CN110458231B - Ceramic product detection method, device and equipment - Google Patents
Ceramic product detection method, device and equipment Download PDFInfo
- Publication number
- CN110458231B CN110458231B CN201910740460.0A CN201910740460A CN110458231B CN 110458231 B CN110458231 B CN 110458231B CN 201910740460 A CN201910740460 A CN 201910740460A CN 110458231 B CN110458231 B CN 110458231B
- Authority
- CN
- China
- Prior art keywords
- ceramic product
- trained
- detection
- image
- result
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method, a device and equipment for detecting a ceramic product, which comprise the following steps: acquiring an image set of a ceramic product to be trained; carrying out color difference calculation on the image of the ceramic product to be trained to obtain a color difference calculation result; when the color difference calculation result is greater than or equal to a first threshold value, the detection result is unqualified, and when the color difference calculation result is smaller than the first threshold value, the detection result is qualified; classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in detection result; inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network, and training the convolutional neural network to obtain a trained convolutional neural network model; and inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result. The invention detects the ceramic product through the convolutional neural network, replaces the traditional manual visual detection method, and solves the technical problem of low detection accuracy of the existing ceramic product.
Description
Technical Field
The invention relates to the field of image detection, in particular to a method, a device and equipment for detecting a ceramic product.
Background
In the sintering, manufacturing and polishing processes of ceramic products, the products have defects due to various reasons, for example, the defects of uneven internal and external contraction of the ceramic products and certain depth are caused by high environmental humidity, high preheating and temperature rising speed and short cooling time in the sintering process, and the defects are shown as breakage, edge breakage, unfilled corners or crack defects on the products; the polishing powder has uneven strength or is doped with larger particle impurities in the polishing process, and factors such as dust contained in the production environment or surface cleaning rag, incomplete grinding in tight grinding and the like are main reasons for defects such as incomplete polishing, small spots and the like on the surface of the ceramic product. Once the product containing defects is missed to be detected and enters an installation link or enters the market, the product performance and the service life are influenced, and the public praise of consumers is also influenced. Therefore, strict appearance inspection of the ceramic product is required.
The existing ceramic product detection has low automation degree, still stays in a manual visual detection stage, and detection personnel detect whether the surface of the ceramic product has defects through the prior knowledge acquired in the past so as to judge whether the ceramic product is qualified or unqualified. However, the detection accuracy is not high due to the fact that the same object is repeatedly detected under the influence of heavy detection workload, detection personnel are prone to fatigue and lacked psychological emotions, and defects and omission easily occur.
Disclosure of Invention
The invention provides a method, a device and equipment for detecting a ceramic product, which are used for solving the technical problem that the existing ceramic product is low in detection accuracy.
In view of the above, the first aspect of the present application provides a method for inspecting a ceramic product, including:
acquiring an image set of a ceramic product to be trained;
performing color difference calculation on the ceramic product image to be trained in the ceramic product image set to be trained to obtain a color difference calculation result;
when the color difference calculation result is larger than or equal to a first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified, and when the color difference calculation result is smaller than the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified;
classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in the detection result;
inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network, and training the convolutional neural network;
calculating the detection accuracy of the ceramic product image, and finishing training when the detection accuracy is higher than a second threshold value to obtain a trained convolutional neural network model;
and inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result.
Preferably, the acquiring an image of a ceramic product to be trained further comprises:
and constructing a detection table with consistent light sources, and acquiring a ceramic product image set with multiple visual angles through the detection table, wherein the ceramic product image set comprises a ceramic product image set to be trained and/or a ceramic product image set to be detected.
Preferably, the classifying and storing the images of the ceramic products with the qualified and unqualified detection results further comprises:
carrying out binarization processing on the ceramic product image to be trained;
performing feature extraction on the binaryzation-processed image of the ceramic product to be trained to obtain a feature vector;
and inputting the feature vector into a trained classifier, and outputting a detection result, wherein the detection result comprises a qualified result or an unqualified result.
