CN115311247A - Defect classification correction method and application of workpiece appearance detection thereof - Google Patents

Defect classification correction method and application of workpiece appearance detection thereof Download PDF

Info

Publication number
CN115311247A
CN115311247A CN202211022984.4A CN202211022984A CN115311247A CN 115311247 A CN115311247 A CN 115311247A CN 202211022984 A CN202211022984 A CN 202211022984A CN 115311247 A CN115311247 A CN 115311247A
Authority
CN
China
Prior art keywords
defect
classification
detection
classification correction
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211022984.4A
Other languages
Chinese (zh)
Inventor
章国川
王罡
童竹勍
潘正颐
侯大为
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changzhou Weiyizhi Technology Co Ltd
Original Assignee
Changzhou Weiyizhi Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Changzhou Weiyizhi Technology Co Ltd filed Critical Changzhou Weiyizhi Technology Co Ltd
Priority to CN202211022984.4A priority Critical patent/CN115311247A/en
Publication of CN115311247A publication Critical patent/CN115311247A/en
Pending legal-status Critical Current

Links

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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • 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
    • 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
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • 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
    • G06T2207/30164—Workpiece; Machine component

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Multimedia (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The invention relates to the technical field of workpiece defect detection, in particular to a defect classification correction method and application of workpiece appearance detection thereof, wherein the defect classification correction method comprises the following steps: acquiring an optical surface image of the appearance of the workpiece through an acquisition module, and transmitting the acquired optical surface image to an image processing module for image processing to finally obtain an image for model detection; a model detection classification step, namely sending the obtained image of the model detection to a deep learning model for detection, and finally obtaining a detection result A of the image of the model detection; and a classification correction step, namely setting a classification correction rule, sending the detection result A to the classification correction rule, reclassifying the detection result A, comparing and analyzing the detection result A with the detection result A, and acquiring a defect result after the comparison of the classification correction rule. The defect classification correction method provided by the invention improves the defect classification accuracy and controls the over-killing and missing-detection indexes.

Description

Defect classification correction method and application of workpiece appearance detection thereof
Technical Field
The invention relates to the technical field of workpiece defect detection, in particular to a defect classification correction method and application of workpiece appearance detection.
Background
The purpose of defect detection is to find defective products in time, and in the field of quality inspection, a deep learning model is generally used for quality inspection of workpiece defects. In a traditional quality inspection mode, a deep learning model is usually trained by aiming at the mode whether a specified optical surface contains defects, in the model training process, the prediction result of the deep learning model is often classified to a certain degree, and the classification of different defects is inconsistent with the subsequently used judgment standard. The judgment standards corresponding to the wrong classification are used when the system is applied due to the wrong classification of the deep learning model, particularly when the product has several types of similar defects, such as the defect of partial different colors and white dots in scratch classification, or the defect of partial scratch and white dots in the different color classification, the judgment standards of the scratch are applied to the defect of the different colors or the white dots, or the judgment standards of the different colors are applied to the scratch and the white dots, and under the condition that the classification result is wrong, the result is often missed detection if the result is over-detected, the requirement of a customer detection index is difficult to meet, the project is difficult to fall on the ground, and the accuracy of the defect classification is particularly important.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: in order to solve the problem that defect classification is wrong during deep learning model detection, or when a product has several types of similar defects, such as defects with partial heterochromia and white spots in a scratch result, or defects with partial scratches and white spots in heterochromia classification, so that the defects are inaccurate during classification detection, the invention provides a defect classification-based correction method, which comprises the following steps:
acquiring an optical surface image of the appearance of the workpiece through an acquisition module, and transmitting the acquired optical surface image to an image processing module for image processing to finally obtain an image for model detection;
a step of model detection and classification, which is to send the obtained image of the model detection to a deep learning model for detection, and finally obtain a detection result A of the image of the model detection, wherein the detection result A comprises a defect A and defect quantitative information corresponding to the defect A;
a classification correction step, namely setting classification correction rules, wherein the classification correction rules comprise classification correction standards and reclassification rules, sending the detection result A to the classification correction rules, comparing and analyzing the defect quantitative information corresponding to the defect A with the classification correction standards, setting the corresponding reclassification rules to reclassify the defects, setting the reclassification result as a defect B, comparing and analyzing the defect B with the defect A, and judging that the defect A is classified correctly if the defect B is compared with the defect A in the same way; if the comparison between the defect B and the defect A is different, judging that the defect A is classified wrongly, and correcting the defect A into the defect B; and acquiring a classification correction rule to judge the compared defect result.
Further, specifically, the model detecting step includes:
a building step, wherein a deep learning model for detecting the defects of the workpiece is designed and built;
extracting, namely extracting the characteristics of the defects in the image detected by the model, filtering partial useless information and reserving effective information of the characteristics of the defects;
classifying, namely classifying the defects of each defect feature effective information in the image detected by the model to obtain the defect A;
and an output step of outputting the detection result a of the image detected by the model.
Further, specifically, the setting of the classification correction rule includes:
step one, setting a classification correction standard, and comparing and analyzing the defect quantitative information corresponding to the defect A with the classification correction standard parameters one by one;
step two, manually setting a corresponding reclassification rule according to an optical surface or an area by combining the analysis result of the step one to serve as a defect classification condition, and reclassifying the defect by adopting the corresponding reclassification rule according to the defect quantification information corresponding to the defect A;
the reclassification rule condition refers to a numerical value of a linear quantization ratio which can be used as a defect classification condition, and the linear quantization numerical value is a numerical value of a classification correction standard parameter.
Further, it is specifically characterized in that: further comprising: and a defect judgment step, namely judging the defects by adopting corresponding judgment standards according to the defect results after the judgment and comparison.
Further, specifically, the method further includes: displaying a defect statistical result, wherein the defect statistical result comprises: defect name, total defect count, detected record, over-detected record, error-detected record, and missed-detected record.
Further, in the step of model detection and classification, the defects a include scratches, white spots, off-colors, sagging and cracks.
Further, specifically, in the classifying and correcting step, the defect quantization information corresponding to the defect a includes: threshold, width, height, length, roundness, area, depth.
Further, specifically, the classification correction criterion parameters include: defect position distribution, production line defect proportion, production line defect judgment standard and classification characteristic analysis.
The application of the workpiece appearance detection applies the defect classification correction method.
The method has the advantages that the detection result of the deep learning model detection is reclassified, the classification correction rule is utilized, the corresponding reclassification rule is set according to the defect characteristic quantization information and the classification correction standards (defect position distribution, production line defect proportion, production line defect judgment standards and the like), the final defect classification can be determined only when the conditions set by the classification correction rule are met, the method improves the accuracy of the defect classification, can effectively provide the defect classification, more effectively align the production line standards, control over-killing and omission inspection indexes and achieve the expected model training and prediction effects.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a flow diagram of a method for defect classification correction in one embodiment.
FIG. 2 is a flowchart illustrating a detection step of a model according to an embodiment.
FIG. 3 is a flowchart illustrating the classification correction procedure according to an embodiment.
Fig. 4 is an internal structural diagram of a computer apparatus according to an embodiment.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings in which the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the invention, as detailed in the appended claims.
As shown in fig. 1, an embodiment of the present invention provides a defect classification based correction method, including:
