Only information corresponding to the MDC defect class or only information corresponding to the mismatch ID may be displayed.
- the defect class assigned by the MDC device 104 (MDCs 1 to N) is read from the storage unit 111 to the defect class comparison unit 112 and compared (S203). A defect number with a matching defect class is identified by comparison, and a defect image corresponding to the identified defect number is selected by the image selection unit 113 (S204).
- the defect image selected by the image selection unit 113 is automatically classified by the image classification unit 115 (S205), and the classification performance evaluation unit 114 determines the defect class obtained by the automatic classification by the image classification unit 115 and the selected image.
- the ADC is evaluated using the MDC defect class (S206). The steps from S200 to S206 are collectively referred to as S210 for reference in FIG. If the defect image selected by the image selection unit 113 has already been automatically classified by the image classification unit 115, S205 can be omitted.
- FIG. 4 shows a first example of ADC performance evaluation executed by the classification performance evaluation unit 114.
- a vertical column 401 is a defect class assigned by the MDC
- a horizontal column 402 is a defect class assigned by the ADC.
- the performance of the defect class “A” by the ADC is Purity performance 403, that is, (True Positive number for the ADC defect class “A”) / (True Positive number and False Positive number for the ADC defect class “A”).
- the defect class is “A” by ADC, this can be interpreted as being 92% reliable, which is an index of reliability.
- the total accuracy rate is displayed.
- the total accuracy rate shown in the column 405 in FIG. 4 is compared with a preset value registered in advance (S207). If the overall accuracy rate is equal to or lower than the set value, that is, if the classification performance is low, the ADC is determined by the image selected in S204. Learning is performed (S208). When the total accuracy rate is equal to or less than the set value, an alarm indicating that the total accuracy rate is equal to or less than the set value is displayed on the display terminal 116 in FIG. The process may proceed to ADC learning update (S208).
- the ADC learning update is adjustment of an ADC processing recipe.
- FIG. 5 shows a method of comparing the first ADC performance value and the set value different from S207 and S208.
- S210 represents steps in the same range as S210 shown in FIG. 2, that is, steps S200 to S206.
- the Purity 403 shown in FIG. 4 is compared with the set value for each defect class (S500), and the defect class corresponding to the Purity below the set value (defect class with low classification performance) is selected from the image selected in S204.
- the learning update of the ADC is performed using only the image of () (S501). If there is a defect class corresponding to the Purity below the set value, the display terminal 116 is notified, and if there is an input instruction from the user, the process may proceed to ADC learning update (S501).
- FIG. 6 shows a method of comparing the second ADC performance value and the set value different from S207 and S208.
- Accuracy 404 shown in FIG. 4 is compared with the set value for each defect class (S600), and the defect class corresponding to Accuracy below the set value (defect class with low classification performance) is selected from the image selected in S204.
- the ADC learning update is performed using only the image of () (S601). If there is a defect class corresponding to Accuracy below the set value, the notification may be issued to the display terminal 116, and if there is an input instruction from the user, the process may proceed to ADC learning update (S601).
- Fig. 7 shows a second example of ADC performance evaluation.
- the table of FIG. 7 is similar to the table described with reference to FIG. 4, in which the vertical column 701 is the defect class due to MDC, the horizontal column 702 is the defect class due to ADC, the bottom column of the horizontal column is Purity 703, the vertical column Accuracy 704 is displayed at the right end of the column, and the total accuracy rate is displayed in the lower right column 705.
- an unkown class (unknown defect class) 7021 is added to the defect class of the ADC in the horizontal column 702 as compared to the table of FIG.
- the unkown class 7021 is a class provided for assignment when the ADC determines a boundary case of the learned defect class or a case that does not correspond to the learned defect class.
- the unkown class 7021 is explicitly identified by the ADC and is not reflected in the calculation of the accuracy 704.
- Fig. 8 shows the ADC learning update method using unkown.
- the number of images determined by the ADC as unkown for each defect class of MDC is compared with a preset value registered (S800), and if there is an MDC defect class whose unkown number is greater than or equal to the set value.
- the learning update of the ADC is performed using the image determined to be unkown in the corresponding MDC defect class from the image selected in S204 (S801).
- the learning update of the ADC may be performed on all the images of the corresponding MDC defect class.
- FIG. 9 shows an example when a new defect class D appears.
- the configuration of the table shown in FIG. 9 is the same as the configuration described with reference to FIG. 7.
- the vertical column 901 has a defect class due to MDC
- the horizontal column 902 has a defect class due to ADC
- Is Purity 903, Accuracy 904 is displayed at the right end of the vertical column, and the total accuracy rate is displayed in the lower right column 905.
- MDC in the vertical column 901 of the table shown in FIG. 9
- a new defect class can be identified as a new class D9011, but since it is not taught in the ADC in the horizontal column 902, it becomes an unkown class 9021.
- the defect class D of the MDC in the vertical column 901 is D9011, and the defect class D is used as the defect class D for the image that is the unkown class 9021 in the ADC of the horizontal column 902, so that the defect class D is also added to the ADC. It becomes possible to register.
- FIG. 10 After performing the processing of S210 described in FIG. 2, the cumulative number of learning images of each defect class is recorded, the cumulative number of images and the set value are compared (S1000), If the number of images is equal to or less than the set value, ADC learning is updated using all or part of the images of the corresponding MDC defect class from the images selected in S204 (S1001). By this method, it is possible to eliminate a defect class having insufficient learning images. If there is a defect class with the cumulative image number equal to or less than the set value, the display terminal 116 is notified, and if there is an input instruction from the user, the process may proceed to ADC learning update (S1001).
- the defect class may be a defect class obtained as a result of majority vote. If the majority result is used, all images determined in S200 can be used.
- Fig. 11 shows a method that considers weighting as a method for comparing defect classes.
- the table shown in FIG. 11 includes a column 1101 for displaying defect IDs, a column 1102 for displaying MDC results for each MDC device, a column 1103 for displaying weights, and a column 1104 for displaying the majority result for each defect ID.
- a column 1105 for displaying the ADC result is provided for each defect ID.
- FIG. 11 shows the classification results of the defect numbers 1, 2, 3, and 4 in the defect ID column 1101 in the MDC devices 1, 2, 3, 4, and 5 in the MDC device column 1102.
- the number of defects is set to 4, but the number of defects is not limited to this.
- A, B, and C in the table represent defect classes.
- the way of viewing the defect classes in the defect ID column 1101 and the MDC device column 1102 is the same as in FIG.
- a weighting column 1103 is displayed taking into account the concept of weighting.
- the defect ID 1 of the defect ID column 1101 there are three determinations of the defect class A in the MDC of the MDC device column 1102, and two B. This determination number is used as it is as a weighting value in the weighting column 1103.
- the defect IDs 2, 3, and 4 can be summarized as shown in FIG.
- the defect class written in the majority decision column 1104 in the table of FIG. 11 is determined by the majority decision method described in FIG. Although the class is uniquely determined, it is not known that defect IDs 1 and 4 are in competition with other defect classes.
- FIG. 12 shows a case where ADC performance is evaluated in consideration of weighting.
- the configuration of the table shown in FIG. 12 is the same as that described with reference to FIG. 4.
- the vertical column 1201 is the defect class assigned by the MDC
- the horizontal column 1202 is the defect class assigned by the ADC
- the bottom The column shows Purity performance 1203, the position version fixture column shows Accuracy performance 1204, and the column 1205 in the lower right corner displays the total accuracy rate.
- the defect ID 1 in the defect ID column 1101 has three MDC defect classes A, two B, and the ADC defect class A
- “MDC defect class A / ADC defect” in the matrix of FIG. 3 votes for “Class A” and 2 votes for “MDC defect class B / ADC defect class A”.