Preferably, the performing feature extraction on the binarized image of the ceramic product to be trained to obtain a feature vector includes:
performing inclination angle extraction and adaptive Canny edge detection on the binaryzation-processed ceramic product image to be trained by adopting fuzzy adaptive Hough transformation to obtain a feature vector;
or
And performing feature extraction on the binaryzation-processed image of the ceramic product to be trained by adopting a SURF algorithm to obtain a feature vector.
Preferably, the classifying and storing the images of the ceramic products with the qualified and unqualified detection results further comprises:
establishing a three-dimensional model of the ceramic product in the image of the ceramic product to be trained;
comparing the size of the three-dimensional model of the ceramic product with that of a standard three-dimensional model, and quantizing the size comparison result to obtain a deformation result of the ceramic product;
and when the deformation result is greater than or equal to a third threshold value, the detection result is unqualified, and when the deformation result is less than the third threshold value, the detection result is qualified.
The present application provides in a second aspect a ceramic product detecting device comprising:
the image acquisition module is used for acquiring an image set of the ceramic product to be trained;
the color difference calculation module is used for performing color difference calculation on the ceramic product image to be trained of the ceramic product image set to be trained to obtain a color difference calculation result;
the color difference detection module is used for determining that the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified when the color difference calculation result is greater than or equal to a first threshold value, and determining that the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified when the color difference calculation result is smaller than the first threshold value;
the classified storage module is used for classifying and storing the images of the ceramic products to be trained, wherein the detection results of the images are qualified and unqualified;
the training module is used for inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network and training the convolutional neural network;
the calculation module is used for calculating the detection accuracy of the ceramic product image, and when the detection accuracy is higher than a second threshold value, the training is completed to obtain a trained convolutional neural network model;
and the first detection module is used for inputting the image of the ceramic product to be detected into the trained convolutional neural network model and outputting a detection result.
Preferably, the method further comprises the following steps:
the detection table building module is used for building a detection table with consistent light sources, and the detection table is used for collecting a ceramic product image set with multiple visual angles, wherein the ceramic product image set comprises a ceramic product image set to be trained and/or a ceramic product image set to be detected.
Preferably, the method further comprises the following steps:
the binarization processing module is used for carrying out binarization processing on the ceramic product image to be trained;
the characteristic extraction module is used for extracting the characteristics of the ceramic product image to be trained after the binarization processing to obtain a characteristic vector;
and the second detection module is used for inputting the feature vectors into the trained classifier and outputting detection results, wherein the detection results comprise pass or fail.
Preferably, the method further comprises the following steps:
the three-dimensional model establishing module is used for establishing a three-dimensional model of the ceramic product in the image of the ceramic product to be trained;
the size comparison module is used for comparing the size of the three-dimensional model of the ceramic product with that of a standard three-dimensional model and quantizing the size comparison result to obtain a deformation result of the ceramic product;
and the size detection module is used for judging that the detection result is unqualified when the deformation result is greater than or equal to a third threshold value, and judging that the detection result is qualified when the deformation result is smaller than the third threshold value.
The third aspect of the present application provides a ceramic product inspection apparatus comprising: the apparatus includes a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the ceramic product detection method of any one of the first aspect according to instructions in the program code.
According to the technical scheme, the embodiment of the application has the following advantages:
the application provides a method for detecting a ceramic product, which comprises the following steps: acquiring an image set of a ceramic product to be trained; performing color difference calculation on the ceramic product image to be trained of the ceramic product image set to be trained to obtain a color difference calculation result; when the color difference calculation result is larger than or equal to a first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified, and when the color difference calculation result is smaller than the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified; classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in detection result; inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network, and training the convolutional neural network; calculating the detection accuracy of the ceramic product image, and finishing training when the detection accuracy is higher than a second threshold value to obtain a trained convolutional neural network model; and inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result.
The ceramic product detection method provided by the application is characterized in that when the ceramic product is subjected to color difference detection, detection personnel are easy to make a detection error, particularly, the detection personnel are difficult to detect fine color difference, the color difference calculation result is obtained through color difference calculation, the color difference calculation result is compared with a first threshold value, the obtained detection result is qualified ceramic product or unqualified ceramic product, the detection accuracy is higher than that of a manual visual detection method, the obtained data set is more accurate than a manually classified data set, an accurate data set is favorable for training a better convolutional neural network model and is favorable for improving the detection accuracy of the convolutional neural network model, therefore, the convolutional neural network is adopted for ceramic product detection, the traditional manual visual detection method is replaced, and the problem that the detection personnel are easy to have defect omission is solved, thereby leading to the technical problem of low detection accuracy.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a first embodiment of a method for inspecting a ceramic product according to the present invention;
FIG. 2 is a schematic flow chart of a ceramic product inspection method according to a second embodiment of the present invention;
FIG. 3 is a schematic flow chart of a method for inspecting a ceramic product according to a third embodiment of the present invention;
fig. 4 is a schematic structural diagram of an embodiment of a ceramic product detection apparatus provided by the present invention.
Detailed Description
The invention provides a method, a device and equipment for detecting a ceramic product, which are used for solving the technical problem that the existing ceramic product is low in detection accuracy.
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the embodiments described below are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For easy understanding, referring to fig. 1, the present invention provides a first embodiment of a method for inspecting a ceramic product, comprising:
It should be noted that, the obtained image set of the ceramic product to be trained needs to be screened, and some images with low quality are screened out.
And 102, performing color difference calculation on the ceramic product image to be trained in the ceramic product image set to be trained to obtain a color difference calculation result.
It should be noted that, in the embodiment of the present application, color difference calculation may be performed based on the CIELAB color space, that is, the ceramic product image to be trained and the standard ceramic product image are first converted into the CIEXYZ color space, and then converted into the CIELAB color space from the CIEXYZ color space, so as to obtain the ceramic product image to be trained and the standard ceramic product image in the CIELAB color space, and the lightness L between the ceramic product image to be trained and the standard ceramic product image is calculated*And the color index a*And b*Assuming a standard ceramic product image in CIELAB color space as
The image of the ceramic product to be trained in CIELAB color space is
Performing color difference calculation on the image of the ceramic product to be trained to obtain a color difference calculation result
The formula for the color difference is as follows:
and 103, when the color difference calculation result is greater than or equal to the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified, and when the color difference calculation result is less than the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified.
The color difference calculation result is compared with a first threshold, and when the color difference result is greater than or equal to the first threshold, the difference between the surface color of the actually produced ceramic product and the surface color of the standard ceramic product is larger, and the ceramic product is unqualified; and when the color difference result is smaller than a first threshold value, the surface color of the actually produced ceramic product is relatively close to the surface color of the standard ceramic product, and the ceramic product is qualified.
The setting of the first threshold is determined according to the actual situation, and when the color difference of the ceramic product has little influence on the use of the ceramic product, the first threshold can be properly set to be a larger value; when the color difference of the ceramic seriously affects the use value and the ornamental value of the ceramic product, the first threshold value may be appropriately set to a small value.
And 104, classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in detection result.
It should be noted that, the ceramic product image is detected in step 103 to obtain detection results, that is, the ceramic product is qualified and the ceramic product is unqualified, and the ceramic product image corresponding to the qualified ceramic product and the unqualified ceramic product is classified and stored, and is divided into two classes, namely, qualified ceramic product and unqualified ceramic product, so that the subsequent convolutional neural network model is subjected to two-class training.
The color difference detection is that detection personnel are easy to make a detection error, particularly, for slight color difference, the detection personnel are not easy to detect, but the embodiment requires color difference calculation to obtain a color difference calculation result, the color difference calculation result is compared with a first threshold value, the obtained detection result is that a ceramic product is qualified or the ceramic product is unqualified, the detection accuracy is higher than that of a manual visual detection method, the obtained data set is more accurate than a manually classified data set, and an accurate data set is favorable for training a better convolutional neural network model and is favorable for improving the detection accuracy of the model.
