Real-time automatic acquisition, identification and adjustment method and system for secondary inflation of PVC (polyvinyl chloride) film
By collecting and processing film images in real time and automatically adjusting the width of the PVC film, the film quality instability caused by manual visual inspection in traditional production methods is solved, efficient and accurate film production is achieved, and production efficiency and product quality are improved.
Patent Information
- Application Number
- CN202411308291.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-19
- Publication Date
- 2024-11-15
AI Technical Summary
In the traditional PVC film production method, the thickness and width adjustment of the film depends on manual visual inspection, lacking real-time and high precision, resulting in unstable film quality and affecting subsequent slitting and printing quality.
Using PVC film secondary blowing real-time automatic acquisition, identification and adjustment method, by obtaining real-time film images, filtering, color difference balance and edge feature extraction, input the trained width adjustment decision model, realize automatic adjustment of the clamp lifting device and real-time optimization Film width.
It realizes rapid and precise adjustment of film width, reduces manual intervention, improves production efficiency and automation level, ensures consistency of film specifications, reduces product abandonment caused by unqualified width, and improves the overall product quality.
Smart Images

Figure CN118952628A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of film production, and in particular to a method and system for real-time automatic collection, identification and adjustment of secondary inflation of a PVC film. Background Art
[0002] In the production process of PVC film, the quality of the film directly affects its application effect. In the traditional film production method, the adjustment of film thickness and width mainly depends on manual experience, that is, the existing method is to adjust by manual visual inspection, which has the following disadvantages:
[0003] (1) It can only be adjusted periodically and is not real-time;
[0004] (2) Manual visual inspection has large deviations, which increases labor costs and makes it difficult to achieve high-precision control, resulting in unstable film quality. Summary of the invention
[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a real-time automatic collection, identification and adjustment method and system for secondary inflation of PVC film, which solves the problem that the secondary inflation of PVC changes according to manual visual inspection, the manual adjustment is cumbersome, the straightness changes frequently in the production process, the manual adjustment workload is large, and it is impossible to achieve real-time monitoring, the film straightness is poor, and the subsequent slitting and printing quality are seriously affected.
[0006] To achieve the above object, the present invention provides the following solutions:
[0007] A real-time automatic collection, identification and adjustment method for secondary inflation of PVC film, comprising:
[0008] Get real-time production film images;
[0009] Performing filtering processing on the film image to obtain a filtered image;
[0010] Performing color difference balance on the filtered image to obtain a color difference balanced image;
[0011] Extracting edge features according to the color difference balance image to obtain width features;
[0012] Inputting the width feature into a trained width adjustment decision model to obtain an adjustment decision;
[0013] The clamping plate lifting device is adjusted according to the adjustment decision to adjust the width of the film.
[0014] Preferably, filtering the film image to obtain a filtered image comprises:
[0015] Performing wavelet decomposition on the film image to obtain a plurality of wavelet coefficients;
[0016] constructing an adaptive wavelet threshold according to the number of decomposition layers of the film image;
[0017] Constructing a wavelet coefficient denoising model according to the adaptive wavelet threshold;
[0018] The film image is filtered using the wavelet coefficient denoising model to obtain the filtered image.
[0019] Preferably, constructing an adaptive wavelet threshold according to the number of decomposition layers of the film image includes:
[0020] The noise standard deviation is calculated using the wavelet coefficients at each decomposition level;
[0021] The adaptive wavelet threshold is constructed using the noise standard deviation; wherein the adaptive wavelet threshold is:
[0022]
[0023] Among them, λ j represents the wavelet threshold at the jth decomposition level, σ j represents the standard deviation of noise at the jth decomposition level, N j represents the signal length at the jth decomposition level, Represents the coefficient of the i-th wavelet at the j-th decomposition level.
[0024] Preferably, the formula of the wavelet coefficient denoising model is:
[0025]
[0026] Among them, sign is the sign function, a is the adjustment parameter, Represents the wavelet coefficients after denoising.
[0027] Preferably, performing color difference balancing on the filtered image to obtain a color difference balanced image comprises:
[0028] Calculate the pixel mean of the filtered image in the three RGB channels; wherein the pixel mean calculation formula is:
[0029]
[0030] Where D represents the pixel mean, represents the pixel mean of the filtered image in the red channel, represents the pixel mean of the filtered image in the green channel, Represents the pixel mean of the filtered image in the blue channel;
[0031] Calculating a balance coefficient for each color channel according to the pixel mean;
[0032] The color difference of the filtered image is balanced by using the balance coefficient to obtain the color difference balanced image.
[0033] Preferably, performing color difference balance on the filtered image by using the balance coefficient to obtain the color difference balanced image comprises:
[0034] Using the formula:
[0035]
[0036] Perform color difference balance on the filtered image to obtain the color difference balanced image; wherein K R Indicates the balance coefficient of the red channel, K G Indicates the balance coefficient of the green channel, K B Represents the balance coefficient of the blue channel, R′ represents the pixel value of the red channel after balance, G′ represents the pixel value of the green channel after balance, B′ represents the pixel value of the blue channel after balance, R represents the pixel value of the filtered image in the red channel, G represents the pixel value of the filtered image in the green channel, and B represents the pixel value of the filtered image in the blue channel.
[0037] Preferably, edge feature extraction is performed according to the color difference balance image to obtain width features, including:
[0038] Performing grayscale processing on the color difference balance image to obtain a grayscale image;
[0039] Based on the Canny edge detection algorithm, edge detection is performed on the grayscale image to extract an edge image;
[0040] Based on the Hough transform method, the spatial points in the edge image are converted into a parameter space to obtain a boundary line;
[0041] Screening is performed among the boundary straight lines to obtain two optimal straight lines;
[0042] The width feature is calculated according to the optimal straight line; the calculation formula of the width feature is: Among them, ω is a coefficient adaptively determined according to the edge strength during the edge detection process, A and B are both directional coefficients of the optimal straight line, C1 and C2 are respectively the intercept parameters of the two optimal straight lines; W is the width feature.
[0043] Preferably, the width adjustment decision model comprises: an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer connected in sequence.
[0044] Preferably, the formula of the loss function of the width adjustment decision model is:
[0045]
[0046] in, for overall loss; is the film width predicted by the model, the estimated value calculated by the model; y is the actual film width; Δa is the rate of change of the adjustment action; is the mean square error, |Δa| is the absolute action change; Var(Δa) is the variance of the action change; α, β and γ are all hyperparameters; n is the sample size.
[0047] A real-time automatic collection, identification and adjustment system for secondary inflation of PVC film, comprising:
[0048] An image acquisition module, used to acquire real-time produced film images;
[0049] A filtering module, used for filtering the film image to obtain a filtered image;
[0050] A balancing module, used for performing color difference balance on the filtered image to obtain a color difference balanced image;
[0051] A feature extraction module, used for extracting edge features according to the color difference balance image to obtain width features;
[0052] A decision module, used for inputting the width feature into a trained width adjustment decision model to obtain an adjustment decision;
[0053] The adjustment module is used to adjust the clamping plate lifting device according to the adjustment decision to adjust the width of the film.
[0054] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0055] The present invention provides a method and system for real-time automatic acquisition, identification and adjustment of secondary inflation of PVC film, the method comprising: acquiring a real-time produced film image; filtering the film image to obtain a filtered image; performing color balance on the filtered image to obtain a color balance image; extracting edge features based on the color balance image to obtain width features; inputting the width features into a trained width adjustment decision model to obtain an adjustment decision; adjusting the clamp lifting device based on the adjustment decision to adjust the width of the film. The present invention realizes rapid adjustment of the film width by real-time acquisition and processing of film images, reduces the need for manual intervention, thereby improving production efficiency and automation level, and can effectively ensure the consistency of film specifications through precise width feature extraction and adjustment, reduce product waste due to unqualified width, and improve the overall quality of the product. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0057] Figure 1 A flow chart of a method provided by an embodiment of the present invention;
[0058] Figure 2 A system structure diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION
[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0060] The purpose of the present invention is to provide a real-time automatic collection, identification and adjustment method and system for secondary inflation of PVC film. By introducing real-time automation technology, the intelligence level and production efficiency of PVC film production are improved, effectively meeting the needs of modern industry for high-quality and low-cost production.
[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0062] Figure 1A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a real-time automatic collection, identification and adjustment method for secondary inflation of a PVC film, comprising:
[0063] Step 100: Acquire a real-time produced film image;
[0064] Step 200: filtering the film image to obtain a filtered image;
[0065] Step 300: performing color difference balance on the filtered image to obtain a color difference balanced image;
[0066] Step 400: extracting edge features according to the color difference balance image to obtain width features;
[0067] Step 500: inputting the width feature into a trained width adjustment decision model to obtain an adjustment decision;
[0068] Step 600: adjusting the clamping plate lifting device according to the adjustment decision to adjust the width of the film.
[0069] Optionally, in this embodiment, the acquisition of real-time production film images is mainly achieved by a high-resolution industrial camera installed on the production line. The camera captures the image of the film in the production process in real time through high-speed shutter and appropriate light source (such as LED light). The camera is connected to the computer system via a USB or Ethernet interface, and the image data is transmitted and processed in real time through dedicated software. The software will periodically obtain image frames from the camera and store them in memory for subsequent filtering, color difference balance, and edge feature extraction. In this way, the system can ensure that the status of the film is continuously monitored during the production process and high-quality image data is obtained in a timely manner.
[0070] Furthermore, the video and photo taking device of this embodiment is usually installed in the film forming area of the production line, and the specific location should be selected after the film is inflated and cooled and before entering the cutting or winding stage. The selection of this location is based on several key factors: first, to ensure that the camera can clearly capture the full view of the film to avoid image distortion caused by viewing angle limitations; second, considering the spatial layout of the production line, the device should be installed in a place that does not hinder the production process to reduce interference with operations; finally, the installation height of the device should match the operating height of the film to ensure the best shooting angle and lighting conditions. Through these considerations, this embodiment can effectively improve the quality and accuracy of image acquisition.
[0071] Preferably, filtering the film image to obtain a filtered image comprises:
[0072] Performing wavelet decomposition on the film image to obtain a plurality of wavelet coefficients;
[0073] constructing an adaptive wavelet threshold according to the number of decomposition layers of the film image;
[0074] Constructing a wavelet coefficient denoising model according to the adaptive wavelet threshold;
[0075] The film image is filtered using the wavelet coefficient denoising model to obtain the filtered image.
[0076] Preferably, constructing an adaptive wavelet threshold according to the number of decomposition layers of the film image includes:
[0077] The noise standard deviation is calculated using the wavelet coefficients at each decomposition level;
[0078] The adaptive wavelet threshold is constructed using the noise standard deviation; wherein the adaptive wavelet threshold is:
[0079]
[0080] Among them, λ j represents the wavelet threshold at the jth decomposition level, σ j represents the standard deviation of noise at the jth decomposition level, N j represents the signal length at the jth decomposition level, Represents the coefficient of the i-th wavelet at the j-th decomposition level.
[0081] Preferably, the formula of the wavelet coefficient denoising model is:
[0082]
[0083] Among them, sign is the sign function, a is the adjustment parameter, Represents the wavelet coefficients after denoising.
[0084] Specifically, this embodiment uses a wavelet coefficient denoising model to perform adaptive threshold transformation on the wavelet coefficients at each scale, which can highlight the detailed features of the image at different decomposition levels, suppress the noise of the image, and enhance the layering of the image.
[0085] Preferably, performing color difference balancing on the filtered image to obtain a color difference balanced image comprises:
[0086] Calculate the pixel mean of the filtered image in the three RGB channels; wherein the pixel mean calculation formula is:
[0087]
[0088] Where D represents the pixel mean, represents the pixel mean of the filtered image in the red channel, represents the pixel mean of the filtered image in the green channel, Represents the pixel mean of the filtered image in the blue channel;
[0089] Calculating a balance coefficient for each color channel according to the pixel mean;
[0090] The color difference of the filtered image is balanced by using the balance coefficient to obtain the color difference balanced image.
[0091] Preferably, performing color difference balance on the filtered image by using the balance coefficient to obtain the color difference balanced image comprises:
[0092] Using the formula:
[0093]
[0094] Perform color difference balance on the filtered image to obtain the color difference balanced image; wherein K R Indicates the balance coefficient of the red channel, K G Indicates the balance coefficient of the green channel, K B Represents the balance coefficient of the blue channel, R′ represents the pixel value of the red channel after balance, G′ represents the pixel value of the green channel after balance, B′ represents the pixel value of the blue channel after balance, R represents the pixel value of the filtered image in the red channel, G represents the pixel value of the filtered image in the green channel, and B represents the pixel value of the filtered image in the blue channel.
[0095] Specifically, this embodiment takes into account that when the external lighting conditions change, the surface color of the object remains unchanged, but the color of the object captured by the imaging device will deviate with the change of lighting and gradually deviate from the true color of the object. The present invention uses a color difference balance algorithm to compensate for this defect of the imaging device and correct the color difference of the image.
[0096] Preferably, edge feature extraction is performed according to the color difference balance image to obtain width features, including:
[0097] Performing grayscale processing on the color difference balance image to obtain a grayscale image;
[0098] Based on the Canny edge detection algorithm, edge detection is performed on the grayscale image to extract an edge image;
[0099] Based on the Hough transform method, the spatial points in the edge image are converted into a parameter space to obtain a boundary line;
[0100] Screening is performed among the boundary straight lines to obtain two optimal straight lines;
[0101] The width feature is calculated according to the optimal straight line; the calculation formula of the width feature is: Among them, ω is a coefficient adaptively determined according to the edge strength during the edge detection process, A and B are both directional coefficients of the optimal straight line, C1 and C2 are respectively the intercept parameters of the two optimal straight lines; W is the width feature.
[0102] Specifically, the step of edge feature extraction in this embodiment includes:
[0103] Convert the color-balanced image to grayscale to reduce the data dimension and focus on intensity changes;
[0104] Apply edge detection algorithms, such as the Canny edge detection algorithm, to identify significant edges in the image. The Canny algorithm steps include: Gaussian filtering to reduce noise, calculating gradient magnitude and direction, applying non-maximum suppression to refine edges, and applying double thresholds to connect strong and weak edges.
[0105] Use techniques such as Hough transform or contour detection to identify the boundary line of the film. Hough transform converts points in image space to parameter space and detects straight lines.
[0106] The distance between the boundary lines perpendicular to the image or a given reference axis is calculated as the width feature of the film.
[0107] Furthermore, the Hough transform of this embodiment detects points that meet the straight line characteristics by converting points in the image space into the parameter space. Among the multiple fitted straight lines, two straight lines that best represent the film boundary are selected. This embodiment is achieved by analyzing the relative position, direction and length of the straight lines, as follows:
[0108] Select two straight lines that are relatively parallel and far apart in the image.
[0109] Make sure the orientation of the two lines matches the expected boundary orientation of the film.
[0110] It is preferred to choose straight lines with longer length and better continuity to improve stability.
[0111] Through these steps, the present embodiment can accurately determine the two boundary lines of the film and use these boundary lines to calculate the width characteristics of the film.
[0112] Furthermore, in the edge detection process, the coefficients adaptively determined according to the edge strength are usually implemented by analyzing the gradient amplitude of the edge in the image. Specifically, an edge detection algorithm (such as the Canny algorithm) is first applied to the filtered image to calculate the gradient amplitude of each pixel. Then, based on these gradient amplitudes, a dynamic threshold is set, which can be adjusted according to the overall contrast and edge strength distribution of the image. By statistically analyzing the distribution of edge strength in the image, a suitable percentile (for example, the 90% percentile) is selected as a reference value for edge strength, thereby generating an adaptive coefficient ω. This coefficient of the present embodiment is used in the width calculation formula to weight edge features, ensuring that the model can pay more attention to areas with higher edge strength, thereby improving the accuracy and robustness of width feature extraction.
[0113] Preferably, the width adjustment decision model comprises: an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer connected in sequence.
[0114] Preferably, the formula of the loss function of the width adjustment decision model is:
[0115]
[0116] in, for overall loss; is the film width predicted by the model, the estimated value calculated by the model; y is the actual film width; Δa is the rate of change of the adjustment action; is the mean square error, |Δa| is the absolute action change; Var(Δa) is the variance of the action change; α, β and γ are all hyperparameters; n is the sample size.
[0117] Specifically, the parameters of the loss function of the width adjustment decision model of this embodiment are explained as follows:
[0118] : Overall loss, combining prediction error, adjustment details, and stability considerations.
[0119] : The film width predicted by the model, the estimated value calculated by the model.
[0120] y: Actual film width, that is, the target width value.
[0121] Δa: The rate of change of the adjustment action, that is, the difference between the current and last adjustment. For example, if the amplitude of the last lifting and lowering operation of the splint is a t , this time is a t+1 , then Δa=a t+1 -a t .
[0122] : Mean square error, used to measure the gap between the predicted width and the actual width, is defined as:
[0123]
[0124] Here, n is the number of samples.
[0125] |Δa|: Absolute action change, penalizing large changes to encourage smooth adjustments and avoid system oscillations.
[0126] Var(Δa): Variance of action changes, which suppresses overly frequent adjustments and improves the control stability and response smoothness of the system.
[0127] α, β, γ: These are hyperparameters that control the weights of prediction error, adjustment amplitude, and adjustment stability, respectively, and can be adjusted according to the requirements of the actual application.
[0128] Optionally, in the Beijing South embodiment, the adjustment decision refers to generating specific operation instructions for the clamping plate lifting device based on the film width characteristics obtained in real time and the trained width adjustment decision model. This process includes:
[0129] The width feature data obtained from the edge feature extraction stage is input into the trained decision model.
[0130] The decision model is based on the input width features, combined with historical data and learned patterns, to reason and calculate the required adjustment magnitude and direction.
[0131] Based on the output of the model, specific adjustment instructions are generated, such as "the clamp rises by 5 mm" or "the clamp drops by 3 mm" to achieve precise control of the film width.
[0132] After executing the adjustment instruction, the width change of the film continues to be monitored, and the new width characteristics are fed back to the decision model for real-time adjustment and optimization to form a closed-loop control.
[0133] Through this series of steps, adjustment decisions ensure real-time optimization of film width, improving the automation and accuracy of the production process.
[0134] Furthermore, after executing the adjustment instruction, the present embodiment continuously captures real-time images of the film through a high-resolution industrial camera installed on the production line. The captured images are immediately processed by the computer system, and new width feature data is extracted again. These data are sent back to the width adjustment decision model, and the model re-evaluates the current film state in real time based on the new input, and evaluates the previous adjustment effect. If the deviation still exists, the model will generate a new adjustment instruction to correct any deviation. At the same time, the present embodiment uses a PID (proportional-integral-differential) control algorithm, combined with feedback width data, to continuously optimize control parameters and achieve more accurate adjustments. This uninterrupted monitoring and instant feedback form a closed-loop control system, which enables the production line to quickly respond to any abnormal fluctuations in width, keep the film width within the desired tolerance range, and optimize the adjustment strategy to reduce material waste and production stagnation. Through this mechanism, the present embodiment can gradually improve the control accuracy in each iteration, and adjust the model to adapt to dynamically changing production conditions, so that the production process remains efficient and stable.
[0135] In addition, the present embodiment is also designed with an integrated human-machine interface, which realizes efficient collaboration between the automation system and the operator through the integrated human-machine interface (HMI). The HMI interface is usually installed on the control console of the production line, providing an intuitive graphical display, and presenting key information such as film width, adjustment status, historical data trends, etc. in real time. The operator can interact with the system through a touch screen or a control panel to view real-time data and adjustment plans recommended by the system. When an abnormality is detected or manual intervention is required, an alarm or prompt will be issued through the interface, and the operator can make a decision quickly based on the prompt. In addition, the HMI interface allows the operator to manually enter adjustment parameters or select preset adjustment plans to cope with special production needs or emergencies. The system also provides a log recording function to help operators review and analyze adjustment history to optimize future operating strategies.
[0136] Through this human-machine collaboration mechanism, this embodiment not only improves the level of automation, but also retains the flexibility of manual intervention, ensuring that in complex or emergency situations, operators can quickly intervene and make effective adjustments, thereby improving the overall efficiency and reliability of the production line.
[0137] Corresponding to the above method, such as Figure 2 As shown, this embodiment also provides a real-time automatic acquisition, identification and adjustment system for secondary inflation of PVC film, including:
[0138] An image acquisition module, used to acquire real-time produced film images;
[0139] A filtering module, used for filtering the film image to obtain a filtered image;
[0140] A balancing module, used for performing color difference balance on the filtered image to obtain a color difference balanced image;
[0141] A feature extraction module, used for extracting edge features according to the color difference balance image to obtain width features;
[0142] A decision module, used for inputting the width feature into a trained width adjustment decision model to obtain an adjustment decision;
[0143] The adjustment module is used to adjust the clamping plate lifting device according to the adjustment decision to adjust the width of the film.
[0144] The beneficial effects of the present invention are as follows:
[0145] (1) The present invention realizes rapid adjustment of film width by real-time acquisition and processing of film images, thereby reducing the need for manual intervention and improving production efficiency and automation level.
[0146] (2) The present invention can effectively ensure the specification consistency of the film through precise width feature extraction and adjustment, reduce product waste due to unqualified width, and improve the overall quality of the product.
[0147] (3) The present invention can monitor the production process in real time, identify and respond to changes in a timely manner, ensure the stability of the production process, and thus reduce the impact caused by deviations.
[0148] (4) The present invention utilizes a trained width adjustment decision model, and the system can make intelligent decisions based on real-time data, thereby improving the accuracy and reliability of the adjustment and reducing the risks in the adjustment process.
[0149] (5) The present invention can effectively avoid the waste of raw materials, reduce production costs and environmental impact, and promote sustainable development through precise width control.
[0150] (6) The design of the present invention can flexibly cope with PVC films of different thicknesses, colors and types, thereby improving its universality and application range.
[0151] (7) Although the present invention achieves a high degree of automation, it can still collaborate efficiently with operators and provide a user-friendly adjustment interface when necessary through real-time data feedback.
[0152] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0153] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A real-time automatic collection, identification and adjustment method for secondary inflation of PVC film, characterized in that: include: Get real-time production film images; Performing filtering processing on the film image to obtain a filtered image; Performing color difference balance on the filtered image to obtain a color difference balanced image; Extracting edge features according to the color difference balance image to obtain width features; Inputting the width feature into a trained width adjustment decision model to obtain an adjustment decision; The clamping plate lifting device is adjusted according to the adjustment decision to adjust the width of the film.
2. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 1 is characterized in that: Performing filtering processing on the film image to obtain a filtered image includes: Performing wavelet decomposition on the film image to obtain a plurality of wavelet coefficients; constructing an adaptive wavelet threshold according to the number of decomposition layers of the film image; Constructing a wavelet coefficient denoising model according to the adaptive wavelet threshold; The film image is filtered using the wavelet coefficient denoising model to obtain the filtered image.
3. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 2 is characterized in that: Constructing an adaptive wavelet threshold according to the number of decomposition layers of the film image, including: The noise standard deviation is calculated using the wavelet coefficients at each decomposition level; The adaptive wavelet threshold is constructed using the noise standard deviation; wherein the adaptive wavelet threshold is: Among them, λ j represents the wavelet threshold at the jth decomposition level, σ j represents the standard deviation of noise at the jth decomposition level, N j represents the signal length at the jth decomposition level, Represents the coefficient of the i-th wavelet at the j-th decomposition level.
4. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 3 is characterized in that: The formula of the wavelet coefficient denoising model is: Among them, sign is the sign function, a is the adjustment parameter, Represents the wavelet coefficients after denoising.
5. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 1 is characterized in that: Performing color difference balance on the filtered image to obtain a color difference balanced image includes: Calculate the pixel mean of the filtered image in the three RGB channels; wherein the pixel mean calculation formula is: Where D represents the pixel mean, represents the pixel mean of the filtered image in the red channel, represents the pixel mean of the filtered image in the green channel, Represents the pixel mean of the filtered image in the blue channel; Calculating a balance coefficient for each color channel according to the pixel mean; The color difference of the filtered image is balanced by using the balance coefficient to obtain the color difference balanced image.
6. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 5 is characterized in that: Performing color difference balance on the filtered image by using the balance coefficient to obtain the color difference balanced image includes: Using the formula: Perform color difference balance on the filtered image to obtain the color difference balanced image; wherein K R Indicates the balance coefficient of the red channel, K G Indicates the balance coefficient of the green channel, K B Represents the balance coefficient of the blue channel, R′ represents the pixel value of the red channel after balance, G′ represents the pixel value of the green channel after balance, B′ represents the pixel value of the blue channel after balance, R represents the pixel value of the filtered image in the red channel, G represents the pixel value of the filtered image in the green channel, and B represents the pixel value of the filtered image in the blue channel.
7. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 5 is characterized in that: Extracting edge features according to the color difference balance image to obtain width features includes: Performing grayscale processing on the color difference balance image to obtain a grayscale image; Based on the Canny edge detection algorithm, edge detection is performed on the grayscale image to extract an edge image; Based on the Hough transform method, the spatial points in the edge image are converted into a parameter space to obtain a boundary line; Screening is performed among the boundary straight lines to obtain two optimal straight lines; The width feature is calculated according to the optimal straight line; the calculation formula of the width feature is: Among them, ω is a coefficient adaptively determined according to the edge strength during the edge detection process, A and B are both directional coefficients of the optimal straight line, C1 and C2 are respectively the intercept parameters of the two optimal straight lines; W is the width feature.
8. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 1 is characterized in that: The width adjustment decision model includes: an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer connected in sequence.
9. The real-time automatic acquisition, identification and adjustment method for secondary inflation of PVC film according to claim 1 is characterized in that: The formula of the loss function of the width adjustment decision model is: in, for overall loss; is the film width predicted by the model, the estimated value calculated by the model; y is the actual film width; Δa is the rate of change of the adjustment action; is the mean square error, |Δa| is the absolute action change; Var(Δa) is the variance of the action change; α, β and γ are all hyperparameters; n is the sample size.
10. A real-time automatic collection, identification and adjustment system for secondary inflation of PVC film, characterized in that: include: An image acquisition module, used to acquire real-time produced film images; A filtering module, used for filtering the film image to obtain a filtered image; A balancing module, used for performing color difference balance on the filtered image to obtain a color difference balanced image; A feature extraction module, used for extracting edge features according to the color difference balance image to obtain width features; A decision module, used for inputting the width feature into a trained width adjustment decision model to obtain an adjustment decision; The adjustment module is used to adjust the clamping plate lifting device according to the adjustment decision to adjust the width of the film.
Citation Information
Cited By
Detection method and system of metering box
CN119648657A
