CN116777911A - Double-substrate radiator surface defect detection system based on image recognition - Google Patents
Double-substrate radiator surface defect detection system based on image recognition Download PDFInfo
- Publication number
- CN116777911A CN116777911A CN202311045856.6A CN202311045856A CN116777911A CN 116777911 A CN116777911 A CN 116777911A CN 202311045856 A CN202311045856 A CN 202311045856A CN 116777911 A CN116777911 A CN 116777911A
- Authority
- CN
- China
- Prior art keywords
- defect
- image
- pixel point
- value
- homogeneity
- 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
Links
- 230000007547 defect Effects 0.000 title claims abstract description 301
- 239000000758 substrate Substances 0.000 title claims abstract description 65
- 238000001514 detection method Methods 0.000 title claims abstract description 43
- 238000004364 calculation method Methods 0.000 claims description 22
- 230000008439 repair process Effects 0.000 claims description 18
- 230000002950 deficient Effects 0.000 claims description 17
- 238000000034 method Methods 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 15
- 238000012417 linear regression Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 230000003647 oxidation Effects 0.000 claims description 6
- 238000007254 oxidation reaction Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000002708 enhancing effect Effects 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 239000003795 chemical substances by application Substances 0.000 claims description 3
- 238000005554 pickling Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 abstract description 14
- 238000012545 processing Methods 0.000 abstract description 6
- 230000009286 beneficial effect Effects 0.000 description 8
- 239000002699 waste material Substances 0.000 description 7
- 230000017525 heat dissipation Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000001965 increasing effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 229910052751 metal Inorganic materials 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 229910052782 aluminium Inorganic materials 0.000 description 1
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 229910052802 copper Inorganic materials 0.000 description 1
- 239000010949 copper Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000002355 dual-layer Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000010410 layer Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000007769 metal material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011158 quantitative evaluation Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- 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
- G06V10/40—Extraction of image or video features
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a double-substrate radiator surface defect detection system based on image recognition, which relates to the technical field of image data processing, and comprises the steps of shooting and collecting multi-frame image data of the surface of a double-substrate radiator, and reconstructing the multi-frame image to obtain a pixel point reconstruction coefficientAnd reconstructing the feature valuesAnd calculate and obtain homogeneous difference characteristicsAnd pixel point homogeneous density valueThe defect evolution index Qx and the loss area value Sy are calculated. And the defect compensation coefficient Mb is calculated and obtained by fitting the defect evolution index Qx and the loss area value Sy, a detailed defect report is generated by the system according to the defect type, the defect evolution index Qx and the defect compensation coefficient Mb judged by the judging module, and the defect detection system based on image recognition can monitor the defect condition of the surface of the double-substrate radiator in real time and feed back to an operator in time by matching the corresponding compensation scheme. Compared with manual detection, the production efficiency can be improved, and the requirement for human resources can be reduced.
Description
Technical Field
The invention relates to the technical processing field of image data, in particular to a double-substrate radiator surface defect detection system based on image recognition.
Background
A dual-substrate heatsink is a heatsink with a particular design and structure relative to a conventional single-substrate heatsink. The following are some features and advantages of a dual-substrate heatsink: by increasing the heat dissipation area and improving the heat conduction path, a higher heat dissipation performance is proposed, a dual-substrate heat sink: two suitable metal substrates are selected as the double substrate of the heat sink. These substrates generally have good thermal conductivity and heat dissipation properties, such as metallic materials like copper or aluminum. Double-layer structure: the dual-substrate heat sink adopts a dual-layer structure, and is generally composed of two metal substrates. The structure makes the heat dissipation area of the heat sink larger, and can provide better heat dissipation effect.
In the installation of double-substrate radiator substrate, the surface of the double-substrate radiator is required to be detected, especially under the condition that the surface has defects, most of the prior art is to reduce waste products and improve product quality through manual detection and repair, however, the defects are not repaired in time due to manual production line monitoring and repair, fatigue is easy, waste of raw materials and resources is caused, and the manual detection is carried out, so that the production efficiency is reduced.
Disclosure of Invention
(one) solving the technical problems
Aiming at the defects of the prior art, the invention provides a double-substrate radiator surface defect detection system based on image recognition, which is used for shooting and collecting multi-frame image data of the double-substrate radiator surface, and acquiring pixel point reconstruction coefficients by reconstructing multi-frame images through learning and recognizing different types of defect modes by a feature extraction unit and an analysis unit in a model building moduleAnd reconstructing the characteristic value +.>And calculate to obtain the homogeneity difference feature +.>And pixel point homogeneous density valueAnd use the homogeneity difference feature +.>And pixel homogeneity density value +.>The defect evolution index Qx and the loss area value Sy are calculated. These indicators enable quantitative assessment of the extent of evolution of defects and the magnitude of surface loss, providing a deeper understanding of defect conditions. And the defect compensation coefficient Mb is calculated and obtained by fitting the defect evolution index Qx and the loss area value Sy, a detailed defect report is generated by the system according to the defect type, the defect evolution index Qx and the defect compensation coefficient Mb judged by the judging module, and the defect detection system based on image recognition can monitor the defect condition of the surface of the double-substrate radiator in real time and feed back to an operator in time by matching the corresponding compensation scheme. The problems can be quickly perceived and solved, the production flow can be adjusted in time, the defects can be corrected, and the defect detection system based on image recognition can realize high speed and large scaleCompared with manual detection, the automatic detection of the scale can improve the production efficiency and reduce the requirement of human resources. In addition, by early finding and disposing of defects, the system can reduce waste production and production costs.
(II) technical scheme
In order to achieve the above purpose, the invention is realized by the following technical scheme:
preferably, the image preprocessing module comprises an image denoising unit and a data enhancement unit;
the image denoising unit is used for eliminating noise by adopting one or more of mean value filtering, median filtering and Gaussian filtering, carrying out pixel value averaging or median calculation on a local area of an image, adjusting brightness by adopting histogram equalization, stretching an image range and adjusting the contrast of the image;
the data enhancement unit is used for enhancing the number and diversity of continuous image frames on the surface of the double-substrate radiator by rotation, scaling and translation, identifying the defect part and marking.
Preferably, along the time axis, images of x1, x2, x3, & gt, xn are acquired, the image at the T-th time is taken for calculation, the center position i of the defective pixel point is identified in the surface image of the dual-substrate radiator, and the gray data at the center of the defective pixel point isPixel point i is taken as center to be +.>Wherein Z extends according to the size of the defect to form a defect area, pixels at different positions in the defect area are analyzed, and the pixel area is set to be in the same reconstruction interval as the pixel value at the i position of the central pixel point +.>The number of pixels is recorded as +.>And calculate and obtain the position of the pixel point i by the following formulaReconstruction coefficient->：
Wherein: scale represents the normalization function, ii represents the pixel maximum, ji represents the pixel minimum,in the defective area denoted by i, remove +.>The number of pixels; d represents the numerical values of different pixel points in the corresponding defect area; s represents the numerical average area of the pixel points.
Preferably, the eigenvalues are reconstructedThe pixel point value at the position of the pixel point i is reconstructed, and an acquisition formula is as follows:
in the formula, when the pixel point i is positioned in the defect area of the surface of the double-substrate radiator, the pixel point distribution condition of the defect area in the center of the pixel point i is calculated, and when the difference between the defect local area and other pixel points with the center of the pixel point i as the center is obvious, the reconstructed characteristic value of the corresponding pixel point i is calculatedWill increase accordingly, indicating that the defect is more pronounced.
Preferably, the analysis unit is configured to calculate a homogeneity-difference characteristicThe homogeneity difference feature->The calculation steps comprise:
extending eight areas to the periphery by using the defect area where the pixel point i is located, further calculating and analyzing the numerical characteristics of the pixel points in the areas, and calculating the reconstruction characteristic values of the extended areas to obtain corresponding reconstruction characteristic sequences, wherein the reconstruction characteristic sequences are as follows:、、/>、...、/>calculating the reconstruction characteristics to obtain homogeneity difference characteristics +.>And homogeneity Density number->：
Wherein 8 areas are represented by eight areas extending to the periphery of a defect area where a pixel point i is located, the center of each area is represented by C, and similarity is calculated by C, normalized and then a homogeneity difference characteristic is obtained>The larger the representative difference is smaller, the smaller the representative difference is larger; m is represented as a density factor;
if reconstruction centered on pixel i obtains homogeneous difference characteristicsIs a scratch and expandsSmall differences in 8 areas, which are indicative of a scratch defect, are indicated if the homogeneity is characterized by +.>If the difference is large, judging that the scratch defect is absent;
if the homogeneity density is a valueM is within the density range of the scratch, if the homogeneity density value is +.>If the defect density value is higher, the defect density value representing the scratch is higher, and the defect degree is higher; if the homogeneity density value +.>If the defect density value is lower, the defect degree is lower, and the corresponding repairing means are selected differently.
Preferably, the defect evolution index Qx is calculated by the following formula:
wherein: nf represents the number of defects in the final image,representing the number of defects in the initial image.
Preferably, the calculation module comprises a fitting unit and a calculation unit;
the fitting unit is used for fitting the defect evolution index Qx and the loss area value Sy, and after correlation is carried out by adopting a linear regression mode, a linear regression equation is expressed as follows:
where β0 is the intercept and β1 is the slope fitting factor; training a linear regression model using the sample data to find optimal β0 and β1 values so that the model can best fit the sample data; after training is completed, the best fitting parameters beta 0 and beta 1 are obtained, the fitting factor is beta 1, and the slope of the linear relation between the defect evolution index Qx and the loss area value Sy is represented;
the calculating unit is configured to calculate the defect evolution index Qx and the loss area value Sy to obtain a defect compensation coefficient Mb:
the meaning of the formula is that assuming a linear relationship between Qx and Sy, the defect-compensating coefficient Mb is the result of the product of Qx and Sy.
Preferably, the judging module is used for judging the defect type according to the defect evolution index Qx, and comprises a threshold unit and a judging unit;
the threshold unit is used for setting a rule threshold to determine the defect type as a judging condition;
the judging unit is used for comparing the evolution index Qx of the defect with a rule threshold value, judging the type of the defect and generating a corresponding defect report;
the defect types include the following classifications: if Qx is greater than the threshold A, judging that the crack is generated;
if Qx is smaller than threshold A and greater than threshold B, judge to wear;
if Qx is smaller than the threshold B and larger than the threshold C, judging that the scratch is generated;
if Qx is smaller than the threshold C, it is determined that oxidation is occurring.
Preferably, the repair module comprises a defect library and a scheme unit;
the defect library is used for collecting historical defect data and big data double-substrate radiator surface defect data and comparing the historical defect data and the big data double-substrate radiator surface defect data with the defect report;
the scheme unit is used for acquiring the defect report and generating a corresponding compensation scheme according to the information in the defect report; the repairing scheme provides a repairing method according to different defect types, and comprises the steps of filling cracks, smearing repairing agent and pickling the surface oxidized surface.
(III) beneficial effects
The invention provides a double-substrate radiator surface defect detection system based on image recognition. The beneficial effects are as follows:
(1) The system can learn and identify different types of defect modes, and obtain pixel point reconstruction coefficients by reconstructing multi-frame imagesAnd reconstructing the characteristic value +.>And calculate to obtain the homogeneity difference feature +.>And pixel homogeneity density value +.>And use the homogeneity difference feature +.>And pixel point homogeneous density valueThe defect evolution index Qx and the loss area value Sy are calculated. These indicators enable quantitative assessment of the extent of evolution of defects and the magnitude of surface loss, providing a deeper understanding of defect conditions. And the defect compensation coefficient Mb is calculated and obtained by fitting the defect evolution index Qx and the loss area value Sy, a detailed defect report is generated by the system according to the defect type, the defect evolution index Qx and the defect compensation coefficient Mb judged by the judging module, and the defect detection system based on image recognition can monitor the defect condition of the surface of the double-substrate radiator in real time and feed back to an operator in time by matching the corresponding compensation scheme.The defect detection system based on image recognition can realize high-speed and large-scale automatic detection, and compared with manual detection, the defect detection system based on image recognition can improve production efficiency and reduce the requirement of human resources. In addition, by early finding and disposing of defects, the system can reduce waste production and production costs.
(2) According to the system for detecting the surface defects of the double-substrate radiator based on image recognition, the defects can be accurately positioned and located by recognizing the center position i of the defective pixel point in the surface image of the double-substrate radiator. This facilitates subsequent analysis and computation, enabling the system to accurately process and determine specific defects; according to the center position i of the defective pixel point, the system can extend according to the size of the defect to form a defect area. This helps to gather the relevant pixels together to form a complete defect region for subsequent analysis and calculation, and by analyzing the pixels at different locations in the defect region, the system can calculate the reconstruction coefficients at the i-location of the pixel. This coefficient may reflect the numerical characteristics and importance of the pixel point in the defective area. By reconstructing coefficients->The severity and extent of the defect can be further assessed.
(3) According to the system for detecting the surface defects of the double-substrate radiator based on image recognition, a judging module judges the types of defects according to the defect evolution index Qx and the rule threshold value, and a defect report is generated. And the repair module compares the defect library according to the information in the defect report and generates a corresponding compensation scheme to repair the defect. Thus, the selection of the repair method aiming at different defect types can be realized, and the beneficial effect is achieved.
Drawings
FIG. 1 is a block diagram and schematic flow chart of a dual-substrate radiator surface defect detection system based on image recognition;
fig. 2 is a schematic diagram showing the image being divided into eight areas extending to the periphery by the defect area where the pixel point i is located according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the installation of double-substrate radiator base plate, need detect the surface of double-substrate radiator, especially under the circumstances that the surface exists the defect, most in the prior art is through manual detection in-process and repair for reduce the waste product and produce with improving product quality, however the manual monitoring and the repair of carrying out the assembly line, tired easily can lead to the defect not in time to be repaired, leads to extravagant raw and other materials and the waste of resource, and manual detection reduces production efficiency.
Example 1
In order to achieve the above purpose, the invention is realized by the following technical scheme: referring to fig. 1, the system for detecting surface defects of a dual-substrate radiator based on image recognition comprises an image acquisition module, an image preprocessing module, a model building module, a calculation module, a judgment module and an adjustment module;
the image acquisition module is used for shooting and acquiring continuous image frames of the surface of the double-substrate radiator by adopting image acquisition equipment;
the image preprocessing module is used for denoising, brightness adjustment and contrast adjustment of continuous image frames on the surface of the bipolar plate radiator;
the model building module is used for analyzing the characteristics by using a defect recognition mode algorithm and training to learn different types of defect modes; the modeling module comprises a feature extraction unit and an analysis unit; by establishing the feature extraction unit and the analysis unit in the model module, the system is able to learn and analyze different types of defect patterns. This enables the system to more accurately identify and classify surface defects such as cracks, wear, scratches, and oxidation;
the feature extraction unit is used for identifying preprocessed continuous image frames of the double-substrate radiator, sequencing time axis image data, marking serial numbers as x1, x2, x3, & gt, xn, quantifying key features in a plurality of continuous image frames, calculating surface features of different pixel points in the surface image of the double-substrate radiator, and obtaining a pixel point reconstruction coefficientAnd reconstructing the characteristic value +.>The method comprises the steps of carrying out a first treatment on the surface of the Reconstructing coefficients according to pixel points>And reconstructing the characteristic value +.>Further calculating to obtain homogeneity difference characteristics>The method comprises the steps of carrying out a first treatment on the surface of the The analysis unit is used for analyzing the difference characteristic according to the homogeneity>Optimizing and selecting a clustering center to perform clustering to obtain a clustering result, and calculating to obtain a pixel point homogeneity density value +.>The method comprises the steps of carrying out a first treatment on the surface of the Several homogeneity-difference features in x1, x2, x3, & gt, xn images>Comparing, obtaining defect evolution index Qx, and adding several homogeneous density values +.>Comparing to obtain a loss area value Sy; the computing module is used for processingFitting the defect evolution index Qx with the loss area value Sy, wherein the indexes can provide quantitative evaluation of defects and are used for judging the severity degree and the evolution condition of the defects; a defect-compensating coefficient Mb is calculated and obtained, which can be used to evaluate the remedy and repair of the defect.
