A polypropylene film defect detection method, device and system

Abstract

The present application relates to the technical field of defect detection, in particular to a polypropylene film defect detection method, device and system. A polypropylene film multi-row pixel point fringe image is obtained and stored in a row-level buffer, and is processed by frequency domain and signal analysis to obtain a refractive phase diagram, a fringe density diagram and a transmittance diagram; a crystal point defect forms an optical micro-lens effect, and has a high-frequency local frequency mutation on the fringe image; the fringe density diagram is filtered and compared to obtain a micro-lens distortion diagram, thereby improving the sensitivity of the defect detection; a dynamic gain coefficient is calculated by using a low-frequency density diagram and a refractive phase diagram data dynamic response relationship, and an anti-shake phase diagram is corrected by modifying the refractive phase diagram; a target row phase curvature diagram is calculated by analyzing the difference between different row anti-shake phase diagrams, thereby revealing the bending condition of the film surface; finally, the target row transmittance diagram, the phase curvature diagram and the micro-lens distortion diagram are fused to detect defects, thereby improving the detection accuracy and comprehensiveness.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and specifically to a method, equipment and system for detecting defects in polypropylene films. Background Technology

[0002] Polypropylene film, as a key basic industrial material, is widely used in many fields such as packaging, electronics, and optics due to its excellent physical properties, chemical stability, and good optical properties. During the production process of polypropylene film, defects, such as crystal point defects, may appear on the film surface or inside. These defects not only affect the appearance quality of the film but also have a serious impact on its performance.

[0003] Existing technologies for defect detection of polypropylene films typically rely solely on the film's transmittance or simple grayscale characteristics. However, non-lethal dust or minute foreign objects often adhere to the surface of polypropylene films. Different defects offer varying degrees of relevance for optimizing subsequent slitting processes. Single-dimensional detection methods struggle to effectively distinguish between lethal refractive crystal points and non-lethal obstructive dust from an imaging mechanism perspective, significantly reducing the reliability and accuracy of defect detection. Summary of the Invention

[0004] To address the problem that non-lethal dust or micro-defects often adhere to the surface of polypropylene films, and that different defects have varying degrees of relevance for optimizing subsequent slitting processes, while single-dimensional detection methods struggle to effectively distinguish between lethal refractive crystal points and non-lethal obstructive dust from the imaging mechanism, leading to a significant reduction in the reliability and accuracy of defect detection, this invention aims to provide a method, equipment, and system for detecting defects in polypropylene films. The specific technical solution adopted is as follows:

[0005] A method for detecting defects in polypropylene films, comprising:

[0006] The stripe images of multiple rows of pixels on the polypropylene film are acquired sequentially and stored in a row-level buffer. Frequency domain processing and signal analysis are performed on the stripe images of each row of pixels to obtain the refractive phase map, stripe density map and transmittance map of each row of pixels.

[0007] The fringe density map is filtered to obtain a low-frequency density map; the differences between the fringe density map and the low-frequency density map are compared to obtain a microlens distortion map; the dynamic gain coefficient is calculated using the change response characteristics between the data values ​​in the low-frequency density map and the data values ​​in the refractive phase map, thereby adaptively correcting the refractive phase map to obtain an anti-jitter phase map; the changes in the anti-jitter phase maps of different rows are analyzed to calculate the phase curvature map of the target row.

[0008] By fusing the transmittance map, phase curvature map, and microlens distortion map of the target row, a comprehensive defect score is calculated for each pixel on the target row. Based on the comprehensive defect score and the numerical characteristics of the transmittance map, defect detection is performed on the target row on the film.

[0009] Furthermore, the method for obtaining the microlens distortion map includes:

[0010] In the low-frequency density map and stripe density map corresponding to each row of pixels in the polypropylene film, the absolute value of the difference between the data value of the pixel at the same position in the stripe density map and the data value in the low-frequency density map is used as the density deviation factor.

[0011] The microlens distortion map for each row of pixels is constructed based on the density deviation factor corresponding to each row of pixels in the polypropylene film.

