CN111558542A - Ceramic tile surface quality online detection sorting system and method - Google Patents
Ceramic tile surface quality online detection sorting system and method Download PDFInfo
- Publication number
- CN111558542A CN111558542A CN202010586640.0A CN202010586640A CN111558542A CN 111558542 A CN111558542 A CN 111558542A CN 202010586640 A CN202010586640 A CN 202010586640A CN 111558542 A CN111558542 A CN 111558542A
- Authority
- CN
- China
- Prior art keywords
- ceramic tile
- module
- sorting
- conveying belt
- detection
- 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.)
- Withdrawn
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/02—Measures preceding sorting, e.g. arranging articles in a stream orientating
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/34—Sorting according to other particular properties
- B07C5/342—Sorting according to other particular properties according to optical properties, e.g. colour
- B07C5/3422—Sorting according to other particular properties according to optical properties, e.g. colour using video scanning devices, e.g. TV-cameras
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/36—Sorting apparatus characterised by the means used for distribution
- B07C5/361—Processing or control devices therefor, e.g. escort memory
- B07C5/362—Separating or distributor mechanisms
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Sorting Of Articles (AREA)
Abstract
The invention discloses an online detection and sorting system and method for surface quality of ceramic tiles, wherein the detection and sorting system comprises a conveying belt, an imaging module arranged on the conveying belt, an identification prediction module connected with the imaging module, and a sorting module connected with the identification prediction module, wherein the sorting module is positioned on the output side of the imaging module; the imaging module is used for acquiring a ceramic tile image and transmitting the ceramic tile image to the recognition and prediction module, the recognition and prediction module is used for recognizing the ceramic tile surface type, color and/or surface flaw of the ceramic tile image, classifying the ceramic tile according to the recognition result and transmitting the classification result to the sorting module; and the sorting module unloads the ceramic tiles from the conveying belt to corresponding classifications according to the received classification results. The ceramic tile color sorting machine has the advantages that the color, the surface type and the flaw of the ceramic tile can be rapidly detected, the ceramic tile can be automatically sorted, the production efficiency of an enterprise line can be favorably improved, and the like.
Description
Technical Field
The invention relates to the technical field of ceramic tile surface quality detection and sorting, in particular to a ceramic tile surface quality on-line detection and sorting system and method.
Background
The ceramic tile is a kind of roof building material, and is generally rectangular, its size is about 400mmX300mm, and the front surface of the tile body has longitudinal grooves, and its tile form and colour are various, and the tile form includes wave form or step form, and its front surface has different textures. The production of the ceramic tile needs to be carried out by processes of forming, sintering, glazing and the like as other ceramic products, and defects such as unfilled corners, cracks, scratches, bulges, unglazed glaze, overglaze and the like can be caused to the ceramic tile due to various reasons, so that the quality of the product is seriously influenced. Before being packed down on line, the ceramic tiles must be inspected for defects on their surfaces and sorted into rejects, excellences and first-class products according to the type, number and extent of the defects. In addition, in order to prevent the aliasing and packaging phenomena of ceramic tiles with different colors and different surface types from occurring during packaging, the color and surface type identification and judgment of the ceramic tiles are also required to be completed before packaging, and the ceramic tiles which do not belong to the surface type or the color are judged to be unqualified and do not participate in subsequent packaging.
In the traditional ceramic tile production and manufacturing process, due to the limitation of the technical level, a manual detection and sorting mode is usually adopted, the quality of the surface of the ceramic tile is identified by only visual observation, and then manual sorting and classification are carried out. There are significant limitations to this approach: the ceramic tile detection speed is low and the efficiency is low; labor is intensive and labor cost is high; the subjectivity is strong, the fatigue is easy, and the omission factor is high; the detection precision is low, quantification cannot be realized, and the false detection rate is high; the enterprise investment and the actual output are unbalanced; the repetitive labor is difficult to recruit workers.
At present, chinese patent publication CN110296997A, whose publication (publication) number is CN110296997A, discloses a method and an apparatus for detecting defects of ceramic tiles based on machine vision, which uses a conventional image processing method to perform object segmentation and defect detection by using a gray scale image of an obtained image. The difference of the surface reflectivity of the ceramic tiles is caused by the difference of the colors of the ceramic tiles, so that the difference of the brightness during imaging brings the difference of gray level images of different ceramic tiles, different algorithms and threshold values need to be set artificially aiming at the ceramic tiles with different colors during segmentation and defect detection, and the method has poor robustness and low speed. In addition, the multi-CPU in the industrial personal computer is adopted for parallel calculation, and the image filtering processing speed of the GPU is far higher than that of the CPU because the image is multidimensional matrix floating point data.
In addition, in the patent, a light source and a camera are simultaneously used for imaging the surface of the ceramic tile, a large number of light reflecting and light blocking areas (see attached figure 1) appear in an imaging result, the maximum gray value is brought by the light reflecting areas, if the areas are defective, the areas cannot be detected, and similarly, light blocking dark areas cannot be detected. Therefore, a large number of defect missing detection phenomena exist, and the accuracy is low. In addition, only qualified ceramic tiles and waste ceramic tiles can be detected by the method, the classification precision is low, and the method is not beneficial to enterprises to reduce the production cost.
In addition, Chinese patent with publication number CN110715935 discloses 'a ceramic defect detection device and method', the method is only suitable for ceramic tiles with the size not less than 1mX1m, and the method cannot be compatible with the detection of small-size ceramic tiles; in addition, images of the ceramic tile are acquired in a time sequence in a linear array camera and encoder synchronous mode, the images need to be spliced and fused to complete defect detection, on one hand, the processing speed is relatively low, and the accuracy of subsequent defect detection can be influenced by image splicing errors.
Meanwhile, the existing detection and sorting system is low in integration level and poor in field applicability. And the colors and the face types cannot be detected, the automatic packaging on the production line is easy to cause the aliasing phenomenon, and the field applicability is poor.
Disclosure of Invention
Aiming at the defects of the prior art, the technical problems to be solved by the invention are as follows: how to provide a structural design is reasonable, can carry out short-term test to colour, face type and the flaw of ceramic tile, can carry out automatic sorting to ceramic tile, is favorable to promoting enterprise's line production efficiency's ceramic tile surface quality on-line measuring letter sorting system.
In order to solve the technical problems, the invention adopts the following technical scheme:
the ceramic tile surface quality online detection sorting system is characterized by comprising a conveying belt for conveying ceramic tiles, an imaging module, an identification prediction module and a sorting module, wherein the imaging module is installed on the conveying belt in an online mode; the imaging module is used for taking a picture to obtain a ceramic tile image and transmitting the obtained ceramic tile image to the recognition and prediction module, the recognition and prediction module is used for recognizing the ceramic tile surface type, color and/or surface flaw of the ceramic tile image, classifying the ceramic tile according to the recognition result and transmitting the classification result to the sorting module; and the sorting module unloads the ceramic tiles from the conveying belt to corresponding classifications according to the received classification results.
By adopting the system, after the ceramic tiles are identified and classified by the identification and prediction module, the ceramic tiles can be unloaded from the conveying belt according to the classification of the ceramic tiles corresponding to the image by the sorting module according to the corresponding classification, so that the automatic sorting of the ceramic tiles is realized, the sorting efficiency of the ceramic tiles can be greatly improved, and the production efficiency of an enterprise line is improved. Meanwhile, the color and the surface type of the ceramic tile are identified when the identification prediction module identifies the ceramic tile image, so that the aliasing phenomenon in packaging can be avoided.
Furthermore, the imaging module comprises a position detection unit for detecting the ceramic tile, a light source for supplementing light, an X + imaging unit for taking a picture and an X-imaging unit; the X + imaging unit and the X-imaging unit are respectively arranged on two sides of the width direction of the conveying belt and both obliquely and downwards face the conveying belt, so that the photographing areas of the X + imaging unit and the X-imaging unit are mutually connected; the position detection unit and the light source are located between the X + imaging unit and the X-imaging unit, the position detection unit is located in the middle of the conveying belt in the width direction, and the light source is located right above the position detection unit.
By adopting the structure, the position detection unit is arranged between the X + imaging unit and the X-imaging unit, and in the middle of the conveying belt, once the ceramic tile is detected, the light source, the X + imaging unit and the X-imaging unit can be synchronously triggered to carry out flash and exposure imaging. Because the light source is positioned right above the position detection unit, and the X + imaging unit and the X-imaging unit shoot the ceramic tile on the conveying belt from two sides obliquely downwards, the detection blind area caused by the reflection of light on the surface of the ceramic tile caused by illumination can be effectively avoided, and the detection accuracy and the detection precision can be improved. In addition, the photographing areas of the X + imaging unit and the X-imaging unit are mutually connected, so that the imaging unit can be ensured to acquire a complete ceramic tile image, and the area missing detection can be avoided.
Furthermore, the input side of the imaging module is provided with an X + centering mechanism and an X-centering mechanism which are symmetrically arranged on two sides of the conveying belt in the width direction, one side of the X + centering mechanism opposite to the X-centering mechanism forms an eight-shaped guide part, the width of one side of the guide part, which is deviated from the imaging module, is matched with the width of the conveying belt, and the width of one side, which faces the imaging module, is matched with the maximum width of the ceramic tile.
Therefore, the ceramic tiles on the conveying belt can be completely corrected to the middle of the conveying belt through the splayed guide part, and once the ceramic tiles are conveyed to the imaging module by the conveying belt, the X + imaging unit and the X-imaging unit can uniformly image the ceramic tiles due to the symmetrical arrangement of the X + imaging unit and the X-imaging unit, so that the rapid and accurate identification of the later-stage identification and prediction module is facilitated.
Furthermore, the sorting module comprises an unqualified product sorting module and an equal product sorting module which are arranged on the side part of the conveying belt, the unqualified product sorting module and the equal product sorting module respectively comprise a telescopic mechanism arranged along the width direction of the conveying belt, and the height from the telescopic mechanism to the conveying belt is smaller than the minimum thickness of the ceramic tile.
Like this, unqualified product letter sorting module and first-class product letter sorting module just can be according to the categorised result of receiving, push out conveyor belt through telescopic machanism with the ceramic tile of unqualified ceramic tile and first-class product on the conveyor belt respectively, realize the letter sorting to ceramic tile. Because the height from the telescopic mechanism to the conveying belt is less than the minimum thickness of the ceramic tile, the ceramic tile can be reliably pushed out of the conveying belt by the telescopic mechanism.
Furthermore, the unqualified product sorting module and the first-class product sorting module further comprise an unqualified product position detection unit and a first-class product position detection unit which are arranged in the width direction of the conveying belt and correspond to the telescopic mechanism, wherein the unqualified product position detection unit and the first-class product position detection unit are used for detecting the ceramic tiles and are arranged in the middle of the width direction of the conveying belt.
Like this, discernment prediction module sends classified structure to letter sorting module after, letter sorting module just can let correspond nonconforming article position detecting element or the equal grade position detecting element on categorised and go up the electricity, prepares to detect the ceramic tile on the conveyer belt, in case sense after the ceramic tile reachs, just can push out conveyer belt with ceramic tile through the telescopic machanism that corresponds.
Furthermore, the output side of imaging module has set gradually nonconforming quality sorting module and first-class quality sorting module, the side of releasing of first-class quality sorting module links up and is provided with the conveyor belt who is used for carrying first-class quality ceramic tile, two conveyor belt sets up side by side.
Furthermore, the identification prediction module comprises an industrial personal computer and a display, and the imaging module is in data connection with the identification prediction module through a gigabit network interface; the unqualified product sorting module and the first-class product sorting module are in data communication and control connection with the identification prediction module through an industrial bus; the ceramic tile surface comprises a wave shape or a step shape; the ceramic tile color comprises black, cyan, red, blue or gray; the surface flaws include unfilled corners, cracks, scratches, bulges, unglazed glaze and multiple glazes.
The ceramic tile surface quality online detection sorting method is characterized in that the ceramic tile surface quality online detection sorting system is obtained firstly, and then the detection sorting is carried out by adopting the following steps:
s1: acquiring an image of the ceramic tile through an imaging module, and manually marking the color, the surface type and the surface flaw of the ceramic tile;
s2: constructing a multi-scale detection network model for identifying and predicting colors, surface types and surface flaws by using a single-step multi-frame detector of a deep learning model (SSD);
s3: repeatedly training the structure parameters of the optimization model by combining the labels and the prediction model obtained in the steps S1 and S2 to obtain an optimal prediction network model;
s4: combining the step S3, carrying out real-time parallel detection on the ceramic tile images collected by the imaging module, and simultaneously finding out the optimal positioning position and classification confidence value of the color, the surface type and the flaw;
s5: extracting the ceramic tile as a recognition key area by using a key area detection algorithm;
s6: combining the step S4 and the step S5, eliminating the color, the surface type and the flaw outside the key area, directly outputting the color, the surface type and the flaw positioning position and the classification result on the ceramic tile, and identifying the unqualified ceramic tile;
s7: and combining the qualified ceramic tiles identified in the step S6, and automatically classifying the qualified ceramic tiles into first-class tiles and superior tiles by using the set classification threshold.
Further, the SSD single-step multi-frame detector adopts VGG16 as a basic model, a convolution layer is added to obtain a multi-scale feature map for classification and positioning, a non-maximum suppression algorithm is adopted to remove a predicted positioning frame with large overlapping degree and frame positioning error, the remaining predicted positioning frame is used as a target detection frame, and a confidence value corresponding to the features in the target detection frame is used as a classification result.
Furthermore, the imaging module also comprises a side wall which is arranged in a rectangular surrounding manner, and the inner surface of the side wall is in a dark color; the conveying belt is provided with a background bottom plate for supporting ceramic tiles, and the surfaces of the conveying belt and the background bottom plate are light color; the key area detection algorithm comprises the following steps: the background bottom plate and the conveying belt are firstly separated from the deep color side wall, and then the ceramic tile is separated from the light color background of the background bottom plate and the conveying belt to be used as a key area.
Compared with the prior art, the invention has the beneficial effects that: the invention combines the ceramic tile surface quality detection and the sorting machine by using the depth network model, simultaneously acquires a ceramic tile image by two imaging units respectively, effectively solves the influence of a light reflecting area and a light blocking area, improves the image quality of the ceramic tile image, forms the ceramic tile quality detection, sorting and packaging device and method which can be used before packaging on an enterprise production line, has good real-time performance, high speed, high accuracy, wide applicability and high detection precision, can define multi-classification and sorting, is innovative by intelligent detection and automatic sorting, and can operate unattended. The existing production bottleneck of enterprises is solved, the production cost of the enterprises is reduced, and the market competitiveness of the products of the enterprises is improved.
Drawings
FIG. 1 is a schematic view of an embodiment of an online detection and sorting apparatus for surface quality of ceramic tiles according to the present invention;
FIG. 2 is a schematic structural diagram of an imaging module according to an embodiment of the invention;
FIG. 3 is a top view of FIG. 2;
FIG. 4 is a flowchart of an embodiment of the ceramic tile surface quality on-line detection and sorting method of the present invention;
FIGS. 5 and 6 are exemplary images of a ceramic tile obtained by an X + imaging unit and an X-imaging unit, respectively, in an embodiment of the present invention;
FIGS. 7 and 8 are images of off-grade ceramic tiles according to embodiments of the present invention;
FIGS. 9 and 10 are images of superior ceramic tiles in accordance with embodiments of the present invention;
FIGS. 11 and 12 are images of an equivalent ceramic tile in accordance with embodiments of the present invention;
reference numerals: 1, conveying a belt; 10 an imaging module; 20 identifying a prediction module; 30 unqualified product sorting modules; 40 a first-class sorting module; 50, preparing a ceramic tile to be detected; 11 a light source; a 12X + imaging unit; 13X-imaging unit; 14 a position detection unit; 15X + centering mechanism; a 16X-centering mechanism; 31 unqualified grade position detection unit; 32 unqualified ceramic tiles; 41 a first-class product position detection unit; 42 an first-class ceramic tile.
Detailed Description
The present invention will be described in further detail with reference to examples.
In the specific implementation: as shown in fig. 1, an online detection and sorting device for surface quality of ceramic tile is characterized by comprising: imaging module 10, identification and prediction module 20, reject sorting module 30, and first-class sorting module 40. The imaging module 10 is installed on the automatic ceramic tile packaging and conveying belt 1 on line, acquires an in-place signal of the ceramic tile 50 to be detected through the position detection unit 14, and synchronously triggers the flash and exposure imaging of the light source 11, the X + imaging unit 12 and the X-imaging unit 13, so that a complete visual image of the ceramic tile 50 to be detected running on line is acquired, and the image is transmitted to the identification and prediction module 20 in real time; the recognition and prediction module 20 realizes feature positioning and classification through parallel detection of a color classifier, a face classifier and various defect classifiers, recognizes ceramic tile types and defects, and transmits results to the unqualified product sorting module 30 and the first-class product sorting module 40 in real time; the unqualified product sorting module 30 rapidly executes horizontal movement in the X-direction to push the unqualified product ceramic tile 32 out of the conveying belt 1 according to the result output by the identification and prediction module 20 and the unqualified product position detection unit 31 detecting the unqualified product ceramic tile 32 in-place signal; the first-class product sorting module 40 rapidly executes horizontal movement in the X-direction to push and move the first-class product ceramic tile 42 from the conveying belt 1 to the first-class product packaging conveying belt 1 according to the result output by the identification and prediction module 20 and the in-place signal detected by the first-class position detection unit 41; the ceramic tile surface quality on-line detection sorting device sorts out unqualified ceramic tiles 32 and first-class ceramic tiles 42 on the automatic packing conveying belt 1 through the unqualified product sorting module 30 and the first-class product sorting module 40, keeps the excellent ceramic tiles to continue to run on the conveying belt 1, and finally completes excellent product packing.
Fig. 2 is a schematic diagram of an arrangement position of imaging modules in an embodiment of the invention, wherein fig. 2 is a front view thereof, and fig. 3 is a top view thereof. The imaging module 10 includes a position detecting unit 14, a light source 11, an X + imaging unit 12, and an X-imaging unit 13. The detection unit 14 and the light source 11 are in the Y + direction of the movement of the conveyor belt and are located at the very center of the imaging module 10. In the embodiment, the detection unit 14 and the light source 11 are collinear and are located at the center of the X-axis and the Y-axis of the imaging module 10, the detection unit 14 is located at the bottom, and the light source 11 is located at the top. The X + imaging unit 12 and the X-imaging unit 13 are respectively located in the imaging module 10 in the Y direction perpendicular to the X direction of the conveyor belt 1, are arranged in mirror symmetry, and keep a certain included angle with the X direction, so that the ceramic tile 50 to be detected is just obtained by the X + imaging unit 12 and the X-imaging unit 13 in an even imaging manner, and a detection blind area caused by surface light reflection of the ceramic tile due to illumination is effectively avoided. In the embodiment, included angles between the X + imaging unit 12 and the X-imaging unit 13 and an X axis are both 50 degrees, and the X-axis imaging unit is composed of a color area array CCD and an industrial lens. The resolution ratio of the color area array CCD is as follows: 2448X2048 with a pixel size of 3.45 μm. The detection unit 14, the light source 11, the X + imaging unit 12, and the X-imaging unit 13 are located in the same plane. The focal length of the industrial lens is 12mm, the aperture is 1:1.6-22, the imaging module 10 can be used for imaging the ceramic tile with the size of 400mmX300mm or so, and the imaging module further comprises a ceramic tile X + centering mechanism 15 and an X-centering mechanism 16 which respectively comprise 5 nylon guide wheels and are in an eight-shaped structure, the opening size is 480mm, and the closing size is 420 mm. The ceramic tile 50 to be inspected on the conveyor belt 1 is ensured to enter in the central posture of the imaging module 10X direction.
The recognition and prediction module 20 is composed of an industrial personal computer and a display, and mainly comprises feature (color, surface type and flaw) positioning and feature (color, surface type and flaw) classification in function. Image data transmission is carried out with the imaging module 10 through a gigabit network interface, data communication and control are carried out with the unqualified product sorting module 30 and the first-class product sorting module 40 through an industrial bus, and signal communication is realized in an RS485 mode in the embodiment. The parallel detection is realized by using a plurality of CUDA modules in the GPU of the industrial personal computer, wherein the GPU of the industrial personal computer adopts Nvidia GTX1050 Ti. The defects of the ceramic tile mainly comprise unfilled corners, cracks, scratches, bulges, unglazed glaze and multiple glazes. The ceramic tile surface mainly relates to a tile shape, and specifically comprises a wavy shape or a step shape; the ceramic tile colors include, but are not limited to, mainly black, cyan, red, blue, gray.
The unqualified product sorting module 30 rapidly executes horizontal movement in the X-direction to push out the ceramic tiles 32 which are detected and identified as unqualified, so as to achieve the sorting function. In the embodiment, a container is provided beside the conveyor belt 1 for receiving the defective ceramic tiles 32. Wherein the off-spec ceramic tile 32 is defined as: the color and the surface type of the ceramic tile do not belong to the same category or the ceramic tile belongs to the same category but the flaw characteristic scale exceeds the defined threshold range, and the flaw threshold values of the unqualified products defined in the embodiment are respectively:
cracking and scratching: the length is more than 5mm, and the width is more than 1 mm;
② corner lack, bulge, glaze lack and multiple glaze with area more than 75mm2。
The manner of performing the horizontal movement in the X-direction may be pneumatic or electric, such as an electric push rod or an air cylinder; in the embodiment, the running speed of the conveying belt 1 is 700mm/s, the 34-type double-acting air cylinder is adopted to directly output the linear motion in the X direction, and the stroke is 450 mm. The quick reversing is realized by the double-coil 5-position 2-way electromagnetic valve, the quick action of the piston rod of the air cylinder is realized by the speed regulation of the throttle valve, and the action speed is 750 mm/s.
The first-class product sorting module 40 quickly executes horizontal movement in the X-direction to push the porcelain tiles 42 which are detected and identified as first-class products onto a first-class product packaging and conveying belt, so that a sorting function is achieved. The first-class product is defined as a ceramic tile belonging to the same class but the flaw characteristic scale exceeds a defined threshold range, and the first-class product flaw threshold values defined in this embodiment are:
③ cracking and scratching: the length is more than 2mm and less than 5mm, and the width is more than 0.5mm and less than 1 mm;
④ corner lacking, bulge, glaze lacking and multiple glaze 10mm2Area < 75mm2。
The manner in which the horizontal movement in the X-direction is performed may be pneumatic or electric. In the embodiment, the running speed of the conveying belt 1 is 700mm/s, the 34-type double-acting air cylinder is adopted to directly output the linear motion in the X direction, and the stroke is 450 mm. The quick reversing is realized by the double-coil 5-position 2-way electromagnetic valve, and the quick action of the piston rod of the air cylinder is realized by additionally arranging a throttle valve for speed regulation, so that the action speed is 750 mm/s.
Fig. 4 is a flowchart of an embodiment of the online detection and sorting method for the surface quality of the ceramic tile, and the online detection and sorting method for the surface quality of the ceramic tile is characterized by comprising the following specific steps:
step 1: acquiring images of the ceramic tiles through the imaging module 10, and manually marking the colors, the surface types and various defects of the ceramic tiles;
step 2: constructing a multi-scale detection network model for identifying and predicting colors, surface types and flaws by using a single-step multi-frame detector of a deep learning model (SSD);
and step 3: repeatedly training the structural parameters of the optimization model by combining the labels and the prediction models obtained in the steps 1 and 2 to obtain an optimal prediction network model;
and 4, step 4: combining the step 3, carrying out real-time parallel detection on the ceramic tile images collected by the imaging module 10, and simultaneously finding out the optimal positioning positions and classification confidence values of colors, surface shapes and defects;
and 5: extracting the ceramic tile as a recognition key area by using a key area detection algorithm;
step 6: combining the steps 4 and 5, eliminating the color, the surface type and the flaw outside the key area, directly outputting the color, the surface type and the flaw positioning position and the classification result on the ceramic tile, and identifying the unqualified ceramic tile;
and 7: and (4) automatically classifying the qualified ceramic tiles identified in the step (6) into first-class products and superior products by utilizing a set classification threshold.
In the embodiment, the running speed of the conveying belt 1 is 700mm/s, a global exposure industrial high-speed CCD is adopted for capturing a ceramic tile high-speed moving image, the exposure time of the X + imaging unit 12 and the X-imaging unit 13 is 1.2ms when the image is collected, and the gain is 2 times. The light source 11 is an array type high-brightness LED, the brightness exceeds 4000Lx @0.5m, the divergence angle of the light source is 45 degrees, the diameter of the lighting coverage at the distance of 0.5m is ensured to exceed 700mm, and the whole area of the ceramic tile can be illuminated by the light source.
The collected ceramic tile image is composed of two parts, namely an X + imaging unit 12 image and an X-imaging unit 13 image. Fig. 5 and 6 are exemplary images of a ceramic tile obtained by the X + imaging unit 12 and the X-imaging unit 13, respectively, in an embodiment of the present invention, where fig. 5 is the image obtained by the X + imaging unit 12 and fig. 6 is the image obtained by the X-imaging unit 13. It is apparent that: the 2 images are combined into a one-piece ceramic tile image. Colors of the marking ceramic tile include, but are not limited to, black, cyan, red, blue, gray; the surface type of the marking ceramic tile comprises grains and a cross-sectional shape of the front surface; the flaws of the marked ceramic tile comprise unfilled corners, cracks, bulges, unglazed glaze and multiple glazes, and various flaw characteristic dimensions such as unfilled corner areas, crack widths and lengths, bulge areas, glaze-less areas and glaze-more areas are quantized. In this embodiment, label of 600 photos in total of 300 ceramic tiles is completed by using label img, image labels are stored in a PASCAL VOC format as an XML file, and the specific surface types of the labels include a wavy shape, a step shape, a vertical stripe shape and a cross grid texture, for example, 501a in fig. 5 and 6 is a vertical stripe shape, 501b is a wavy cross section, and fig. 5 and 6 are a step shape cross section and a cross grid texture.
The SSD single-step multi-frame detector adopts VGG16 as a basic model, and is additionally provided with a convolutional layer to obtain a multi-scale feature map for classification and positioning. And eliminating predicted positioning frames with large overlapping degree and frame positioning errors by adopting a non-maximum suppression algorithm, wherein the final residual frames are target detection frames, and the confidence degrees corresponding to the features in the predicted positioning frames are classification results. In the embodiment, the VGG16 model is simplified, specifically, the full connection layers FC6 and FC7 are converted into 3X3 convolution layers conv6 and 1X1 convolution layers conv7, and the pool layer pool5 is changed from 2X2 of original stride-2 to 3X3 of stride-1, so that the size of the feature diagram is simplified and the calculation speed is increased. 6 convolution layers are added later, and 6 feature maps with different scales are extracted respectively and used for classification and positioning prediction. From the perspective of the convolution view, it is understood that the deeper the network is, the larger the convolution operation is relative to the size of the convolution view of the original input image, and the features of small objects are slowly attenuated, leaving behind larger object features. Therefore, the method of inputting the feature maps of 6 different scales into the classification network has practical effects, and both the real-time performance and the detection accuracy are greatly improved.
The key region detection algorithm realizes the segmentation of the ceramic tile image from the background, in the embodiment, the imaging module 10 further comprises a side wall which is arranged in a rectangular surrounding manner, and the inner surface of the side wall is dark, specifically black; the surfaces of the conveying belt and the background bottom plate are light-colored, particularly white; in the obtained ceramic tile image: the ceramic tile background bottom plate (conveying belt) is white, and the peripheral side walls are black. The background bottom plate and the peripheral side walls are firstly segmented through positioning segmentation, then the ceramic tiles are used for carrying out secondary foreground extraction on the background bottom plate and the conveying belt, and the ceramic tiles are segmented from the white bottom plate to be used as key areas. In this embodiment, when the image is subjected to binary segmentation, the segmentation threshold th is 127, so as to obtain a segmentation binary image bin 0; then carrying out reverse color treatment on the bin0 to obtain a binary segmentation map bin 1; white connected region calculation is carried out on bin0 and bin1 respectively, the white bottom plate contour is located and found out, and then the ceramic tile contour is located and found out.
bin1＝255-bin0
The artificially set classification threshold is quantitatively defined according to the areas of unfilled corners, bulges, unglazed and multiple glazes, the width and the length of cracks, and is defined as an first-class product when the characteristic dimensions are respectively set to be larger than a certain threshold; and when the value is less than a certain threshold value, defining the product as a superior product. 7-12 are images of defective, premium, and first-class ceramic tiles detected in accordance with an embodiment of the present invention, and the ceramic tile defects detected in FIGS. 7 and 8 are cracks, 88.12mm in length and 12.74mm in width, exceeding a set threshold, and outputting a result as a defective; the ceramic tile in fig. 9 and 10 has no defect detected and the output is a qualified product; the ceramic tile defects detected in fig. 11 and 12 were scratches having a length of 3.72mm and a width of 0.67mm, and the results were output as an equal grade.
The above description is only exemplary of the present invention and should not be taken as limiting, and any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The ceramic tile surface quality online detection and sorting system is characterized by comprising a conveying belt (1) for conveying ceramic tiles, an imaging module (10) installed on the conveying belt (1) online, a recognition and prediction module (20) connected with the imaging module (10), and a sorting module connected with the recognition and prediction module (20), wherein the sorting module is positioned on the output side of the imaging module (10); the imaging module (10) is used for taking a picture to obtain a ceramic tile image and transmitting the obtained ceramic tile image to the recognition and prediction module (20), the recognition and prediction module (20) is used for recognizing the ceramic tile surface type, color and/or surface flaw of the ceramic tile image, classifying the ceramic tile according to the recognition result and transmitting the classification result to the sorting module; the sorting module unloads the ceramic tiles from the conveyor belt (1) to corresponding classifications according to the received classification results.
2. The ceramic tile surface quality on-line detection sorting system of claim 1, wherein the imaging module (10) comprises a position detection unit (14) for detecting ceramic tiles, a light source (11) for supplementing light, an X + imaging unit (12) and an X-imaging unit (13) for taking pictures; the X + imaging unit (12) and the X-imaging unit (13) are respectively arranged on two sides of the width direction of the conveying belt (1) and both obliquely downwards face the conveying belt (1), so that the photographing areas of the X + imaging unit (12) and the X-imaging unit (13) are mutually connected; the position detection unit (14) and the light source (11) are located between the X + imaging unit (12) and the X-imaging unit (13), the position detection unit (14) is located in the middle of the conveying belt (1) in the width direction, and the light source (11) is located right above the position detection unit (14).
3. The ceramic tile surface quality online detection sorting system according to claim 2, characterized in that the input side of the imaging module (10) is provided with an X + centering mechanism (15) and an X-centering mechanism (16) which are symmetrically arranged at two sides of the conveying belt (1) in the width direction, one side of the X + centering mechanism (15) opposite to the X-centering mechanism (16) forms a guide part in a shape of Chinese character 'ba', the width of the guide part at the side departing from the imaging module (10) is matched with the width of the conveying belt (1), and the width of the guide part at the side facing to the imaging module (10) is matched with the maximum width of the ceramic tile.
4. The ceramic tile surface quality on-line detection and sorting system of claim 1, characterized in that the sorting modules comprise a defective sorting module (30) and an equal sorting module (40) which are arranged at the side parts of the conveying belt (1), each of the defective sorting module (30) and the equal sorting module (40) comprises a telescopic mechanism arranged along the width direction of the conveying belt (1), and the height from the telescopic mechanism to the conveying belt (1) is smaller than the minimum thickness of the ceramic tile.
5. The on-line ceramic tile surface quality detecting and sorting system according to claim 4, wherein the defective sorting module (30) and the first-class sorting module (40) further comprise a defective position detecting unit (31) and a first-class position detecting unit (41) which are arranged corresponding to the telescopic mechanism in the width direction of the conveyor belt (1), and the defective position detecting unit (31) and the first-class position detecting unit (41) are used for detecting the ceramic tiles and are both arranged in the middle of the width direction of the conveyor belt (1).
