Automatic Ceramic Glazing Control System and Method Integrating Machine Vision and Trajectory Planning

Abstract

This invention discloses an automatic ceramic glazing control system and method integrating machine vision and trajectory planning, belonging to the field of ceramic production automation technology. The method includes acquiring basic information and glazing requirements of the ceramic to be glazed. The automatic ceramic glazing control system and method of this invention, integrating machine vision and trajectory planning, acquires three-dimensional data by capturing ceramic images from multiple angles and scanning them. This allows for precise capture of the three-dimensional features of surface contours, textures, and minute defects, and can even identify depressions and protrusions. When planning the glazing path, the nozzle angle is adjusted in real time according to the ceramic surface morphology to ensure that the glaze is sprayed perpendicular to curved and irregularly shaped surfaces. Simultaneously, the nozzle movement speed and glaze output are matched to avoid deviations in material usage per unit area. These factors work together to improve the uniformity of glaze thickness and the accuracy of coverage area, solving the problems of missed coating on irregularly shaped ceramics and large thickness fluctuations in curved products, significantly improving the pass rate of complex ceramic products.

Description

Technical Field

[0001] This invention relates to the field of ceramic production automation technology, and in particular to an automatic ceramic glazing control system and method that integrates machine vision and trajectory planning. Background Technology

[0002] In the ceramics manufacturing industry, the glazing process is a crucial step that determines the appearance, texture, and performance of a product. While various non-manual glazing methods exist in the market, they still have significant technical shortcomings in adapting to complex products, responding to environmental changes, and ensuring quality stability. These shortcomings make it difficult to meet the demands of high-precision and high-efficiency production. Specific problems are particularly prominent in the following mainstream glazing methods:

[0003] Fixed-track glazing is commonly used in large-scale production lines for daily-use ceramics. It involves controlling the glazing nozzle along a fixed path using a pre-set mechanical motion trajectory. However, this method has significant limitations: First, the trajectory is completely fixed and cannot adapt to changes in product size. When switching product specifications, the mechanical components must be disassembled and the trajectory parameters readjusted, a time-consuming process that can lead to production line downtime and affect order delivery efficiency. Second, it cannot recognize differences in product surface morphology. Glazing areas with curved transitions using a flat trajectory can easily result in uneven glaze thickness, reducing product yield. Third, it lacks adaptability to surface defects. If the ceramic body has minor dents or other defects, fixed-track glazing can cause glaze buildup in the defective areas, creating additional defects that require subsequent manual repair, increasing production costs.

[0004] Some high-end ceramic production lines have attempted to introduce vision technology, using industrial cameras to capture two-dimensional images of ceramics to assist in adjusting glazing parameters. However, this method still has limitations: First, it can only acquire two-dimensional information and cannot construct a three-dimensional model of the ceramic surface. For irregularly shaped products, it cannot identify the three-dimensional shape of surface protrusions and depressions, which can easily lead to problems such as the nozzle colliding with raised areas or missing coating in recessed areas during glazing. Second, the vision function is limited, only used to identify the product position, without combining glaze characteristics and motion parameters for linkage optimization. After identifying product surface defects, it cannot automatically adjust the glaze output of the corresponding area, affecting product quality. Third, it lacks the ability to reuse historical data. Every time the product type changes, technicians need to readjust the glazing parameters based on the new two-dimensional images, resulting in a long adjustment cycle and an inability to quickly respond to the production needs of small batches and multiple varieties.

[0005] While existing glazing methods have freed them from complete reliance on manual labor, they all suffer from technical shortcomings: traditional mechanical fixed-track glazing has poor adaptability, simple automated glazing has low precision and weak anti-interference capabilities, and single vision-assisted glazing lacks three-dimensional recognition and linkage optimization capabilities. These problems result in low pass rates for complex products, low changeover efficiency, and poor environmental adaptability in ceramic glazing production, failing to meet the current ceramic industry's demands for high-end, customized, and flexible development. To address these issues, we propose an automated ceramic glazing control system and method that integrates machine vision and trajectory planning. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides an automated ceramic glazing control system and method that integrates machine vision and trajectory planning, achieving full automation, precision, and intelligence in the ceramic glazing process, improving the stability of glazing quality, reducing labor costs, and shortening the production cycle.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] An automated ceramic glazing control method integrating machine vision and trajectory planning, comprising:

[0011] Obtain the basic information and glazing requirements of the ceramic to be glazed. The basic information reflects the appearance and surface condition of the ceramic, including its dimensions, surface shape, and surface condition. The glazing requirements are the glazing-related standards determined according to production requirements, including the target glazing thickness and target coverage area.

[0012] The technology uses image acquisition and analysis to perform visual processing on ceramics. This technology acquires ceramic images through specialized equipment and extracts useful information through calculation and analysis. After processing, it obtains the ceramic's feature data and glazing parameters, and at the same time, it establishes a three-dimensional model of the ceramic surface. This model can display the ceramic surface morphology in three dimensions, and realize the three-dimensional visualization of ceramic features through the model.

[0013] Based on the characteristic data of ceramics, glazing parameters and three-dimensional models, combined with the influence analysis of real-time environmental temperature and humidity data on trajectory planning, trajectory planning is the process of designing the movement path and related parameters of the equipment performing the glazing operation. Based on the analysis and design of the movement path and movement parameters of the glazing equipment, a glazing trajectory scheme that can be adjusted according to the actual situation is formed.

[0014] The glazing equipment is controlled to perform automatic glazing operation on ceramics according to the glazing trajectory scheme. During the process, the temperature change data of the glazing area is collected in real time, and the material output and movement speed of the glazing equipment are dynamically adjusted according to the temperature change data.

[0015] Visual inspection and thickness measurement are performed on the glazed ceramics. The actual glaze thickness data is obtained through a laser thickness measuring device. This device uses laser technology to measure the thickness of the glaze layer and combines the visual inspection results to form a glazing quality result that includes multiple evaluation aspects.

[0016] If the glazing quality result meets the preset quality standard, the glazing parameters, trajectory scheme, and quality result are stored in the database to form a case library. The case library is classified and searched according to ceramic type, glazing requirements, and environmental conditions. If the result does not meet the preset quality standard, the system uses successful glazing data with similar ceramic characteristics in the case library to perform intelligent matching based on the cause of the quality defect, generates parameter adjustment suggestions, optimizes the glazing trajectory scheme, and re-executes the trajectory scheme until the glazing quality result meets the preset quality standard.

[0017] Preferably, the steps of obtaining the basic information and glazing requirements of the ceramic to be glazed include:

[0018] The ceramic to be glazed is placed on a preset positioning platform. This platform is specifically designed to fix the ceramic and help determine its position. Through dual calibration using positioning sensors and visual positioning, the positioning sensors can sense the position of the ceramic, and the visual positioning determines the position of the ceramic through image analysis. After dual calibration, the precise position of the ceramic on the platform is determined, and the positioning error is controlled within a high-precision range.

[0019] The image acquisition device consisting of a high-definition industrial camera and a structured light scanner is activated. The high-definition industrial camera is used in industrial scenarios and has high shooting accuracy. The structured light scanner obtains the three-dimensional information of the object by emitting structured light and receiving reflected light. It captures ceramic images from three or more different angles, including the front, side, and top, and scans the ceramic surface. At the same time, it acquires two-dimensional image data reflecting the planar shape of the ceramic and three-dimensional point cloud data composed of a large number of three-dimensional coordinate points.

[0020] Two-dimensional image data and three-dimensional point cloud data are fused together. This process combines the useful information from both types of data and extracts the external dimensions of the ceramic, including height, diameter, length and width.

[0021] The morphology of ceramic surfaces is analyzed by three-dimensional modeling to accurately determine whether they are flat, curved, arc-shaped, or irregularly shaped. The transition angle and radius of curvature between different surfaces are marked. The transition angle is the angle between the connecting parts of adjacent surfaces, and the radius of curvature reflects the degree of curvature of the surface.

[0022] Combining image grayscale analysis and 3D point cloud comparison to detect the surface condition of ceramics, image grayscale analysis obtains information by analyzing the brightness of different areas in the image, while 3D point cloud comparison compares the actual 3D point cloud data of the ceramic with standard data. Both methods are used to detect whether there are problems such as cracks, depressions, protrusions or stains on the ceramic surface, while quantifying parameters such as the depth and width of defects.

[0023] Based on the glazing requirements document, determine the specific value of the target glazing thickness and the specific location range of the target coverage area on the ceramic surface. Then, associate this range with the coordinate system of the ceramic 3D model. The coordinate system is used to determine the position of each point in the 3D model. Through association, the precise positioning of the coverage area is achieved.

[0024] Preferably, the step of visually processing the ceramic using image acquisition and analysis technology to obtain the ceramic's feature data and glazing parameters includes:

[0025] The acquired ceramic images and 3D point cloud data are preprocessed. Gaussian filtering is used to remove image noise points, which can reduce useless interference points in the image. Histogram equalization is used to enhance image contrast, which can make the difference between light and dark in the image more obvious. The 3D data is optimized by combining point cloud denoising algorithm, which can reduce useless points in 3D point cloud data. After preprocessing, high-quality images and 3D model data are obtained.

[0026] Extracting feature data of ceramics from high-quality data, specifically including curve equation parameters describing the contour curve of ceramic surface, texture gray value distribution law reflecting the distribution characteristics of light and dark on ceramic surface, and three-dimensional coordinates and volume size to determine the location and scale of defects;

[0027] Based on the shape and state of the ceramic surface, a deep learning algorithm is used to identify key areas of interest for glazing. The deep learning algorithm can simulate the human learning process and extract patterns from the data. The key areas identified are specific areas around surface defects and areas with large transition angles of irregular surfaces.

[0028] Combining the target glazing thickness with the characteristics of the ceramic surface material, material data is obtained through a material detection module. This module is specifically designed to detect the relevant properties of the ceramic surface material. After obtaining the data, a correlation model is established that reflects the relationship between the output of the glazing material and the final glazing thickness. This model is used to calculate the output of the glazing material in different areas of the glazing equipment. At the same time, the distance parameter between the glazing equipment and the ceramic surface is dynamically determined based on the curvature change of the ceramic surface to ensure that the glazing thickness meets the target requirements.

[0029] The feature data, glazing material output, and distance parameters are summarized to form a set of glazing parameters containing all necessary information, and a mapping table that clearly shows the correspondence between each parameter and the 3D model is generated.

[0030] Preferably, the step of designing the motion path and motion parameters of the glazing equipment based on the characteristic data and glazing parameters of the ceramic, and forming a glazing trajectory scheme, specifically includes:

[0031] Based on the surface shape of the ceramic, the target glazing area, and the 3D model, the A* path planning algorithm is used to plan the movement path of the glazing equipment. The A* path planning algorithm can efficiently find the optimal movement route. The path must completely cover all target areas, and the path overlap rate should be controlled at a low level to avoid uneven glazing thickness.

[0032] For different shaped areas of the ceramic surface, the movement speed of the glazing equipment is calculated in combination with the flowability parameters of the glazing material. The flowability parameters are obtained from the material property database, which stores the relevant attribute data of various glazing materials. The movement speed of the planar area is set to a higher level, the curved area to a medium level, and the irregular surface area to a lower level to ensure uniform glazing.

[0033] The inverse kinematics algorithm is used to determine the angle adjustment method of the glazing equipment. The inverse kinematics algorithm can calculate the motion angle of each component of the equipment according to the target position. When the ceramic surface is a curved surface or an irregular surface, the spraying angle of the glazing equipment is adjusted in real time so that the glazing material is sprayed perpendicular to the ceramic surface, and the angle adjustment reaches the standard corresponding to a high precision.

[0034] Based on the output of glazing material in the glazing parameters, a flow rate and speed matching model is established in combination with the movement speed of the glazing equipment. This model reflects the matching relationship between the flow rate of glazing material and the movement speed of the equipment, ensuring that the amount of glazing material per unit area meets the target thickness requirements and the thickness error is controlled within a small range.

[0035] The motion path, motion speed, and angle adjustment methods are integrated in chronological order. The influence of real-time environmental temperature and humidity on the material curing speed is taken into account. The material curing speed is the speed at which the glazing material changes from a liquid to a solid state. The trajectory scheme is dynamically modified according to the influence to form a glazing trajectory scheme that clearly defines the position, speed, and angle of the glazing equipment at each time node.

[0036] Preferably, the step of controlling the glazing equipment to perform automatic glazing operation on the ceramic according to the glazing trajectory scheme specifically includes:

[0037] The motion path, speed, and angle adjustment method in the glazing trajectory scheme are converted into control commands that the glazing equipment can recognize. The control commands enable the glazing equipment to perform corresponding actions, and the command transmission delay is kept at a low level.

[0038] Based on the output amount of glazing material in the glazing parameters, the material flow rate and spraying pressure of the glazing equipment are set through the flow control valve. The flow control valve is used to adjust the output amount of glazing material. After the setting is completed, a pre-spraying test is carried out. The pre-spraying test is a small-scale spraying operation before formal glazing. The test ensures that the material output is stable.

[0039] Start the glazing equipment and execute motion and glazing operations according to the control instructions. At the same time, turn on the real-time monitoring device composed of a high-speed camera and a temperature sensor. The high-speed camera has a fast shooting speed and can capture the dynamic process, while the temperature sensor can sense temperature changes. The device acquires image data and equipment operation data during the glazing process at high frequency.

[0040] Based on the real-time monitored image data, the image segmentation algorithm is used to determine whether there is any omission or excessive thickness in the glazing area. The image segmentation algorithm can divide the image into different regions and identify specific regions. If there is omission or excessive thickness, the material flow rate and movement speed of the glazing equipment are adjusted in real time through the PID control algorithm. The PID control algorithm is adjusted according to the deviation between the actual situation and the target, and the adjustment response time is controlled in a short time.

[0041] The real-time monitored equipment operation data is compared with the preset data in the trajectory plan. The equipment operation data includes equipment joint angles, motor speeds, etc. If there is a deviation, the motion parameters of the equipment are adjusted through the servo control system. The servo control system can accurately control the movement of the equipment and ensure motion accuracy through adjustment.

[0042] Preferably, the step of visually inspecting the glazed ceramic to obtain the glazing quality result includes:

[0043] The detection device, consisting of a high-definition camera and a laser thickness gauge, is activated. The high-definition camera has high shooting accuracy, and the laser thickness gauge uses laser technology to measure thickness. The device captures an all-round image of the glazed ceramic and simultaneously obtains the actual thickness data of different areas on the ceramic surface, thus obtaining the image data and thickness data after glazing.

[0044] By comparing the image data after glazing with the ceramic image data before glazing, the image registration algorithm is used to analyze whether the glazed area completely covers the target area. The image registration algorithm can align different images to facilitate comparative analysis, identify the location and range of the missed areas, and achieve a high accuracy in the corresponding identification standard.

[0045] By comparing the laser thickness measurement data with the target thickness, the uniformity of the glaze layer thickness distribution is calculated. Uniformity is the degree of consistency of the thickness of each area of ​​the glaze layer. It is used to determine whether the glaze thickness in different areas meets the target thickness requirements and to identify the location and degree of areas that are too thick or too thin.

[0046] Defect detection algorithms are used to detect whether there are glazing defects such as bubbles, runs, and pinholes on the surface of the glaze layer. The defect detection algorithm can identify problems on the surface of the glaze layer and record the type, location, and size of the defects.

[0047] The overall glazing quality is scored by summarizing the coverage integrity, thickness uniformity, and surface defects. The fuzzy comprehensive evaluation method can comprehensively evaluate multiple factors and form a glazing quality result that includes the score and defect details.

[0048] An automated ceramic glazing control system integrating machine vision and trajectory planning includes the following modules:

[0049] The information acquisition module is used to collect ceramic-related information, obtain basic information and glazing requirements of the ceramic to be glazed. The basic information reflects the appearance and surface condition of the ceramic and includes its size, surface shape and surface condition. The glazing requirements are determined according to production requirements and include the target thickness and target coverage area of ​​the glaze. This module also integrates a material detection unit, which is specifically used to detect the material properties of the ceramic surface and can detect the material characteristics of the ceramic surface and output material data.

[0050] The visual analysis module is used for visual processing and analysis of ceramics. It obtains the characteristic data and glazing parameters of ceramics through image acquisition and analysis technology. This technology acquires ceramic images through specialized equipment and extracts useful information through calculation and analysis. At the same time, it builds a three-dimensional model of the ceramic surface. This model can display the surface morphology of ceramics in three dimensions and realize the three-dimensional visualization of features through the model.

[0051] The trajectory planning module is used to design motion-related schemes for the glazing equipment. Based on the ceramic characteristic data, glazing parameters, and 3D model, combined with the real-time temperature and humidity data obtained by the environmental perception module (which acquires ambient temperature and humidity information), the module designs the motion path and motion parameters of the glazing equipment to form a glazing trajectory scheme that can be adjusted according to actual conditions.

[0052] The glazing control module is used to control the glazing equipment to perform operations. It controls the glazing equipment to perform automatic glazing operations on ceramics according to the glazing trajectory plan. It integrates a real-time monitoring unit and a PID control unit. The real-time monitoring unit can acquire glazing process data in real time, and the PID control unit uses the PID control algorithm to make adjustments. Through these two units, the operating parameters of the equipment are adjusted in real time to ensure glazing accuracy.

[0053] The quality inspection module is used to evaluate the glazing quality. It performs visual inspection and actual thickness measurement on the glazed ceramics. The actual thickness data is obtained through a laser thickness measuring device. This device uses laser technology to measure the thickness of the glaze layer and combines the visual inspection results to form a glazing quality result that includes multiple evaluation aspects.

[0054] The judgment and feedback module is used to judge the glazing quality and provide adjustment solutions. It judges whether the glazing quality results meet the preset quality standards. If they do, the relevant data are stored in the case database according to the classification and retrieval labels. If they do not, the module performs intelligent retrieval and parameter matching of similar cases based on the case database, generates an optimization solution, and restarts the glazing and testing process.

[0055] The case database module is used to store glazing-related data, including parameters, trajectory schemes, and quality results of each successful glazing process. It supports multi-dimensional classification retrieval and intelligent recommendation of similar cases, providing data support for parameter adjustment.

[0056] Preferably, the information acquisition module specifically includes the following units:

[0057] The ceramic positioning unit is used to place the ceramic to be glazed on a preset positioning platform. This platform is specifically designed to fix the ceramic and help determine its position. Through dual calibration of positioning sensor and visual positioning, the positioning sensor can sense the position of the ceramic, and the visual positioning determines the position of the ceramic through image analysis. After dual calibration, the precise position of the ceramic on the platform is determined.

[0058] The multi-dimensional data acquisition unit is used to activate the image acquisition device composed of a high-definition industrial camera and a structured light scanner. The high-definition industrial camera is used in industrial scenarios and has high shooting accuracy. The structured light scanner obtains the three-dimensional information of the object by emitting structured light and receiving reflected light. It captures ceramic images from multiple angles and scans the surface to obtain two-dimensional image data reflecting the planar shape of the ceramic and three-dimensional point cloud data composed of a large number of three-dimensional coordinate points.

[0059] The data fusion processing unit is used to fuse two-dimensional image data and three-dimensional point cloud data. This process combines the useful information of the two types of data and extracts the external dimensions of the ceramic after processing. The dimensions reach a high accuracy standard.

[0060] The surface morphology analysis unit is used to analyze the morphology of ceramic surfaces through a 3D model, and to annotate the transition angles and radii of curvature between different surfaces. The transition angle is the angle between the connecting parts of adjacent surfaces, and the radius of curvature reflects the degree of curvature of the surface.

[0061] The surface defect quantification unit is used to detect ceramic surface defects by combining image grayscale analysis and 3D point cloud comparison. Image grayscale analysis obtains information by analyzing the brightness of different areas in the image, while 3D point cloud comparison compares the actual 3D point cloud data of the ceramic with standard data. Defects are detected and defect parameters are quantified through these two methods.