And 105, inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network, and training the convolutional neural network.
The convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and an output layer, stored qualified and unqualified ceramic product images are used as input and input into a convolutional neural network model through the input layer, feature extraction and feature selection are carried out through a series of convolutional layers and pooling layers, abstract and deep features are extracted, secondary classification is carried out through a Softmax classifier, and finally a detection result is output through the output layer.
And 106, calculating the detection accuracy of the ceramic product image, and finishing training when the detection accuracy is higher than a second threshold value to obtain a trained convolutional neural network model.
It should be noted that the detection accuracy is obtained by calculating the ratio of the number of correctly detected training images to the number of all training images, and when the detection accuracy is higher than a preset second threshold, the training is considered to be completed, and the training is stopped, so as to obtain a trained convolutional neural network model.
And 107, inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result.
It should be noted that, the convolutional neural network model is adopted to detect the ceramic product, so that excessive manual interference is not needed, and the labor cost is reduced.
The ceramic product detection method provided by the embodiment of the application is easy for detection personnel to make mistakes when the ceramic product is subjected to color difference detection, particularly for slight color difference, the detection personnel are difficult to detect, the color difference calculation result is obtained through color difference calculation, the color difference calculation result is compared with the first threshold value, the detection result is qualified ceramic product or unqualified ceramic product, the detection accuracy is higher than that of a manual visual detection method, the obtained data set is more accurate than a manually classified data set, an accurate data set is favorable for training a better convolutional neural network model and is favorable for improving the detection accuracy of the convolutional neural network model, therefore, the convolutional neural network is adopted for ceramic product detection, the traditional manual visual detection method is replaced, and the problem that the detection personnel are easy to have defect omission is solved, thereby leading to the technical problem of low detection accuracy.
For easy understanding, please refer to fig. 2, which is a schematic flow chart of a ceramic product detection method according to a second embodiment of the present invention.
It should be noted that, the selection of the light source is the key to ensure the consistency of the light source, and the light source selection techniques in different detection scenes: the method has the advantages that the contrast is created by using colors, the fast moving object is irradiated by using flat flash, reflection is eliminated by using infrared rays, color change is reduced by using infrared rays, and the corresponding light source is selected according to actual needs, so that subsequent feature extraction and three-dimensional modeling are facilitated.
And acquiring a ceramic product image set from a plurality of visual angles through a plurality of depth cameras of the detection table, wherein the ceramic product image set comprises a ceramic product image set to be trained and/or a ceramic product image set to be detected, and preparing for subsequent feature extraction and three-dimensional modeling.
And 203, performing color difference calculation on the ceramic product image to be trained in the ceramic product image set to be trained to obtain a color difference calculation result.
And 204, when the color difference calculation result is greater than or equal to the first threshold, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified, and when the color difference calculation result is less than the first threshold, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified.
And step 205, performing binarization processing on the ceramic product image to be trained.
It should be noted that the binarization processing is a step in image preprocessing, and is intended to distinguish between foreground and background to obtain an interested part of an image, i.e., a feature point, and the binarization processing on the image can reduce feature dimensions and improve the operation efficiency of an algorithm.
And step 206, performing feature extraction on the binaryzation processed image of the ceramic product to be trained to obtain a feature vector.
The method comprises the steps of performing fuzzy self-adaptive Hough transformation on a ceramic product image to be trained after binarization processing to extract an inclination angle and perform self-adaptive Canny edge detection to obtain a feature vector related to the outline information of the ceramic product; or performing feature extraction on the ceramic product image to be trained after binarization processing by adopting an SURF algorithm to obtain a feature vector.
The method for extracting the appearance characteristics of the ceramic product can extract the defect characteristics of the surface of the ceramic product, and is beneficial to subsequent detection of a classifier, so that the detection accuracy is improved.