6. The on-line ceramic tile surface quality detecting and sorting system according to claim 4, characterized in that the output side of the imaging module (10) is provided with the unqualified product sorting module (30) and the first-class product sorting module (40) in sequence, the pushing side of the first-class product sorting module (40) is provided with a conveying belt (1) for conveying first-class product ceramic tiles in a joint mode, and the two conveying belts (1) are arranged side by side.
7. The ceramic tile surface quality online detection sorting system of claim 4, wherein the identification and prediction module (20) comprises an industrial personal computer and a display, and the imaging module (10) is in data connection with the identification and prediction module (20) through a gigabit network interface; the unqualified product sorting module (30) and the first-class product sorting module (40) are in data communication and control connection with the identification and prediction module (20) through an industrial bus; the ceramic tile surface comprises a wave shape or a step shape; the ceramic tile color comprises black, cyan, red, blue or gray; the surface flaws include unfilled corners, cracks, scratches, bulges, unglazed glaze and multiple glazes.
8. An online detection and sorting method for the surface quality of ceramic tiles, which is characterized in that the online detection and sorting system for the surface quality of the ceramic tiles as claimed in any one of claims 1 to 7 is obtained, and then the steps are adopted for detection and sorting:
s1: acquiring a ceramic tile image through an imaging module (10), and manually marking the color, the surface type and the surface flaw of the ceramic tile;
s2: constructing a multi-scale detection network model for identifying and predicting colors, surface types and surface flaws by using a single-step multi-frame detector of a deep learning model (SSD);
s3: repeatedly training the structure parameters of the optimization model by combining the labels and the prediction model obtained in the steps S1 and S2 to obtain an optimal prediction network model;
s4: combining the step S3, carrying out real-time parallel detection on the ceramic tile images collected by the imaging module (10), and simultaneously finding out the optimal positioning position and classification confidence value of the color, the surface type and the flaw;
s5: extracting the ceramic tile as a recognition key area by using a key area detection algorithm;
s6: combining the step S4 and the step S5, eliminating the color, the surface type and the flaw outside the key area, directly outputting the color, the surface type and the flaw positioning position and the classification result on the ceramic tile, and identifying the unqualified ceramic tile;
s7: and combining the qualified ceramic tiles identified in the step S6, and automatically classifying the qualified ceramic tiles into first-class tiles and superior tiles by using the set classification threshold.
9. The on-line ceramic tile surface quality detecting and sorting method of claim 8, wherein the SSD single-step multi-frame detector uses VGG16 as a basic model, adds convolution layers to obtain multi-scale feature maps for classification and positioning, removes predicted positioning frames with large overlap and frame positioning errors by using a non-maximum suppression algorithm, uses the remaining predicted positioning frames as target detection frames, and uses confidence values corresponding to features in the target detection frames as classification results.
10. The ceramic tile surface quality on-line detection sorting method according to claim 8, characterized in that the imaging module (10) further comprises a side wall in a rectangular enclosure arrangement, wherein the inner surface of the side wall is dark; the surfaces of the conveying belt and the background bottom plate are light color; the key area detection algorithm comprises the following steps: the background bottom plate and the conveying belt are firstly separated from the deep color side wall, and then the ceramic tile is separated from the light color background of the background bottom plate and the conveying belt to be used as a key area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010586640.0A CN111558542A (en) | 2020-06-24 | 2020-06-24 | Ceramic tile surface quality online detection sorting system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010586640.0A CN111558542A (en) | 2020-06-24 | 2020-06-24 | Ceramic tile surface quality online detection sorting system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111558542A true CN111558542A (en) | 2020-08-21 |
Family
ID=72073962
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010586640.0A Withdrawn CN111558542A (en) | 2020-06-24 | 2020-06-24 | Ceramic tile surface quality online detection sorting system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111558542A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112710669A (en) * | 2020-12-09 | 2021-04-27 | 北方华锦化学工业股份有限公司 | Method for rapidly evaluating crystal points of hard elastic diaphragm of homo-polypropylene lithium battery |
| CN113843180A (en) * | 2021-09-16 | 2021-12-28 | 大连华信计算机技术股份有限公司 | Wood veneer color sorting system and device based on deep learning method |
| CN113976452A (en) * | 2021-07-30 | 2022-01-28 | 广东省威汇智能科技有限公司 | Failure detection method and device based on flexible flat cable |
| CN114043614A (en) * | 2021-11-08 | 2022-02-15 | 冷水江市鑫达耐火材料制造有限公司 | Firebrick selects positioner and firebrick selects system |
| CN114769145A (en) * | 2022-04-27 | 2022-07-22 | 合肥名德光电科技股份有限公司 | MDX multiple-radiation-source sorting system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109078869A (en) * | 2018-04-09 | 2018-12-25 | 常州轻工职业技术学院 | Automatic sorting device based on image recognition |
| CN109344772A (en) * | 2018-09-30 | 2019-02-15 | 中国人民解放军战略支援部队信息工程大学 | Ultrashort wave signal specific reconnaissance method based on spectrogram and depth convolutional network |
| CN110296997A (en) * | 2019-07-03 | 2019-10-01 | 三峡大学 | The defect inspection method and device of ceramic tile based on machine vision |
| CN110687121A (en) * | 2019-09-19 | 2020-01-14 | 湖北三江航天万峰科技发展有限公司 | Intelligent online detection and automatic grading method and system for ceramic tiles |
| CN111065468A (en) * | 2017-08-25 | 2020-04-24 | 锡克拜控股有限公司 | Assembly and method for producing an optical effect layer comprising oriented, non-spherical, flat magnetic or magnetizable pigment particles |
| CN210701229U (en) * | 2019-10-17 | 2020-06-09 | 郑州电力职业技术学院 | Electronic commerce article sorting device |
| CN210754504U (en) * | 2019-10-13 | 2020-06-16 | 福州鑫奋祥金属制品有限公司 | Rotary feeding device for plate rolling machine |
-
2020
- 2020-06-24 CN CN202010586640.0A patent/CN111558542A/en not_active Withdrawn
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111065468A (en) * | 2017-08-25 | 2020-04-24 | 锡克拜控股有限公司 | Assembly and method for producing an optical effect layer comprising oriented, non-spherical, flat magnetic or magnetizable pigment particles |
| CN109078869A (en) * | 2018-04-09 | 2018-12-25 | 常州轻工职业技术学院 | Automatic sorting device based on image recognition |
| CN109344772A (en) * | 2018-09-30 | 2019-02-15 | 中国人民解放军战略支援部队信息工程大学 | Ultrashort wave signal specific reconnaissance method based on spectrogram and depth convolutional network |
| CN110296997A (en) * | 2019-07-03 | 2019-10-01 | 三峡大学 | The defect inspection method and device of ceramic tile based on machine vision |
| CN110687121A (en) * | 2019-09-19 | 2020-01-14 | 湖北三江航天万峰科技发展有限公司 | Intelligent online detection and automatic grading method and system for ceramic tiles |
| CN210754504U (en) * | 2019-10-13 | 2020-06-16 | 福州鑫奋祥金属制品有限公司 | Rotary feeding device for plate rolling machine |
| CN210701229U (en) * | 2019-10-17 | 2020-06-09 | 郑州电力职业技术学院 | Electronic commerce article sorting device |
Non-Patent Citations (2)
| Title |
|---|
| 刘国成: "《人群异常行为数字图像处理与分析》", 31 January 2020 * |
| 汪宋: "基于深度学习的目标检测算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112710669A (en) * | 2020-12-09 | 2021-04-27 | 北方华锦化学工业股份有限公司 | Method for rapidly evaluating crystal points of hard elastic diaphragm of homo-polypropylene lithium battery |
| CN113976452A (en) * | 2021-07-30 | 2022-01-28 | 广东省威汇智能科技有限公司 | Failure detection method and device based on flexible flat cable |
| CN113843180A (en) * | 2021-09-16 | 2021-12-28 | 大连华信计算机技术股份有限公司 | Wood veneer color sorting system and device based on deep learning method |
| CN114043614A (en) * | 2021-11-08 | 2022-02-15 | 冷水江市鑫达耐火材料制造有限公司 | Firebrick selects positioner and firebrick selects system |
| CN114769145A (en) * | 2022-04-27 | 2022-07-22 | 合肥名德光电科技股份有限公司 | MDX multiple-radiation-source sorting system |
| CN114769145B (en) * | 2022-04-27 | 2024-03-26 | 合肥名德光电科技股份有限公司 | MDX multi-source sorting system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111558542A (en) | Ceramic tile surface quality online detection sorting system and method | |
| KR102579783B1 (en) | Vision inspection system by using remote learning of product defects image | |
| CN106409711B (en) | A kind of solar energy silicon crystal chip defect detecting system and method | |
| CN110296997B (en) | Method and device for detecting defects of ceramic tiles based on machine vision | |
| CN109376792B (en) | Photovoltaic cell appearance defect classification method based on multi-channel residual error neural network | |
| CN206139527U (en) | Panoramic vision potato is selected separately and defect detecting device | |
| CN109724984A (en) | A kind of defects detection identification device and method based on deep learning algorithm | |
| CN107944504B (en) | Board recognition and machine learning method and device for board recognition and electronic equipment | |
| CN104574389A (en) | Battery piece chromatism selection control method based on color machine vision | |
| CN110246122A (en) | Small size bearing quality determining method, apparatus and system based on machine vision | |
| KR20220051430A (en) | Surface modeling system and method using polarization cue | |
| Liu et al. | Automatic detection technology of surface defects on plastic products based on machine vision | |
| CN112893159B (en) | Coal gangue sorting method based on image recognition | |
| CN114119526A (en) | Steel plate surface defect detection and identification system and method based on machine vision | |
| CN109035232A (en) | A kind of intermediate plate defect inspection method merged based on deep learning and machine vision | |
| CN113129304B (en) | Part detection method based on machine vision | |
| CN110554052A (en) | artificial board surface defect detection method and system | |
| CN111551559A (en) | LCD (liquid Crystal display) liquid crystal screen defect detection method based on multi-view vision system | |
| WO2023134286A1 (en) | Online automatic quality testing and classification method for cathode copper | |
| CN109693140A (en) | A kind of intelligent flexible production line and its working method | |
| CN113706496B (en) | Aircraft structure crack detection method based on deep learning model | |
| CN109701890A (en) | Magnetic tile surface defect detection and method for sorting | |
| CN109387525A (en) | On a kind of film in film defect determination method | |
| CN107246841A (en) | A kind of multiple views grouping system device of auto parts and components | |
| CN113538323A (en) | Defect detection method and system based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200821 |