- FIG. 12 shows the result obtained by performing the same processing on the defect IDs 2, 3, and 4.
- FIG. 13 shows the result of evaluation using the defect class determined as a result of majority vote.
- the configuration of the table shown in FIG. 13 is the same as that described with reference to FIG. 4.
- the vertical column 1301 is a defect class assigned by the MDC
- the horizontal column 1302 is a defect class assigned by the ADC
- the bottom row The Purity performance 1303 is displayed in the column
- the Accuracy performance 1304 is displayed in the column of position version fixtures
- the total accuracy rate is displayed in the column 1305 in the lower right corner.
- the ADC is 100% correct.
- the execution timing during production application of the ADC performance evaluation and learning data update shown in FIGS. 2 to 10 is not only explicitly started by the operator, but also by a method based on the number of classified images into the ADC unkown class, or periodically There are automatic startup methods such as implementation.
- FIG. 14 a processing method in which the ADC performance evaluation is not performed and the MDC result is reflected in the learning of the ADC is shown in FIG. 14 as the second embodiment. This will be described with reference to the drawings.
- ADC is evaluated for an image selected based on the MDC result, and the ADC performance is evaluated.
- the image is selected based on the MDC result.
- the ADC was learned and updated based on the image.
- ADC performance evaluation is not performed, and the ADC is immediately learned and updated based on the image selected based on the MDC result.
- the configuration of the automatic defect classification apparatus in the present embodiment is the configuration described with reference to FIG. 1 in the first embodiment except for the classification performance evaluation unit 114, and the other configuration is the same as described in FIG. Since this is the same, the description is omitted.
- the defect number with the matching defect class is identified by the comparison method described with reference to FIG. 3, and the defect image corresponding to the defect number identified in S1404 is selected by the image selection unit 113.
- the learning update unit 116 performs ADC learning update using the image selected by the image selection unit 113, and the ADC processing recipe executed by the image classification unit 115 is adjusted.
- the result of the majority decision from the plurality of defect classification means or the plurality of defect classification results is automatically reflected in the database of the automatic defect classification unit, so that the conventional defect determination result is correct.
- the database of the automatic defect classification device is based on the result of majority decision based on the exact opposite view that there are individual differences in humans and the results vary from person to person.
- processing is performed on the premise that MDC is first performed on a defect image.
- ADC is first performed on the defect image as a preceding stage, and as a result, the unknown cannot be classified.
- a process of executing MDC only when the number of defective defects is equal to or greater than a set value, and periodically will be described.
- the configuration of the automatic defect classification apparatus in the present embodiment is the same as the configuration described with reference to FIG.
- FIG. 15 shows a processing flow in the case where activation is performed according to the number of classified images into the unkown class of the ADC in this embodiment.
- the defect image is received by the ADC device 100 from the defect image capturing device 102 or the yield management system 103 (S1500), and the received defect image is classified by the image classification unit 115 of the ADC device 100 (S1501).
- the number of images of the unkown class obtained as a result of the classification is compared with a preset value registered (S1502). If the number of images of the unkown class is equal to or larger than the set value, FIGS.
- the performance evaluation and additional learning of the ADC described above are performed (S1503).
- the ADC process in S205 of FIG. 2 may be skipped. Further, S1503 may be executed using an image different from the image received in S1500.
- Fig. 16 shows the processing flow for regular startup.
- the current time managed by the ADC device is compared with a preset time registered in advance (S1600), and when the current time reaches the set time, the defect image is picked up from the defect image pickup device 102 or the yield management system 103.
- Is received by the ADC device 100 S1601
- the ADC performance evaluation and additional learning described in the first embodiment with reference to FIGS. 2 to 10 are performed (S1602).
- the set time to be registered the elapsed time from the previous ADC performance evaluation / additional learning is registered, and the registered elapsed time is added to the previous ADC performance evaluation / additional learning time as the set time Also good. Thereby, periodic performance evaluation and learning update of ADC can be performed.
- the defect IDs 1, 4, 7, and 8 whose MDC defect classes do not match can be understood as variations in the work level of the visual classification operator.
- such an image is returned to the MDC apparatus again so that the work level of the visual classification operator is leveled.
- a defect image is determined (S1700).
- This defect image is the same as the image determined in S200 described with reference to FIG.
- the image may be an image that is intentionally selected for operator training.
- the defect image is transmitted to the plurality of MDC devices 104 (S1701).
- MDC 1 to N the defect images are classified by an operator, and a defect class is assigned.
- the assigned defect class information is received from the MDC device 104 via the data transmission / reception unit 110 and stored in the storage unit 111 (S1702).
- the defect class assigned by the MDC device 104 (MDC 1 to N) is read from the storage unit 111 to the defect class comparison unit 112, and the comparison is executed (S1703).
- a defect number with a mismatched defect class is identified by comparison, and a defect image corresponding to the identified defect number is selected (S1704).
- the method for extracting the mismatch of the defect class by comparison is as described in FIG. 3 in the first embodiment.
- defect class information is also transmitted for image data in which the selected MDC defect class does not match (S1705). If the defect class determined by the expert is not disclosed to the worker as an answer (NO in S1705)
- the defect image selected again is transmitted to the MDC 1 to NDC to request the operator to perform the MDC operation (S1606), and the defect class determined by the expert is disclosed to the operator as an answer (in the case of YES in S1705)
- the selected defect image and the defect class information of the selected image are transmitted to the MDCs 1 to N, and the operator is confirmed. (S1607).
- the ADC performance is evaluated according to the processing flow of FIG. 2 described in the first embodiment, and the ADC learning update is performed.
- the ADC learning update is performed using the defect image selected according to the processing flow of FIG.
- DESCRIPTION OF SYMBOLS 100 Automatic defect classification device, 101 ... Communication network, 102 ... Defect image imaging device, 103 ... Yield management system, 104 ... Visual defect classification unit, 110 ... Data transmission / reception part, 111 ... Memory
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Pathology (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Chemical & Material Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Molecular Biology (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Analysis (AREA)
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
Abstract
Description
分類性能評価部で評価した結果が予め設定した基準に達しない場合には画像選択部で選択した画像を用いて学習更新部で画像分類部の分類レシピを更新するようにした。
以下に、本発明の実施例を、図を用いて説明する。
Claims (20)
- 欠陥の画像を分類する装置であって、
他の撮像手段で撮像して得られた欠陥の画像を記憶する記憶部と、
他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて前記記憶部に記憶した欠陥の画像の中から画像を選択する画像選択部と、
前記画像選択部で選択した画像を分類レシピに基づいて分類する画像分類部と、
前記画像を分類した結果に基づいて前記画像分類部の分類性能を評価する分類性能評価部と、
前記分類性能評価部で評価した結果が予め設定した基準に達しない場合には前記画像選択部で選択した画像を用いて前記画像分類部の分類レシピを更新する学習更新部と
を備えたことを特徴とする欠陥画像分類装置。 - 請求項1記載の欠陥画像分類装置であって、前記他の複数の欠陥分類手段で分類した欠陥クラスを比較する欠陥クラス比較部を更に備え、前記欠陥クラス比較部で比較した情報に基づいて前記画像選択部で前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類装置。
- 請求項1記載の欠陥画像分類装置であって、前記学習更新部は、前記分類性能評価部で評価した結果が予め設定した基準に達しない場合には前記画像選択部で選択した画像のうち前記画像分類部で未知欠陥クラスと判定された欠陥の画像を用いて前記画像分類部の分類レシピを更新することを特徴とする欠陥画像分類装置。
- 請求項1記載の欠陥画像分類装置であって、前記画像分類部は、前記記憶部に記憶した前記他の撮像手段で撮像して得られた欠陥の画像を前記分類レシピに基づいて分類し、前記画像分類部で前記分類した結果未知欠陥クラスと判定された欠陥の数が予め設定した数以上になった場合に、前記画像選択部は、前記他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類装置。
- 請求項1記載の欠陥画像分類装置であって、前記他の撮像手段および前記他の複数の欠陥分類手段と通信回線を介して接続されていることを特徴とする欠陥画像分類装置。
- 欠陥の画像を分類する装置であって、
他の撮像手段で撮像して得られた欠陥の画像を記憶する記憶部と、
他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて前記記憶部に記憶した欠陥の画像の中から画像を選択する画像選択部と、
前記記憶部に記憶した画像を分類レシピに基づいて分類する画像分類部と、
前記画像選択部で選択した画像を用いて前記画像分類部の分類レシピを更新する学習更新部と
を備えたことを特徴とする欠陥画像分類装置。 - 請求項6記載の欠陥画像分類装置であって、前記他の複数の欠陥分類手段で分類した欠陥クラスを比較する欠陥クラス比較部を更に備え、前記欠陥クラス比較部で比較した情報に基づいて前記画像選択部で前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類装置。
- 請求項6記載の欠陥画像分類装置であって、前記学習更新部は、前記分類性能評価部で評価した結果が予め設定した基準に達しない場合には前記画像選択部で選択した画像のうち前記画像分類部で未知欠陥クラスと判定された欠陥の画像を用いて前記画像分類部の分類レシピを更新することを特徴とする欠陥画像分類装置。
- 請求項6記載の欠陥画像分類装置であって、前記画像分類部は、前記記憶部に記憶した前記他の撮像手段で撮像して得られた欠陥の画像を前記分類レシピに基づいて分類し、前記画像分類部で前記分類した結果未知欠陥クラスと判定された欠陥の数が予め設定した数以上になった場合に、前記画像選択部は、前記他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類装置。
- 請求項6記載の欠陥画像分類装置であって、前記他の撮像手段および前記他の複数の欠陥分類手段と通信回線を介して接続されていることを特徴とする欠陥画像分類装置。
- 欠陥の画像を分類する方法であって、
他の撮像手段で撮像して得られた欠陥の画像を記憶部に記憶し、
他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて画像選択部で前記記憶部に記憶した欠陥の画像の中から画像を選択し、
前記画像選択部で選択した画像を画像分類部で分類レシピに基づいて分類し、
前記画像を分類した結果に基づいて分類性能評価部で前記画像分類部の分類性能を評価し、
前記分類性能評価部で評価した結果が予め設定した基準に達しない場合には前記画像選択部で選択した画像を用いて学習更新部で前記画像分類部の分類レシピを更新する
ことを特徴とする欠陥画像分類方法。 - 請求項11記載の欠陥画像分類方法であって、前記他の複数の欠陥分類手段で分類した欠陥クラスを欠陥クラス比較部で比較する工程を更に備え、前記欠陥クラス比較部で比較した情報に基づいて前記画像選択部で前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類方法。
- 請求項11記載の欠陥画像分類装置であって、前記学習更新部で行う分類レシピの更新が、前記分類性能評価部で評価した結果が予め設定した基準に達しない場合には前記画像選択部で選択した画像のうち前記画像分類部で未知欠陥クラスと判定された欠陥の画像を用いて前記画像分類部の分類レシピを更新することを特徴とする欠陥画像分類方法。