[0012] Furthermore, the method for obtaining the dynamic gain coefficient includes:

[0013] In the low-frequency density map of each row of pixels in the polypropylene film, the median value within the preset window is used as a reference value.

[0014] In each row of pixels of the polypropylene film, the sum of the data value of each pixel in the low-frequency density map and the preset constant is used as the denominator, and the reference value is used as the numerator, which serves as the dynamic gain coefficient corresponding to each pixel.

[0015] Furthermore, the method for obtaining the anti-jitter phase map includes:

[0016] The refractive phase map of each row of pixels in the polypropylene film is subjected to mean filtering to obtain a low-frequency phase map. In the low-frequency phase map and the refractive phase map of each row of pixels, the difference between the data value of the pixel at the same position in the refractive phase map and the data value in the low-frequency phase map is calculated to obtain the local phase map of each row of pixels.

[0017] In each row of pixels of the polypropylene film, the product of the data value of each pixel in the local phase map and the corresponding dynamic gain coefficient is used as the anti-jitter phase of each pixel.

[0018] The anti-jitter phase diagram is composed of the anti-jitter phase of each row of pixels in the polypropylene film.

[0019] Furthermore, the method for obtaining the phase curvature map includes:

[0020] In the row-level buffer, rows acquired after the target row's acquisition time are designated as the latest rows at preset intervals; rows acquired before the target row's acquisition time are designated as historical comparison rows at preset intervals.

[0021] In the anti-jitter phase map of the target row, the latest row, and the historical comparison row, for any pixel at the same position in each row, the sum of the data values ​​of the pixel in the latest row and the historical comparison row is subtracted by twice the data value in the target row, and the absolute value of the difference is used as the phase curvature of the pixel in the target row.

[0022] The target curvature of all pixels in the target row is used to construct the phase curvature map of the target row.

[0023] Furthermore, the method for obtaining the comprehensive defect score includes:

[0024] For any pixel in the target row, the data value of the pixel in the microlens distortion map is multiplied by a preset balance factor, and then added to the data value of the pixel in the phase curvature map to obtain the superimposed morphological feature value of the pixel.

[0025] The superimposed morphological feature value of the pixel is multiplied by the data value of the pixel in the transmittance map and then normalized to obtain the comprehensive defect score of the pixel.

[0026] Furthermore, the defect detection of the target row on the film based on the numerical characteristics of the comprehensive defect score and the transmittance map includes:

[0027] On the target row, for any pixel, if the overall defect score of the pixel is greater than the preset crystal point alarm threshold, then it is determined that there is a crystal point defect at the pixel.

[0028] If the overall defect score of the pixel is less than or equal to the preset crystal point alarm threshold, and if the data value of the pixel in the transmittance map is less than the preset transmittance threshold, then it is determined that there is a dust defect at the pixel.

[0029] Furthermore, the method for obtaining the low-frequency density map includes:

[0030] After applying median filtering to the fringe density map of each row, the low-frequency density map of each row is obtained.

[0031] A polypropylene film defect detection system, comprising:

[0032] The data acquisition module is used to acquire stripe images of multiple rows of pixels on a polypropylene film in a time sequence and store them in a row-level buffer; it performs frequency domain processing and signal analysis on the stripe images of each row of pixels to obtain the refractive phase map, stripe density map and transmittance map of each row of pixels.

[0033] The feature analysis module is used to filter the fringe density map to obtain a low-frequency density map; compare the differences between the fringe density map and the low-frequency density map to obtain a microlens distortion map; use the change response characteristics between the data values ​​in the low-frequency density map and the data values ​​in the refractive phase map to calculate the dynamic gain coefficient, thereby adaptively correcting the refractive phase map to obtain an anti-jitter phase map; analyze the changes in the anti-jitter phase maps of different rows, and calculate the phase curvature map of the target row.

[0034] The defect detection module is used to fuse the transmittance map, phase curvature map and microlens distortion map of the target row, calculate the comprehensive defect score of each pixel on the target row, and perform defect detection on the target row on the film based on the comprehensive defect score and the numerical characteristics of the transmittance map.

[0035] A polypropylene film defect detection device includes a processor and a memory. The memory stores at least one instruction, at least one program, code set, or instruction set. When the processor loads and executes the at least one instruction, at least one program, code set, or instruction set, it implements the steps of a polypropylene film defect detection method.

[0036] The present invention has the following beneficial effects:

[0037] By acquiring fringe images of multiple rows of pixels on a polypropylene film sequentially and storing them in a row-level buffer, the continuity and integrity of the image data can be ensured. Simultaneously, frequency domain processing and signal analysis are performed on the fringe images of each row of pixels to obtain a refraction phase map, fringe density map, and transmittance map. This multi-dimensional information acquisition method can more accurately capture various features of the film. Crystal point defects, due to their tiny spherical structures, form optical microlens effects, manifesting as high-frequency local frequency abrupt changes in the fringe image. Therefore, filtering the fringe density map to obtain a low-frequency density map and comparing the differences between the two to obtain a microlens distortion map makes this microlens distortion feature in the film more obvious, thereby improving the subsequent detection sensitivity for such defects. Furthermore, in a transmission-type fringe deflection imaging system, there is a significant engineering principle: when the film is displaced due to jitter, the fringe frequency in the imaging field of view changes, and the sensitivity of phase measurement also changes accordingly. Therefore, by utilizing the dynamic response characteristics between the data values ​​in the low-frequency density map and the data values ​​in the refraction phase map, a dynamic gain coefficient is calculated, and the refraction phase map is adaptively corrected to obtain an anti-jitter phase map. This step effectively eliminates phase errors caused by equipment vibration, film jitter, and other factors, improving the accuracy of phase information. Next, the differences in the anti-jitter phase maps of different rows are analyzed, and the phase curvature map of the target row is calculated, revealing the degree and trend of film surface curvature from the perspective of phase change. Finally, the transmittance map, phase curvature map, and microlens distortion map of the target row are fused to calculate the comprehensive defect score for each pixel in the target row. Defect detection is then performed based on the comprehensive defect score and the numerical characteristics of the transmittance map. By fully considering various film characteristics, this method can more comprehensively and accurately determine whether each pixel has defects, as well as the type and severity of those defects, greatly improving the accuracy and comprehensiveness of defect detection. Attached Figure Description

[0038] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a method for detecting defects in a polypropylene film according to an embodiment of the present invention;

[0040] Figure 2 This is a system block diagram of a polypropylene film defect detection system provided in one embodiment of the present invention;

[0041] Figure 3This is a schematic diagram of the structure of a polypropylene film defect detection device according to an embodiment of the present invention. Detailed Implementation

[0042] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a polypropylene film defect detection method, device, and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0044] The following description, in conjunction with the accompanying drawings, details the specific solution of the polypropylene film defect detection method, equipment, and system provided by the present invention.

[0045] Please see Figure 1 The diagram illustrates a flowchart of a method, device, and system for detecting defects in polypropylene films according to an embodiment of the present invention. The method includes the following steps:

[0046] Step S1: Acquire stripe images of multiple rows of pixels on the polypropylene film in a time sequence and store them in a row-level buffer; perform frequency domain processing and signal analysis on the stripe images of each row of pixels to obtain the refraction phase map, stripe density map and transmittance map of each row of pixels.

[0047] Because the polypropylene film production line adopts a continuous high-speed transmission mode, and subsequent defect topology analysis requires spatial neighborhood information in the longitudinal (time axis direction), a W×N storage space can be allocated in memory to construct a first-in-first-out (FIFO) row-level buffer to transform the one-dimensional time-stream data output by the line scan camera into local image patches with two-dimensional spatial relationships. Here, W is the lateral pixel resolution of the line scan camera (e.g., 4096 or 8192 pixels), and N is the buffer depth, the value of which is determined by the longitudinal sampling step size k. In this embodiment of the invention, to solve the scale matching problem between micrometer-level pixel sampling and millimeter-level macroscopic defects, a longitudinal sampling step size k is introduced. The value of this step size k is set to be related to the number of rows corresponding to the smallest diameter of the crystal point to be detected; for example, k=50, and the buffer depth N is set to at least 2k+1 rows to ensure that data from the current time, the past k time points, and the past 2k time points can be accessed simultaneously.

[0048] Therefore, by continuously acquiring stripe images of multiple rows of pixels on a polypropylene film in a time sequence using a high-frequency linear array camera, each time a new row of image data is acquired, it is stored at the top of the row-level buffer, and the old data at the bottom of the row-level buffer is removed and discarded.

[0049] Because the original striped image contains a mixture of background light intensity distribution caused by light source inhomogeneity, light field modulation information carried by sinusoidal stripes, and high-frequency imaging noise, a Fast Fourier Transform (FFT) technique can be used to perform frequency domain analysis on the striped image of each row of pixels in the row-level buffer in order to separate independent physical properties from single-row pixels. Furthermore, to suppress edge spectral leakage in the subsequent frequency domain transformation, a window function can be applied to the striped image before performing the FFT transformation. In this embodiment of the invention, the Hanning window is used as the weighting function, that is, each data value in the striped image is multiplied by the Hanning window function to obtain the windowed image.

[0050] Then, a one-dimensional Fast Fourier Transform (FFT) is performed on the windowed image of each row of pixels to obtain its frequency domain spectrum distribution. Based on the optical characteristics of fringe projection, the spectrum contains a zero-frequency component carrying illumination information and a carrier frequency band carrying phase information. Therefore, low-pass and band-pass filters are designed for extraction: the low-pass filter extracts the low-frequency spectrum centered at zero frequency, performs an inverse Fourier transform (IFFT), and takes the modulus value to obtain the background light intensity distribution of each row of pixels. Using a band-pass filter with a center frequency equal to the fringe carrier frequency, the positive first-order spectral component is extracted, and an inverse Fourier transform yields the complex analytic signal.

[0051] Then, based on the above analytical signal and background light intensity distribution map, the following three sets of physical feature maps can be solved synchronously for each row of pixels: (1) Refraction phase map. When light passes through the polypropylene film, the refractive index distribution of the medium will change the optical path, causing the stripe phase to shift. In order to quantify this refraction deflection, the argument of the analytical signal is calculated to obtain the wrapping phase. The unwrap algorithm is used to recover the continuous phase. Finally, the linear phase component introduced by the preset stripe carrier frequency is subtracted to obtain the refraction phase map caused only by film refraction.

[0052] in, This represents the refraction phase value of the pixel at position x in the nth row; This represents the imaginary part of the pixel at the x-th position in the n-th row of the analytic signal. This represents the real part of the pixel at the x-th position in the n-th row of the analytic signal; This represents the arctangent function, used to calculate the argument. This represents the phase unwrapping operator, used to restore discontinuous phases to continuous phases; Indicates the preset stripe carrier frequency; This indicates a linear phase component introduced by a preset stripe carrier frequency.

[0053] (2) Transmittance Map: To accurately distinguish between transparent films and opaque impurities (such as dust), a physical quantity that can stably characterize light flux is needed. This requires obtaining a pre-calibrated standard background light intensity constant under defect-free conditions. Using the data value of each pixel in the previously calculated background light intensity distribution map as the numerator, and the sum of the standard background light intensity constant and the pre-calibrated constant as the denominator, we obtain the data value representing the light path transmittance of each pixel. The larger the value, the better the transmittance; thus, the transmittance map is obtained. The standard background light intensity is calibrated based on an unobstructed state, so the value is usually large. Therefore, the data value in the transmittance map ranges from 0 to 1. The closer to 1, the better the transmittance; the closer to 0, the worse the transmittance, and the more likely it corresponds to dust.

[0054] It should be noted that the preset constant here is used to prevent the denominator from being 0, and the value can be 0.001. The specific value can be adjusted according to the implementation scenario, and is not limited here.

[0055] (3) Stripe density map: The change in distance (jitter) between the thin film and the camera will cause a slight perturbation in the stripe frequency within the imaging field of view, while the microlens effect of the crystal point will cause local frequency distortion. In order to capture this frequency change, the first-order partial derivative of the total phase function containing the carrier with respect to the horizontal coordinate x can be obtained to obtain the stripe density map of each row of pixels. The data values ​​in the stripe density map reflect the spatial rate of phase change.

[0056]

[0057] in, This represents the stripe density of the pixel at position x in the nth row; Represents partial differential operators; This represents the refraction phase value of the pixel at position x in the nth row; Indicates the preset stripe carrier frequency; This indicates a linear phase component introduced by a preset stripe carrier frequency; This indicates the unit conversion factor.

[0058] It should be noted that the Fast Fourier Transform and Inverse Fourier Transform in the aforementioned process are well-known technologies, and the specific process will not be described in detail here; and the preset stripe carrier frequency in this embodiment of the present invention is set to make the period of the stripes on the imaging surface between 8 and 16 pixels, which can be adjusted according to the implementation scenario, and is not limited here.

[0059] Step S2: Filter the fringe density map to obtain a low-frequency density map; compare the differences between the fringe density map and the low-frequency density map to obtain a microlens distortion map; use the correlation characteristics between the data values ​​in the low-frequency density map and the data values ​​in the refractive phase map to calculate the dynamic gain coefficient, thereby adaptively correcting the refractive phase map to obtain an anti-jitter phase map; analyze the differences in the changes of the anti-jitter phase maps of different rows, and calculate the phase curvature map of the target row.

[0060] Mechanical vibration in polypropylene films is a macroscopic, low-frequency change. Crystal point defects possess tiny spherical structures, which create an optical microlens effect. In fringe patterns, this effect manifests as high-frequency local frequency abrupt changes. In other words, the region containing crystal point defects causes rapid and drastic changes in the density of the fringes, creating a significant difference from the fringe characteristics of the surrounding normal areas. Therefore, filtering the fringe density map yields a low-frequency density map. This filtering operation removes high-frequency noise and interference from the fringe density map, preserving its low-frequency components. Essentially, it extracts the system vibration state, which can serve as an intrinsic benchmark. By comparing the differences between the fringe density map and the low-frequency density map, the high-frequency local frequency abrupt changes caused by the optical microlens effect of crystal point defects can be highlighted, thus obtaining a microlens distortion map.

[0061] Preferably, in one embodiment of the present invention, the method for obtaining the microlens distortion map includes:

[0062] First, the stripe density map is filtered to obtain the low-frequency density map: After median filtering of the stripe density map of each row, the low-frequency density map of each row is obtained. The low-frequency density map at this time can be used as an endogenous reference to reflect the low-frequency smooth frequency reference of each row of pixels in the polypropylene film caused by the jitter of the film position, and filter out the high-frequency narrow pulse frequency change caused by the microlens effect of the crystal point.

[0063] It should be noted that median filtering is a well-known technique, and the specific process will not be described in detail here. In this embodiment of the present invention, the width of the filtering window for median filtering must be significantly greater than the number of pixels corresponding to the maximum expected diameter of the crystal point to be detected. For example, if the maximum expected diameter of the crystal point is 15 pixels, then the width of the filtering window in this embodiment of the present invention can be set to a size of 32 to 64 pixels.

[0064] Microlens distortion causes high-frequency variations in the fringe density map in localized areas, differing from the overall trend of the low-frequency density map. Therefore, in the low-frequency and fringe density maps corresponding to each row of pixels in the polypropylene film, the absolute value of the difference between the data values ​​in the fringe density map and the low-frequency density map for the same pixel location is used as a density deviation factor. This density deviation factor can more accurately capture the local variations caused by microlens distortion, that is, the degree of deviation of the fringe density at each location relative to the overall low-frequency background. The larger the value, the more drastic the local refraction direction of light has occurred at that location, and the stronger the lens effect. Finally, by arranging the density deviation factors corresponding to each row of pixels in the polypropylene film according to the pixel arrangement order, a microlens distortion map corresponding to each row of pixels can be obtained, clearly reflecting the distribution of lens distortion in the polypropylene film.

[0065] In transmission-type fringe deflection imaging systems, a significant engineering principle exists: when a polypropylene film shifts due to vibration, the fringe frequency within the imaging field of view changes. This is because the film's movement alters the light propagation path, affecting the density of the fringes; simultaneously, the sensitivity of phase measurement also changes. Phase measurement calculates information such as the film's refractive phase based on fringe deformation. Changes in fringe frequency interfere with the accuracy of phase measurement, leading to errors. Furthermore, there is a certain change response characteristic between the data values ​​in the low-frequency density map and the data values ​​in the refractive phase map. This relationship is used to calculate a dynamic gain coefficient, which can adjust the correction degree of the refractive phase map in real time according to the film's vibration. By adaptively correcting the refractive phase map—that is, adjusting the data in the refractive phase map according to the dynamic gain coefficient—phase errors caused by equipment vibration, film vibration, and other factors can be eliminated, resulting in an anti-vibration phase map. This step ensures the accuracy of phase information, providing a reliable data foundation for subsequent defect detection based on phase information.

[0066] Preferably, in one embodiment of the present invention, the method for obtaining the anti-jitter phase map includes:

[0067] First, the dynamic gain coefficient is obtained: To address the differences between different production batches, film tensions, or camera mounting heights, a reference point needs to be dynamically determined. Therefore, the median within a preset window is used as the reference value in the low-frequency density map of each row of pixels on the polypropylene film. In this embodiment of the invention, the preset window is selected as [0.05W, 0.95W], where W represents the horizontal pixel resolution of the line scan camera. The preset window can exclude interference from the edge regions on both the left and right sides of the image, which account for 5% of the total edge area.

[0068] If the low-frequency density value at a pixel is greater than the reference value, it indicates that the stripes at that pixel are denser, meaning the measurement response (sensitivity) is higher; conversely, if the low-frequency density value at that pixel is less than the reference value, it indicates that the stripes at that pixel are sparser, and the response is weaker. Therefore, in each row of pixels on the polypropylene film, the reference value is used as the numerator, and the sum of the data value of each pixel in the low-frequency density map and the preset constant is used as the denominator, which serves as the dynamic gain coefficient for each pixel. Based on the aforementioned logic, a dynamic gain coefficient greater than 1 indicates that the low-frequency density value is less than the reference value, requiring the attenuated weak phase signal to be amplified in subsequent processes; conversely, a dynamic gain coefficient less than 1 indicates that the low-frequency density value is greater than the reference value, requiring the amplified phase signal to be suppressed in subsequent processes. It should be noted that the preset constant here is used to prevent the denominator from being 0, and can be set to 0.001. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0069] Because the overall tilt, bending, or large shaking and surging of the polypropylene film will form a huge, gradually changing background "slope" or "wave" on the refractive phase map, it is necessary to eliminate the interference of this factor. Therefore, the refractive phase map of each row of pixels of the polypropylene film is subjected to mean filtering, and the window size is set to 4 times the filtering window of the aforementioned median filtering. The specific value can be adjusted according to the implementation scenario to obtain a low-frequency phase map. Then, in the low-frequency phase map and the refractive phase map of each row of pixels, the difference between the data value of the pixel in the refractive phase map and the data value in the low-frequency phase map is calculated, which is to subtract the influence of the aforementioned factors, thereby obtaining the local phase map of each row of pixels.

[0070] Next, in each row of pixels of the polypropylene film, the data value of each pixel in the local phase map is multiplied by the corresponding dynamic gain coefficient to compensate for the data value in the local phase map. The resulting product is used as the anti-jitter phase of each pixel. Based on this, the anti-jitter phases of each row of pixels of the polypropylene film can be combined to form an anti-jitter phase map.

[0071] Although the aforementioned process eliminates large-scale nonlinear gain fluctuations caused by polypropylene film jitter, a crystal point is a macroscopic solid defect that typically spans multiple pixel rows. Therefore, the distribution characteristics of the defect signal in the longitudinal direction (time axis) have not yet been extracted. Thus, in this embodiment of the invention, the differences in the anti-jitter phase map of different rows are analyzed, and the phase curvature map of the target row is calculated to reflect the morphological polarity of the defect.

[0072] Preferably, in one embodiment of the present invention, the method for obtaining the phase curvature map includes:

[0073] Because the sampling frequency of linear scan cameras is typically extremely high, with row spacing at the micrometer level, while crystal point defects are at the millimeter level, the data changes between adjacent rows are usually minimal, mainly dominated by noise. Therefore, in the row-level buffer, rows with a preset step size are used as the latest rows after the acquisition time of the target row, and rows with a preset step size are used as historical comparison rows before the acquisition time of the target row. This forces the physical spacing of subsequent analyses to match the physical size of the defects. It should be noted that in this embodiment of the invention, the target row index is set to nk, and the preset step size is set to k-1, where k is the preset sampling step size in step S1.

[0074] In the anti-jitter phase maps of the target row, the latest row, and the historical comparison row, for any pixel at the same position in each row, a second-order spatial differentiation operation is performed: the sum of the data values ​​of this pixel in the latest row and the historical comparison row is subtracted by twice the data value in the target row. The absolute value of the difference is taken as the phase curvature of this pixel in the target row. The larger the phase curvature, the more significant the bulge or depression structure at this pixel in the target row, and the more drastic the abrupt change in refractive index. Finally, the target curvatures of all pixels in the target row are used to construct the phase curvature map of the target row.

[0075] Step S3: Fuse the transmittance map, phase curvature map, and microlens distortion map of the target row, calculate the comprehensive defect score of each pixel on the target row, and perform defect detection on the target row on the film based on the comprehensive defect score and the numerical characteristics of the transmittance map.

[0076] Since transparent crystal point defects optically manifest as both topological protrusions on the phase wavefront (refractive properties) and frequency distortion of local fringes (zoom properties), while opaque impurities only block light intensity, in order to accurately distinguish between these two types of defects, the features obtained from different physical dimensions can be integrated into a unified evaluation index. That is, the transmittance map, phase curvature map, and microlens distortion map of the target row can be integrated to calculate the comprehensive defect score of each pixel on the target row.

[0077] Preferably, in one embodiment of the present invention, the method for obtaining the comprehensive defect score includes:

[0078] Based on the analysis in step S2, the larger the data value in the phase curvature map, the more significant the convex or concave structure at that pixel in the target row. The larger the data value in the microlens distortion map, the more drastic the local deflection direction of the light at that location, and the stronger the lens effect, indicating that it is more likely to be an abnormal defect. Therefore, transparent crystal points have high curvature and high distortion. However, since the data values ​​in the phase curvature map and the data values ​​in the microlens distortion map belong to completely different physical dimensions, and the values ​​are greatly affected by the optical path structure and camera resolution, for any pixel in the target row, the data value of the pixel in the microlens distortion map is multiplied by a preset balance factor, and then added to the data value of the pixel in the phase curvature map to obtain the superimposed morphological feature value of the pixel. Based on the aforementioned analysis, the larger the superimposed morphological feature value, the more likely the pixel is to be a crystal point defect.

[0079] It should be noted that in this embodiment of the present invention, the preset balance factor is the ratio of the average phase curvature of the normal polypropylene film to the average microlens distortion, which can be obtained according to the normal film region. If there is a special case where the denominator is 0, a preset constant of 0.001 is added to the denominator as an addend to prevent the denominator from being 0.

[0080] For transparent crystalline points, their data values ​​in the transmittance map will be relatively large. For opaque dust, although its edges may produce distorted signals, its data values ​​in the transmittance map will be relatively small due to light obstruction. Therefore, for any pixel in the target row, the superimposed morphological feature value of the pixel is multiplied by its data value in the transmittance map, and then normalized to obtain the comprehensive defect score of the pixel. The higher the comprehensive defect score, the higher the probability that the pixel is a transparent crystalline point. Normalization is a technique well-known to those skilled in the art. The normalization function can be linear normalization or standard normalization, such as maximum-minimum normalization. The specific normalization method is not limited here.

[0081] After obtaining the comprehensive defect score for each pixel on the target row, defects on the target on the film can be detected based on the score and the numerical characteristics of light transmittance.

[0082] Preferably, in one embodiment of the present invention, defect detection is performed on the target row on the film based on the numerical characteristics of the comprehensive defect score and transmittance, including:

[0083] Based on the above analysis, the higher the comprehensive defect score, the higher the probability that the pixel is a transparent crystal point defect. