Method for identifying pattern formation mode on ceramic

Abstract

The invention belongs to the technical field of pattern identification, and particularly relates to a method for identifying a pattern forming mode on ceramic, which comprises the following steps of: 1, acquiring a digital image of a pattern on the surface of the ceramic; 2, calculating pattern similarity indexes of different areas in the image; 3, color space data of the image are analyzed, and the number of main colors and the brightness standard deviation are counted; 4, detecting microscopic features of the image, wherein the microscopic features comprise at least one of dot arrangement, ink mark accumulation and pen wielding marks; and 5, in combination with the similarity index, the color space data and the microscopic characteristics, judging that the pattern formation mode is decal, printing or drawing. According to the method, a three-layer identification framework of image acquisition, quantitative analysis and comprehensive judgment is constructed, and the limitation of single image extraction is broken through; process traces are directly identified through microscopic feature detection, so that misjudgment caused by environment illumination and glaze reflection is avoided, and reduction of segmentation precision under a complex background is prevented; and a basic framework is provided for automatic identification.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ornament identification, and particularly relates to a method for identifying the forming mode of an ornament on a ceramic. BACKGROUND

[0002] As a traditional art product, the ornament process of a ceramic mainly includes three types of pasting, printing and painting. The pasting process is to print a pattern on a flower paper, then paste the flower paper on a ceramic body or glaze, and then perform firing; the printing process is to print a pattern on a body by using an impression with a decorative pattern, or to leave a pattern by using a mold; the painting process is to draw a pattern on a ceramic by a painter with pure handwork. The three processes have respective characteristics, but still have certain similarities in appearance.

[0003] In the prior art, the three types of ornament processes of pasting, printing and painting have certain similarities in appearance, which leads to difficulties in actual identification. At present, there is a lack of a systematic and accurate distinguishing method in the market, and most of the identification relies on subjective experience of experts, and there is a lack of an objective and reproducible method, and human errors are prone to exist.

[0004] According to the search, the patent for invention with the publication number CN111814550A discloses a ceramic ornament extraction method based on a convolutional neural network and image processing, and the method is implemented according to the following steps: obtaining an original ceramic image data set; obtaining a ceramic ornament area by using a U-2-Net convolutional neural network; performing computer vision image algorithm processing on the ceramic ornament area and the original image respectively to obtain all ornaments; positioning core ornaments by using a search algorithm; saving and outputting the core ornaments; and inputting a ceramic image by combining a neural convolutional network and a computer vision image processing method, so that the ceramic ornament pattern can be automatically and losslessly extracted. The method has a rigorous process, is less affected by image backgrounds and environmental factors, and has a remarkable effect. The method can be used in fields of ceramic age identification, ceramic appreciation and ceramic design, and has important significance for establishing a ceramic ornament library and protecting and inheriting Chinese traditional art culture.

[0005] In the above-mentioned disclosed technology, the focus ornament area segmentation only realizes ornament extraction, and does not solve the process identification problem.

[0006] Meanwhile, the technology depends on gray scale conversion and geometric weighting, only processes pixel position information, ignores process characteristics such as color levels and repetition rules, and lacks a process characteristic quantization model.

[0007] And the background technology points out that environmental factors cause extraction damage, but the solution only depends on image algorithm optimization (such as optimization operators), but does not process the influence of the physical properties of a body and a glaze on an ornament.

[0008] Therefore, the application aims to provide a method capable of quickly and accurately distinguishing between pasting, printing and painting ornaments on a ceramic. SUMMARY

[0009] In order to solve the above problems in the prior art, the application provides a method for identifying the forming mode of ceramic surface decoration, aiming to solve the technical problem that there is no systematic and accurate distinguishing method in the prior art, and the identification mostly depends on the subjective experience of experts, and there is a lack of objective and reproducible method.

[0010] In order to achieve the above-mentioned purpose, the application provides the following technical scheme: a method for identifying the forming mode of ceramic surface decoration, comprising the following steps: step one: obtaining a digital image of ceramic surface decoration; step two: calculating the pattern similarity index of different regions in the image; step three: analyzing the color space data of the image, counting the number of main colors and the standard deviation of brightness; step four: detecting microscopic features of the image, including at least one of dot arrangement, ink accumulation and pen trace; step five: combining the similarity index, color space data and microscopic features to determine the decoration forming mode as decal, printing or painting.

[0011] Further, the similarity index is calculated by structural similarity index SSIM and cosine similarity algorithm, and is realized based on image block comparison.

[0012] Further, the color space data is counted by using clustering algorithm to count the number of main colors after converting the RGB image to HSV space.

[0013] Further, the dot arrangement is realized by detecting frequency distribution through Fourier transform, and the ink accumulation is identified through image roughness analysis.

[0014] Further, the pen trace is extracted by microscopic imaging to extract the directional change characteristics of pen lifting, pen lowering or pen moving.

[0015] Further, when the similarity index is greater than or equal to 0.9, the number of main colors is less than or equal to 3, and the standard deviation of brightness is less than 0.1, it is determined as decal decoration.

[0016] Further, when the similarity index is 0.7 to 0.9, the number of main colors is 3 to 6, and the ink accumulation exists, it is determined as printing decoration.

[0017] Further, when the similarity index is less than 0.6, the number of main colors is greater than or equal to 8, and the pen trace is detected, it is determined as painting decoration.

[0018] Further, the step four further comprises detecting the roughness data of ceramic glaze surface, and the roughness is obtained by a three-dimensional scanner.

[0019] Further, the step four further comprises detecting the symmetry data of the body, and the symmetry is quantitatively analyzed by an image measuring instrument.

[0020] Compared with the prior art, the present application has the beneficial effects that:

[0021] 1、In the present application, a three-layer identification framework of image acquisition, quantitative analysis and comprehensive judgment is constructed, similarity index calculation, color space analysis and microscopic trace detection are carried out, and finally a comprehensive decision mechanism is adopted to output the process type by fusing three types of features; thereby breaking through the limitation of single image extraction, filling the technical gap of lack of systematic process distinguishing method in the field of ceramic identification; at the same time, the process trace is directly identified through microscopic feature detection, avoiding misjudgment caused by environmental light and glaze reflection, and preventing the decrease of segmentation accuracy in complex background; and a basic framework is provided for automatic identification, and the three-step quantitative process can be directly embedded into an intelligent identification system to replace the identification relying on experience.

[0022] 2、In the present application, the HSV space quantitative model converts the RGB image to the HSV space, counts the number of main colors through a clustering algorithm, calculates the standard deviation of brightness, establishes the mapping relationship between color hierarchy and process, and solves the subjective experience misjudgment problem according to a threshold value judgment system; the traditional identification relies on expert experience, and the description of color hierarchy and line smoothness is fuzzy, the present model converts the subjective description into a calculable parameter to reduce the influence of human factors; and the threshold value rule is used to realize efficient judgment of small samples and improve the adaptability of small sample scenes; at the same time, the threshold value can be flexibly adjusted to adapt to the process differences of different kiln openings, and can be compatible with multi-level identification requirements.

[0023] 3、In the present application, an anti-interference verification mechanism is realized through multi-source feature fusion to solve the problem of "old tire new color" fraud; traditional scientific instrument method may be deceived by "old porcelain piece grinding fine as tire soil", but the present scheme verifies through the linkage of glaze and decoration features to identify old methods such as bottom connection and head replacement; the image analysis of the prior art is easy to fail in the reflection and pollution environment, the present scheme supplements the physical structure features (such as the symmetry data of the body are not affected by the surface stains), guarantees the identification stability, and enhances the robustness in complex scenes; at the same time, structural evidence chain is provided for part of porcelain (such as the body symmetry of official kiln painting is significantly higher than that of folk kiln applique), solving the dilemma of "isolated product without standard reference" in typology, thereby supporting the identification of high-value cultural relics. BRIEF DESCRIPTION OF DRAWINGS

[0024] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, which together with the embodiments of the present application serve to explain the present application, and do not constitute a limitation of the present application. In the drawings:

[0025] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0026] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of the present application.

[0027] Embodiment 1

[0028] Please refer to Figure 1 The embodiment provides the following technical solutions: a ceramic decoration forming method, comprising the following steps: step 1, obtaining a digital image of a ceramic surface decoration; step 2, calculating a pattern similarity index of different regions in the image; step 3, analyzing color space data of the image, and counting the number of main colors and the standard deviation of brightness; step 4, detecting microscopic features of the image, including at least one of dot arrangement, ink accumulation and pen trace; and step 5, combining the similarity index, color space data and microscopic features to determine whether the decoration forming method is decal, printing or painting, so as to realize non-destructive identification of the ceramic decoration forming method, avoid physical contact damage to cultural relics by obtaining a digital image, establish a systematic identification path, solve the problem of lack of multi-dimensional analysis framework in the prior art by comprehensively analyzing the similarity index, color space data and microscopic features, and cover the identification ability of all process types by detecting dot arrangement, ink accumulation and pen trace (step 4) to distinguish three types of decal, printing and painting processes.

[0029] The similarity index is calculated by a structural similarity index SSIM and a cosine similarity algorithm, and is realized based on image block comparison, improves the quantification accuracy of pattern repeatability, solves the problem of poor consistency of artificial observation by the structural similarity index (SSIM) or the cosine similarity algorithm, enhances the algorithm compatibility, and realizes similarity calculation based on image block comparison (right 2) to adapt to irregular ceramic curved surface decoration analysis.

[0030] The color space data is converted from an RGB image to an HSV space, and the number of main colors is counted by using a clustering algorithm, so as to accurately quantify the color level difference, overcome the defect of insufficient sensitivity to brightness in the traditional RGB space by converting the HSV space and counting the number of main colors by using the clustering algorithm, and establish a color and process mapping relationship, and distinguish the color step transition characteristics of decal (standard deviation <0.1) and painting (standard deviation >0.2) by the brightness standard deviation parameter.

[0031] The dot arrangement is realized by Fourier transform detection frequency distribution, the ink accumulation is identified by image roughness analysis, and the pen trace is extracted by microscopic imaging to change the directionality of the pen lifting, pen lowering or line drawing, so as to directionally identify the process fingerprint characteristics, the dot arrangement is detected by Fourier transform, and the mechanical printing trace of the decal process is accurately captured; the ink accumulation is identified by image roughness analysis, and the ink penetration characteristics of the printing process are locked; the pen direction change is extracted by microscopic imaging, and the handcrafting trajectory of the drawing process is restored; meanwhile, the microscopic trace missing detection problem is solved, and through multi-modal feature parallel detection, the failure of a single analysis method in a contaminated sample is avoided.

[0032] When the similarity index is greater than or equal to 0.9, the number of main colors is less than or equal to 3, and the brightness standard deviation is less than 0.1, it is determined to be a decal ornament; when the similarity index is 0.7 to 0.9, the number of main colors is 3 to 6, and the ink accumulation exists, it is determined to be a printing ornament; when the similarity index is less than 0.6, the number of main colors is greater than or equal to 8, and the pen trace is detected, it is determined to be a drawing ornament, the decal threshold (similarity greater than or equal to 0.9 + main color less than or equal to 3 + brightness standard deviation less than 0.1) is used to identify the highly standardized production characteristics; the printing threshold (similarity 0.7-0.9 + ink accumulation) is used to capture the common defects of semi-mechanical processes; the drawing threshold (similarity less than 0.6 + pen trace) is used to verify the uniqueness of handcrafting, and a quantifiable process determination rule is constructed; at the same time, by clearly defining the numerical boundary, non-expert users can perform a standardized identification process, and the professional threshold of identification is reduced.

[0033] Step four further includes detecting ceramic glaze roughness data, the roughness being obtained by a three-dimensional scanner, and step four further includes detecting body symmetry data, the symmetry being quantitatively analyzed by an image measuring instrument, the glaze roughness data being obtained by a three-dimensional scanner to assist in verifying the physical trace characteristics of the decal and printing processes; solving the problem of microscopic feature disappearance caused by surface wear of old ceramics; verifying the overall coordination of the drawing process by quantitatively analyzing the body symmetry by an image measuring instrument; solving the fraud loophole of local imitation of pen trace of counterfeit products; at the same time, by fusing image features and physical structure parameters, the technical requirements of cross-validation for cultural relics judicial identification are met, and multi-source evidence chain support is provided.

[0034] Embodiment 2:

[0035] Please refer to Figure 1 In this embodiment, the staff used the method disclosed in the present application in a ceramic museum for a collection digitalization project, and needed to identify the process of a certain cover pot and enter the file. The pot is 35 centimeters high, with a peony ribbon bird ornament on the abdomen, and the glaze is well preserved but has local kiln change phenomenon.

[0036] Image acquisition was performed using a Hasselblad H6D-400c medium format digital camera with HC 120mm f / 4.0 macro lens at 300 dpi resolution. Three-dimensional scanning was performed using a Tianyuan OKIO 5M blue light three-dimensional scanner with a point distance of 0.01 mm. Microscopic imaging was performed using an Olympus DSX1000 digital microscope with 500x optical zoom. Body measurement was performed using a Zhonggu instrument SJ6000 laser interferometer.

[0037] The technician placed the covered jar on a constant temperature and humidity operating platform and used a Hasselblad camera to capture multispectral images to obtain raw image data. The Tianyuan scanner synchronously collected the glaze three-dimensional point cloud, and the average roughness Ra value was measured to be 0.7 microns, and the corrugation amplitude of the kiln transformation area reached 12 microns. The laser interferometer detected the body to show a maximum eccentricity of 0.05 mm, which met the characteristics of hand-pulled blanks;

[0038] The image processing system used an improved U-Net to segment the decoration area, which was divided into forty 60x60 pixel units. The average value of the repeatability calculated by the cosine similarity calculation unit was 0.38, and the HSV space analysis identified nine main color tones such as carmine, jadeite green, and egg white. The brightness standard deviation was 0.23. The Olympus microscope captured key evidence at the decoration transition: the edges of the peony petals showed a color transition formed by a brush pressing, and the brush direction change variance reached 18.7, and the leaf part detected a "clam light" effect formed by the deposition of natural mineral materials.

[0039] Therefore, the comprehensive similarity 0.38 is lower than the painting process threshold 0.6, the number of main colors 9 exceeds the upper limit 3 of the decal, the microscopic evidence clearly shows the brush marks and mineral material characteristics, and the body eccentricity 0.05 mm meets the hand-made artifact standard. According to claim 8, it is determined to be a painted decoration, and the digital archive is marked as "a certain hand-drawn genuine product", with 500x microscopic images as evidence.

[0040] In this case, multispectral imaging accurately restores the lost ancient color formula color, and laser interferometer data confirms the traditional hand-pulled blank process. The technical team found two places where the veins were connected, combined with historical data, and inferred that two artists cooperated to complete it. This discovery is included in the museum digital guide system as a key explanation. When visitors scan the exhibit cabinet two-dimensional code, the phone automatically displays a three-dimensional animation demonstration of the decoration formation process.

[0041] When the blue laser line of the Tianyuan scanner slowly scans the curved surface of the jar, the motorized stage of the Olympus microscope is tracking the brush stroke trajectory at 5 microns per step. The laser interferometer probe of Zhonggu instrument detects 0.001 mm level body vibration, which immediately triggers the anti-shock platform compensation system. These data are real-time imported into the central processor, and a color-coded process feature heat map is displayed on the Hasselblad camera viewfinder screen - the brush stroke is cyan, the mineral material deposition point is gold, and the modern repair area is marked as a sharp red alert.

[0042] The working principle of the application is: by acquiring the digital image of the ceramic surface decoration, the pattern similarity index between image blocks is calculated by using structural similarity index or cosine similarity algorithm, and after the image is converted to HSV color space, the number of main colors is counted and the brightness standard deviation is calculated by using clustering algorithm; combined with Fourier transform to detect dot arrangement characteristics, image roughness analysis to identify ink accumulation characteristics, microscopic imaging to extract pen direction change characteristics, and synchronous collection of glaze roughness and body symmetry physical parameters; finally, according to the preset process judgment threshold system, when the similarity is greater than or equal to 0.9, the main color is less than or equal to 3, and the brightness standard deviation is less than 0.1, it is judged as decal; when the similarity is 0.7-0.9, the main color is 3-6, and there is ink accumulation, it is judged as printing; when the similarity is less than 0.6, the main color is greater than or equal to 8, and the pen trace is detected, it is judged as painting; multi-source cross verification of image features and physical structure data is realized to realize accurate identification of ceramic decoration formation mode against environmental interference.

[0043] Finally, it should be noted that: the above only describes the preferred embodiments of the application and is not used to limit the application, although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or make equivalent replacement for part of the technical features. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. A method of discriminating between ceramic decoration forming methods, characterized by: The method comprises the following steps: Step 1: obtaining a digital image of the ceramic surface decoration; Step 2: calculating a pattern similarity index of different regions in the image; Step 3: analyzing color space data of the image, counting the number of main colors and the standard deviation of brightness; Step 4: detecting microscopic features of the image, including at least one of dot arrangement, ink accumulation, and pen trace; Step 5: combining the similarity index, color space data, and microscopic features to determine the formation method of the decoration as decal, printing, or painting.

2. A method of discriminating between ceramic decoration forming methods according to claim 1, characterized in that: The similarity index is calculated by structural similarity index SSIM and cosine similarity algorithm, and is realized based on image block comparison.

3. The method according to claim 1, wherein the method is characterized by: The color space data is counted by clustering algorithm after converting the RGB image to HSV space.

4. The method according to claim 1, wherein the method is characterized by: The dot arrangement is detected by Fourier transform to analyze frequency distribution, and the ink accumulation is identified by image roughness analysis.

5. The method according to claim 1, wherein the method is characterized by: The pen trace is extracted by microscopic imaging to detect the directional change characteristics of pen lifting, pen lowering, or pen stroke.

6. The method according to claim 1, wherein the method is characterized by: When the similarity index is greater than or equal to 0.9, the number of main colors is less than or equal to 3, and the standard deviation of brightness is less than 0.1, it is determined as decal decoration.

7. The method according to claim 1, wherein the method is characterized by: When the similarity index is 0.7 to 0.9, the number of main colors is 3 to 6, and the ink accumulation exists, it is determined as printing decoration.

8. The method according to claim 1, wherein the method is characterized by: When the similarity index is less than 0.6, the number of main colors is greater than or equal to 8, and the pen trace is detected, it is determined as painting decoration.

9. The method according to claim 1, wherein the method is characterized by: Step 4 also includes detecting ceramic glaze roughness data, which is obtained by a three-dimensional scanner.

10. The method according to claim 1, wherein the method is characterized by: Step 4 also includes detecting body symmetry data, which is quantitatively analyzed by an image measuring instrument.

Images