an image obtaining step 100, collecting an optical surface image of the appearance of the workpiece (the shot optical surface image is a defect image) through a collecting module, transmitting the obtained optical surface image to an image processing module for image processing, specifically, the image processing module removes the duplicate of the same image, and performs algorithm calculation on the optical surface image, so as to improve the quality of the image, and finally obtain an image for model detection;
and a step 200 of model detection classification, wherein the obtained image of the model detection is sent to a deep learning model for detection, and finally a detection result A returned by the image of the model detection is obtained.
In the embodiment of the present invention, as shown in fig. 2, the model detecting step includes:
a building step 202, designing and building a deep learning model for detecting the defects of the workpiece;
an extraction step 203, which is to extract the characteristics of the defects in the image detected by the model, filter part of useless information and reserve effective information of the characteristics of the defects, wherein the effective information of the characteristics of the defects comprises: a plurality of defect features and each defect feature, the effective information of the defect features comprises threshold values (position information x coordinate, y coordinate), width, roundness, height, length, height, area and the like.
A classification step 204, performing defect judgment on each defect feature effective information in the image detected by the model to be learned to obtain a defect classification result; and the classification step is used for classifying the defects according to the defect characteristic values. The classification step comprises a plurality of category judgment rules, and effective information of the defect characteristics of the image detected by each model is input into the category judgment rules for judgment to obtain the corresponding defect A.
Outputting 205 a detection result a returned by the image detected by the model, wherein the detection result a comprises a defect a and defect quantitative information corresponding to the defect a; the defect A comprises scratch, white point, heterochrosis, collapsed edge, crack and the like, the defect quantitative information corresponding to the defect A is effective information of defect characteristics corresponding to the image detected by the model, and the defect quantitative information corresponding to the defect A comprises the following steps: threshold (position information x-coordinate, y-coordinate), width, height, length, roundness, area, depth information, etc.
In an embodiment of the present invention, the model detecting step further comprises: and a deep learning model training step, namely training the image subjected to model detection of characteristic information class classification to obtain a trained deep learning model, and performing defect detection on the image of the workpiece appearance by using the trained deep learning model.
A classification correction step, namely setting classification correction rules, reclassifying and correcting the defect A, wherein the classification correction rules comprise a classification correction standard and a reclassification rule, sending a detection result A to the classification correction rules, setting a corresponding reclassification rule for reclassification after comparing and analyzing defect quantitative information corresponding to the defect A with the classification correction standard, setting a reclassification result as a defect B, comparing and analyzing the defect B with the defect A, and judging that the defect A is classified correctly if the defect B is compared with the defect A in the same way; if the comparison between the defect B and the defect A is different, judging that the defect A is classified wrongly, and correcting the defect A into the defect B;
in the embodiment of the present invention, as shown in fig. 3, the classification correcting step includes:
an obtaining step 301, obtaining a detection result A detected by a deep learning model;
the step 302 of setting classification correction rules includes:
step one, setting a classification correction standard, wherein the classification correction standard is an industrial quality inspection parameter of manual quality inspection, and the industrial quality inspection parameter of the manual quality inspection comprises the following steps:
defect position distribution: the defects have fixed area distribution and morphological distribution according to different process flows. For example: white spots only exist corner surfaces, heterochromatic colors only appear large surfaces, cracks only vertically split from side edges or hole edges, and the position distribution relation of the edge collapse defects on the edges is the same.
Defect proportion of a production line: the defect of learning training model output is the mar, and the defect has scratch, white point and heterochrosis respectively after artifical quality control, and statistics defect accounts for the proportion condition that distributes in each position, if: scratch accounts for 30% (60% of scratch on the large surface, 35% of the side surface and 5% of the corner), heterochromatic 20% (98% of the large surface, 0% of the corner and 2% of the side surface) and white point 10% (75% of the corner, 20% of the side surface and 5% of the large surface).
Judging a defect judgment standard by a production line: the production line quality inspection personnel judge whether the workpieces are good or not according to corresponding standards, such as a degree standard (such as a scratch depth D is more than 20) and a size standard (such as a scratch length L is more than 2 mm);
and (3) classification characteristic analysis, namely classifying and analyzing strong correlation characteristic information of different defects, such as scratch: the strong correlation characteristic information includes (length, depth, threshold, etc.), heterochromatic strong correlation characteristic information (area, mao Caodu, threshold, etc.), and white point strong correlation characteristic information (roundness, threshold, etc.).
Step two, setting a reclassification rule, wherein different reclassification rules are manually set as reclassification conditions due to different products, different defects and different parameter conditions of the defects corresponding to the judgment of different parts, and the reclassification rules comprise a classification rule I, a classification rule II, a classification rule III and the like;
the reclassification rule condition refers to a numerical value of a linear quantization ratio which can be used as a condition, in the embodiment of the invention, in order to improve the accuracy of defect classification, the linear quantization numerical value is a numerical value of an industrial quality inspection parameter of artificial quality inspection, and the condition comprises the following physical quantity characteristic information: quality control defect threshold, quality control defect length, quality control defect width, quality control defect area, make up a reclassification rule through setting up one or more rules, if heterochrosis rule: smoothness >50& area >300& roundness <50.
Specifically, in the embodiment of the present invention, after the defect quantization information corresponding to the defect a and the classification correction standard parameters are combined and compared one by one, the corresponding reclassification rule is manually set in combination with the analysis result, and the linear quantization value in the reclassification rule is used to compare with the defect quantization information corresponding to the defect a. For example, white spots and different colors can be distinguished according to data such as roundness of corresponding quality inspection defects and area of the quality inspection defects; and distinguishing the heterochromatic color from the scratch by using data such as the length-width ratio of the quality inspection defect, the width of the quality inspection defect and the like.
In the embodiment of the present invention, preferably, the set reclassification rule includes:
the optical surface comprises:
a first classification rule: length >200px & width <20px corrected for scratches;
and (2) classification rule II: roundness >80& depth >200& area <100 square pixels corrected to white point;
and (3) classification rules are three: depth <120& smoothness >50& area >300& roundness <50 is corrected to heterochromatic.
The method comprises the following steps according to regions (combined with defect position distribution):
a first classification rule: white points exist on corner surfaces;
and (2) classification rule II: the heterochromatic color only appears on a large surface;
and (3) classification rules are three: cracks can only crack vertically from the side edges or hole edges.
In the embodiment of the present invention, when the deep learning model is used to classify the appearance defects of the workpiece, it is found that the workpiece is globally classified when the deep learning model is detected, and when the defects are classified, the classification result has a deviation, which easily causes over-killing, missing detection and false detection, for example, if the determination of a defect type in the deep learning model detection is determined by the defect length and the defect width, for example, 20cm < defect length <100cm, and 5cm < defect width <20cm, the defect type is determined as defect a, but since the states of the defects presented under different conditions may be different, some defects may satisfy the above conditions, but the defect type may be another type. Therefore, in order to improve the accuracy of defect type judgment, the detection result detected by the deep learning model needs to be classified again by combining with a classification correction rule, and the final defect classification can be determined only when the condition set by the classification correction rule is met, for example, after the corresponding defect quantization information of the defect a is compared and analyzed with each standard (industrial quality inspection parameter) in the classification correction standards one by one, the corresponding re-classification rule is manually set by combining with the comparison and analysis result, the re-classification result is a defect B, and compared with the detection result defect a detected by the deep learning model, whether the detection result detected by the deep learning model is correct is judged, the defect B is compared and analyzed with the defect a, and if the defect B is the same as the defect a in comparison, the defect a classification is correct; if the comparison between the defect B and the defect A is different, judging that the defect A is classified wrongly, correcting the defect A into the defect B, and improving the accuracy of defect type classification.
Specifically, for example, the detection result defect a detected by the deep learning model is a different color, the corresponding defect quantization information of the different color is compared with each standard in the classification correction standard one by one for analysis, for example, the threshold of the different color is compared with the defect position distribution, and whether the threshold of the defect is in the defect position distribution is judged; counting the defect proportion of the production line of the artificial quality inspection of the different colors to obtain the error detection rate of the learning training model; judging whether the different colors meet the non-defective standard or not by combining with a production line to judge a defect judgment standard; after the corresponding defect quantitative information of the different colors and the corresponding strong correlation characteristic information are contrastively analyzed, a plurality of corresponding re-classification rules are manually set in combination with the contrastive analysis result, wherein the plurality of re-classification rules comprise scratch classification rules, white point classification rules, different color classification rules and the like, the defect A is reclassified, the re-classification result is a defect B, the defect B is compared with the detection result defect A (different color) detected by the deep learning model, whether the detection result detected by the deep learning model is correct or not is judged, if the defect B is the different color and the defect A (different color) is the same, the defect A is judged to be correctly classified, and if the defect B is the scratch and the defect A (different color) are different, the defect A is judged to be wrongly classified, and the defect A (different color) is corrected to be a defect B (scratch).
In another embodiment of the present invention, the step 302 of setting classification correction rules further includes: and generating corresponding classification rules through a combined optimization classification algorithm, wherein the generation rules of the combined optimization classification algorithm comprise defect local optimal classification rules aiming at different optical surfaces or regional subdivided particle dimensions, optimally combining all the granularity local optimal rules, and finally generating the rules according to the optimal combination. Compared with the manually set reclassification rule, the rules corresponding to the combined optimization classification algorithm have higher processing speed when the combined optimization classification algorithm generates the corresponding classification rules.
And an output step 303, of outputting a classification correction rule to judge the compared defect result, wherein the compared defect result comprises a correct defect result (defect A) or a corrected defect result (defect B).
In the embodiment of the present invention, the method further includes: and a defect judgment step, namely judging the defects by adopting corresponding judgment standards according to the defect results after judgment and comparison.
In the embodiment of the present invention, the method further includes: displaying a defect statistical result, wherein the defect statistical result comprises: defect name, total defect count, detected record, past kill record, error check record, missed check record, and the like.
Defect name: the name of a defect of some kind, such as scratch, heterochrosis, white spot, etc.
The total defect number: it is the number of defects of a certain type.
And (4) killing and recording: it is the number of overkill of the defect.
And (4) missing detection recording: it is the number of missed defects.
And (4) error detection recording: it is the number of false detections of the defect corrected by the classification correction step.
The application of the industrial visual appearance detection applies the defect classification correction method.
In summary, according to the defect classification correction method and the application of the workpiece appearance inspection thereof provided by the invention, the inspection result of the deep learning model inspection is reclassified, the classification correction rule is utilized, and the corresponding reclassification rule is set according to the defect characteristic quantization information and the classification correction standards (defect position distribution, production line defect proportion, production line defect judgment standard and the like), so that the final defect classification can be determined only when the conditions set by the classification correction rule are met.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a non-volatile computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), rambus (Rambus) direct RAM (RDRAM), direct bused dynamic RAM (DRDRAM), and bused dynamic RAM (RDRAM).
Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.
It will be understood that the invention is not limited to the precise arrangements that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the invention is limited only by the appended claims.