[0062] The requirement analysis and mapping unit is used to determine the target parameters based on the glazing requirement document, and associate the target coverage area with the coordinate system of the ceramic 3D model. The coordinate system is used to determine the position of each point in the 3D model, and the accurate positioning of the coverage area is achieved through association.

[0063] Preferably, the visual analysis module specifically includes the following units:

[0064] The multi-source data preprocessing unit is used to preprocess ceramic images and 3D point cloud data. It employs various algorithms to optimize data quality by reducing useless interference points in the image, enhancing the difference between light and dark areas in the image, and optimizing the 3D point cloud data to obtain high-quality image and 3D model data.

[0065] The deep learning feature extraction unit is used to extract feature data of ceramics from high-quality data. It includes mathematical parameters describing the contour curve of the ceramic surface, texture gray value rules reflecting the distribution characteristics of the brightness of the ceramic surface, and three-dimensional coordinates and volume parameters to determine the location and scale of defects, providing data support for subsequent glazing parameter calculations.

[0066] The key area intelligent recognition unit is used to identify specific areas around surface defects and key glazing areas with large transition angles on irregular surfaces based on the shape and state of the ceramic surface, using computational methods that can simulate the human learning process, thus ensuring that the glazing process is targeted at key areas.

[0067] The material and parameter correlation calculation unit is used to combine the target glazing thickness with the material characteristics of the ceramic surface. It obtains material data through a component that specifically detects the ceramic surface material, establishes a correlation model that reflects the relationship between the output of glazing material and the final glazing thickness, calculates the output of glazing material in different areas of the glazing equipment through this model, and dynamically determines the distance parameter between the glazing equipment and the ceramic surface based on the curvature change of the ceramic surface to ensure that the glazing thickness in each area meets the target requirements.

[0068] The parameter and model mapping unit is used to summarize feature data, glazing material output, and distance parameters to form a set of glazing parameters containing all necessary information. It also generates a mapping table that clearly shows the correspondence between each parameter and the 3D model, providing intuitive data basis for the trajectory planning module to call parameters.

[0069] Preferably, the trajectory planning module specifically includes the following units:

[0070] The intelligent path design unit is used to plan the movement path of the glazing equipment based on the surface shape of the ceramic, the target glazing area, and the three-dimensional model, using a calculation method that can efficiently find the optimal movement path. The path must completely cover all target areas and the overlap rate must be controlled at a low level to avoid uneven glazing thickness due to path repetition.

[0071] The material properties and speed matching unit is used to obtain flow parameters from the database storing glazing material property data for different shaped areas of the ceramic surface, and calculate the movement speed of the glazing equipment based on these parameters. The speed is set to a higher level for planar areas, a medium level for curved areas, and a lower level for irregularly shaped areas to ensure uniform glazing.

[0072] The adjustment unit is used to determine the angle adjustment method of the glazing equipment by using a calculation method that can calculate the movement angle of the equipment parts based on the target position. When the ceramic surface is a curved or irregular surface, the spray angle is adjusted in real time to make the glazing material sprayed perpendicular to the ceramic surface, and the angle adjustment reaches a high precision standard.

[0073] The flow rate and speed dynamic matching unit is used to establish a flow rate and speed matching model based on the material output in the glazing parameters and the movement speed of the glazing equipment, to ensure that the amount of glazing material per unit area meets the target thickness requirements and the thickness error is controlled within a small range.

[0074] The environmental adaptive trajectory correction unit integrates the motion path, motion speed, and angle adjustment method in chronological order. It also incorporates the influence of real-time environmental temperature and humidity data on the material curing speed, which is the speed at which the glazing material changes from a liquid to a solid state. Based on this influence, the initially integrated trajectory scheme is dynamically corrected to ultimately form a glazing trajectory scheme. This scheme needs to clearly define the position, speed, and angle of the glazing equipment at each time point to ensure that the trajectory scheme can adapt to environmental changes and guarantee the stability of glazing quality under different temperature and humidity conditions.

[0075] (III) Beneficial Effects

[0076] 1. By capturing ceramic images from multiple angles and scanning them to obtain 3D data, it can accurately capture the 3D features of surface contours, textures, and minute defects, even identifying millimeter-level depressions and protrusions. When planning the glazing path, it adjusts the nozzle angle in real time according to the ceramic surface morphology to ensure that the glaze is sprayed perpendicular to curved and irregularly shaped surfaces. Simultaneously, it matches the nozzle movement speed with the glaze output to avoid deviations in material usage per unit area. This synergy improves the uniformity of glaze thickness and the accuracy of coverage, completely solving the problems of missed coating on irregularly shaped ceramics and large thickness fluctuations on curved products. The pass rate of ceramic products has been significantly improved. In addition, when planning the glazing scheme, environmental temperature and humidity data are collected in real time to analyze their impact on the curing speed of the glaze. For example, when the temperature is too low, the nozzle movement speed is slowed down to prevent the glaze from flowing, and when the humidity is too high, the spray pressure is finely adjusted to ensure stable glaze adhesion. Moreover, for ceramics with different shapes and surface conditions, there is no need to adjust the hardware. It can be quickly adapted by analyzing the product's three-dimensional data, so as to achieve production change without interruption. This dual adaptability to the environment and products has greatly reduced the defect rate caused by production fluctuations, and its adaptability far exceeds that of existing glazing methods.

[0077] 2. Successful glazing parameters and path schemes are categorized and stored by product type and environmental conditions. When changing products, parameters from similar cases can be directly called upon and fine-tuned according to the current product characteristics, eliminating the need for starting from scratch. Furthermore, the product status and equipment deviations are monitored in real time during the glazing process, and issues such as missed coatings and uneven thickness are corrected immediately to avoid batch defects. These designs reduce changeover debugging time from several days to less than half an hour, allowing for problem correction during production without waiting for post-production inspection, significantly increasing the daily capacity of a single production line. Secondly, the planned path achieves full automation of the glazing process, eliminating the need for manual operation and greatly reducing labor input. Simultaneously, real-time monitoring and immediate correction functions allow defects to be resolved in their early stages, avoiding batch rework and significantly reducing glaze waste. In addition, the ability to reuse historical cases avoids the trial-and-error costs of each changeover, allowing for the determination of accurate parameters without repeated trial coatings. Attached Figure Description

[0078] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0079] Figure 1 This is an overall flowchart of an embodiment of the present invention. Detailed Implementation

[0080] This application provides an automated ceramic glazing control system and method that integrates machine vision and trajectory planning. This achieves full automation, precision, and intelligence in the ceramic glazing process, improving glazing quality stability, reducing labor costs, and shortening the production cycle. By capturing ceramic images from multiple angles and scanning them to obtain 3D data, it can accurately capture the 3D features of surface contours, textures, and minute defects, even identifying millimeter-level depressions and protrusions. When planning the glazing path, it adjusts the nozzle angle in real time according to the ceramic surface morphology to ensure that the glaze is sprayed perpendicular to curved and irregular surfaces. Simultaneously, it matches the nozzle movement speed with the glaze output to avoid deviations in material usage per unit area. These factors work together to improve the uniformity of glaze thickness and the accuracy of the coverage area, completely solving the problems of missed coating on irregularly shaped ceramics and large thickness fluctuations in curved products, significantly improving the pass rate of complex ceramic products.

[0081] Example: The technical solution in this application aims to achieve full automation, precision, and intelligence in the ceramic glazing process, thereby improving the stability of glazing quality, reducing labor costs, and shortening the production cycle. The overall approach is as follows:

[0082] Ceramic glazing is a core process that determines the appearance, texture, mechanical strength, and corrosion resistance of ceramic products, and its quality directly affects the product's market competitiveness. Currently, ceramic glazing production mainly relies on two methods: manual glazing and semi-automatic equipment. However, both methods have significant limitations. Manual glazing requires operators to control the spray gun angle, distance, and movement speed based on experience, leading to some uncontrollability in glaze thickness deviation, coverage area misalignment rate, and single-piece production cycle. While semi-automatic equipment can reduce some manual intervention, it mainly supports regular-shaped ceramics and cannot adapt to irregularly shaped artistic ceramics or ceramics with surface defects. Furthermore, its trajectory planning scheme is fixed and does not consider the impact of ambient temperature and humidity on the glaze curing speed. For example, when the temperature fluctuates by ±5℃ or the humidity fluctuates by ±10%, the glazing pass rate will decrease by 10% to 15%. In addition, both methods lack a mechanism for reusing historical production data. For the same type of ceramic, parameters need to be readjusted for each production run, resulting in a certain degree of glaze waste and labor time loss.

[0083] In recent years, the application of machine vision and trajectory planning technology in ceramic glazing has gradually increased, but existing research still suffers from technological gaps: some studies use monocular cameras to acquire two-dimensional images of ceramics, which can only identify planar dimensions and surface textures, but cannot capture three-dimensional features such as surface curvature and irregular transition angles, resulting in uneven glazing thickness on curved ceramics; some studies plan glazing paths based on the A* algorithm, which improves coverage integrity, but does not combine glaze viscosity to adjust the movement speed, nor does it consider the influence of temperature and humidity on curing speed, making it prone to sagging defects in high temperature and high humidity environments; other studies have designed machine vision-based quality inspection systems that can identify defects such as missed coating and bubbles, but lack real-time feedback from the glazing execution module, cannot dynamically correct parameters, and require manual readjustment, resulting in low efficiency.

[0084] To address the problems existing in the prior art, this invention provides an automatic ceramic glazing control method that integrates machine vision and trajectory planning. This glazing control method includes:

[0085] I. Architecture:

[0086] (a) Data interaction:

[0087] The system architecture employs a combination of modular and distributed approaches, primarily consisting of seven core modules. These modules interact in real-time via industrial Ethernet using the Profinet protocol. The overall architecture is as follows: Figure 1 As shown; the functions and data flow of each module are as follows:

[0088] The information acquisition module, acting as the data input, collects basic ceramic information, glazing requirements, and environmental data, transmitting them to the visual analysis module. Basic ceramic information includes dimensions, shape, and defects; glazing requirements include target thickness and coverage area; and environmental data primarily includes temperature and humidity. The visual analysis module preprocesses the collected data, extracts features, calculates parameters such as glaze flow rate and equipment distance, and transmits this data to the trajectory planning module. The trajectory planning module then optimizes the path, speed, and angle based on the environmental data, generating a trajectory plan and transmitting it to the glazing control module. The glazing control module converts the trajectory plan into equipment instructions, executes the glazing operation, and simultaneously transmits real-time monitoring data to the quality inspection module. The quality inspection module evaluates the glazing quality, including coverage integrity, thickness uniformity, and surface defects, generating quality results and transmitting them to the judgment feedback module. The judgment feedback module stores qualified data in the case database and retrieves parameter adjustment suggestions from the case database for unqualified data, feeding them back to the glazing control module. The case database module stores historical production data, providing similar case retrieval services for the judgment feedback module and parameter recommendations for new production requirements. Specific details are as follows:

[0089] (II) Information Collection Module:

[0090] The system is responsible for acquiring accurate raw data, encompassing three key stages: ceramic positioning, multi-dimensional data acquisition, and data fusion processing. Ceramic positioning employs dual calibration using a laser displacement sensor and a high-definition industrial camera. During operation, the ceramic is placed on a platform with positioning marks. The laser sensor detects the distance deviation between the bottom of the ceramic and the platform, while the camera captures the relative position of the positioning marks and the ceramic edge. A coordinate calibration algorithm based on the least squares method controls the ceramic coordinate error to within ≤0.1mm, ensuring a consistent benchmark for subsequent image acquisition and glazing operations. In the multi-dimensional data acquisition stage, the high-definition industrial camera captures images from five angles: front, side, top, 45° oblique, and 135° oblique. For irregularly shaped ceramics, the shooting angle is set and added according to the ceramic shape to acquire two-dimensional images of the ceramic surface texture and defects. The structured light scanner achieves a point cloud density of 100 points / mm². 2 The process involves scanning the ceramic surface to generate 3D point cloud data, reflecting the ceramic's three-dimensional morphology, such as the radius of curvature of curved surfaces and the angle of transition between irregular shapes. In the data fusion processing stage, a coordinate mapping algorithm is used to associate the 2D image with the 3D point cloud. A coordinate system is established with the center point of the positioning platform as the origin, converting the pixel coordinates of the 2D image into 3D spatial coordinates. Then, a feature matching algorithm is used to extract the ceramic dimensions, namely height, diameter, length, and width, identifying the surface morphology as planar, curved, or irregular. The radius of curvature for curved surfaces is identified within a range of 50 to 500 mm. Defect parameters such as crack length are identified up to 100 mm, and depression depth is identified within a range of 0.1 to 1 mm. Simultaneously, the glazing requirement document is parsed, such as a target thickness of 0.1 to 0.2 mm, covering the entire surface or a local area. The covered area is mapped to the 3D coordinate system, generating a data table corresponding to the ceramic features and glazing requirements.

[0091] (III) Visual Analysis Module:

[0092] The raw data is transformed into executable glazing parameters, and the implementation steps are as follows: In the data preprocessing stage, Gaussian filtering with 3×3 to 5×5 kernels is used to remove random noise for 2D images, and the standard deviation of the filter is set to 0.5 to 1.0, which is adjusted according to the noise intensity of the image. Histogram equalization with gray levels ranging from 0 to 255 is used to enhance the contrast between surface texture and defects. For 3D point cloud data, statistical filtering is used to remove outliers, with the number of neighborhood points set to 50 and the distance threshold set to 1.5 times the standard deviation to ensure data quality. In the feature extraction stage, the ceramic surface contour curve parameters are extracted based on a convolutional neural network, using a ResNet-50 network structure. The training dataset contains image data of more than 1000 different ceramics. For circular ceramics, a circular arc equation is fitted; for irregularly shaped ceramics, a piecewise cubic B-spline curve is fitted, with the node spacing set to 5mm. The 3D coordinates of defects are extracted based on the PointNet algorithm, with the number of input point clouds ranging from 10,000 to 20,000 points.

[0093] The glazing parameter calculation step is the core of the visual analysis module. It requires establishing a quantitative correlation based on the characteristics of the ceramic material, the target glazing thickness, and the equipment's movement speed. This is derived through fitting experimental data to ensure that the amount of glaze used per unit area accurately corresponds to the target thickness, avoiding drips due to excessive glaze or missed areas due to insufficient glaze. The core calculation formula is: Q = k × d × v, where Q is the glaze output rate in mL / min, representing the volume of glaze sprayed from the glazing gun per unit time. This directly determines the amount of glaze accumulated per unit area, with a value ranging from 0.5 to 5.0 mL / min, which needs to be matched with the adjustment range of the high-precision flow valve. Its value must balance the thickness requirement with the equipment's capacity; for example, for everyday ceramics... When the target thickness of the porcelain bowl is 0.1 mm and the speed is 10 mm / s, Q needs to be controlled at around 1.1 mL / min. Too high a speed can easily lead to sagging, while too low a speed can easily lead to missed coating. k is the material coefficient, reflecting the ceramic surface's adsorption capacity and penetration characteristics of the glaze. Its value ranges from 1.0 to 1.5 and needs to be determined by measuring the ceramic's water absorption rate and porosity using a near-infrared spectrometer: k = 1.0 to 1.1 when the water absorption rate is 2% to 5%, and k = 1.2 to 1.3 when the water absorption rate is 5% to 8%. If there are defects such as cracks or depressions on the ceramic surface, the defective areas need to be filled with additional glaze, and the k value needs to be increased by 0.2. For example, k = 1.1 for normal areas and k = 1.3 for defective areas, ensuring that the glaze thickness in the defective areas meets the standard. d is the upper... The target glaze thickness, in mm, is determined based on the ceramic product design standards and usage scenarios, ranging from 0.1 to 0.2 mm: For everyday ceramics, such as bowls and plates, d = 0.1 to 0.12 mm; too thick a glaze will reduce gloss, while too thin a glaze will affect surface smoothness; for artistic ceramics, d = 0.12 to 0.15 mm, balancing decorative effect and glaze flatness; for industrial ceramics, d = 0.15 to 0.2 mm, improving wear resistance and corrosion resistance. The accuracy of d setting needs to reach 0.01 mm to match the measurement accuracy of the laser thickness gauge, ensuring that the deviation between the target value and the measured value can be quantified and assessed; v is the equipment movement speed, in mm / s, referring to the linear velocity of the multi-axis robotic arm driving the spray gun. The dwell time of the spray gun on a unit area of ​​ceramic surface is determined, ranging from 3.0 to 15.0 mm / s. This needs to be adjusted based on the ceramic surface morphology and glaze viscosity: for flat areas with smooth surfaces, v = 10 to 15 mm / s; for curved areas with large curvature variations, v = 5 to 10 mm / s; for irregularly shaped areas with complex morphology, v = 3 to 5 mm / s; for glaze viscosities of 500 to 600 mPa·s (low viscosity), v can be appropriately increased; for 600 to 800 mPa·s (high viscosity), v needs to be decreased to prevent glaze accumulation. Simultaneously, v must match the path in the trajectory planning. For spiral paths, v remains stable; for segmented paths, v transitions smoothly between segments, with acceleration controlled at ≤5 mm / s². 2 This avoids speed fluctuations caused by impacts from the robotic arm.

[0094] The glazing parameters are calculated based on the characteristics of the ceramic material and the target thickness: For example, for a daily-use ceramic bowl with a water absorption rate of 5% (k=1.1), a target thickness of 0.1mm, and a speed of 10mm / s in a flat area, Q=1.1×0.1×10=1.1mL / min; for an industrial ceramic tube with a cracked area and a water absorption rate of 7% (k=1.5), a target thickness of 0.15mm, and a speed of 5mm / s, Q=1.5×0.15×5=1.125mL / min; simultaneously, the distance between the device and the ceramic is adjusted according to the curvature of the ceramic surface: 9 to 10mm in a flat area, 7 to 8mm in a curved area, and 5 to 6mm in an irregularly shaped area. The distance is adjusted by the movement of the robotic arm along the Z-axis; finally, the text is generated. A standard parameter mapping table with 3D coordinates is used, where the key is the coordinate range, such as x∈[0,10]mm, y∈[0,10]mm, z∈[50,60]mm, and the value is the corresponding flow rate Q and distance h, which facilitates the trajectory planning module to call parameters by coordinate. During fault handling, if the contrast of the preprocessed image is still insufficient, check the brightness of the ring light source, which should be between 500 and 1000 lux. If the feature extraction error exceeds the threshold, add 50 to 100 similar ceramic samples to retrain the neural network or adjust the number of point cloud samples to 25,000 points. If the actual thickness after parameter calculation deviates from the target by more than ±0.02mm, correct the material coefficient k, adjusting it by 0.05 each time and recalculating.

[0095] (iv) Trajectory Planning and Glazing Control Module:

[0096] Collaborative path optimization and precise execution are achieved. The trajectory planning employs the A* path planning algorithm, using the surface coordinates of the ceramic 3D model as nodes. The objective function is set to minimize path length and overlap rate, with Euclidean distance as the heuristic function. During planning, the path is ensured to completely cover the target area, with a coverage integrity ≥99% and an overlap rate ≤5% to avoid uneven thickness caused by repeated glazing. For irregularly shaped ceramics, a segmented path design is used, with each segment corresponding to a surface morphology such as a planar segment, curved segment, or transition segment. The speed transition between segments is smooth, with an acceleration set to 5 mm / s². 2To avoid impact, the trajectory is corrected based on environmental data. Using a temperature of 25℃ and humidity of 60% as a baseline, the speed increases by 10% for every 5℃ increase in temperature (e.g., from 10mm / s to 11mm / s); the spray pressure increases by 0.05MPa for every 10% increase in humidity (e.g., from 0.3MPa to 0.35MPa). This results in a final trajectory scheme in XML format, including the robotic arm position (x, y, z), speed v, spray angle θ, and spray pressure P at each 0.1s time point, ensuring the scheme can be directly parsed by the glazing control module. Glazing control... A 6-axis robotic arm equipped with a high-precision flow valve, with a flow rate adjustment range of 0 to 5 mL / min, was selected. First, the trajectory scheme was converted into commands recognizable by the robotic arm and transmitted via the Profinet protocol. Then, the flow rate and distance were set according to the mapping table, and a pre-spraying test was initiated. The spraying time was 1 to 2 seconds, and the thickness of the test glaze was collected. If the deviation exceeded ±0.01 mm, readjustment was performed. During execution, a high-speed camera captured real-time images of the glazed area. A U-Net-based semantic segmentation algorithm was used to determine whether there were any missed areas or excessive thickness. The criterion for missed areas was an area > 0.5 mm. 2 The standard for judging excessive thickness is 10% over the target thickness; the temperature sensor monitors the surface temperature. If the temperature rises and the curing speed is accelerated, the speed or flow rate is adjusted in real time using a PID control algorithm with a proportional coefficient Kp=2.0, an integral coefficient Ki=0.5, and a derivative coefficient Kd=0.1; at the same time, the joint angle of the robotic arm and the motor speed are monitored. When the deviation exceeds the threshold, it is corrected by the servo control system to ensure motion accuracy.