And step 207, inputting the feature vectors into the trained classifier, and outputting a detection result, wherein the detection result comprises a qualified result or an unqualified result.
The obtained feature vectors are input into a trained two-classifier to be detected, so that appearance detection results are obtained, and when the detection results are qualified, the surface of the ceramic product is free of defects; and when the detection result is unqualified, the surface of the ceramic product is indicated to have defects. Wherein, the classifier can be an SVM classifier or a KNN classifier.
In the actual detection, the manual visual detection cannot completely detect whether the appearance of the ceramic product has defects, particularly some slight surface defects, and the ceramic product with the appearance defects is easily detected as a qualified ceramic product; the data set obtained through the appearance detection is more accurate than the data set obtained through manual classification, and an accurate data set is beneficial to training a better convolutional neural network model and improving the detection accuracy of the model.
In this embodiment, the sequence relationship between steps 205 to 207 and steps 203 to 204 does not exist, steps 205 to 207 and steps 203 to 204 may be performed simultaneously, and steps 205 to 207 may be before or after steps 203 to 204.
And step 208, classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in detection result.
And step 209, inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network, and training the convolutional neural network.
And step 210, calculating the detection accuracy of the ceramic product image, and finishing training when the detection accuracy is higher than a second threshold value to obtain a trained convolutional neural network model.
And step 211, inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result.
It should be noted that steps 208 to 211 in the present embodiment are the same as steps 104 to 107 in the previous embodiment, and are not described herein again.
For easy understanding, please refer to fig. 3, which is a schematic flow chart of a ceramic product detection method according to a third embodiment of the present invention.
And 303, performing color difference calculation on the ceramic product image to be trained in the ceramic product image set to be trained to obtain a color difference calculation result.
And 304, when the color difference calculation result is greater than or equal to the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified, and when the color difference calculation result is less than the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified.
And 305, performing binarization processing on the ceramic product image to be trained.
And step 306, performing feature extraction on the binaryzation-processed image of the ceramic product to be trained to obtain a feature vector.
And 307, inputting the feature vectors into the trained classifier, and outputting a detection result, wherein the detection result comprises a qualified result or an unqualified result.
It should be noted that steps 301 to 307 in this embodiment are the same as steps 201 to 207 in the previous embodiment, and are not described herein again.
It should be noted that, according to the ceramic product images and depth information of multiple viewing angles acquired by the detection table, a three-dimensional model of the ceramic product is established by using three-dimensional modeling software.
And 309, comparing the size of the three-dimensional model of the ceramic product with that of the standard three-dimensional model, and quantizing the size comparison result to obtain the deformation result of the ceramic product.
It should be noted that the standard three-dimensional model size of the ceramic product is subtracted from the three-dimensional model size of the ceramic product obtained through modeling, and the obtained difference is the deformation result, and the deformation result reflects the deformation degree of the ceramic product.
And 310, when the deformation result is greater than or equal to the third threshold value, the detection result is unqualified, and when the deformation result is less than the third threshold value, the detection result is qualified.
The deformation result is compared with a third threshold value, and when the deformation result is greater than or equal to the third threshold value, the deviation between the size of the actually produced ceramic product and the size of the standard ceramic product is larger, and the ceramic product is unqualified; and when the deformation result is smaller than the third threshold value, the deviation between the size of the actually produced ceramic product and the size of the standard ceramic product is smaller, and the ceramic product is qualified.
In the actual detection, whether the size of the ceramic product has deviation or not and the deviation amount cannot be completely detected through manual visual detection, the ceramic product with the size defect can be easily detected as the qualified ceramic product, in the embodiment, the ceramic product with the size defect is subjected to size detection, the ceramic product image with unqualified size and qualified size is obtained, the ceramic product with the size defect is prevented from being detected as the qualified ceramic product, and the detection accuracy of the ceramic product is improved; the data set obtained by the size comparison method is more accurate than the data set obtained by manual classification, and an accurate data set is beneficial to training a better convolutional neural network model and improving the detection accuracy of the model.
In this embodiment, there is no chronological sequence relationship between steps 308 to 310 and steps 303 to 404 and 305 to 307, for example, steps 308 to 310 may be performed simultaneously with steps 303 to 404 and 305 to 307, or before or after steps 303 to 404, or before or after steps 305 to 307, or after steps 303 to 404.
And 311, classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in detection result.
And step 312, inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into the convolutional neural network, and training the convolutional neural network.
And 313, calculating the detection accuracy of the ceramic product image, and finishing training when the detection accuracy is higher than a second threshold value to obtain a trained convolutional neural network model.
And step 314, inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result.
It should be noted that steps 311 to 314 in the present embodiment are the same as steps 104 to 107 in the first embodiment, and are not repeated herein.
For easy understanding, referring to fig. 4, an embodiment of a ceramic product detecting apparatus according to the present invention includes:
the image acquisition module 401 is used for acquiring an image set of the ceramic product to be trained;
the color difference calculating module 402 is used for performing color difference calculation on the ceramic product image to be trained of the ceramic product image set to be trained to obtain a color difference calculation result;
the color difference detection module 403 is configured to determine that the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified when the color difference calculation result is greater than or equal to the first threshold, and determine that the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified when the color difference calculation result is less than the first threshold;
a classification storage module 404, configured to classify and store the images of the ceramic products to be trained, which are qualified and unqualified in the detection result;
the training module 405 is used for inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into the convolutional neural network, and training the convolutional neural network;
the calculation module 406 is used for calculating the detection accuracy of the ceramic product image, and when the detection accuracy is higher than a second threshold, the training is completed to obtain a trained convolutional neural network model;
the first detection module 407 is configured to input the image of the ceramic product to be detected into the trained convolutional neural network model, and output a detection result.
Further, still include:
and the detection table building module 408 is used for building a detection table with consistent light sources, and acquiring a ceramic product image set with multiple visual angles through the detection table, wherein the ceramic product image set comprises a ceramic product image set to be trained and/or a ceramic product image set to be detected.
Further, still include:
a binarization processing module 409 for performing binarization processing on the ceramic product image to be trained;
the feature extraction module 410 is configured to perform feature extraction on the ceramic product image to be trained after binarization processing to obtain a feature vector;
the second detection module 411 is configured to input the feature vector into the trained classifier, and output a detection result, where the detection result includes pass or fail.
Further, still include:
a three-dimensional model building module 412, configured to build a three-dimensional model of a ceramic product in an image of the ceramic product to be trained;
the size comparison module 413 is used for comparing the size of the three-dimensional model of the ceramic product with that of the standard three-dimensional model and quantizing the size comparison result to obtain a deformation result of the ceramic product;
and the size detection module 414 is configured to determine that the detection result is unqualified when the deformation result is greater than or equal to the third threshold, and determine that the detection result is qualified when the deformation result is less than the third threshold.
The application also provides a ceramic product detection device, which comprises a processor and a memory:
the memory is used for storing the program codes and transmitting the program codes to the processor;
the processor is configured to execute the ceramic product inspection method of the foregoing embodiments of the ceramic product inspection method according to instructions in the program code.
The above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (7)
1. A method of inspecting a ceramic product, comprising:
acquiring an image set of a ceramic product to be trained;
performing color difference calculation on the ceramic product image to be trained in the ceramic product image set to be trained to obtain a color difference calculation result;
when the color difference calculation result is larger than or equal to a first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified, and when the color difference calculation result is smaller than the first threshold value, the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified;
carrying out binarization processing on the ceramic product image to be trained;
performing feature extraction on the ceramic product image to be trained after the binarization processing to obtain a feature vector, specifically, performing inclination angle extraction and adaptive Canny edge detection on the ceramic product image to be trained after the binarization processing by adopting fuzzy adaptive Hough transformation to obtain the feature vector; or, performing feature extraction on the binaryzation-processed image of the ceramic product to be trained by adopting an SURF algorithm to obtain a feature vector;
inputting the feature vector into a trained classifier, and outputting a detection result, wherein the detection result comprises a qualified result or an unqualified result, and the classifier is an SVM classifier or a KNN classifier;
classifying and storing the images of the ceramic products to be trained, which are qualified and unqualified in the detection result;
inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network, and training the convolutional neural network;
calculating the detection accuracy of the ceramic product image, and finishing training when the detection accuracy is higher than a second threshold value to obtain a trained convolutional neural network model;
and inputting the image of the ceramic product to be detected into the trained convolutional neural network model, and outputting a detection result.
2. The method for inspecting ceramic products according to claim 1, wherein the step of obtaining the image of the ceramic product to be trained further comprises:
and constructing a detection table with consistent light sources, and acquiring a ceramic product image set with multiple visual angles through the detection table, wherein the ceramic product image set comprises a ceramic product image set to be trained and/or a ceramic product image set to be detected.
3. The method for detecting ceramic products according to claim 1, wherein the step of classifying and storing the images of the ceramic products with the qualified and unqualified detection results further comprises the following steps:
establishing a three-dimensional model of the ceramic product in the image of the ceramic product to be trained;
comparing the size of the three-dimensional model of the ceramic product with that of a standard three-dimensional model, and quantizing the size comparison result to obtain a deformation result of the ceramic product;
and when the deformation result is greater than or equal to a third threshold value, the detection result is unqualified, and when the deformation result is less than the third threshold value, the detection result is qualified.
4. A ceramic product testing apparatus, comprising:
the image acquisition module is used for acquiring an image set of the ceramic product to be trained;
the color difference calculation module is used for performing color difference calculation on the ceramic product image to be trained of the ceramic product image set to be trained to obtain a color difference calculation result;
the color difference detection module is used for determining that the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is unqualified when the color difference calculation result is greater than or equal to a first threshold value, and determining that the detection result of the ceramic product corresponding to the image of the ceramic product to be trained is qualified when the color difference calculation result is smaller than the first threshold value;
the binarization processing module is used for carrying out binarization processing on the ceramic product image to be trained;
the characteristic extraction module is used for extracting the characteristics of the ceramic product image to be trained after the binarization processing to obtain a characteristic vector;
the feature extraction module is specifically used for performing inclination angle extraction and adaptive Canny edge detection on the binaryzation-processed ceramic product image to be trained by adopting fuzzy adaptive Hough transformation to obtain a feature vector; or, performing feature extraction on the binaryzation-processed image of the ceramic product to be trained by adopting an SURF algorithm to obtain a feature vector;
the second detection module is used for inputting the feature vectors into a trained classifier and outputting detection results, wherein the detection results comprise pass or fail, and the classifier is an SVM classifier or a KNN classifier;
the classified storage module is used for classifying and storing the images of the ceramic products to be trained, wherein the detection results of the images are qualified and unqualified;
the training module is used for inputting the qualified and unqualified ceramic product images to be trained which are stored in a classified manner into a convolutional neural network and training the convolutional neural network;
the calculation module is used for calculating the detection accuracy of the ceramic product image, and when the detection accuracy is higher than a second threshold value, the training is completed to obtain a trained convolutional neural network model;
and the first detection module is used for inputting the image of the ceramic product to be detected into the trained convolutional neural network model and outputting a detection result.
5. The ceramic product detecting device according to claim 4, further comprising:
the detection table building module is used for building a detection table with consistent light sources, and the detection table is used for collecting a ceramic product image set with multiple visual angles, wherein the ceramic product image set comprises a ceramic product image set to be trained and/or a ceramic product image set to be detected.
6. The ceramic product detecting device according to claim 4, further comprising:
the three-dimensional model establishing module is used for establishing a three-dimensional model of the ceramic product in the image of the ceramic product to be trained;
the size comparison module is used for comparing the size of the three-dimensional model of the ceramic product with that of a standard three-dimensional model and quantizing the size comparison result to obtain a deformation result of the ceramic product;
and the size detection module is used for judging that the detection result is unqualified when the deformation result is greater than or equal to a third threshold value, and judging that the detection result is qualified when the deformation result is smaller than the third threshold value.
7. A ceramic product testing apparatus, comprising: the apparatus includes a processor and a memory:
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the ceramic product detection method of any one of claims 1-3 according to instructions in the program code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910740460.0A CN110458231B (en) | 2019-08-12 | 2019-08-12 | Ceramic product detection method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910740460.0A CN110458231B (en) | 2019-08-12 | 2019-08-12 | Ceramic product detection method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110458231A CN110458231A (en) | 2019-11-15 |
| CN110458231B true CN110458231B (en) | 2022-05-10 |
Family
ID=68485922
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910740460.0A Active CN110458231B (en) | 2019-08-12 | 2019-08-12 | Ceramic product detection method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110458231B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113554716A (en) * | 2021-07-28 | 2021-10-26 | 广东工业大学 | Knowledge distillation-based tile color difference detection method and device |
| CN113344919B (en) * | 2021-08-02 | 2021-11-23 | 苏州大学 | Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919899A (en) * | 2017-12-13 | 2019-06-21 | 香港纺织及成衣研发中心有限公司 | The method for evaluating quality of image based on multispectral imaging |
| CN108038853B (en) * | 2017-12-18 | 2020-05-26 | 浙江工业大学 | Ceramic tile surface defect identification method based on convolutional neural network and active learning |
| CN208091372U (en) * | 2018-01-22 | 2018-11-13 | 广东理工学院 | A kind of tile detection device based on RGB-D cameras |
-
2019
- 2019-08-12 CN CN201910740460.0A patent/CN110458231B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110458231A (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
| US10964004B2 (en) | Automated optical inspection method using deep learning and apparatus, computer program for performing the method, computer-readable storage medium storing the computer program, and deep learning system thereof | |
| US11132786B2 (en) | Board defect filtering method based on defect list and circuit layout image and device thereof and computer-readable recording medium | |
| CN108355981B (en) | Battery connector quality detection method based on machine vision | |
| CN111383209B (en) | Unsupervised flaw detection method based on full convolution self-encoder network | |
| CN111833306B (en) | Defect detection method and model training method for defect detection | |
| CN107486415B (en) | Thin bamboo strip defect online detection system and detection method based on machine vision | |
| CN110286124B (en) | Machine vision-based refractory brick measuring system | |
| CN112088387B (en) | System and method for detecting defects in an imaged article | |
| CN110610475B (en) | Visual defect detection method of deep convolutional neural network | |
| CN116188475B (en) | Intelligent control method, system and medium for automatic optical detection of appearance defects | |
| CN111582294A (en) | Method for constructing convolutional neural network model for surface defect detection and application thereof | |
| CN112233067A (en) | Hot rolled steel coil end face quality detection method and system | |
| CN110473201A (en) | A kind of automatic testing method and device of disc surface defect | |
| CN110458231B (en) | Ceramic product detection method, device and equipment | |
| CN113516619B (en) | Product surface flaw identification method based on image processing technology | |
| CN111161237A (en) | Fruit and vegetable surface quality detection method, storage medium and sorting device thereof | |
| CN115147363A (en) | Image defect detection and classification method and system based on deep learning algorithm | |
| CN104952754A (en) | Coated silicon chip sorting method based on machine vision | |
| CN115471476A (en) | Method, device, equipment and medium for detecting component defects | |
| CN113019973A (en) | Online visual inspection method for manufacturing defects of ring-pull cans | |
| CN111178405A (en) | Similar object identification method fusing multiple neural networks | |
| Liu et al. | Application of statistical modeling of image spatial structures to automated visual inspection of product quality | |
| Zhang et al. | Fabric defect detection based on visual saliency map and SVM | |
| CN117173172B (en) | Machine vision-based silica gel molding effect detection method and 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |