CN113627435A - Method and system for detecting and identifying flaws of ceramic tiles - Google Patents
Method and system for detecting and identifying flaws of ceramic tiles Download PDFInfo
- Publication number
- CN113627435A CN113627435A CN202110788961.3A CN202110788961A CN113627435A CN 113627435 A CN113627435 A CN 113627435A CN 202110788961 A CN202110788961 A CN 202110788961A CN 113627435 A CN113627435 A CN 113627435A
- Authority
- CN
- China
- Prior art keywords
- ceramic tile
- picture
- tile
- network
- flaw
- 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
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
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for detecting and identifying ceramic tile flaws, which are used for acquiring ceramic tile flaw pictures and position and category information thereof in a real scene; the acquired information is transmitted to a pre-constructed ceramic tile flaw detection and recognition network training network based on deep learning to obtain trained ceramic tile flaw detection and recognition network parameters, and the trained ceramic tile flaw detection and recognition network parameters are loaded into a ceramic tile flaw detection and recognition neural network based on deep learning which is manually pre-designed; acquiring tile picture information acquired from a production line, inputting the tile picture information into a trained tile flaw detection and identification network, judging whether a current tile has flaws or not, and outputting the position and category information of the flaws of the tile if the flaws exist. The advantages are that: the speed, accuracy and robustness of detecting and identifying the ceramic tile flaws are improved.
Description
Technical Field
The invention belongs to the technical field of flaw detection and identification of ceramic tiles, and particularly relates to a flaw detection and identification method and system of ceramic tiles.
Background
In the traditional ceramic tile production links, a large amount of labor is needed, such as raw material purchasing, detection, batching, intermittent ball milling, kiln entering and exiting, glaze spraying and printing, quality inspection and sorting, packaging and warehousing and the like, and each link needs a large amount of skilled operators. When an enterprise is in a busy production season or needs a large amount of staff are required to continuously work, and people are guaranteed to be in a position for 24 hours. In recent years, tile enterprises have begun to use fully automated production lines, completing one-stop work from production to assembly. Most of enterprises in the post with high operation repeatability and high labor intensity such as packaging, brick loading, brick unloading, brick picking, stacking and the like start to realize automation, and particularly, the ink-jet printing is widely applied, so that the labor amount of the conventional procedures such as fancy glaze preparation, printing and the like is greatly reduced. The automatic production of ceramic tiles is an advanced production mode and has the characteristics of rapidness, parallelism and high efficiency. The product structure and the production process of the ceramic tile produced automatically are relatively stable, the ceramic tile manufacturing process can be divided into a plurality of procedures such as packaging, brick loading, brick unloading, brick picking, stacking and the like, and certain procedures can be properly combined and decomposed according to procedure time sequence to improve the production benefit. The ceramic tile produced automatically has high production yield and high resource utilization rate in unit time. However, the time interval in the production process of the automatic production is fast, and the interval between the processes is faster, so that higher requirements on the time complexity of the algorithm are provided.
The traditional ceramic tile flaw detection and identification technology mainly adopts manual design characteristics to carry out flaw detection and identification on the flaws of the ceramic tiles. Firstly, a tile image is obtained, then, the feature points of tile flaws in the image are calculated by using an artificially designed filter operator, and then, the flaw area position and flaw category information of the tile are calculated by an exhaustive search method based on a sliding window or a selective search based on the image area similarity. The traditional ceramic tile flaw detection and identification method is low in speed, low in accuracy and poor in robustness, and particularly in a real complex scene, the defects are more obvious.
Disclosure of Invention
The invention aims to overcome the defects that the existing ceramic tile flaw detection and identification technology uses artificial design characteristics, has low identification accuracy and low speed and cannot meet the requirement of high-precision real-time identification, and provides a rapid, accurate and robust ceramic tile flaw detection and identification method and system.
