Water jet type glaze spraying method and system

Abstract

The invention relates to the technical field of water jet type glaze spraying, in particular to a water jet type glaze spraying method and system.The water jet type glaze spraying method comprises the steps that image data of a workpiece on a conveying device is collected; preprocessing the acquired image data; performing feature extraction on the image data; identifying a workpiece in the image data; the position information of the workpiece in the image data is output, edge detection is carried out on the image, shape features of the workpiece in the image data are extracted, template matching is carried out on the shape features extracted from the image and a template image in a database, and then the workpiece in the collected image data is positioned; the water jet cutter can uniformly glaze the surface of a workpiece on the conveying device, so that uneven glazing of the surface of the workpiece caused by overlarge relative position deviation between the workpiece and the water jet cutter is avoided, the overall attractiveness of a product is improved, uniform thickness of a glaze layer on the surface of the workpiece is guaranteed, and the product performance is improved.

Description

Technical Field

[0001] The invention relates to the technical field of water-jet glaze spraying, and in particular to a water-jet glaze spraying method and system. Background Art

[0002] Spray glazing is one of the modern ceramic glazing techniques. It uses a spray gun or sprayer to atomize the glaze slurry and then spray it onto the surface of the workpiece. This method is particularly suitable for large vessels, complex shapes or thin-body products, because it can spray glaze multiple times to achieve multi-color glazing and thicker glaze layer effects;

[0003] Using a water jet glaze sprayer to glaze the workpiece surface can ensure that the glaze layer is uniform and smooth, and improve the anti-slip performance and aesthetics of the product. At the same time, due to the high efficiency and stability of the water jet glaze sprayer, it can greatly improve production efficiency and reduce production costs;

[0004] Due to the differences in the size and position of the workpieces transported by the conveying device to the water jet glaze spraying machine, the glaze sprayed on the surface of the workpiece will appear uneven when the water jet glaze spraying machine glazes the workpiece, so that the color of the workpiece surface varies, affecting the overall appearance of the product. Summary of the invention

[0005] The main purpose of the present invention is to provide a water jet glaze spraying method and system to solve the problems raised in the related art that the current calculation is mainly based on manual verification or through customer feedback, but the cost of manual verification is high, the result collection speed is slow, the large amount of data is easy to miss, and it is impossible to test in a specific environment for a long time, which limits the efficiency, accuracy and real-time performance of quality testing.

[0006] In order to achieve the above object, according to one aspect of the present invention, a water jet glazing method is provided, comprising:

[0007] S1: Collecting image data of the workpiece on the conveying device;

[0008] S2: preprocessing the collected image data;

[0009] S3: extract features from image data;

[0010] S4: identifying artifacts in the image data;

[0011] S5: Output the position information of the workpiece in the image data.

[0012] In a preferred embodiment, in S1, a database is established and known image templates are stored in the database, and a machine vision sensor is used to capture the workpiece transported by the conveying device, thereby acquiring image data of the workpiece on the conveying device.

[0013] In a preferred embodiment, in S2, the image is smoothed by a Gaussian filter to reduce image noise, and the edge contour of the workpiece in the image data is identified based on the Canny algorithm.

[0014] The kernel function expression of the Gaussian filter is:

[0015]

[0016] Where (x, y) represents the coordinates of the pixel in the image and σ represents the standard deviation of the Gaussian kernel.

[0017] In a preferred embodiment, the Canny algorithm calculates the gradient strength and direction of each pixel in the image data by using the Sobel operator, and the calculation expression of the gradient strength and direction is:

[0018] G z =Sobel z *I,

[0019] G y =Sobel y *I,

[0020]

[0021] Among them, I represents the input image, * represents the convolution operation, Sobel z and Sobel y are the Sobel operators in the x and y directions respectively.

[0022] In a preferred embodiment, in S3, the shape of the workpiece on the conveying device is fixed, so the shape feature of the workpiece in the image data is extracted, and the expression is:

[0023]

[0024] Among them, f(x, y) represents the image function, m pq Express the (p+q)-order moment.

[0025] In a preferred embodiment, in S4, based on a known template image, a template matching algorithm is used to identify and locate the position of the workpiece in the collected image data. The calculation expression of the template matching algorithm is:

[0026]

[0027] Wherein, T represents the known template image, I represents the acquired image data, (x, y) represents the position offset in the acquired image data, and They represent the pixel mean of the known image template and the acquired image template at the current position respectively. By moving the known template image pixel by pixel in the acquired image data and calculating the NCC value, the position with the largest NCC value is the most likely position of the object.

[0028] In a preferred embodiment, in S5, the position information of the workpiece in the collected image data is transmitted to the water jet control system, so that after the water jet control system obtains the position information of the workpiece, the control system drives the execution module to spray glaze on the surface of the workpiece on the conveying device, and at the same time, the machine vision sensor is used to collect images of the workpiece after spraying glaze again, and the spraying effect of the workpiece surface is evaluated by extracting features from the collected image data of the workpiece after spraying glaze and calculating the glossiness of the glazed surface of the workpiece after spraying glaze. The expression of glossiness is:

[0029]

[0030] Among them, G represents glossiness, R sample Indicates the reflectivity of the workpiece surface after glazing in the collected image data, R standard Represents the reflectivity of the workpiece surface in the known template image.

[0031] In a preferred embodiment, a neuron model is established based on the spraying effect of the workpiece surface after glazing, and then the defect detection of the workpiece after glazing is performed. The expression is:

[0032]

[0033] Among them, x i The input from the ith neuron, w i represents the corresponding weight, b represents the bias term, n represents the number of input neurons, and z represents the result of weighted summation;

[0034] The activation function f is applied to z to generate the neuron output a, which is expressed as:

[0035] a=f(z),

[0036]

[0037] Here, e represents the base of natural logarithms.

[0038] In a preferred embodiment, the difference between the output value z and the true label is quantified by a loss function in the neuron model, and the loss expression of the loss function for a single sample is:

[0039]

[0040] Among them, y represents the true label, is the predicted output of the neural network, m is the number of output nodes, and the gradient of the loss function is calculated by the back propagation algorithm and Use gradient descent to update weights and biases, the expression is:

[0041]

[0042] Among them, a represents the learning rate, which controls the step size of weight update.

[0043] A water jet glaze spraying method, comprising:

[0044] Building a module, establishing a database, storing known template images in the database, and collecting image data of the workpiece on the conveying device through a machine vision sensor;

[0045] The preprocessing module removes noise from the collected image data and smoothes the image based on a Gaussian filter;

[0046] An extraction module detects the edge of the workpiece in the denoised image data based on the Canny algorithm, and is used to identify the edge contour of the workpiece in the image data;

[0047] The recognition module performs template matching on the edge contour of the workpiece identified in the image data and the known template image in the database to locate the position of the workpiece in the image;

[0048] The output module outputs the position information of the workpiece in the image data to the water jet control system, and the actuator sprays glaze on the workpiece under the control of the control system;

[0049] The feedback module establishes a neuron model and trains the neural network through the image data in the database to evaluate the glazing effect on the surface of the workpiece after spraying and feed back to the control system, so as to improve the spraying effect of the workpiece next time.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. By performing edge detection on the collected image, the shape features of the workpiece in the image data are extracted, and the shape features extracted from the image are used to perform template matching with the template image in the database, and then the workpiece in the collected image data is positioned, so that the water jet can evenly glaze the surface of the workpiece on the conveying device, avoiding uneven glazing on the workpiece surface caused by excessive relative position deviation between the workpiece and the water jet, improving the overall beauty of the product, ensuring the uniform thickness of the glaze layer on the workpiece surface, and improving product performance;

[0052] 2. By storing the collected images in a database and training the neural network model through the database, the neural network can evaluate and provide feedback on the uniformity of the glaze layer on the surface of the workpiece after glazing, and then correct the next spraying trajectory of the water jet, further improving the glazing effect of the water jet on the workpiece and further improving the quality of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention is a schematic flow chart of a water jet glaze spraying method.

[0054] Figure 2 The present invention is a schematic diagram of the process of a water jet glaze spraying system. DETAILED DESCRIPTION

[0055] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0056] Example 1

[0057] See also Figure 1 As shown, the purpose of this embodiment is to provide a water jet glazing method, comprising:

[0058] S1: Collecting image data of the workpiece on the conveying device;

[0059] S2: preprocessing the collected image data;

[0060] S3: extract features from image data;

[0061] S4: identifying artifacts in the image data;

[0062] S5: Output the position information of the workpiece in the image data.

[0063] In S1, a database is established and known image templates are stored in the database, and a machine vision sensor is used to capture the workpiece transported by the conveying device, thereby collecting image data of the workpiece on the conveying device.

[0064] In S2, the image is smoothed by a Gaussian filter to reduce image noise, and the edge contour of the workpiece in the image data is identified based on the Canny algorithm.

[0065] The kernel function expression of the Gaussian filter is:

[0066]

[0067] Among them, (x, y) represents the coordinates of the pixel in the image, σ represents the standard deviation of the Gaussian kernel, and the Gaussian filter can retain the edge information of the image while reducing noise. The weight distribution is smooth without sudden truncation, thus avoiding obvious blurring effects at the edges, maintaining the details in the collected image data, and improving the effect of subsequent processing and algorithm efficiency.

[0068] The Canny algorithm calculates the gradient strength and direction of each pixel in the image data by using the Sobel operator. The calculation expression of the gradient strength and direction is:

[0069] G z =Sobel z *I,

[0070] G y =Sobel y *I,

[0071]

[0072] Among them, I represents the input image, * represents the convolution operation, Sobel z and Sobel y They are Sobel operators in the x and y directions, which quantitatively represent the strength of the edge in the image through the gradient intensity, making edge detection more objective and accurate. Calculating the gradient intensity can enhance the edge information in the image, making the edge more prominent visually, which is convenient for subsequent image analysis and processing;

[0073] For a pixel point (x, y) in the image, its horizontal gradient value Gx(x, y) can be calculated by weighted summing the convolution kernel with the pixel values ​​in the 3x3 neighborhood centered at (x, y), expressed as:

[0074] G x (x,y)=I(x+1,y-1)+2I(x+1,y)+I(x+1,y+1)-I(x-1,y-1)-2I(x-1,y)-I(x-1,y+1).

[0075] In S3, the shape of the workpiece on the conveying device is fixed, so the shape feature of the workpiece in the image data is extracted, and the expression is:

[0076]

[0077] Among them, f(x, y) represents the image function, m pq Express the (p+q)-order moment.

[0078] In S4, based on the known template image, the template matching algorithm is used to identify and locate the workpiece position in the collected image data. The calculation expression of the template matching algorithm is:

[0079]

[0080] Wherein, T represents the known template image, I represents the acquired image data, (x, y) represents the position offset in the acquired image data, and They represent the pixel mean of the known image template and the acquired image template at the current position respectively. By moving the known template image pixel by pixel in the acquired image data and calculating the NCC value, the position with the largest NCC value is the most likely position of the object. The template matching algorithm is robust to interference factors such as illumination changes and noise in the image, and can offset the impact, thereby more accurately identifying the position of the object.

[0081] In S5, the position information of the workpiece in the collected image data is transmitted to the water jet control system, so that after the water jet control system obtains the position information of the workpiece, the control system drives the execution module to spray glaze on the surface of the workpiece on the conveying device, and at the same time, the machine vision sensor is used to collect images of the workpiece after spraying glaze again. By extracting features from the collected image data of the workpiece after spraying glaze, and calculating the glossiness of the glazed surface of the workpiece after spraying glaze, the spraying effect of the workpiece surface is evaluated. The expression of glossiness is:

[0082]

[0083] Among them, G represents glossiness, R sample Indicates the reflectivity of the workpiece surface after glazing in the collected image data, R standard It represents the reflectivity of the workpiece surface in the known template image. By transmitting the workpiece position information to the water jet control system, manual intervention is reduced, the degree of automation of the production line is improved, thereby accelerating the production speed and improving the overall production efficiency. By accurately obtaining the position information of the workpiece, the water jet control system can accurately drive the execution module to spray glaze on the workpiece, avoiding misalignment and waste in the glazing process, and improving the accuracy and quality of glazing. By calculating the glossiness of the glazed surface of the workpiece after glazing and comparing it with the reflectivity of the workpiece surface in the known template image, the spraying effect of the workpiece surface is objectively evaluated. Through real-time evaluation and timely adjustment of the spraying effect, it is ensured that each workpiece reaches the expected spraying quality and improves the overall quality of the product.

[0084] Based on the spraying effect of the workpiece surface after glazing, a neural model is established to detect defects on the workpiece after glazing. The expression is:

[0085]

[0086] Among them, x i The input from the ith neuron, w i represents the corresponding weight, b represents the bias term, n represents the number of input neurons, and z represents the result of weighted summation;

[0087] The activation function f is applied to z to generate the neuron output a, which is expressed as:

[0088] a=f(z),

[0089]

[0090] Here, e represents the base of natural logarithms.

[0091] In the neuron model, the difference between the output value z and the true label is quantified by the loss function. The loss expression of the loss function for a single sample is:

[0092]

[0093] Among them, y represents the true label, is the predicted output of the neural network, m is the number of output nodes, and the gradient of the loss function is calculated by the back propagation algorithm and Use gradient descent to update weights and biases, the expression is:

[0094]

[0095] Among them, a represents the learning rate, which controls the step size of weight update;

[0096] Through learning and training, the neural model can capture subtle feature changes on the surface of the workpiece after glazing. Through high-precision detection capabilities, it can promptly detect and eliminate unqualified products, improve product quality, avoid the influence of personal subjective factors on the accuracy and consistency of the test results during manual inspection, and further ensure the quality of the product.

[0097] Example 2

[0098] See also Figure 2 As shown, the part not described in detail in this embodiment can be seen from the description of embodiment 1, and a water jet glazing system for a water jet glazing method is provided, comprising:

[0099] Building a module, establishing a database, storing known template images in the database, and collecting image data of the workpiece on the conveying device through a machine vision sensor;

[0100] The preprocessing module removes noise from the collected image data and smoothes the image based on a Gaussian filter;

[0101] An extraction module detects the edge of the workpiece in the denoised image data based on the Canny algorithm, and is used to identify the edge contour of the workpiece in the image data;

[0102] The recognition module performs template matching on the edge contour of the workpiece identified in the image data and the known template image in the database to locate the position of the workpiece in the image;

[0103] The output module outputs the position information of the workpiece in the image data to the water jet control system, and the actuator sprays glaze on the workpiece under the control of the control system;

[0104] The feedback module establishes a neuron model and trains the neural network through the image data in the database to evaluate the glazing effect on the surface of the workpiece after spraying and feed back to the control system, so as to improve the spraying effect of the workpiece next time.

[0105] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A water jet glazing method, characterized in that: include: S1: Collecting image data of the workpiece on the conveying device; S2: preprocessing the collected image data; S3: extract features from image data; S4: identifying artifacts in the image data; S5: Output the position information of the workpiece in the image data.

2. A water jet glazing method according to claim 1, characterized in that: In S1, a database is established and known image templates are stored in the database, and a machine vision sensor is used to capture the workpiece transported by the conveying device, thereby collecting image data of the workpiece on the conveying device.

3. A water jet glazing method according to claim 1, characterized in that: In S2, the image is smoothed by a Gaussian filter to reduce image noise, and the edge contour of the workpiece in the image data is identified based on the Canny algorithm. The kernel function expression of the Gaussian filter is: Where (x, y) represents the coordinates of the pixel in the image and σ represents the standard deviation of the Gaussian kernel.

4. A water jet glazing method according to claim 3, characterized in that: The Canny algorithm calculates the gradient strength and direction of each pixel in the image data by using the Sobel operator. The calculation expressions of the gradient strength and direction are: G z =Sobwl z *I, G y =Sobel y *I, Among them, I represents the input image, * represents the convolution operation, Sobel z and Sobel y are the Sobel operators in the x and y directions respectively.

5. The water jet glazing method according to claim 1, characterized in that: In S3, the shape of the workpiece on the conveying device is fixed, so the shape feature of the workpiece in the image data is extracted, and the expression is: Among them, f(x, y) represents the image function, m pq Express the (p+q)-order moment.

6. A water jet glazing method according to claim 2, characterized in that: In S4, based on the known template image, the template matching algorithm is used to identify and locate the position of the workpiece in the collected image data. The calculation expression of the template matching algorithm is: Wherein, T represents the known template image, I represents the acquired image data, (x, y) represents the position offset in the acquired image data, and They represent the pixel mean of the known image template and the acquired image template at the current position respectively. By moving the known template image pixel by pixel in the acquired image data and calculating the NCC value, the position with the largest NCC value is the most likely position of the object.

7. The water jet glazing method according to claim 1, characterized in that: In S5, the position information of the workpiece in the collected image data is transmitted to the water jet control system, so that after the water jet control system obtains the position information of the workpiece, the control system drives the execution module to spray glaze on the surface of the workpiece on the conveying device, and at the same time, the machine vision sensor is used to collect images of the workpiece after spraying glaze again, and the spraying effect of the workpiece surface is evaluated by extracting features from the collected image data of the workpiece after spraying glaze and calculating the glossiness of the glazed surface of the workpiece after spraying glaze. The expression of glossiness is: Among them, G represents glossiness, R sample Indicates the reflectivity of the workpiece surface after glazing in the collected image data, R standard Represents the reflectivity of the workpiece surface in the known template image.

8. A water jet glazing method according to claim 7, characterized in that: Based on the spraying effect of the workpiece surface after glazing, a neuron model is established to detect defects of the workpiece after glazing. The expression is: Among them, x i The input from the ith neuron, w i represents the corresponding weight, b represents the bias term, n represents the number of input neurons, and z represents the result of weighted summation; The activation function f is applied to z to generate the neuron output a, which is expressed as: a=f(z), Here, e represents the base of natural logarithms.

9. A water jet glazing method according to claim 8, characterized in that: In the neuron model, the difference between the output value z and the true label is quantified by the loss function. The loss expression of the loss function for a single sample is: Among them, y represents the true label, is the predicted output of the neural network, m is the number of output nodes, and the gradient of the loss function is calculated by the back propagation algorithm and Use gradient descent to update weights and biases, the expression is: Among them, a represents the learning rate, which controls the step size of weight update.

10. A water jet glazing system, based on a water jet glazing method according to any one of claims 1 to 9, characterized in that: include: Building a module, establishing a database, storing known template images in the database, and collecting image data of the workpiece on the conveying device through a machine vision sensor; The preprocessing module removes noise from the collected image data and smoothes the image based on a Gaussian filter; An extraction module detects the edge of the workpiece in the denoised image data based on the Canny algorithm, and is used to identify the edge contour of the workpiece in the image data; The recognition module performs template matching on the edge contour of the workpiece identified in the image data and the known template image in the database to locate the position of the workpiece in the image; The output module outputs the position information of the workpiece in the image data to the water jet control system, and the actuator sprays glaze on the workpiece under the control of the control system; The feedback module establishes a neuron model and trains the neural network through the image data in the database to evaluate the glazing effect on the surface of the workpiece after spraying and feed back to the control system, so as to improve the spraying effect of the workpiece next time.

Images