- 請求項11記載の欠陥画像分類方法であって、前記画像分類部で前記記憶部に記憶した前記他の撮像手段で撮像して得られた欠陥の画像を前記分類レシピに基づいて分類し、前記画像分類部で前記分類した結果未知欠陥クラスと判定された欠陥の数が予め設定した数以上になった場合に、前記画像選択部で前記他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類方法。
- 請求項11記載の欠陥画像分類方法であって、前記他の撮像手段および前記他の複数の欠陥分類手段からの情報を通信回線を介して取得することを特徴とする欠陥画像分類方法。
- 欠陥の画像を分類する方法であって、
他の撮像手段で撮像して得られた欠陥の画像を記憶部に記憶し、
他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて画像選択部で前記記憶部に記憶した欠陥の画像の中から画像を選択し、
前記記憶部に記憶した画像を画像分類部で分類レシピに基づいて分類し、
前記画像選択部で選択した画像を用いて学習更新部で前記画像分類部の分類レシピを更新する
ことを特徴とする欠陥画像分類方法。 - 請求項16記載の欠陥画像分類方法であって、前記他の複数の欠陥分類手段で分類した欠陥クラスを欠陥クラス比較部で比較する工程を更に備え、前記欠陥クラス比較部で比較した情報に基づいて前記画像選択部で前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類方法。
- 請求項16記載の欠陥画像分類方法であって、前記学習更新部で行う分類レシピの更新が、前記分類性能評価部で評価した結果が予め設定した基準に達しない場合には前記画像選択部で選択した画像のうち前記画像分類部で未知欠陥クラスと判定された欠陥の画像を用いて前記画像分類部の分類レシピを更新することを特徴とする欠陥画像分類装置。
- 請求項16記載の欠陥画像分類方法であって、前記画像分類部で前記記憶部に記憶した前記他の撮像手段で撮像して得られた欠陥の画像を前記分類レシピに基づいて分類し、前記画像分類部で前記分類した結果未知欠陥クラスと判定された欠陥の数が予め設定した数以上になった場合に、前記画像選択部で前記他の複数の欠陥分類手段で欠陥を分類した欠陥クラスの情報を用いて前記記憶部に記憶した欠陥の画像の中から画像を選択することを特徴とする欠陥画像分類方法。
- 請求項16記載の欠陥画像分類方法であって、前記他の撮像手段および前記他の複数の欠陥分類手段からの情報を通信回線を介して取得することを特徴とする欠陥画像分類方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2015/066244 WO2016194210A1 (ja) | 2015-06-04 | 2015-06-04 | 欠陥画像分類装置および欠陥画像分類方法 |
US15/579,047 US20180174000A1 (en) | 2015-06-04 | 2015-06-04 | Defect image classification device and defect image classification method |
KR1020177034927A KR101978995B1 (ko) | 2015-06-04 | 2015-06-04 | 결함 화상 분류 장치 및 결함 화상 분류 방법 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2015/066244 WO2016194210A1 (ja) | 2015-06-04 | 2015-06-04 | 欠陥画像分類装置および欠陥画像分類方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016194210A1 true WO2016194210A1 (ja) | 2016-12-08 |
Family
ID=57440345
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/066244 WO2016194210A1 (ja) | 2015-06-04 | 2015-06-04 | 欠陥画像分類装置および欠陥画像分類方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US20180174000A1 (ja) |
KR (1) | KR101978995B1 (ja) |
WO (1) | WO2016194210A1 (ja) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108257121A (zh) * | 2018-01-09 | 2018-07-06 | 北京百度网讯科技有限公司 | 产品缺陷检测模型更新的方法、装置、存储介质及终端设备 |
JP2019075553A (ja) * | 2017-09-28 | 2019-05-16 | アプライド マテリアルズ イスラエル リミテッド | 半導体試料の欠陥を分類する方法およびそのシステム |
JP2019152948A (ja) * | 2018-03-01 | 2019-09-12 | 日本電気株式会社 | 画像判定システム、モデル更新方法およびモデル更新プログラム |
JP2020021224A (ja) * | 2018-07-31 | 2020-02-06 | 株式会社Screenホールディングス | 分類器生成方法および分類器生成装置 |
US10769774B2 (en) | 2018-01-09 | 2020-09-08 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and device for detecting a defect in a steel plate, as well as apparatus and server therefor |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI608369B (zh) * | 2016-11-23 | 2017-12-11 | 財團法人工業技術研究院 | 分類方法、分類模組及電腦程式產品 |
