Search

CN-122023331-A - Coating high-precision deviation detection method based on sub-pixel edge fitting

CN122023331ACN 122023331 ACN122023331 ACN 122023331ACN-122023331-A

Abstract

The application relates to the technical field of image processing, in particular to a coating high-precision deviation detection method based on sub-pixel edge fitting. The method comprises the steps of obtaining an initial image of a coating product in a coating production process, determining a target area containing a coating edge from the initial image, carrying out convolution processing on each row of scanning lines in the target area to obtain a gradient field matrix, determining sub-pixel edge coordinates to construct an edge track point set according to the statistical characteristics of local gradient distribution for each row of the gradient field matrix, constructing an edge fitting model, constructing an objective function based on the edge track point set and the edge fitting model, determining model parameters of trend components through the minimum objective function, and calculating the theoretical edge position at the current moment by utilizing the model parameters of the determined trend components to determine the deviation detection result of the coating edge. Through the technical scheme, a more accurate deviation detection result of the coating machine can be obtained.

Inventors

  • Lai Houliang
  • YAO ZONGWEI
  • JIN XIANCHENG

Assignees

  • 青岛三人行机械有限公司

Dates

Publication Date
20260512
Application Date
20260129

Claims (10)

  1. 1. A coating high-precision deviation detection method based on sub-pixel edge fitting is characterized by comprising the following steps: Acquiring an initial image of a coating product in a coating production process, determining a target area containing a coating edge from the initial image, and carrying out convolution treatment on each row of scanning lines in the target area to obtain a gradient field matrix reflecting the energy distribution of the physical position of the coating edge; Calculating sub-pixel edge coordinates by using a space moment method based on the statistical characteristics of local gradient distribution for each row of the gradient field matrix, and constructing an edge track point set distributed along the coating direction by using a plurality of rows of sub-pixel edge coordinates; The method comprises the steps of constructing an edge fitting model, and constructing an objective function based on an edge track point set and the edge fitting model, wherein the edge fitting model is formed by superposing a trend component for representing real deviation and a vibration component for representing mechanical interference; and determining model parameters of the trend component by minimizing an objective function, and calculating a theoretical edge position after removing the vibration component at the current moment by using the determined model parameters of the trend component so as to determine a deviation detection result of the coating edge by using the theoretical edge position.
  2. 2. The method for detecting the coating high-precision deviation based on sub-pixel edge fitting according to claim 1, wherein the convolution processing of each line of scanning lines in the target area is realized by using a smooth differential kernel, and the smooth differential kernel is constructed by the following calculation formula: , wherein, In order to smooth the derivative kernels, The distance variable relative to the convolution center, The standard deviation parameter of the Gaussian distribution, the numerical distribution of the smooth differential kernel corresponds to the Gaussian pulse signal obtained by deriving the S-shaped edge curve, and exp is a natural exponential function.
  3. 3. The method for detecting the coating high-precision deviation based on the sub-pixel edge fitting according to claim 1, wherein the calculating of the sub-pixel edge coordinates by using a spatial moment method based on the statistical characteristics of the local gradient distribution comprises: Searching gradient peak values in each row of the gradient field matrix to determine integer coordinates, and taking the integer coordinates as the center to intercept a local neighborhood region; And determining the edge coordinates of the sub-pixels of each row by using the ratio of the first moment to the zero moment of gradient distribution in the local neighborhood region.
  4. 4. The method for high-precision coating deviation detection based on sub-pixel edge fitting of claim 1, wherein prior to constructing the edge fitting model, the method further comprises: Taking the product of the absolute value of the differential variation of the edge positions of two adjacent rows and the position weight function as a row characteristic value, and taking the average value of the row characteristic values corresponding to all rows as a fluctuation characteristic parameter; and judging that the current edge fluctuation is dominated by high-frequency vibration or low-frequency deviation by utilizing the fluctuation characteristic parameters so as to adjust the initial value of the edge fitting model.
  5. 5. The method for detecting coating high-precision deviation based on sub-pixel edge fitting according to claim 1, wherein the objective function is constructed by the following calculation formula: Where J is the objective function, N is the number of sub-pixel edges, Representing the observed sub-pixel edge coordinates, Respectively representing a base intercept parameter, a dip rate parameter and a curvature parameter, Respectively representing the vibration amplitude, the spatial frequency and the initial phase, The regularization coefficient is set according to the inertia of the physical system of the coater.
  6. 6. The method for high-precision coating deviation detection based on sub-pixel edge fitting of claim 5, further comprising: Calculating the product of the curvature parameter and the row coordinate at the current moment, and taking the sum of the product and the inclination rate parameter as a deviation change rate index at the current moment, wherein the deviation change rate index is used for representing the change speed of the edge position along with the travelling distance; and outputting an early warning signal under the condition that the deviation detection result of the coating edge is within a tolerance range and the deviation change rate index is continuously increased in a neighborhood time period.
  7. 7. The method for high-precision coating deviation detection based on sub-pixel edge fitting of claim 1, wherein before determining model parameters of trend components by minimizing objective functions based on the set of edge trajectory points, the method further comprises: performing a plurality of iterative screening processes, randomly extracting sub-set points from the edge track point set in each iteration, determining a temporary straight line model by using the sub-set points, and counting the number of inner points, the distance from the edge track point set to the temporary straight line model, of which is smaller than a preset distance threshold; And taking the temporary linear model with the largest number of inner points as a reference linear model, determining the distance residual error from each point in the edge track point set to the reference linear model, marking the point with the distance residual error larger than a preset distance threshold as an outlier, and eliminating the outlier or reducing the weight of the outlier when constructing an objective function.
  8. 8. The method for detecting the coating high-precision deviation based on sub-pixel edge fitting according to claim 1, wherein calculating the theoretical edge position after removing the vibration component at the current moment by using the model parameters of the determined trend component comprises: The method comprises the steps of determining image line coordinates corresponding to the current latest acquisition time, extracting basic intercept parameters, inclination rate parameters and curvature parameters from the determined model parameters, constructing a second-order polynomial function only containing trend components, substituting the image line coordinates as independent variables into the second-order polynomial function for calculation, and taking the obtained result as a theoretical edge position.
  9. 9. The method for detecting the deviation of the coating with high precision based on the sub-pixel edge fitting according to claim 1, wherein determining the deviation detection result of the coating edge by using the theoretical edge position comprises: The method comprises the steps of obtaining preset reference line coordinates, calculating the difference value between a theoretical edge position and the preset reference line coordinates to obtain pixel-level position deviation, wherein the preset reference line coordinates are used for representing ideal edge positions when a coated product is qualified; And obtaining camera pixel equivalent, multiplying the pixel-level position deviation by the camera pixel equivalent to obtain a deviation detection result of the coating edge expressed in physical length units, wherein the camera pixel equivalent is used for converting the pixel value of the image domain into the length value of the physical domain.
  10. 10. The method for high-precision coating deviation detection based on sub-pixel edge fitting of claim 1, further comprising: Corresponding algorithm parameter packages are pre-configured for coating products with different specifications, wherein the algorithm parameter packages at least comprise preset datum line coordinates for the products with the specifications, half-width parameters of gradient calculation windows, spatial frequency initial values of vibration components and regularization coefficients; In response to receiving the production work order information, a corresponding algorithm parameter package is automatically invoked to configure an edge fitting model and a smooth derivative kernel.