Claims (9)

1. A defect classification based correction method is characterized in that: the method comprises the following steps:
acquiring an optical surface image of the appearance of a workpiece through an acquisition module, and transmitting the acquired optical surface image to an image processing module for image processing to finally obtain an image for model detection;
a model detection classification step, namely sending the obtained image of the model detection to a deep learning model for detection, and finally obtaining a detection result A of the image of the model detection, wherein the detection result A comprises a defect A and defect quantitative information corresponding to the defect A;
a classification correction step of setting classification correction rules including classification correction standards and reclassification rules, sending the detection result A to the classification correction rules, comparing and analyzing the defect quantitative information corresponding to the defect A with the classification correction standards, setting the corresponding reclassification rules to reclassify the defect, setting the reclassification result as a defect B, and comparing and analyzing the defect B with the defect A,
if the defect B is the same as the defect A in comparison, judging that the defect A is correctly classified;
if the comparison between the defect B and the defect A is different, judging that the defect A is classified wrongly, and correcting the defect A into the defect B; and acquiring a classification correction rule to judge the compared defect result.
2. The method of defect classification correction according to claim 1, wherein the model detection step comprises:
a building step, wherein a deep learning model for detecting the defects of the workpiece is designed and built;
an extraction step, namely performing feature extraction on defects in the image detected by the model, filtering partial useless information and reserving effective defect feature information;
classifying, namely classifying the defects of each defect feature effective information in the image detected by the model to obtain the defect A;
and an output step of outputting the detection result a of the image detected by the model.
3. The method for defect classification correction according to claim 1, wherein the setting of classification correction rules comprises:
step one, setting a classification correction standard, and comparing and analyzing the defect quantitative information corresponding to the defect A with the classification correction standard parameters one by one;
step two, manually setting a corresponding reclassification rule according to an optical surface or an area by combining the analysis result of the step one to serve as a defect classification condition, and reclassifying the defect by adopting the corresponding reclassification rule according to the defect quantification information corresponding to the defect A;
the reclassification rule condition refers to a numerical value of a linear quantization ratio which can be used as a defect classification condition, and the linear quantization numerical value is a numerical value of a classification correction standard parameter.
4. The method of defect classification correction according to claim 1, characterized in that: further comprising: and a defect judging step of judging the defects by adopting corresponding judgment standards according to the defect results after the judgment and comparison.
5. The method of defect classification correction according to claim 1, characterized in that: further comprising: displaying a defect statistical result, wherein the defect statistical result comprises: defect name, total defect count, detected record, over-detected record, error-detected record, and missed-detected record.
6. The method of defect classification correction according to claim 1, characterized in that: in the step of model detection and classification, the defects A comprise scratches, white spots, heterochrosis, collapsed edges and cracks.
7. The method of defect classification correction according to claim 1, characterized in that: in the classification correction step, the defect quantization information corresponding to the defect a includes: threshold, width, height, length, roundness, area.
8. The method of defect classification correction according to claim 3, characterized in that: the classification correction standard parameters comprise: defect position distribution, production line defect proportion, production line defect judgment standard and classification characteristic analysis.
9. The application of the workpiece appearance detection is characterized in that: a method of defect classification correction as claimed in any one of claims 1 to 8 is applied.
CN202211022984.4A 2022-08-25 2022-08-25 Defect classification correction method and application of workpiece appearance detection thereof Pending CN115311247A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211022984.4A CN115311247A (en) 2022-08-25 2022-08-25 Defect classification correction method and application of workpiece appearance detection thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211022984.4A CN115311247A (en) 2022-08-25 2022-08-25 Defect classification correction method and application of workpiece appearance detection thereof

Publications (1)

Publication Number Publication Date
CN115311247A true CN115311247A (en) 2022-11-08

Family

ID=83865152

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211022984.4A Pending CN115311247A (en) 2022-08-25 2022-08-25 Defect classification correction method and application of workpiece appearance detection thereof

Country Status (1)

Country Link
CN (1) CN115311247A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115861315A (en) * 2023-02-27 2023-03-28 常州微亿智造科技有限公司 Defect detection method and device
CN116721098A (en) * 2023-08-09 2023-09-08 常州微亿智造科技有限公司 Defect detection method and defect detection device in industrial detection

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115861315A (en) * 2023-02-27 2023-03-28 常州微亿智造科技有限公司 Defect detection method and device
CN116721098A (en) * 2023-08-09 2023-09-08 常州微亿智造科技有限公司 Defect detection method and defect detection device in industrial detection
CN116721098B (en) * 2023-08-09 2023-11-14 常州微亿智造科技有限公司 Defect detection method and defect detection device in industrial detection

Similar Documents

Publication Publication Date Title
CN110349126B (en) Convolutional neural network-based marked steel plate surface defect detection method
CN111292305B (en) Improved YOLO-V3 metal processing surface defect detection method
CN115311247A (en) Defect classification correction method and application of workpiece appearance detection thereof
CN110570410B (en) Detection method for automatically identifying and detecting weld defects
CN111161243B (en) Industrial product surface defect detection method based on sample enhancement
US6507670B1 (en) System and process for removing a background pattern from a binary image
US8520892B2 (en) Method and apparatus for detecting objects
CN114219794B (en) Method and system for evaluating surface quality of shaving board based on machine vision
TW201514472A (en) Method of optical defect detection through image analysis and data mining integrated
CN106846316A (en) A kind of GIS inside typical defect automatic distinguishing method for image
CN104794491A (en) Fuzzy clustering steel plate surface defect detection method based on pre classification
CN111402236B (en) Hot rolled strip steel surface defect grading method based on image gray value
CN109741314A (en) A kind of visible detection method and system of part
CN111815555A (en) Metal additive manufacturing image detection method and device combining anti-neural network with local binary
DE112017005011T5 (en) Method for generating a taught-in model, device for generating a taught-in model, method for distinguishing signal data, device for distinguishing signal data and program for distinguishing signal data
CN108038838A (en) A kind of cotton fibriia species automatic testing method and system
CN111753794B (en) Fruit quality classification method, device, electronic equipment and readable storage medium
US6415062B1 (en) System and process for repairing a binary image containing discontinuous segments of a character
CN103258218B (en) Masking-out detects skeleton generating method, device, defect inspection method and device
CN117152161A (en) Shaving board quality detection method and system based on image recognition
CN117333467B (en) Image processing-based glass bottle body flaw identification and detection method and system
CN116664540A (en) Rubber sealing ring surface defect detection method based on Gaussian line detection
CN115880481A (en) Curve positioning algorithm and system based on edge profile
CN115759964A (en) Method, system and equipment for verifying universal qualification of bid document
CN115937085A (en) Nuclear cataract image processing method based on neural network learning

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