JP7054436B2 (ja) * | 2017-12-14 | 2022-04-14 | オムロン株式会社 | 検出システム、情報処理装置、評価方法及びプログラム |
JP7144244B2 (ja) * | 2018-08-31 | 2022-09-29 | 株式会社日立ハイテク | パターン検査システム |
JP7203678B2 (ja) | 2019-04-19 | 2023-01-13 | 株式会社日立ハイテク | 欠陥観察装置 |
CN111126487A (zh) * | 2019-12-24 | 2020-05-08 | 北京安兔兔科技有限公司 | 设备性能测试方法、装置及电子设备 |
IL276245A (en) * | 2020-04-24 | 2021-10-31 | Camtek Ltd | A method and system for classifying woofer defects using images of a defective woofer, based on deep learning |
CN114898121B (zh) * | 2022-06-13 | 2023-05-30 | 河海大学 | 基于图注意力网络的混凝土坝缺陷图像描述自动生成方法 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002515650A (ja) * | 1998-05-11 | 2002-05-28 | アプライド マテリアルズ インコーポレイテッド | 製造設備歩留まり向上システム |
JP2004295879A (ja) * | 2003-03-12 | 2004-10-21 | Hitachi High-Technologies Corp | 欠陥分類方法 |
WO2010076882A1 (ja) * | 2008-12-29 | 2010-07-08 | 株式会社日立ハイテクノロジーズ | 画像分類基準更新方法、プログラムおよび画像分類装置 |
WO2012144183A1 (ja) * | 2011-04-20 | 2012-10-26 | 株式会社日立ハイテクノロジーズ | 欠陥分類方法及び欠陥分類システム |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001256480A (ja) | 2000-03-09 | 2001-09-21 | Hitachi Ltd | 画像自動分類方法及び装置 |
US6678404B1 (en) * | 2000-10-31 | 2004-01-13 | Shih-Jong J. Lee | Automatic referencing for computer vision applications |
KR20070104331A (ko) * | 2004-10-12 | 2007-10-25 | 케이엘에이-텐코 테크놀로지스 코퍼레이션 | 표본 상의 결함들을 분류하기 위한 컴퓨터-구현 방법 및시스템 |
JP2006268825A (ja) * | 2005-02-28 | 2006-10-05 | Toshiba Corp | オブジェクト検出装置、学習装置、オブジェクト検出システム、方法、およびプログラム |
JP2011155123A (ja) | 2010-01-27 | 2011-08-11 | Kyocera Corp | 積層セラミックコンデンサ |
-
2015
- 2015-06-04 US US15/579,047 patent/US20180174000A1/en not_active Abandoned
- 2015-06-04 WO PCT/JP2015/066244 patent/WO2016194210A1/ja active Application Filing
- 2015-06-04 KR KR1020177034927A patent/KR101978995B1/ko active IP Right Grant
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002515650A (ja) * | 1998-05-11 | 2002-05-28 | アプライド マテリアルズ インコーポレイテッド | 製造設備歩留まり向上システム |
JP2004295879A (ja) * | 2003-03-12 | 2004-10-21 | Hitachi High-Technologies Corp | 欠陥分類方法 |
WO2010076882A1 (ja) * | 2008-12-29 | 2010-07-08 | 株式会社日立ハイテクノロジーズ | 画像分類基準更新方法、プログラムおよび画像分類装置 |
WO2012144183A1 (ja) * | 2011-04-20 | 2012-10-26 | 株式会社日立ハイテクノロジーズ | 欠陥分類方法及び欠陥分類システム |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019075553A (ja) * | 2017-09-28 | 2019-05-16 | アプライド マテリアルズ イスラエル リミテッド | 半導体試料の欠陥を分類する方法およびそのシステム |
JP7286290B2 (ja) | 2017-09-28 | 2023-06-05 | アプライド マテリアルズ イスラエル リミテッド | 半導体試料の欠陥を分類する方法およびそのシステム |
CN108257121A (zh) * | 2018-01-09 | 2018-07-06 | 北京百度网讯科技有限公司 | 产品缺陷检测模型更新的方法、装置、存储介质及终端设备 |
CN108257121B (zh) * | 2018-01-09 | 2019-01-25 | 北京百度网讯科技有限公司 | 产品缺陷检测模型更新的方法、装置、存储介质及终端设备 |
US10769774B2 (en) | 2018-01-09 | 2020-09-08 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and device for detecting a defect in a steel plate, as well as apparatus and server therefor |
JP2019152948A (ja) * | 2018-03-01 | 2019-09-12 | 日本電気株式会社 | 画像判定システム、モデル更新方法およびモデル更新プログラム |
JP7130984B2 (ja) | 2018-03-01 | 2022-09-06 | 日本電気株式会社 | 画像判定システム、モデル更新方法およびモデル更新プログラム |
JP2020021224A (ja) * | 2018-07-31 | 2020-02-06 | 株式会社Screenホールディングス | 分類器生成方法および分類器生成装置 |
JP7083721B2 (ja) | 2018-07-31 | 2022-06-13 | 株式会社Screenホールディングス | 分類器生成方法および分類器生成装置 |
Also Published As
Publication number | Publication date |
---|---|
KR101978995B1 (ko) | 2019-05-16 |
US20180174000A1 (en) | 2018-06-21 |
KR20170141255A (ko) | 2017-12-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2016194210A1 (ja) | 欠陥画像分類装置および欠陥画像分類方法 | |
KR102218364B1 (ko) | 패턴 검사 시스템 | |
US20210090238A1 (en) | System and method for performing automated analysis of air samples | |
TWI670781B (zh) | 疊代式缺陷濾除製程 | |
JP3834041B2 (ja) | 学習型分類装置及び学習型分類方法 | |
JP5444092B2 (ja) | 検査方法およびその装置 | |
US11450122B2 (en) | Methods and systems for defect inspection and review | |
US8041443B2 (en) | Surface defect data display and management system and a method of displaying and managing a surface defect data | |
WO2010076882A1 (ja) | 画像分類基準更新方法、プログラムおよび画像分類装置 | |
US9747520B2 (en) | Systems and methods for enhancing inspection sensitivity of an inspection tool | |
CN101197301A (zh) | 缺陷检查装置及缺陷检查方法 | |
WO2013153891A1 (ja) | 荷電粒子線装置 | |
JP2000057349A (ja) | 欠陥の分類方法およびその装置並びに教示用データ作成方法 | |
JP7054436B2 (ja) | 検出システム、情報処理装置、評価方法及びプログラム | |
JP2018036226A (ja) | 画像処理プログラム、画像処理方法および画像処理装置 | |
KR20140044395A (ko) | 결함 관찰 방법 및 결함 관찰 장치 | |
US20150362908A1 (en) | Automatic Recipe Stability Monitoring and Reporting | |
CN108445010A (zh) | 自动光学检测方法及装置 | |
JP7062474B2 (ja) | 劣化状況識別装置、劣化状況識別方法及びプログラム | |
TWI553688B (zh) | Charged particle beam device | |
KR101261016B1 (ko) | 평판패널 기판의 자동광학검사 방법 및 그 장치 | |
TW471016B (en) | Defect detection using gray level signatures | |
JP2001127129A (ja) | 試料の欠陥検査システム、および検査方法 | |
JP2019075078A (ja) | 工事現場画像判定装置及び工事現場画像判定プログラム | |
JP5983033B2 (ja) | 位置関係判定プログラム、位置関係判定方法および位置関係判定装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15894243 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15579047 Country of ref document: US |
|
ENP | Entry into the national phase |
Ref document number: 20177034927 Country of ref document: KR Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 15894243 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: JP |