Description

Coating high-precision deviation detection method based on sub-pixel edge fitting Technical Field The application relates to the technical field of image processing, in particular to a coating high-precision deviation detection method based on sub-pixel edge fitting. Background In the industries of adhesive tape manufacture, label printing, sanitary product material compounding, functional film processing and the like, a coating process is a key link for determining the quality of products, and a hot melt adhesive coating machine, a non-setting adhesive coating machine, an adhesive tape coating machine, a hot melt adhesive compounding machine and other coating machines are used, wherein the core working principle is that a mechanical transmission system drives a substrate such as release paper, PET (polyethylene terephthalate) film, non-woven fabric and the like to run at high speed, and a slit die head or an anilox roller is used for uniformly coating a liquid adhesive on the surface of the substrate. The edge uniformity of the adhesive layer and the transverse position accuracy of the coating relative to the substrate are important indexes for measuring the coating quality, if coating deviation occurs, glue can overflow and pollute the roller, the shutdown cleaning frequency is increased, and compound dislocation, insufficient effective width of a product, uneven rolling end face and the like can be caused. In the high-speed operation of the hot melt adhesive coater, the tape coater and other equipment, complex vibration disturbances, such as radial runout of the coating roll, meshing frequency of the gear box and even tension fluctuation of the substrate itself, are inevitably generated. The detection algorithm in the related art cannot distinguish the high-frequency vibration caused by the mechanical factor from the real coating deviation, and the instantaneous vibration peak can be misjudged as a deviation signal, so that a deviation correcting executing mechanism such as a servo motor push rod is triggered to carry out frequent adjustment, and the flatness of the coating edge is seriously affected, therefore, a detection scheme which can adapt to the optical characteristics of special media such as hot melt adhesive and the like and can accurately separate the real deviation trend from the mechanical vibration noise is required. Disclosure of Invention The application provides a coating high-precision deviation detection method based on sub-pixel edge fitting, which comprises the steps of obtaining an initial image of a coating product in a coating production process, determining a target area containing a coating edge from the initial image, and carrying out convolution treatment on each row of scanning lines in the target area to obtain a gradient field matrix reflecting the energy distribution of the physical position of the coating edge; according to each row of the gradient field matrix, calculating sub-pixel edge coordinates by using a space moment method based on the statistical characteristics of local gradient distribution, constructing an edge track point set distributed along the coating direction by using a plurality of rows of sub-pixel edge coordinates, constructing an edge fitting model, and constructing an objective function based on the edge track point set and the edge fitting model, wherein the edge fitting model is formed by superposing trend components for representing real deviation and vibration components for representing mechanical disturbance, the objective function comprises a data fidelity term and a regularization term, the regularization term is used for restraining the severe high-frequency change from being attributed to the trend components based on physical inertia constraint, determining model parameters of the trend components by minimizing the objective function, calculating a theoretical edge position after the vibration components are removed at the current moment by using the model parameters of the determined trend components, and determining a deviation detection result of the coating edge by using the theoretical edge position. Therefore, the interference of mechanical vibration noise on the deviation detection of the coating machine can be avoided, and a more accurate deviation detection result of the coating machine can be obtained. Optionally, the convolution processing of each line of scan lines in the target region is implemented using a smooth derivative kernel, which is constructed by the following calculation formula: , wherein, In order to smooth the derivative kernels,As a distance variable with respect to the convolution center,The numerical distribution of the smooth differentiation kernel corresponds to the gaussian pulse signal obtained by deriving the S-shaped edge curve as standard deviation parameter of the gaussian distribution,As a natural exponential function. In this way, the convolution is carried out by using Gaussian derivative kernel,