Judging the defect type by a judging module according to the defect evolution index Qx, and generating a defect report by combining the defect compensation coefficient Mb; defect types include cracks, wear, scratches, and oxidation; the system generates detailed defect reports. The report includes the location, size, type of defect and the corresponding remedy for the different defects. This provides guidance to the operator so that they can better understand the defect situation and take appropriate corrective action.
The repair module is used for acquiring the defect report and generating a corresponding compensation scheme.
In this embodiment, the system may learn and identify different types of defect modes by establishing the feature extraction unit and the analysis unit in the model module, and obtain the pixel point reconstruction coefficient by reconstructing the multi-frame imageAnd reconstructing the characteristic value +.>And calculate to obtain the homogeneity difference feature +.>And pixel homogeneity density value +.>And use the homogeneity difference feature +.>And pixel homogeneity density value +.>The defect evolution index Qx and the loss area value Sy are calculated. These indexes can quantitatively evaluate the evolution degree of defects and the surface loss, and provideProviding a more thorough understanding of defect conditions. And the defect compensation coefficient Mb is calculated and obtained by fitting the defect evolution index Qx and the loss area value Sy, a detailed defect report is generated by the system according to the defect type, the defect evolution index Qx and the defect compensation coefficient Mb judged by the judging module, and the defect detection system based on image recognition can monitor the defect condition of the surface of the double-substrate radiator in real time and feed back to an operator in time by matching the corresponding compensation scheme. The defect detection system based on image recognition can realize high-speed and large-scale automatic detection, and compared with manual detection, the defect detection system based on image recognition can improve production efficiency and reduce the requirement of human resources. In addition, by early finding and disposing of defects, the system can reduce waste production and production costs.
Embodiment 2, which is an explanation performed in embodiment 1, please refer to fig. 1, specifically, the image preprocessing module includes an image denoising unit and a data enhancing unit;
the image denoising unit is used for eliminating noise by adopting one or more of mean value filtering, median filtering and Gaussian filtering, carrying out pixel value averaging or median calculation on a local area of an image, adjusting brightness by adopting histogram equalization, stretching an image range and adjusting the contrast of the image;
the data enhancement unit is used for enhancing the number and diversity of continuous image frames on the surface of the double-substrate radiator by rotation, scaling and translation, identifying the defect part and marking. The method is favorable for providing accurate labeling information in the model training stage, so that the model learns and recognizes the position and the shape of the defect, and the defect detection accuracy is improved.
In this embodiment, the image denoising unit and the data enhancement unit in the image preprocessing module have the following beneficial effects in the dual-substrate radiator surface defect detection system based on image recognition: noise is removed, and the image quality is improved; adjusting brightness and contrast, and highlighting defect characteristics; the number and the diversity of images are increased, and the generalization capability of the model is improved; and marking the defect part, so that the defect detection accuracy is improved. These effects will help to improve the defect detection capability and reliability of the system.
Embodiment 3, which is an explanation of embodiment 1, please refer to fig. 1-2, specifically, the images of x1, x2, x3, & gt, xn are obtained along the time axis, the image at the T-th time is taken for calculation, the center position i of the defective pixel point is identified in the surface image of the dual-substrate heat sink, and the gray data at the center of the defective pixel point isPixel point i is taken as center to be +.>Wherein Z extends according to the size of the defect to form a defect area, the pixel points at different positions in the defect area are analyzed, and the pixel points are set to be in the same reconstruction interval with the pixel point value at the i position of the central pixel pointThe number of pixels is recorded as +.>And the reconstruction coefficient +.A.at the position of the pixel point i is obtained by calculation using the following formula>：/>
Wherein: scale represents the normalization function, ii represents the pixel maximum, ji represents the pixel minimum,in the defective area denoted by i, remove +.>The number of pixels; d represents the numerical values of different pixel points in the corresponding defect area; s represents the numerical average area of the pixel points.
By identifying the center position i of the defective pixel point in the surface image of the dual-substrate heat sink, the system can accurately locate and position the defect. This facilitates subsequent analysis and computation, enabling the system to accurately process and determine specific defects; according to the center position i of the defective pixel point, the system can extend according to the size of the defect to form a defect area. This helps to gather the relevant pixels together to form a complete defect region for subsequent analysis and calculation, and by analyzing the pixels at different locations in the defect region, the system can calculate the reconstruction coefficients at the i-location of the pixel. This coefficient may reflect the numerical characteristics and importance of the pixel point in the defective area. By reconstructing coefficients->The severity and extent of the defect can be further assessed.
In this embodiment, the reconstruction coefficients are calculated by identifying the center position of the defective pixel pointThe dual-substrate heat sink surface defect detection system based on image recognition can obtain positioning and characteristic information about defects, and further evaluate the severity and influence range of the defects. The beneficial effects are that the system is helpful to carry out more accurate defect analysis and judgment, and the accuracy and reliability of detection are improved.
Example 4, which is an explanation of example 3, please refer to FIGS. 1-2, in particular, reconstructing eigenvaluesThe pixel point value at the position of the pixel point i is reconstructed, and an acquisition formula is as follows:
wherein, the reconstruction threshold t=0.4 is used as a decision criterion for determining which pixels are considered as part of the defect area, when the pixel i is located in the defect area of the surface of the dual-substrate radiator, the distribution of pixels in the defect area at the center of the pixel i is calculated, and when the difference between the pixel i and other pixels with the center pixel i as the center is obvious in the defect local area, the reconstruction characteristic value of the corresponding pixel i is calculatedWill increase accordingly, indicating that the defect is more pronounced.
In the present embodiment, the feature value is reconstructed by calculationAnd a reconstruction threshold value is applied, so that the defect degree and the defect obvious degree can be evaluated by the double-substrate radiator surface defect detection system based on image recognition, the pixel point distribution condition of a defect area is analyzed, and quantitative defect characteristic information is provided. These benefits help to more accurately assess the severity of defects, further analyze the characteristics and properties of defects, and provide guidance for subsequent handling and repair.
Example 5, which is an explanation of example 4, referring to FIGS. 1-2, the analysis unit is specifically configured to calculate the homogeneity-variation characteristicsThe homogeneity difference feature->The calculation steps comprise: extending eight areas to the periphery by using the defect area where the pixel point i is located, further positioning the positions of the defect areas, further calculating and analyzing the numerical characteristics of the pixel points in the areas, and calculating the reconstruction characteristic values of the extended areas to obtain corresponding reconstruction characteristic sequences, wherein the reconstruction characteristic sequences are as follows: />、、/>、...、/>Calculating the reconstruction characteristics to obtain homogeneity difference characteristics +.>And homogeneity Density number->：
Wherein 8 areas are represented by extending eight areas to the periphery of a defect area with a pixel point i, expanding eight areas from the defect area to the periphery to form nine areas (including a central area and eight peripheral areas), wherein the center of each area is represented by C, and the similarity is calculated by C, normalized and obtained as a homogeneous difference characteristicThe larger the representative difference is smaller, the smaller the representative difference is larger; m is represented as a density factor;
if reconstruction centered on pixel i obtains homogeneous difference characteristicsIs a scratch and expands the values of small difference in 8 areas, which indicates that there is still a scratch defect if the homogeneity difference feature +.>If the difference is large, judging that the scratch defect is absent;
if the homogeneity density is a valueM is of the genusWithin the density value range of the scratch, if the homogeneity density value +.>If the defect density value is higher, the defect density value representing the scratch is higher, and the defect degree is higher; if the homogeneity density value +.>If the defect density value is lower, the defect degree is lower, and the corresponding repairing means are selected differently. By using the above obtained reconstructed eigenvalue +.>Calculating to obtain homogeneity difference characteristic +.>And homogeneity Density number->8 areas in the formula represent eight areas except the central area in nine areas, C represents the central area, and the similarity is subjected to normalization processing after calculation by adopting C to obtain the homogeneity difference characteristic. Homogeneity difference characteristic value->The larger indicates the smaller the difference, and the smaller indicates the larger the difference. The above steps and judgment conditions are used to calculate the homogeneity difference feature +.>And evaluating the defect degree, and selecting proper repairing means.
Example 6, which is an explanation made in example 5, specifically, the defect evolution index Qx is calculated by the following formula:
wherein: nf represents the number of defects in the final image, < >>Representing the number of defects in the initial image. The defect evolution index Qx is used to measure the degree of change in the number of defects. The index is calculated by comparing the number of defects in the final image and the initial image. If the value of Qx is a positive number, it indicates that the number of defects in the final image is increased; if the value of Qx is negative, it indicates that the number of defects in the final image is reduced; if the value of Qx is zero, this indicates that the number of defects in the final image remains unchanged.
By calculating the defect evolution index Qx, the evolution condition of the defect can be evaluated, and whether beneficial effects are obtained in the processing process can be judged. If the value of Qx is negative, indicating a decrease in the number of defects during processing, if the value of Qx is positive or zero, indicating an increase or maintenance in the number of defects, further improvements in the processing may be desirable.
In this embodiment, the beneficial effects in the process can be primarily evaluated according to the positive and negative values of the defect evolution index Qx. However, it should be noted that Qx only accounts for variations in the number of defects, and other defect characteristics and image quality may need to be considered to obtain a more comprehensive assessment.
Embodiment 6, which is an explanation made in embodiment 5, specifically, the calculation module includes a fitting unit and a calculation unit;
the fitting unit is used for fitting the defect evolution index Qx and the loss area value Sy, and after correlation is carried out by adopting a linear regression mode, a linear regression equation is expressed as follows:where β0 is the intercept and β1 is the slope fitting factor; training a linear regression model using the sample data to find optimal β0 and β1 values so that the model can best fit the sample data; after the training is completed, the optimal result is obtainedThe fitting factor is beta 1, which represents the slope of the linear relationship between the defect evolution index Qx and the loss area value Sy; the calculating unit is configured to calculate the defect evolution index Qx and the loss area value Sy to obtain a defect compensation coefficient Mb: />The meaning of the formula is that assuming a linear relationship between Qx and Sy, the defect-compensating coefficient Mb is the result of the product of Qx and Sy.
In this embodiment, the relationship between the fitting factor obtained by linear regression fitting and the defect evolution index Qx and the loss area value Sy is used, and then the defect compensation coefficient Mb is calculated. This coefficient can be used to measure the degree of correlation between defect evolution index and area lost, and possibly to evaluate the effect of defect repair.
Embodiment 7, which is an explanation of embodiment 6, referring to fig. 1 specifically, the determining module is configured to determine a defect type according to a defect evolution index Qx, where the determining module includes a threshold unit and a determining unit;
the threshold unit is used for setting a rule threshold to determine the defect type as a judging condition;
the judging unit is used for comparing the evolution index Qx of the defect with a rule threshold value, judging the type of the defect and generating a corresponding defect report; setting a threshold A, a threshold B and a threshold C as the basis for judging the defect type;
the defect types include the following classifications: if Qx is greater than the threshold A, judging that the crack is generated;
if Qx is smaller than threshold A and greater than threshold B, judge to wear;
if Qx is smaller than the threshold B and larger than the threshold C, judging that the scratch is generated;
if Qx is smaller than the threshold C, it is determined that oxidation is occurring.
The repair module comprises a defect library and a scheme unit;
the defect library is used for collecting historical defect data and big data double-substrate radiator surface defect data and comparing the historical defect data and the big data double-substrate radiator surface defect data with the defect report;
the scheme unit is used for acquiring the defect report and generating a corresponding compensation scheme according to the information in the defect report; the repairing scheme provides a repairing method according to different defect types, and comprises the steps of filling cracks, smearing repairing agent and pickling the surface oxidized surface.
In this embodiment, the determining module determines the defect type according to the defect evolution index Qx and the rule threshold, and generates a defect report. And the repair module compares the defect library according to the information in the defect report and generates a corresponding compensation scheme to repair the defect. Thus, the selection of the repair method aiming at different defect types can be realized, and the beneficial effect is achieved.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (9)
1. A double-substrate radiator surface defect detection system based on image recognition is characterized in that: the system comprises an image acquisition module, an image preprocessing module, a model building module, a calculation module, a judgment module and an adjustment module;
the image acquisition module is used for shooting and acquiring continuous image frames of the surface of the double-substrate radiator by adopting image acquisition equipment;
the image preprocessing module is used for denoising, brightness adjustment and contrast adjustment of continuous image frames on the surface of the bipolar plate radiator;
the model building module is used for analyzing the characteristics by using a defect recognition mode algorithm and training to learn different types of defect modes; the modeling module comprises a feature extraction unit and an analysis unit;
the feature extraction unit is used for identifying preprocessed continuous image frames of the double-substrate radiator, sequencing time axis image data, marking serial numbers as x1, x2, x3, and xn, and quantifying one of the continuous image framesKey features are calculated, surface features of different pixel points in the surface image of the double-substrate radiator are calculated, and a pixel point reconstruction coefficient is obtainedAnd reconstructing the feature valuesThe method comprises the steps of carrying out a first treatment on the surface of the Reconstructing coefficients according to pixel points>And reconstructing the characteristic value +.>Further calculating to obtain homogeneity difference characteristics>The method comprises the steps of carrying out a first treatment on the surface of the The analysis unit is used for analyzing the difference characteristic according to the homogeneity>Optimizing and selecting a clustering center to perform clustering to obtain a clustering result, and calculating to obtain a pixel point homogeneity density value +.>The method comprises the steps of carrying out a first treatment on the surface of the Several homogeneity-difference features in x1, x2, x3, & gt, xn images>Comparing, obtaining defect evolution index Qx, and adding several homogeneous density values +.>Comparing to obtain a loss area value Sy; the calculation module is used for fitting the defect evolution index Qx with the loss area value Sy, calculating to obtain a defect compensation coefficient Mb, judging the defect type according to the defect evolution index Qx by the judging module, and generating a defect report by combining the defect compensation coefficient Mb; defect types include cracks, wear, scratches, and oxidation; repair module forAnd obtaining a defect report and generating a corresponding compensation scheme.
2. The dual-substrate heat sink surface defect detection system based on image recognition of claim 1, wherein: the image preprocessing module comprises an image denoising unit and a data enhancement unit;
the image denoising unit is used for eliminating noise by adopting one or more of mean value filtering, median filtering and Gaussian filtering, carrying out pixel value averaging or median calculation on a local area of an image, adjusting brightness by adopting histogram equalization, stretching an image range and adjusting the contrast of the image;
the data enhancement unit is used for enhancing the number and diversity of continuous image frames on the surface of the double-substrate radiator by rotation, scaling and translation, identifying the defect part and marking.
3. The dual-substrate heat sink surface defect detection system based on image recognition of claim 1, wherein: along a time axis, acquiring images of x1, x2, x3, and x, taking the images at the T moment, calculating, identifying the center position i of a defective pixel point in the surface image of the double-substrate radiator, and taking gray data at the center of the defective pixel point as followsPixel point i is taken as center to be +.>Wherein Z extends according to the size of the defect to form a defect area, pixels at different positions in the defect area are analyzed, and the pixel area is set to be in the same reconstruction interval as the pixel value at the i position of the central pixel point +.>The number of pixels is recorded as +.>And the reconstruction coefficient +.A.at the position of the pixel point i is obtained by calculation using the following formula>：/>Wherein: scale represents a normalization function, ii represents a pixel maximum, ji represents a pixel minimum, +.>In the defective area denoted by i, remove +.>The number of pixels; d represents the numerical values of different pixel points in the corresponding defect area; s represents the numerical average area of the pixel points.
4. The dual-substrate heat sink surface defect detection system based on image recognition of claim 1, wherein: reconstructing feature valuesThe pixel point value at the position of the pixel point i is reconstructed, and an acquisition formula is as follows:
in the formula, when the pixel point i is positioned in the defect area of the surface of the double-substrate radiator, the pixel point distribution condition of the defect area in the center of the pixel point i is calculated, and when the difference between the defect local area and other pixel points with the center of the pixel point i as the center is obvious, the reconstruction threshold value T=0.4 corresponds to the reconstruction characteristic value +_of the pixel point i>Will increase accordingly, indicating that the defect is more pronounced.
5. The dual-substrate heat sink surface defect detection system based on image recognition of claim 1, wherein: the analysis unit is used for calculating the homogeneity difference characteristicThe homogeneity difference feature->The calculation steps comprise:
extending eight areas to the periphery by using the defect area where the pixel point i is located, further calculating and analyzing the numerical characteristics of the pixel points in the areas, and calculating the reconstruction characteristic values of the extended areas to obtain corresponding reconstruction characteristic sequences, wherein the reconstruction characteristic sequences are as follows:、/>、、...、/>calculating the reconstruction characteristics to obtain homogeneity difference characteristics +.>And homogeneity Density number->： Wherein 8 regions are represented by eight regions extending to the periphery of the defective region where the pixel point i is located, the center of each region is represented by C, and the similarity is represented byC, after calculation and normalization treatment, obtaining the homogeneity difference characteristic +.>The larger the representative difference is smaller, the smaller the representative difference is larger; m is represented as a density factor; if the reconstruction centered on pixel i obtains a homogeneity difference feature +.>Is a scratch and expands the values of small difference in 8 areas, which indicates that there is still a scratch defect if the homogeneity difference feature +.>If the difference is large, judging that the scratch defect is absent; if the homogeneity density value +.>M is within the density range of the scratch, if the homogeneity density value is +.>If the defect density value is higher, the defect density value representing the scratch is higher, and the defect degree is higher; if the homogeneity density value +.>If the defect density value is lower, the defect degree is lower, and the corresponding repairing means are selected differently.
6. The dual-substrate heat sink surface defect detection system based on image recognition of claim 1, wherein: the defect evolution index Qx is calculated by the following formula:
wherein: nf represents the final imageDefect number in->Representing the number of defects in the initial image.
7. The dual-substrate heat sink surface defect detection system based on image recognition of claim 6, wherein: the computing module comprises a fitting unit and a computing unit;
the fitting unit is used for fitting the defect evolution index Qx and the loss area value Sy, and after correlation is carried out by adopting a linear regression mode, a linear regression equation is expressed as follows:
where β0 is the intercept and β1 is the slope fitting factor; training a linear regression model using the sample data to find optimal β0 and β1 values so that the model can best fit the sample data; after training is completed, the best fitting parameters beta 0 and beta 1 are obtained, the fitting factor is beta 1, and the slope of the linear relation between the defect evolution index Qx and the loss area value Sy is represented;
the calculating unit is configured to calculate the defect evolution index Qx and the loss area value Sy to obtain a defect compensation coefficient Mb:the meaning of the formula is that assuming a linear relationship between Qx and Sy, the defect-compensating coefficient Mb is the result of the product of Qx and Sy.
8. The dual-substrate heat sink surface defect detection system based on image recognition of claim 7, wherein: the judging module is used for judging the defect type according to the defect evolution index Qx, and comprises a threshold unit and a judging unit;
the threshold unit is used for setting a rule threshold to determine the defect type as a judging condition;
the judging unit is used for comparing the evolution index Qx of the defect with a rule threshold value, judging the type of the defect and generating a corresponding defect report;
the defect types include the following classifications: if Qx is greater than the threshold A, judging that the crack is generated;
if Qx is smaller than threshold A and greater than threshold B, judge to wear;
if Qx is smaller than the threshold B and larger than the threshold C, judging that the scratch is generated;
if Qx is smaller than the threshold C, it is determined that oxidation is occurring.
9. The dual-substrate heat sink surface defect detection system based on image recognition of claim 8, wherein: the repair module comprises a defect library and a scheme unit;
the defect library is used for collecting historical defect data and big data double-substrate radiator surface defect data and comparing the historical defect data and the big data double-substrate radiator surface defect data with the defect report;
the scheme unit is used for acquiring the defect report and generating a corresponding compensation scheme according to the information in the defect report; the repairing scheme provides a repairing method according to different defect types, and comprises the steps of filling cracks, smearing repairing agent and pickling the surface oxidized surface.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311045856.6A CN116777911A (en) | 2023-08-18 | 2023-08-18 | Double-substrate radiator surface defect detection system based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311045856.6A CN116777911A (en) | 2023-08-18 | 2023-08-18 | Double-substrate radiator surface defect detection system based on image recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116777911A true CN116777911A (en) | 2023-09-19 |
Family
ID=88011973
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311045856.6A Pending CN116777911A (en) | 2023-08-18 | 2023-08-18 | Double-substrate radiator surface defect detection system based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116777911A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117094609A (en) * | 2023-10-20 | 2023-11-21 | 山东卓越精工集团有限公司 | Intelligent management system for aluminum profile production quality based on machine vision |
| CN117422719A (en) * | 2023-12-19 | 2024-01-19 | 东莞市富其扬电子科技有限公司 | Production quality detection method for high-end chip radiator |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2021173566A (en) * | 2020-04-21 | 2021-11-01 | 興和株式会社 | Defect detection method and defect detection device |
| CN114170227A (en) * | 2022-02-11 | 2022-03-11 | 北京阿丘科技有限公司 | Product surface defect detection method, device, equipment and storage medium |
| CN114372983A (en) * | 2022-03-22 | 2022-04-19 | 武汉市富甸科技发展有限公司 | Shielding box coating quality detection method and system based on image processing |
| JP2022161475A (en) * | 2021-04-09 | 2022-10-21 | レーザーテック株式会社 | Defect detection device, defect detection method, image processing device and image processing program |
| CN115829984A (en) * | 2022-12-14 | 2023-03-21 | 南通恒绮纺织有限公司 | Mercerizing defect identification method for cotton fabric |
| CN116363520A (en) * | 2023-06-02 | 2023-06-30 | 青岛海滨风景区小鱼山管理服务中心 | Landscape ecological detection system for urban green land planning |
-
2023
- 2023-08-18 CN CN202311045856.6A patent/CN116777911A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2021173566A (en) * | 2020-04-21 | 2021-11-01 | 興和株式会社 | Defect detection method and defect detection device |
| JP2022161475A (en) * | 2021-04-09 | 2022-10-21 | レーザーテック株式会社 | Defect detection device, defect detection method, image processing device and image processing program |
| CN114170227A (en) * | 2022-02-11 | 2022-03-11 | 北京阿丘科技有限公司 | Product surface defect detection method, device, equipment and storage medium |
| CN114372983A (en) * | 2022-03-22 | 2022-04-19 | 武汉市富甸科技发展有限公司 | Shielding box coating quality detection method and system based on image processing |
| CN115829984A (en) * | 2022-12-14 | 2023-03-21 | 南通恒绮纺织有限公司 | Mercerizing defect identification method for cotton fabric |
| CN116363520A (en) * | 2023-06-02 | 2023-06-30 | 青岛海滨风景区小鱼山管理服务中心 | Landscape ecological detection system for urban green land planning |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117094609A (en) * | 2023-10-20 | 2023-11-21 | 山东卓越精工集团有限公司 | Intelligent management system for aluminum profile production quality based on machine vision |
| CN117094609B (en) * | 2023-10-20 | 2024-01-12 | 山东卓越精工集团有限公司 | Intelligent management system for aluminum profile production quality based on machine vision |
| CN117422719A (en) * | 2023-12-19 | 2024-01-19 | 东莞市富其扬电子科技有限公司 | Production quality detection method for high-end chip radiator |
| CN117422719B (en) * | 2023-12-19 | 2024-03-19 | 东莞市富其扬电子科技有限公司 | Production quality detection method for high-end chip radiator |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Steel surface defect detection using a new Haar–Weibull-variance model in unsupervised manner | |
| CN116777911A (en) | Double-substrate radiator surface defect detection system based on image recognition | |
| CN109977808B (en) | Wafer surface defect mode detection and analysis method | |
| CN110097034B (en) | Intelligent face health degree identification and evaluation method | |
| CN116758061B (en) | Casting surface defect detection method based on computer vision | |
| CN116843678B (en) | Hard carbon electrode production quality detection method | |
| CN107085846B (en) | Workpiece surface defect image identification method | |
| CN115690108B (en) | Aluminum alloy rod production quality assessment method based on image processing | |
| CN116109644A (en) | Surface defect detection method for copper-aluminum transfer bar | |
| CN115294140A (en) | Hardware part defect detection method and system | |
| CN115294159B (en) | Method for dividing corroded area of metal fastener | |
| CN113935666B (en) | Building decoration wall tile abnormity evaluation method based on image processing | |
| CN116137036B (en) | Gene detection data intelligent processing system based on machine learning | |
| CN116309599B (en) | Water quality visual monitoring method based on sewage pretreatment | |
| CN115984272B (en) | Semitrailer axle defect identification method based on computer vision | |
| CN117095004B (en) | Excavator walking frame main body welding deformation detection method based on computer vision | |
| CN109239073A (en) | A kind of detection method of surface flaw for body of a motor car | |
| CN112862744B (en) | Intelligent detection method for internal defects of capacitor based on ultrasonic image | |
| CN115861307B (en) | Fascia gun power supply driving plate welding fault detection method based on artificial intelligence | |
| CN116958144B (en) | Rapid positioning method and system for surface defect area of new energy connecting line | |
| CN116152242A (en) | Visual detection system of natural leather defect for basketball | |
| CN116843680B (en) | IGBT power module radiator surface defect identification method | |
| CN116778520B (en) | Mass license data quality inspection method | |
| CN117333467A (en) | Image processing-based glass bottle body flaw identification and detection method and system | |
| CN116229438B (en) | Spinning quality visual identification system |
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 |