[0097] (v) Quality Inspection and Case Database Module:

[0098] To achieve quality assessment and data reuse, the quality inspection device consists of a high-definition camera and a laser thickness gauge. The high-definition camera uses the same equipment as the information acquisition module. After glazing, the camera captures images from five angles. The laser thickness gauge sets detection points on the ceramic surface at 2mm intervals, with ≥50 points for regular ceramics and ≥100 points for irregularly shaped ceramics. Quality analysis employs a fuzzy comprehensive evaluation method, establishing a three-level index system: the primary index is glazing quality with a weight of 1.0; the secondary indexes are coverage integrity, thickness uniformity, and surface defects, with weights of 0.4, 0.35, and 0.25 respectively; and in the tertiary index, the area of ​​missed coating receives 100 points for no missed coating, with each 0.5mm area receiving a score. 2 10 points will be deducted for missed coating; 100 points will be awarded if the standard deviation of thickness is ≤5% of the target thickness, and 20 points will be deducted for each 1% exceeding the target thickness; 100 points will be awarded if there are no bubbles, and 5 points will be deducted for each bubble; a comprehensive score of ≥80 points is considered qualified, and a quality report containing the scoring results, defect location map, and thickness distribution table will be generated.

[0099] The case database uses MySQL to store data. The data structure includes a ceramic information table, a glazing requirement table, a parameter table, an environment table, and a quality table. The ceramic information table records type, size, and defects; the glazing requirement table records target thickness and coverage area; the parameter table records flow rate, speed, and pressure. The flow rate parameter is calculated based on the core formula Q=k×d×v, and the specific values ​​of k, d, and v need to be stored together for easy retrieval and reuse. The environment table records temperature and humidity; and the quality table records scores and defects. A three-level index is established based on ceramic type, glazing requirements, and environmental conditions, such as daily-use ceramics, full surface, and 25℃ 60%RH. When searching for cases, the user enters the current... The type, demand, and environmental parameters of the ceramics are used to calculate the similarity with historical cases using a cosine similarity algorithm. The similarity is calculated as (current parameter vector × historical parameter vector) / (||current parameter vector|| × ||historical parameter vector||). The current parameters include the Q value calculated based on the formula and the corresponding k, d, and v, ensuring that the retrieved cases are highly matched in terms of parameter logic. Cases with a similarity of ≥85% are retrieved and the top 3 are recommended in order of similarity. The parameter reuse accuracy is ≥90%. The database supports the storage of 100,000+ cases, with a search response time of ≤1 second. Invalid cases, such as those using outdated glazes, are regularly deleted monthly to ensure data validity.

[0100] II. Complete Implementation Process of Control Methods:

[0101] The automatic ceramic glazing control method integrating machine vision and trajectory planning adopts a six-step process. Each step clearly defines the operational details, parameter standards, and troubleshooting to ensure stable implementation. The specific process is as follows:

[0102] Step 1, Information Acquisition:

[0103] Information acquisition is the foundation for all subsequent operations, and the accuracy and completeness of the data must be ensured. During ceramic positioning, the ceramic to be glazed is placed in the center of a platform with positioning marks. The positioning marks are cross-shaped engravings, 0.1mm deep, 0.2mm wide, spaced 100mm apart, and with an accuracy of ±0.02mm. A laser displacement sensor is activated to detect height deviation from four symmetrical points on the bottom of the ceramic, 20mm from the edge. If the deviation exceeds 0.1mm, it is adjusted using the platform's fine-tuning knob. Simultaneously, a high-definition industrial camera is activated to capture the relative image of the positioning marks and the ceramic edge. The least squares method is used to calculate the coordinate deviation between the ceramic center and the platform origin. If the deviation is >0.1mm, the electric slide of the platform is controlled for correction. The electric slide has a travel of 50mm and an accuracy of 0.005mm, ultimately ensuring that the ceramic coordinate error is ≤0.1mm.

[0104] Multi-dimensional data acquisition determined the acquisition angles based on the ceramic type. For regular ceramics, three angles were captured: front, side, and top. For irregularly shaped ceramics, two additional angles were captured: a 45° angle and a 135° angle, for a total of five angles. The high-definition industrial camera parameters were set to a resolution of 2592×1944, a frame rate of 30fps, and an exposure time of 10 to 20ms. For white ceramics, a 10ms exposure time was used to avoid overexposure, while for dark ceramics, a 20ms exposure time was used to ensure brightness. The gain was adjusted from 1.0 to 1.5dB according to the ambient light intensity, with 1.2dB used under natural light. Five images were captured from each angle, and three images were retained after removing blurry images with a sharpness lower than 0.8 for subsequent processing. The structured light scanner was set to fine mode with a point cloud density of 100 points / mm. 2 The scanning range covers the entire ceramic surface. If a point cloud missing area is found, it should be within 5mm. 2 Adjust the scanner angle and rescan to ensure the integrity of the 3D point cloud data.

[0105] Data fusion and requirements analysis employed a coordinate mapping program written in Python. A three-dimensional coordinate system was established with the platform origin at (0,0,0). The pixel coordinates of the two-dimensional image were converted to three-dimensional spatial coordinates using a camera intrinsic parameter matrix. The focal length in the camera intrinsic parameter matrix was set to 8mm, and the principal point coordinates were (1296,972). The conversion formulas were x=(u-1296)×z / f and y=(v-972)×z / f, where x and y are the three-dimensional spatial coordinates, z is the height value detected by the laser sensor, and f is the distance from the optical center of the camera lens to the image sensor, determining the image magnification. The conversion error was controlled to ≤0.03mm, and the pixel coordinates were (u,v). The ceramic dimensions were extracted using OpenCV's Canny edge detection algorithm, with an edge detection threshold set between 100 and 200. The height was the difference in z-coordinate between the top and bottom of the ceramic, and the diameter was the maximum horizontal x-coordinate of the circular ceramic. The standard deviation value represents the maximum coordinate difference in the x and y directions for length and width of irregularly shaped ceramics. Surface morphology is identified using a PointNet network: a curvature radius change of <5% at any point on the surface indicates a plane, 5% to 30% indicates a curved surface, and >30% indicates an irregularly shaped surface. Surface defects are detected through grayscale difference analysis; a grayscale difference >30 between the pre-glazing image and a standard defect-free image indicates a defect, and parameters such as crack length and indentation depth are recorded. Finally, the glazing requirement document stored in XML format is parsed to extract the target thickness and target coverage area. The boundary coordinates of the coverage area are marked in a three-dimensional coordinate system, generating a CSV format ceramic feature and glazing requirement data table for subsequent module calls. For fault handling, if the positioning deviation repeatedly exceeds the threshold ≥3 times, the levelness of the positioning platform and the calibration status of the laser sensor are checked. If data acquisition is missing, the cleanliness of the camera lens and the working distance of the scanner are checked.

[0106] Step 2, Visual Processing:

[0107] Visual processing requires converting raw data into executable glazing parameters, ensuring precise matching between these parameters and ceramic characteristics and glazing requirements. In the data preprocessing stage, MATLAB is used to implement Gaussian filtering and histogram equalization for 2D images: Gaussian filtering selects either a 3×3 kernel (σ=0.8) or a 5×5 kernel (σ=1.0) based on noise intensity to remove random noise such as dust highlights and uneven lighting spots; histogram equalization adjusts the grayscale distribution from 0 to 255 to a uniform distribution, enhancing the contrast of surface texture and defects. The standard deviation of the grayscale in the processed image must be ≥50; if the standard deviation of the original image is <30, the processing is repeated once. For 3D point cloud data, PCL is used for statistical filtering, setting the number of neighboring points to 50. The average distance between each point and its neighbors is calculated, and points with a distance greater than 1.5 times the standard deviation are identified as outliers and deleted. The point cloud density is maintained at 100 points / mm after processing. 2 This ensures the accuracy of subsequent feature extraction.

[0108] In the refined feature extraction stage, a ResNet-50 network was built based on the TensorFlow framework. The training dataset contained over 1000 images of different types of ceramics, including 300 types of daily-use ceramics, 400 types of artistic ceramics, and 300 types of industrial ceramics. Each sample contained a preprocessed image from 5 angles, labeled with contour parameters and defect locations. The network input was a 256×256 region of interest image, and the output was the contour curve parameters: for circular ceramics, the output was the arc equation (xa). 2 +(yb) 2 =r 2 The center (a, b) and radius r of the circle are calculated, with a center error ≤ 0.02 mm and a radius error ≤ 0.03 mm. The nodal coordinates of the piecewise cubic B-spline curve for irregularly shaped ceramics are output, with a nodal spacing of 5 mm and a curve fitting error ≤ 0.04 mm, and the maximum distance difference from the actual point cloud data. The three-dimensional point cloud data is processed based on the PointNet network, with the input being the coordinates of 10,000 to 20,000 uniformly sampled points, and the output being the three-dimensional coordinates of the defects: for cracks, the starting point (x1, y1, z1) and ending point (x2, y2, z2) are output; for depressions, the coordinates and depth of ≥ 8 boundary points are output, and the difference in z-coordinate between the lowest point of the depression and the surrounding normal surface is calculated, with a depth error ≤ 0.05 mm.

[0109] In the glazing parameter calculation stage, the surface material properties of the ceramic were first detected using a Thermo Scientific Nicoleti S50 near-infrared spectrometer with a wavelength range of 4000-4000 cm⁻¹. -1 32 scans, 4cm resolution -1Based on the spectral curves, the water absorption rate is calculated to be 3% to 8%, and the porosity is 1% to 3%. The formulas for calculating the water absorption rate are: Porosity = (Mass after water absorption - Mass after drying) / Mass after drying × 100% and Porosity = (Volume of water absorbed / Total volume of ceramic) × 100%. Then, the glaze flow rate is calculated based on the core formula Q = k × d × v, where the value of k is determined according to the water absorption rate; for a water absorption rate of 3% to 5%, it is taken as 1.0 to 1.1, and for 5% to 8%, it is taken as 1.2 to 1.3, and d is the upper limit. The target thickness in the glaze requirements document is 0.1 to 0.2 mm, and v is the equipment movement speed determined in the subsequent trajectory planning stage, ranging from 3 to 15 mm / s. For example, for a daily-use ceramic bowl with a water absorption rate of 5%, k=1.1; target thickness 0.1 mm; when the speed in the planar region is 10 mm / s, Q=1.1×0.1×10=1.1 mL / min; if the ceramic has crack defects, the k value in the defect area increases by 0.2, k=1.3, and the speed decreases to 5 mm / s, resulting in Q=1.3× 0.1 × 5 = 0.65 mL / min, ensuring sufficient glaze filling in defective areas; simultaneously, adjust the distance between the equipment and the ceramic according to the surface curvature: 9 to 10 mm for planar areas, 7 to 8 mm for curved areas, and 5 to 6 mm for irregularly shaped areas, with the distance adjusted via the z-axis movement of the robotic arm; finally, generate a dictionary-formatted parameter and three-dimensional coordinate mapping table, where the key is the coordinate range, such as x∈[0,10] mm, y∈[0,10] mm, z∈[50,60] mm, and the value is the corresponding flow rate Q and distance h, facilitating the trajectory planning module to call parameters by coordinate; during fault handling, if the preprocessed image contrast is still insufficient, check the brightness of the ring light source, which should be between 500 and 1000 lux; if the feature extraction error exceeds the threshold, add 50 to 100 similar ceramic samples to retrain the neural network or adjust the point cloud sampling number to 25,000 points; if the actual thickness after parameter calculation deviates from the target by more than ±0.02 mm, correct the material coefficient k and recalculate the Q value, adjusting by 0.05 each time.

[0110] Step 3, Trajectory Planning:

[0111] Trajectory planning requires combining ceramic characteristics, glazing parameters, and environmental data to generate a directly executable dynamic trajectory. In the basic path design phase, the A* path planning algorithm is implemented in C++ on the ROS platform. The map model is preprocessed 3D point cloud data of the ceramic surface, and the nodes are point cloud coordinates selected at 0.5mm intervals to ensure path smoothness. The cost function is a weighted sum of path length and overlap rate, with weights of 0.6 and 0.4 respectively. The heuristic function is the Euclidean distance from the current node to the target node. Planning starts from the initial point on the ceramic surface. For circular ceramics, a spiral path is used, with the spiral radius increasing by 2mm per revolution to match the glaze spray diameter. For irregularly shaped ceramics, segments are divided according to surface morphology: flat segments use straight paths, curved segments use arc paths, and transition segments use spline curve paths. The velocity change rate at the segment junctions is ≤5mm / s. 2To avoid impact from the robotic arm; after path generation, check coverage integrity and overlap rate. If coverage is insufficient, increase path density and node spacing to 0.3mm. If the overlap rate is too high, adjust path direction, such as changing clockwise to counterclockwise.

[0112] In the motion parameter calculation stage, based on the glaze viscosity measured by the NDJ-8S rotational viscometer (500 to 800 mPa·s) at a measurement temperature of 25°C, the basic motion speed v is determined. This speed will serve as the key parameter in the formula Q = k × d × v: for a viscosity of 500 to 600 mPa·s (low viscosity), the speed is 12 to 15 mm / s in a planar region, 8 to 10 mm / s in a curved region, and 4 to 5 mm / s in an irregularly shaped region; for a viscosity of 600 to 800 mPa·s (high viscosity)... The spraying speed is 10 to 12 mm / s in planar areas, 5 to 8 mm / s in curved areas, and 3 to 4 mm / s in irregularly shaped areas. The spraying angle θ is calculated by the inverse kinematics algorithm based on the DH parameter table of the robotic arm, where θ = 90° - α, and α is the angle between the normal of the point on the ceramic surface and the vertical direction. The spraying direction of the glaze is ensured to be perpendicular to the ceramic surface by the point cloud data normal vector calculation. The angle adjustment accuracy is ≤0.5°. The angle of each joint is controlled by the joint angle of the robotic arm, and the angle error of each joint is ≤0.05°.

[0113] In the environmental adaptive correction stage, real-time environmental data is acquired using the SickTH300 temperature and humidity sensor, with 25℃ and 60%RH as the baseline for correction: For every 5℃ increase in temperature, the spray speed v increases by 10%, e.g., from 10mm / s to 11mm / s. Simultaneously, the Q value needs to be recalculated according to the formula Q=k×d×v. For example, if the original Q=1.1mL / min, after the increase in v, Q=1.1×1.1=1.21mL / min, to accelerate the glaze curing speed. For every 5℃ decrease in temperature, v decreases by 8%, e.g., from 10mm / s to 9.2mm / s, and the Q value simultaneously decreases to 1.1×0.92≈1.01mL / min to prevent reduced glaze flowability and missed coating. For every 10% increase in humidity, the spray pressure increases by 0.05MPa, e.g., 0.3MPa. →0.35MPa, achieved through air pressure regulation via flow valve, overcomes the increase in glaze viscosity caused by humidity; for every 10% decrease in humidity, the pressure decreases by 0.04MPa, e.g., 0.3MPa→0.26MPa, to avoid excessive fluidity leading to sagging; after correction, a final trajectory scheme in XML format is generated, including the robotic arm position (x,y,z), movement speed v, spray angle θ, spray pressure P, and corresponding glaze flow rate Q at each 0.1s time node, ensuring that the scheme can be directly parsed by the glazing control module; during fault handling, if the path coverage is insufficient, check the accuracy of the target area coordinate marking; if the robotic arm movement exceeds the joint range after motion parameter calculation, adjust the starting point position; if sagging / missed coating still occurs after environmental correction, increase the correction coefficient and recalculate the Q value and trajectory.

[0114] Step 4, Glazing:

[0115] Glazing is the core step in translating the trajectory plan into actual operation, requiring precise equipment movement and real-time adjustments. In the equipment initialization and command conversion phase, the glazing equipment uses an ABB IBRB1200 6-axis robotic arm with a 0.5mm nozzle diameter glaze spray gun at the end, a spray range of 2-3mm, and a FestoMPPES flow valve with a flow rate adjustment range of 0-5mL / min. A communication connection between the robotic arm and the trajectory plan is established using ABB Robot Studio software. Based on the Profinet protocol, the XML-formatted trajectory plan is converted into joint angle commands recognizable by the robotic arm. Each time point corresponds to 6 joint angle values ​​with an accuracy of ±0.01°. Based on the parameters and three... Using the coordinate mapping table and the formula to calculate the Q value, set the initial flow rate of the flow valve, such as 1.1 mL / min, and set the initial pressure P through the air pressure regulating valve, such as 0.3 MPa. Start the pre-spray test: the robotic arm moves to the test area next to the platform and sprays for 1 to 2 seconds according to the initial parameters. Use a laser thickness gauge to detect the glaze thickness in the test area. If the thickness deviates from the target by more than ±0.01 mm, adjust the flow rate by ±0.1 mL / min each time, or the pressure by ±0.01 MPa each time, until the deviation is ≤ ±0.01 mm. For example, when Q=1.1 mL / min, the measured thickness is 0.09 mm, and the deviation is -0.01 mm. Q needs to be adjusted to 1.2 mL / min, and the measured thickness is 0.10 mm, which meets the standard.

[0116] In the real-time monitoring and dynamic adjustment phase, the robotic arm is activated to perform the glazing operation, while simultaneously turning on the Optronis CP80 high-speed camera and the SickTMT8 temperature sensor. The high-speed camera captures dynamic images of the glazing area. Using a U-Net-based semantic segmentation algorithm, a training dataset containing over 5000 glazing process images is used, labeled as glazed areas, unglazed areas, and excessively thick areas. Areas with unglazed areas exceeding 0.5mm in size are identified. 2If the coating is too thick, exceeding the target thickness by 10%, resulting in missed areas, a PID control algorithm is used with Kp=2.0, Ki=0.5, and Kd=0.1. The flow rate is increased by +0.1 mL / min each time, or the speed is decreased by -0.5 mm / s each time. After adjustment, the matching of the formula Q=k×d×v must be verified simultaneously. For example, when the speed decreases from 10 mm / s to 9.5 mm / s, Q needs to be adjusted from 1.1 mL / min to 1.1×(9.5 / 10)=1.045 mL / min to ensure glaze coverage per unit area. The material usage is stable; if the thickness is too great, reduce the flow rate by -0.1 mL / min each time, or increase the speed by +0.5 mm / s each time, and the Q value must be adjusted accordingly; the temperature sensor monitors the ceramic surface temperature. If the temperature is ≥3℃ higher than the ambient temperature, the speed needs to be increased by 5% or the flow rate reduced by 5%. At this time, the Q value is adjusted accordingly according to the formula. For example, if the speed is increased from 10 mm / s to 10.5 mm / s, Q is adjusted from 1.1 mL / min to 1.1 × 1.05 = 1.155 mL / min to prevent local overheating and uneven curing.

[0117] Meanwhile, the joint angle and motor speed are monitored by the built-in encoder of the robotic arm. When the joint angle deviation exceeds ±0.1° or the speed deviation exceeds ±1r / min, the servo control system is activated to correct it, ensuring motion accuracy and speed stability, and avoiding the imbalance of Q value and v due to speed fluctuations. In the glazing process recording stage, the NIUSB-6211 data acquisition card is used to record key data such as time, robotic arm position, real-time Q value, pressure, temperature, and image recognition results, and store them in CSV format. The data must include the Q value and v value before and after each parameter adjustment to facilitate the subsequent traceability of the accuracy of formula application.

[0118] In the interruption handling process, if there is insufficient glaze, a sudden drop in flow valve pressure >0.05MPa, equipment malfunction alarm, or robotic arm joint temperature exceeding 45℃, the interruption procedure is triggered: the robotic arm immediately stops moving and moves to a safe position, ≥100mm away from the ceramic, closes the flow valve, and records the current glazing progress and the last valid Q and v values; after the fault is cleared, execution resumes from the interruption position, calling the coordinates and parameters in the historical trajectory data, and recalculating the flow rate of the subsequent path based on the formula Q=k×d×v to avoid material waste caused by re-glazing; During troubleshooting, if the pre-spray test thickness deviation repeatedly exceeds the threshold, check for blockage in the flow valve nozzle, clean it with 0.3MPa compressed air, and calibrate it with the laser thickness gauge every two weeks. If the problem persists after real-time adjustment due to missed coating or excessive thickness, check for dirt on the high-speed camera lens, clean it with a lint-free cloth dampened with anhydrous alcohol, or retrain the image segmentation model and add 100 to 200 similar images of glazing abnormalities. If execution cannot resume from the interrupted point after interruption, check the integrity of the historical trajectory data storage path, re-import the trajectory scheme, and restart from the endpoint of the completed area.

[0119] Step 5, Quality Inspection:

[0120] Quality inspection needs to comprehensively evaluate the glazing effect, verify the application accuracy of the formula Q = k × d × v, and provide a quantitative basis for subsequent feedback adjustment; in the detection data acquisition link, after glazing, transfer the ceramic to the detection platform, which has the same accuracy as the glazing platform, and the levelness error ≤ 0.02 mm / m. Start the high-definition camera, which is consistent with the information acquisition module equipment, and the Keyence LK-G80 laser thickness gauge: The high-definition camera takes pictures of the glazed ceramic from 5 angles: the front, side, top, 45° diagonal, and 135° diagonal. Take 3 pictures at each angle to ensure that all glazing areas are covered; the laser thickness gauge sets detection points in a grid pattern. For regular ceramics, use a 5 mm × 5 mm grid, and for irregular ceramics, use a 3 mm × 3 mm grid. The defect area is encrypted to a 1 mm × 1 mm grid. Each detection point is measured 3 times and the average value is taken. Record the coordinates (x, y, z) of each detection point and the actual glazing thickness d_real, and at the same time associate the Q value and v value corresponding to this coordinate to verify the deviation between the calculated value of the formula and the actual effect.

[0121] In the quality index analysis link, for the coverage integrity analysis, use the OpenCV library of Python to register the glazed image and the target coverage area annotation map before glazing based on feature point matching. Through the gray difference method, the gray value of the glazed area is 50 to 100 higher than that of the unglazed area. Identify the uncovered area and calculate the missed coating area. The missed coating area ≤ 0.5 mm 2 is judged as qualified. For every exceedance of 0.5 mm 2 10 points will be deducted, with a full score of 40 points. The missed coating area needs to trace the corresponding Q value and v value. If the Q value is more than 10% lower than the calculated value of the formula, the k value needs to be corrected in subsequent production; for the thickness uniformity analysis, calculate the average thickness d_avg and standard deviation σ of all detection points. σ ≤ d_target × 5%, where d_target is the target thickness, and it is judged as uniform, with a full score of 35 points. When σ > d_target × 5%, 7 points will be deducted for every 1% exceedance. At the same time, calculate the calculated thickness d_calc = (Q × t) / (S × ρ), where t is the residence time of the spray gun in this area, S is the area of the area, and ρ is the density of the glaze, and the deviation from d_real. The deviation ≤ ±z0.01 mm is qualified. For the exceeded area, it is necessary to analyze whether the matching of the Q value and v value is reasonable; for the surface defect analysis, through the defect detection model based on YOLOv5, the training data set contains 500 defect images each of air bubbles, sagging, pinholes, etc. Identify the surface defects of the glazing layer. Air bubbles with a diameter > 0.5 mm, sagging with a length > 2 mm, and pinholes with a diameter > 0.3 mm are judged as unqualified defects. For each defect, 5 points will be deducted, with a full score of 25 points. For the sagging defect, it is necessary to trace whether it is caused by too high Q value or too low v value. For the air bubble defect, it is necessary to check whether the spraying pressure matches the Q value.

[0122] In the quality result generation stage, the scores of three indicators are comprehensively considered. Coverage integrity is 40 points, thickness uniformity is 35 points, and surface defects are 25 points. A total score of ≥80 is considered qualified, and <80 is unqualified. A quality report in PDF format is generated, including the total score, scores of each indicator, defect details, thickness distribution table, and deviation analysis of the formula calculated value and the actual value. The Q value, v value, and k value of the out-of-tolerance area need to be clearly marked in the report to provide a direction for subsequent parameter optimization. During fault handling, if the detection data collection is incomplete, check that the working distance of the laser thickness gauge is between 20 and 100 mm and its alignment with the camera detection area. If the defect detection accuracy rate <90%, add 100 images of each type of defect to retrain the model. If the thickness distribution heat map shows local thickness anomalies, trace back the real-time data of the glazing execution steps to check whether the Q value and v value are adjusted synchronously according to the formula.

[0123] Step 6, Intelligent Feedback:

[0124] Intelligent feedback is the key to realizing the closed-loop optimization of the glazing process and the reuse of formula parameters. In the qualified case storage stage, if the quality report determines that it is qualified, the complete data of this production will be classified and stored in the MySQL case database according to ceramic information, glazing parameters, environmental data, and quality results: Ceramic information includes type, size, surface morphology, and defect conditions; Glazing parameters include the target thickness d_target, coverage area coordinates, Q value calculated based on the formula and the corresponding k value, v value, equipment distance, and spraying pressure; Environmental data includes the temperature and humidity during glazing; Quality results include the total score, scores of each indicator, thickness standard deviation, and average deviation between the formula calculated value and the actual value. At the same time, a four-level index is established according to ceramic type, d_target, environmental temperature, and environmental humidity, such as household ceramics, 0.1 mm, 25 °C, 60% RH. The index uses a B+ tree structure to ensure that the retrieval time ≤1 s, the data storage capacity supports 100,000+ cases, and the calculation process of the formula parameters needs to be associated and stored for each case for subsequent reuse and reference.

[0125] In the unqualified parameter optimization stage, if the quality report determines that it is unqualified, start the similar case retrieval: First, input the retrieval keywords, including the current ceramic type, d_target, environmental temperature and humidity, and the main defect types, such as missed coating, over-thickness, and sagging; Then use the cosine similarity algorithm to calculate the similarity with the cases in the database. After standardizing the current parameters and the historical case parameters, calculate the cosine value of the vector angle. Cases with a similarity ≥85% are included in the recommended range; Finally, generate optimization suggestions, extract the parameter adjustment rules related to the current defect from the recommended cases. For example, in historical cases, missed coating defects were resolved by reducing v by 5% + increasing Q by 8%, and after adjustment, it satisfies Q = k × d × v. This time, the same adjustment direction is recommended. If the current missed coating area exceeds 1 mm 2The adjustment range can be fine-tuned to decrease v by 8% and increase Q by 10%, and the adjusted Q value can be verified to conform to the formula logic. For example, if v decreases from 10 mm / s to 9.2 mm / s, k=1.1, and d_mesh=0.1 mm, then Q should be adjusted to 1.1×0.1×9.2=1.012 mL / min, ensuring that the parameter adjustment is always based on the formula correlation.

[0126] Re-execute the manual warning step, update the optimized k, v, and Q values ​​to the parameter and 3D coordinate mapping table, regenerate the trajectory scheme, and return to step 4 to repeat the operation; if the same ceramic fails to meet the standard after 3 consecutive optimizations and the total score is <80 points, the system will trigger a manual warning: a warning window will pop up, showing 3 consecutive failures, and suggesting the following checks: 1. Is the glaze viscosity abnormal? The current viscosity is in mPa·s, and the standard range is 500 to 800 mPa·s. Abnormal viscosity will cause the v value to be mismatched, affecting the accuracy of the Q=k×d×v calculation; 2. Is the precision of the robotic arm joints up to standard? The current joint error is 1°, and the standard is ≤0.05°. Insufficient precision will cause the v value to fluctuate; 3. Is the laser thickness gauge up to standard? Calibration affects the actual measurement of d, leading to deviations in formula verification. A warning SMS is sent to the equipment administrator's mobile phone to prevent invalid duplicate production. During troubleshooting, if qualified case storage fails, check the database connection parameters: IP address, port number, username, password, and data format, ensuring that numerical data such as k, v, and Q are free of garbled characters. If the number of similar case search results is less than 3, lower the similarity threshold to 80% or manually enter historical experience parameters. The debugging parameters for similar ceramics provided by the equipment administrator must satisfy the formula Q=k×d×v. If the problem persists after a manual warning, contact the equipment manufacturer to calibrate the robotic arm or replace the glaze batch, and re-measure the glaze viscosity to adjust the matching relationship between the v and Q values.

[0127] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for automatic ceramic glazing control integrating machine vision and trajectory planning, characterized in that: The glazing control method includes the following steps: Obtaining the basic information and glazing requirements of the ceramic to be glazed. The basic information includes the ceramic's dimensions, surface shape, and surface condition. The glazing requirements include the target glazing thickness and target coverage area. The process of obtaining the basic information and glazing requirements of the ceramic to be glazed includes: The ceramic to be glazed is placed on a testing platform. The position of the ceramic on the platform is determined by a dual calibration using a positioning sensor that senses the position of the ceramic and a visual positioning system that determines the position through image analysis. Images of the ceramic are captured from multiple angles to obtain two-dimensional image data reflecting the planar shape of the ceramic and three-dimensional point cloud data reflecting the three-dimensional shape of the ceramic. The two-dimensional image data and the three-dimensional point cloud data are then fused. From the fused data, the external dimensions of the ceramic, including its height, diameter, length, and width, are extracted. The surface morphology of ceramics is analyzed using a 3D model, and the transition angle between adjacent surfaces and the radius of curvature reflecting the curvature of the surface are marked. The ceramic surface morphology can be flat, curved, arc-shaped, or irregular. Image grayscale analysis and 3D point cloud comparison are combined. The target thickness value and target coverage area are determined according to the glazing requirements, and the area is associated with the coordinates of the ceramic 3D model. Visual processing of ceramics is performed through image acquisition and analysis technology. This technology acquires ceramic images and extracts feature data and glazing parameters, while simultaneously establishing a three-dimensional model of the ceramic surface. Based on the ceramic's characteristic data, glazing parameters, and 3D model, and considering the impact of real-time environmental temperature and humidity on trajectory planning, trajectory planning is the process of setting the motion path and parameters of the glazing equipment, forming a dynamically adjustable glazing trajectory scheme. The process of designing the motion path and parameters of the glazing equipment based on the ceramic's characteristic data to form the glazing trajectory scheme includes: Based on the ceramic surface shape, the target glazing area, and the 3D model, A... The path planning algorithm plans the motion path of the glazing equipment; the flowability parameters of the glazing material are obtained, and the motion speed of the glazing equipment is determined by combining the shape of the ceramic surface, with high speed used in planar areas, medium speed in curved areas, and low speed in irregularly shaped areas; the inverse kinematics algorithm is used to determine the angle adjustment method of the glazing equipment to keep the glazing material perpendicular to the ceramic surface; based on the output of the glazing material and the motion speed of the glazing equipment, a flow rate and speed matching model reflecting the matching relationship between the material flow rate and the motion speed of the glazing equipment is established; the motion path, speed, and angle adjustment method are integrated in time sequence, and the trajectory scheme is corrected by combining the influence of real-time environmental temperature and humidity on the material curing speed. The glazing equipment is controlled to perform the glazing operation according to the glazing trajectory scheme. During the process, the temperature change data of the glazing area is collected in real time, and the material output and movement speed of the glazing equipment are dynamically adjusted according to the temperature change data. Visual inspection and thickness measurement are performed on the glazed ceramics to obtain glazing quality results. Determine whether the glazing quality meets the preset standard: if it does, store the glazing data in the database to form a case library; otherwise, collect the successful ceramic data in the case library, adjust the parameters and trajectory scheme, and re-execute the glazing operation until it meets the standard.

2. The automatic ceramic glazing control method integrating machine vision and trajectory planning according to claim 1, characterized in that, The process of visually processing ceramics using image acquisition and analysis technology to obtain feature data and glazing parameters includes: Preprocessing is performed on ceramic images and 3D point cloud data, including Gaussian filtering for noise reduction, histogram equalization to enhance contrast, and point cloud denoising algorithm optimization. From the preprocessed data, ceramic feature data including surface contour curve equation parameters, surface texture grayscale value distribution patterns, and 3D coordinates and volume of surface defects are extracted. Based on the ceramic surface shape and state, a deep learning algorithm is used to identify key glazing areas. Combining the target glazing thickness and the ceramic surface material characteristics, a digital correlation model between glazing material output and thickness is established. The model calculates the material output in different areas, and dynamically adjusts the distance parameters between the glazing device and the ceramic surface based on changes in ceramic surface curvature. The feature data, material output, and distance parameters are summarized to form a complete set of glazing parameters, and a mapping table between the parameters and the three-dimensional model is generated.

3. The automatic ceramic glazing control method integrating machine vision and trajectory planning according to claim 1, characterized in that, The process of controlling the glazing equipment to perform automatic glazing according to the glazing trajectory scheme includes: The motion path, speed, and angle adjustment methods in the glazing trajectory scheme are converted into control commands that the glazing equipment can recognize; the material flow rate and spray pressure of the glazing equipment are set through the flow control valve, and a pre-spray test is performed after setting; the glazing equipment is started to execute motion and glazing operations according to the control commands, and at the same time, the real-time monitoring device is turned on to acquire images of the glazing process and the operating data of the glazing equipment; Image segmentation algorithms are used to analyze real-time images to determine whether there are any omissions or excessive coating thickness. If so, the material flow rate and movement speed of the glazing process are adjusted in real time using a PID control algorithm based on the actual deviation from the target. Otherwise, no response is made. The operating data of the glazing equipment, including joint angles and motor speed, are compared with the preset data of the trajectory scheme. If there is a deviation, the motion parameters are adjusted.

4. The automatic ceramic glazing control method integrating machine vision and trajectory planning according to claim 1, characterized in that, The process of visually inspecting glazed ceramics to obtain quality results includes: The process involves capturing omnidirectional images of the glazed ceramic and simultaneously acquiring actual thickness data for different areas of the surface. By comparing images before and after glazing, the integrity of the glazed coverage area is analyzed, and the location and extent of any missed areas are identified. Laser thickness measurement data is compared with the target thickness to calculate whether the glaze thickness in each area meets the target requirements, and areas that are too thick or too thin are identified. A defect detection algorithm is then used to inspect the glazed surface to determine the presence of defects, including bubbles, runs, and pinholes, and the type, location, and size of these defects are recorded. Finally, the integrity, thickness, and surface defect status are summarized, and a fuzzy comprehensive evaluation method is used to score the overall glazing quality, resulting in a quality assessment that includes both the score and defect details.

5. An automatic ceramic glazing control system integrating machine vision and trajectory planning, characterized in that, The glazing control system includes: The information collection module collects basic information and glazing requirements information of the ceramic to be glazed. The basic information includes the outer dimensions, surface shape, and surface condition, while the glazing requirements information includes the target thickness and target coverage area. The visual analysis module processes ceramics using image acquisition and analysis technology to obtain feature data and glazing parameters, and simultaneously builds a three-dimensional model that displays the surface morphology of the ceramics. The trajectory planning module, based on ceramic data, glazing parameters and 3D model, and by acquiring real-time temperature and humidity data, designs the motion path and parameters of the glazing equipment to form a dynamically adjustable glazing trajectory scheme. The glazing control module controls the glazing equipment to perform automatic glazing according to the glazing trajectory scheme. The module integrates real-time acquisition of glazing process data and uses PID algorithm to adjust the parameters of the glazing equipment in real time. The quality inspection module performs visual inspection and thickness measurement on glazed ceramics. It obtains actual thickness data through a laser thickness measuring device and combines the visual inspection results to form a comprehensive quality result. The feedback module determines whether the quality results meet the preset standards. If they do, the data is stored in the case database according to the classification identifier. Otherwise, similar cases are retrieved from the case database, and the parameters and solutions are adjusted accordingly before restarting the glazing and testing process. The case database module stores the glazing parameters, trajectory schemes, and quality results of each successful glazing process.

6. The automatic ceramic glazing control system integrating machine vision and trajectory planning according to claim 5, characterized in that, The information collection module includes: The ceramic positioning unit places the ceramic to be glazed on a positioning platform and performs dual calibration through a positioning sensor that senses the position of the ceramic and a visual positioning system that determines the position through image analysis. The multi-dimensional data acquisition unit activates the industrial camera and structured light scanner to capture ceramic images from multiple angles and scan the surface, obtaining two-dimensional images reflecting the planar morphology of the ceramic and three-dimensional point cloud data reflecting the three-dimensional morphology of the ceramic. The data fusion processing unit merges two-dimensional images and three-dimensional point cloud data to extract ceramic shape and size data. The surface morphology analysis unit analyzes the surface morphology of ceramics through a three-dimensional model, and marks the transition angle between adjacent surfaces and the radius of curvature that reflects the degree of curvature of the surface. The surface defect quantification unit combines image grayscale analysis with 3D point cloud comparison to detect ceramic surface defects and quantify defect parameters. The requirements analysis and mapping unit determines the target parameters based on the glazing requirements document and associates the target coverage area with the 3D model coordinate system that determines the position of each point on the model.

7. The automatic ceramic glazing control system integrating machine vision and trajectory planning according to claim 5, characterized in that, The visual analytics module includes: The preprocessing unit preprocesses the ceramic image and 3D point cloud data by reducing useless interference points in the image, enhancing the difference in brightness and darkness of the image, and optimizing the 3D point cloud data. The feature extraction unit extracts ceramic features from the data, including surface contour parameters, texture distribution patterns, and three-dimensional defect parameters. The recognition unit, based on the shape and state of the ceramic surface, uses a deep learning algorithm to identify key glazing areas; The associated calculation unit, combined with the ceramic material characteristics obtained from the glazing target thickness, establishes a correlation model between material output and thickness. Through the model, the material output of the equipment in different areas is calculated, and the distance between the equipment and the ceramic surface is dynamically determined. The model mapping unit summarizes feature data, material output, and distance parameters to form a complete set of glazing parameters, and generates a mapping table between parameters and the 3D model, providing data association basis for trajectory planning.

8. The automatic ceramic glazing control system integrating machine vision and trajectory planning according to claim 5, characterized in that, The trajectory planning module includes: The path design unit, based on the ceramic surface shape, target coverage area, and 3D model, uses A... Algorithm planning of equipment movement path; The material property speed matching unit retrieves the flowability parameters of the glazing material from the database and determines the equipment movement speed according to the ceramic surface shape, wherein the planar area is high-speed, the curved area is medium-speed, and the irregular surface area is low-speed. The adjustment unit uses an inverse kinematics algorithm to adjust the angle of the glazing equipment so that the material is perpendicular to the ceramic surface; The flow rate and velocity dynamic matching unit establishes a flow rate and velocity matching model based on the material output and equipment movement speed, and adjusts the material usage per unit area to meet the target thickness. The trajectory correction unit integrates path, speed, and angle adjustment methods in chronological order, and corrects the trajectory scheme by taking into account the influence of ambient temperature and humidity on the material curing speed.

Images