In order to solve the above technical problem, a first aspect of the present invention provides a method for detecting and identifying flaws of a tile, including:
acquiring tile template pictures, tile flaw pictures on a production line and position and category information of the tile flaw pictures;
transmitting the acquired information to a pre-constructed ceramic tile flaw detection and identification network based on deep learning, and training the network to obtain trained ceramic tile flaw detection and identification network parameters; the ceramic tile flaw detection and identification network adopts a branch type network structure, wherein the left part is a feature extraction network, the middle part is a feature fusion network, and the right part is a flaw detection network; the feature extraction network is used for extracting image features in the current data set to obtain deep image features and shallow image features; the feature fusion network is used for fusing the deep image features extracted by the feature extraction network with the shallow image features, and meanwhile, an image feature pyramid is constructed by utilizing the position information of the shallow image features and the semantic information of the deep image features to obtain the defective image features; the flaw detection network is used for carrying out flaw classification prediction and flaw position prediction on the tile image characteristics to obtain a final flaw detection result;
the method comprises the steps of acquiring a tile picture collected from a production line by using an industrial camera, inputting the tile picture into a trained tile flaw detection and identification network, judging whether flaws exist in the current tile picture or not, and outputting flaw position and category information if the flaws exist.
With reference to the first aspect, further, in order to obtain more accurate training data, the process of obtaining the tile defect picture and the position and category information thereof on the production line is as follows:
collecting a ceramic tile defect picture under a real scene on a production line, screening and filtering the ceramic tile defect picture, removing an invalid picture, and marking position information and category information of the defect; the invalid tile defect picture comprises a tile dislocation picture, a tile missing picture and a tile type error picture.
With reference to the first aspect, further, the feature extraction network of the tile defect detection and identification network is formed by serially stacking four convolution calculation units, wherein each convolution calculation unit comprises a convolution layer, a maximum pooling layer and a batch normalization layer, and the four serially stacked convolution calculation units enable the scale of the tile defect detection and identification network to be large and the depth of the whole network to be deep; the network structure can better adapt to images with high image resolution, and can improve the extraction capability of the model to the image characteristics of the tile images.
The feature fusion network consists of three parts, namely a bottom-up link, a top-down link and a transverse connection bypass, wherein the bottom-up link is used for constructing an image feature pyramid by using image features output by each stable image scale output layer, and the stable image scale output layer is a level which does not change the size of an input feature map and exists in the bottom-up link; the top-down link is used for sampling the deep image feature map with more semantic information through bilinear interpolation to change the high-level image feature map into image features with the same size as the previous-level image feature scale; the transversely connected bypass fuses the image features of different layers and adds 1 × 1 convolution, so that the advantages of reducing network parameters and saving the overall calculation amount of the network are achieved;
the network structure fuses ceramic tile defect image features of different levels together, and meanwhile, the shallow layer feature containing position information in the ceramic tile defect image and the deep layer feature containing semantic information in the ceramic tile defect image are utilized to construct an image feature pyramid, and the mixed image features are used for detecting the ceramic tile defects, so that the effect of detecting the ceramic tile defects is improved.
The flaw detection network consists of a regional candidate frame network and a classifier, wherein the classifier uses two fully connected layers and additionally uses a 1 × 1 convolution and a Spatial random selection technology Spatial Dropout.
The network structure can enable the position information and the category information of the ceramic tile flaws to mutually influence the gain training effect in the training process, and if the ceramic tile flaws exist in the prediction process, the position information and the category information of the flaws are respectively calculated to accelerate the calculation speed.
With reference to the first aspect, further, in order to obtain more accurate network parameters, the process of obtaining the trained tile defect detection and network parameter identification is as follows:
initializing ceramic tile flaw detection and identification network parameters;
preprocessing the tile template picture and the tile defect picture and the position and type information thereof, wherein the preprocessing refers to using a random image cutting algorithm for the tile template picture and the tile defect picture; because the resolution of the tile image is high, a large amount of hardware resources are consumed and the learning efficiency is not high when the original image is directly used by the model in a single learning mode, and the random image cropping algorithm is used for segmenting the original image with high resolution into a plurality of images with low resolution and simultaneously ensuring that the defect labeling target of the original image is not segmented. Therefore, the use efficiency of hardware resources is improved, the original flaw data set is further subjected to data expansion, and the accuracy of the model result is not influenced. Meanwhile, the random image cropping algorithm is also a method for weakening image data noise and increasing the stability of the characteristics of the model learning image;
inputting the preprocessed ceramic tile template picture and ceramic tile flaw picture information into a ceramic tile flaw detection and identification network to obtain image characteristics, performing down-sampling and characteristic fusion on the image characteristics, and performing logistic regression prediction on the category and position information of the ceramic tile flaw picture to obtain a prediction result;
calculating the overall network Loss by using a focus Loss function local according to the real data labeling and prediction result, wherein the overall network Loss comprises the position Loss of a primary extraction frame, the position Loss of a final prediction flaw frame and the classification Loss of the final prediction flaw frame, sequentially updating network parameters by using a back propagation algorithm and a gradient descent algorithm according to the Loss value, and continuously repeating the above processes until a preset number of rounds is reached, so that the training is finished;
and storing the trained ceramic tile flaw detection and identification network parameters.
With reference to the first aspect, in order to obtain a more accurate result, the determining whether the current tile image has a flaw and outputting the position and type information of the flaw if the current tile image has the flaw includes:
reading the stored ceramic tile flaw detection and identification network parameters, inputting ceramic tile pictures collected from an automatic production line in real time, and carrying out noise filtration and contrast adjustment pretreatment operations on the images;
inputting the preprocessed picture into a read ceramic tile flaw detection and identification network, judging whether the current picture has flaws or not by using the read network, and outputting position and category information of the flaws if the current picture has flaws.
In a second aspect, a system for detecting and identifying flaws of ceramic tiles is provided, which comprises a training information acquisition module, a network framework, a network parameter training module and a detection and identification result processing module;
the training information acquisition module is used for acquiring a ceramic tile flaw picture and position and category information thereof in a real scene;
the network parameter training module is used for transmitting the acquired information to a pre-constructed ceramic tile flaw detection and identification network based on deep learning, training the network and obtaining trained ceramic tile flaw detection and identification network parameters;
and the detection and recognition result processing module is used for acquiring the ceramic tile production picture information acquired from the automatic production line, inputting the ceramic tile defect information into a trained ceramic tile defect detection and recognition network, judging whether a current picture has defects or not and outputting the position and category information of the defects if the current picture has the defects.
With reference to the second aspect, further, the training information obtaining module includes an acquisition module, a processing module, and a labeling module;
the acquisition module is used for acquiring a ceramic tile flaw picture in a real scene by using an industrial camera;
the processing module is used for screening and filtering the collected ceramic tile flaw pictures and removing invalid pictures;
and the marking module is used for marking the position information of the flaws and the category information of the flaws of the processed ceramic tile flaw picture.
With reference to the second aspect, further, the network parameter training module includes an initialization module, a preprocessing module, a logistic regression prediction module, and a cycle training module;
the initialization module is used for initializing the ceramic tile flaw detection and identification network parameters;
the preprocessing module is used for preprocessing a ceramic tile defect picture and position and category information thereof, wherein the preprocessing refers to using a random image cutting algorithm on the ceramic tile defect picture;
the logistic regression prediction module is used for inputting the preprocessed ceramic tile flaw picture information into a ceramic tile flaw detection and identification network to obtain a feature map, performing down-sampling and feature fusion on the feature map, and then performing logistic regression prediction on the category and position information of the ceramic tile flaw picture to obtain prediction data;
and the cyclic training module is used for calculating the overall network Loss by using a Focal Loss function local Loss according to the real data label and the predicted data, sequentially updating network parameters by using a back propagation algorithm and a gradient descent algorithm according to the Loss value, continuously repeating the processes until a preset number of turns is reached, finishing training and storing the trained ceramic tile flaw detection and recognition training model network parameters.
With reference to the second aspect, further, the detection and identification result processing module includes a reading module, a preprocessing operation module, and a result obtaining module;
the reading module is used for reading the stored ceramic tile flaw detection and identification network parameters;
the preprocessing operation module is used for inputting ceramic tile production pictures collected from an automatic production line in real time and carrying out noise filtering and contrast adjustment preprocessing operations on the images;
and the result acquisition module is used for inputting the preprocessed picture into the read ceramic tile defect detection and identification network, judging whether the current picture has defects or not by using the read network, and outputting the position and the category information of the defects if the current picture has the defects.
The invention achieves the following beneficial effects:
the ceramic tile flaw detection and identification neural network based on the deep learning is designed to be trained by using the ceramic tile flaw detection and identification data set under the real scene, and ceramic tile flaw detection and identification neural network parameters based on the deep learning are obtained, so that the ceramic tile flaw detection and identification neural network can judge whether flaws exist in the current ceramic tile and position and category information of the existing flaws, and the speed, accuracy and robustness of the ceramic tile flaw detection and identification are improved.
Drawings
The foregoing and/or other advantages of the invention will become further apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings.
FIG. 1a is a picture of a collected tile defect;
FIG. 1b is a labeled information of a picture corresponding to a defective tile;
fig. 2 is a diagram of a tile defect detection and identification network structure based on deep learning according to an embodiment of the present application;
FIG. 3 is a flowchart of a tile flaw detection and identification algorithm training process based on deep learning according to an embodiment of the present application;
FIG. 4 is a flowchart of a tile flaw detection and identification algorithm prediction based on deep learning according to an embodiment of the present application;
fig. 5 is a graph of the prediction effect of the tile flaw detection and identification algorithm based on deep learning according to the embodiment of the present application.
Detailed Description
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.
The technical scheme of the invention is further explained by the specific implementation mode in combination with the attached drawings.
The first embodiment of the invention discloses a method for detecting and identifying flaws of a ceramic tile, which comprises the following steps:
acquiring tile template pictures, tile flaw pictures on a production line and position and category information of the tile flaw pictures;
transmitting the acquired information to a pre-constructed ceramic tile flaw detection and identification network based on deep learning, and training the network to obtain trained ceramic tile flaw detection and identification network parameters; the ceramic tile flaw detection and identification network adopts a branch type network structure, wherein the left part is a feature extraction network, the middle part is a feature fusion network, and the right part is a flaw detection network; the feature extraction network is used for extracting image features in the current data set to obtain deep image features and shallow image features; the feature fusion network is used for fusing the deep image features extracted by the feature extraction network with the shallow image features, and meanwhile, an image feature pyramid is constructed by utilizing the position information of the shallow image features and the semantic information of the deep image features to obtain the defective image features; the flaw detection network is used for carrying out flaw classification prediction and flaw position prediction on the tile image characteristics to obtain a final flaw detection result;
the method comprises the steps of acquiring a tile picture collected from a production line by using an industrial camera, inputting the tile picture into a trained tile flaw detection and identification network, judging whether flaws exist in the current tile picture or not, and outputting flaw position and category information if the flaws exist.
In a first embodiment, the process of acquiring the tile defect picture and the position and category information thereof on the production line includes:
collecting a ceramic tile defect picture on a production line, screening and filtering the ceramic tile defect picture, removing an invalid ceramic tile defect picture, and labeling position information of the ceramic tile defect and category information of the ceramic tile defect; the invalid tile defect picture comprises a tile dislocation picture, a tile missing picture and a tile type error picture.
FIGS. 1a and 1b illustrate tile defect detection and identification data labeling information based on deep learning. For the collected tile picture, as shown in fig. 1a, the flaw marking information of the tile is composed of 2 parts, and the first parts 'defect _ name' and 'defect _ id' are the categories to which the flaws belong; there are 6 classes of 'defect _ name' that are: edge anomaly, corner anomaly, white point defect, light color block defect, dark color block defect, and aperture defect are corresponding to 'defect _ id' as: 1. 2, 3, 4, 5 and 6. The second part 'bbox' is composed of 4 sets of numbers x1,y1,x2,y2Composition is the location information of the flaw, where x1X-axis minimum coordinate representing flaw, where y1Y-axis minimum coordinate representing flaw, where x2X-axis maximum coordinate representing flaw, where y2Y-axis maximum coordinates representing flaws; FIG. 1b is the corresponding label information, which is composed of defect type information and defect location information { x1,y1,x2,y2And (9) composition.
In a first embodiment, as shown in fig. 2, a structure diagram of a tile defect detecting and identifying network based on deep learning is shown, wherein a feature extraction network of the tile defect detecting and identifying network is formed by serially and overlapping four convolution computing units, each of which includes a convolution layer, a maximum pooling layer and a batch normalization layer;
the feature fusion network consists of three parts, namely a bottom-up link, a top-down link and a transverse connection bypass, wherein the bottom-up link is used for constructing an image feature pyramid by using image features output by each stable image scale output layer, and the stable image scale output layer is a level which does not change the size of an input feature map and exists in the bottom-up link; the top-down link is used for sampling the deep image feature map with more semantic information through bilinear interpolation to change the high-level image feature map into image features with the same size as the previous-level image feature scale; the transversely connected bypass fuses the image features of different layers and adds 1 × 1 convolution;
the flaw detection network consists of a regional candidate frame network and a classifier, wherein the classifier uses two fully connected layers and additionally uses a 1 × 1 convolution and a Spatial random selection technology Spatial Dropout.
Table 1 shows a network structure for detecting and identifying flaws in ceramic tiles based on deep learning. The feature extraction network comprises four layers of convolution residual modules; the feature fusion network comprises four layers of feature fusion modules; the classifier of the fault detection network comprises two layers of fully connected modules.
TABLE 1
In a first embodiment, as shown in fig. 3, a flowchart of a tile defect detection and recognition algorithm training based on deep learning is shown. The training process is described as follows: when training begins, firstly, initializing the structural parameters of the ceramic tile flaw detection and identification network;
inputting a tile template picture, a tile flaw picture and position and category mark information of the tile template picture and the tile flaw picture, and then preprocessing the tile picture, wherein the preprocessing refers to using a random image cutting algorithm on the tile template picture and the tile flaw picture, and comprises random rotation, random cutting, noise addition and standardization;
inputting the preprocessed ceramic tile template picture and ceramic tile defect picture as well as position and category information thereof into a ceramic tile defect detection and identification network to obtain image characteristics; performing down-sampling and feature fusion on image features, and then performing logistic regression prediction on the categories and position information of flaws to obtain a prediction result;
and calculating the overall network Loss by using a focus Loss function local Loss according to the real data label and the predicted data, wherein the overall network Loss comprises the position Loss of a primary extraction frame, the position Loss of a final predicted defective frame and the classification Loss of the final predicted defective frame, and sequentially updating the detection network parameters by using a back propagation algorithm and a gradient descent algorithm according to the Loss value. And continuously repeating the processes until the preset number of rounds is reached, finishing training and storing the detection network parameters.
In a first embodiment, as shown in fig. 4, a flowchart of a tile flaw detection and identification algorithm prediction based on deep learning is shown. The prediction process is described as follows: when the prediction is started, firstly, loading the trained structural parameters of the ceramic tile flaw detection and recognition network into a designed ceramic tile flaw detection and recognition network model;
then, ceramic tile pictures are acquired from an automatic production line in real time, and noise filtering and contrast adjustment processing operations are carried out on the pictures;
inputting the processed picture into a ceramic tile flaw detection and identification network, outputting a picture detection result by a ceramic tile flaw detection and identification network model, and outputting specific types and specific position information of flaws if the flaws exist.
FIG. 5 is a graph showing the predicted effect of the tile defect detection and identification algorithm based on deep learning, wherein a white rectangular frame is the predicted defect position of the defect detection and identification network of the tile; the white characters 3 are the flaw class information of the tile predicted by the tile flaw detection and recognition network, and are white point flaws.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The second embodiment of the invention discloses a flaw detection and recognition system of a ceramic tile, which comprises a training information acquisition module, a network frame, a network parameter training module and a detection and recognition result processing module;
the training information acquisition module is used for acquiring a ceramic tile flaw picture and position and category information thereof in a real scene;
the network parameter training module is used for transmitting the acquired information to a pre-constructed ceramic tile flaw detection and recognition network training network based on deep learning to obtain trained ceramic tile flaw detection and recognition network parameters;
and the detection and recognition result processing module is used for acquiring the ceramic tile production picture information acquired from the automatic production line, inputting the ceramic tile defect information into a trained ceramic tile defect detection and recognition network, judging whether a current picture has defects or not and outputting the position and category information of the defects if the current picture has the defects.
In a second embodiment, the training information acquisition module comprises an acquisition module, a processing module and a labeling module;
the acquisition module is used for acquiring a ceramic tile flaw picture in a real scene by using an industrial camera;
the processing module is used for screening and filtering the collected ceramic tile flaw pictures and removing invalid pictures;
and the marking module is used for marking the position information of the flaws and the category information of the flaws of the processed ceramic tile flaw picture.
In a second embodiment, the network framework and network parameter training module includes an initialization module, a preprocessing module, a logistic regression prediction module, and a cycle training module;
the initialization module is used for initializing the ceramic tile flaw detection and identification network parameters;
the preprocessing module is used for preprocessing a ceramic tile defect picture and position and category information thereof, wherein the preprocessing refers to using a random image cutting algorithm on the ceramic tile defect picture;
the logistic regression prediction module is used for inputting the preprocessed ceramic tile flaw picture information into a ceramic tile flaw detection and identification network to obtain a feature map, performing down-sampling and feature fusion on the feature map, and then performing logistic regression prediction on the category and position information of the ceramic tile flaw picture to obtain prediction data;
and the cyclic training module is used for calculating the overall network Loss by using a Focal Loss function local Loss according to the real data marking and prediction data, sequentially updating network parameters by using a back propagation algorithm and a gradient descent algorithm according to the Loss value, continuously repeating the processes until a preset number of turns is reached, finishing training and storing the trained ceramic tile flaw detection and identification network parameters.
In a second embodiment, the detection and recognition result processing module includes a reading module, a preprocessing operation module and a result obtaining module;
the reading module is used for reading the stored ceramic tile flaw detection and identification network parameters;
the preprocessing operation module is used for inputting ceramic tile production pictures collected from an automatic production line in real time and carrying out noise filtering and contrast adjustment preprocessing operations on the images;
and the result acquisition module is used for inputting the preprocessed picture into the read ceramic tile defect detection and identification network, judging whether the current picture has defects or not by using the read network, and outputting the position and the category information of the defects if the current picture has the defects.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting 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 (9)
1. A method for detecting and identifying flaws of ceramic tiles is characterized in that firstly, ceramic tile template pictures and ceramic tile flaw pictures on a production line as well as position and category information of the ceramic tile flaw pictures are obtained;
transmitting the acquired information to a pre-constructed ceramic tile flaw detection and identification network based on deep learning, and training the network to obtain trained ceramic tile flaw detection and identification network parameters; the ceramic tile flaw detection and identification network adopts a branch type network structure, wherein the left part is a feature extraction network, the middle part is a feature fusion network, and the right part is a flaw detection network; the feature extraction network is used for extracting image features in the current data set to obtain deep image features and shallow image features; the feature fusion network is used for fusing the deep image features extracted by the feature extraction network with the shallow image features, and meanwhile, an image feature pyramid is constructed by utilizing the position information of the shallow image features and the semantic information of the deep image features to obtain the defective image features; the flaw detection network is used for carrying out flaw classification prediction and flaw position prediction on the tile image characteristics to obtain a final flaw detection result;
the method comprises the steps of acquiring a tile picture collected from a production line by using an industrial camera, inputting the tile picture into a trained tile flaw detection and identification network, judging whether flaws exist in the current tile picture or not, and outputting flaw position and category information if the flaws exist.
2. The method for detecting and identifying the flaws of ceramic tiles as claimed in claim 1, wherein the process of obtaining the picture of the flaws of ceramic tiles on the production line and the location and category information thereof is as follows:
collecting a ceramic tile defect picture on a production line, screening and filtering the ceramic tile defect picture, removing an invalid ceramic tile defect picture, and labeling position information of the ceramic tile defect and category information of the ceramic tile defect; the invalid tile defect picture comprises a tile dislocation picture, a tile missing picture and a tile type error picture.
3. The method for detecting and identifying the flaws of the ceramic tile according to claim 1, wherein the feature extraction network of the ceramic tile flaw detection and identification network is formed by serially and superpositioning four convolution calculation units, wherein each convolution calculation unit comprises a convolution layer, a maximum pooling layer and a batch normalization layer;
the feature fusion network consists of three parts, namely a bottom-up link, a top-down link and a transverse connection bypass, wherein the bottom-up link is used for constructing an image feature pyramid by using image features output by each stable image scale output layer, and the stable image scale output layer is a level which does not change the size of an input feature map and exists in the bottom-up link; the top-down link is used for sampling the deep image feature map with more semantic information through bilinear interpolation to change the high-level image feature map into image features with the same size as the previous-level image feature scale; the transversely connected bypass fuses the image features of different layers and adds 1 × 1 convolution;
the flaw detection network consists of a regional candidate frame network and a classifier, wherein the classifier uses two fully connected layers and additionally uses a 1 × 1 convolution and a Spatial random selection technology Spatial Dropout.
4. The method for detecting and identifying flaws of ceramic tiles as claimed in claim 1, wherein the process of obtaining trained parameters of the network for detecting and identifying flaws of ceramic tiles is as follows:
initializing ceramic tile flaw detection and identification network parameters;
preprocessing the tile template picture and the tile defect picture and the position and type information thereof, wherein the preprocessing refers to using a random image cutting algorithm for the tile template picture and the tile defect picture;
inputting the preprocessed tile template picture and tile flaw picture as well as position and category information thereof into a tile flaw detection and identification network to obtain image characteristics; down-sampling and feature fusion are carried out on the image features; performing logistic regression prediction on the category and position information of the ceramic tile flaw picture to obtain a prediction result;
calculating the overall network Loss by using a focus Loss function local according to the real data labeling and prediction result, wherein the overall network Loss comprises the position Loss of a primary extraction frame, the position Loss of a final prediction flaw frame and the classification Loss of the final prediction flaw frame, sequentially updating network parameters by using a back propagation algorithm and a gradient descent algorithm according to the Loss value, and continuously repeating the above processes until the number of iteration rounds reaches the preset number of rounds, and finishing the training;
and storing the trained ceramic tile flaw detection and identification network.
5. The method for detecting and identifying defects of ceramic tiles as claimed in claim 1, wherein the process of determining whether defects exist in the current tile picture and outputting the position and type information of the defects if defects exist is as follows:
reading the stored ceramic tile flaw detection and identification network, inputting a ceramic tile picture collected from a production line in real time, and carrying out noise filtering and contrast adjustment processing operations on the picture;
inputting the processed picture into a ceramic tile flaw detection and identification network, judging whether the current picture has flaws by using the trained network parameters and using the ceramic tile flaw detection and identification network, and outputting the position and category information of the flaws if the current picture has flaws.
6. A ceramic tile flaw detection and recognition system is characterized by comprising a training information acquisition module, a network parameter training module and a detection and recognition result processing module;
the training information acquisition module is used for acquiring a ceramic tile flaw picture and position and category information thereof in a real scene;
the network parameter training module is used for transmitting the acquired information to a pre-constructed ceramic tile flaw detection and identification network based on deep learning, training the network and obtaining trained ceramic tile flaw detection and identification network parameters;
and the detection and recognition result processing module is used for acquiring the ceramic tile production picture information acquired from the production line, inputting the ceramic tile defect information into the trained ceramic tile defect detection and recognition network, judging whether the current picture has defects or not and outputting the position and category information of the defects if the current picture has the defects.
7. The system for detecting and identifying flaws of ceramic tiles of claim 6, wherein the training information acquisition module comprises an acquisition module, a processing module and a labeling module;
the acquisition module is used for acquiring a ceramic tile flaw picture in a real scene by using an industrial camera;
the processing module is used for screening and filtering the collected ceramic tile flaw pictures and removing invalid pictures;
and the marking module is used for marking the position information of the flaws and the category information of the flaws of the processed ceramic tile flaw picture.
8. The system of claim 6, wherein the network parameter training module comprises an initialization module, a preprocessing module, a logistic regression prediction module, and a cycle training module;
the initialization module is used for initializing the ceramic tile flaw detection and identification network parameters;
the preprocessing module is used for preprocessing a ceramic tile defect picture and position and category information thereof, wherein the preprocessing refers to using a random image cutting algorithm on the ceramic tile defect picture;
the logistic regression prediction module is used for inputting the preprocessed ceramic tile defect picture information into the ceramic tile defect detection and identification network to obtain a feature map, performing down-sampling and feature fusion on the feature map, and then performing logistic regression prediction on the category and position information of the ceramic tile defect picture to obtain a prediction result;
and the cyclic training module is used for calculating the overall network Loss by using a Focal Loss function local Loss according to the real data marking and prediction data, sequentially updating network parameters by using a back propagation algorithm and a gradient descent algorithm according to the Loss value, continuously repeating the processes until a preset number of turns is reached, finishing training and storing the trained ceramic tile flaw detection and identification network parameters.
9. The system for detecting and identifying flaws in ceramic tiles of claim 6, wherein the detection and identification result processing module comprises a reading module, a preprocessing operation module and a result acquisition module;
the reading module is used for reading the stored ceramic tile flaw detection and identification network parameters;
the preprocessing operation module is used for inputting ceramic tile production pictures collected from a production line in real time and carrying out noise filtering and contrast adjustment processing operations on the images;
and the result acquisition module is used for inputting the preprocessed picture into the read ceramic tile defect detection and identification network, judging whether the current picture has defects or not by using the read network, and outputting the position and the category information of the defects if the current picture has the defects.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110788961.3A CN113627435A (en) | 2021-07-13 | 2021-07-13 | Method and system for detecting and identifying flaws of ceramic tiles |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110788961.3A CN113627435A (en) | 2021-07-13 | 2021-07-13 | Method and system for detecting and identifying flaws of ceramic tiles |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113627435A true CN113627435A (en) | 2021-11-09 |
Family
ID=78379643
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110788961.3A Pending CN113627435A (en) | 2021-07-13 | 2021-07-13 | Method and system for detecting and identifying flaws of ceramic tiles |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113627435A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114782418A (en) * | 2022-06-16 | 2022-07-22 | 深圳市信润富联数字科技有限公司 | Detection method and device for ceramic tile surface defects and storage medium |
| CN114841915A (en) * | 2022-03-14 | 2022-08-02 | 阿里巴巴（中国）有限公司 | Tile flaw detection method and system based on artificial intelligence and storage medium |
| CN117274249A (en) * | 2023-11-20 | 2023-12-22 | 江西省中鼐科技服务有限公司 | Ceramic tile appearance detection method and system based on artificial intelligent image technology |
-
2021
- 2021-07-13 CN CN202110788961.3A patent/CN113627435A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114841915A (en) * | 2022-03-14 | 2022-08-02 | 阿里巴巴（中国）有限公司 | Tile flaw detection method and system based on artificial intelligence and storage medium |
| CN114782418A (en) * | 2022-06-16 | 2022-07-22 | 深圳市信润富联数字科技有限公司 | Detection method and device for ceramic tile surface defects and storage medium |
| CN114782418B (en) * | 2022-06-16 | 2022-09-16 | 深圳市信润富联数字科技有限公司 | Detection method and device for ceramic tile surface defects and storage medium |
| CN117274249A (en) * | 2023-11-20 | 2023-12-22 | 江西省中鼐科技服务有限公司 | Ceramic tile appearance detection method and system based on artificial intelligent image technology |
| CN117274249B (en) * | 2023-11-20 | 2024-03-01 | 江西省中鼐科技服务有限公司 | Ceramic tile appearance detection method and system based on artificial intelligent image technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110992317B (en) | PCB defect detection method based on semantic segmentation | |
| CN110543878B (en) | Pointer instrument reading identification method based on neural network | |
| CN113627435A (en) | Method and system for detecting and identifying flaws of ceramic tiles | |
| CN110111331B (en) | Honeycomb paper core defect detection method based on machine vision | |
| CN109683360B (en) | Liquid crystal panel defect detection method and device | |
| CN110599445A (en) | Target robust detection and defect identification method and device for power grid nut and pin | |
| CN112037219B (en) | Metal surface defect detection method based on two-stage convolutional neural network | |
| CN111932501A (en) | Seal ring surface defect detection method based on semantic segmentation | |
| CN110927171A (en) | Bearing roller chamfer surface defect detection method based on machine vision | |
| CN105388162A (en) | Raw material silicon wafer surface scratch detection method based on machine vision | |
| CN111523540A (en) | Metal surface defect detection method based on deep learning | |
| CN111738994B (en) | Lightweight PCB defect detection method | |
| CN115272204A (en) | Bearing surface scratch detection method based on machine vision | |
| CN116612106A (en) | Method for detecting surface defects of optical element based on YOLOX algorithm | |
| CN111027538A (en) | Container detection method based on instance segmentation model | |
| CN113158969A (en) | Apple appearance defect identification system and method | |
| CN114743102A (en) | Furniture board oriented flaw detection method, system and device | |
| CN116029979A (en) | Cloth flaw visual detection method based on improved Yolov4 | |
| CN116205876A (en) | Unsupervised notebook appearance defect detection method based on multi-scale standardized flow | |
| CN115866502A (en) | Microphone part surface defect online detection process | |
| CN114663382A (en) | Surface defect detection method for electronic component based on YOLOv5 convolutional neural network | |
| CN110458231B (en) | Ceramic product detection method, device and equipment | |
| CN115587989B (en) | Workpiece CT image defect detection segmentation method and system | |
| CN116206148A (en) | Intelligent detection, identification and classification method for decoration waste | |
| CN108986090A (en) | A kind of depth convolutional neural networks method applied to the detection of cabinet surface scratch |
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 |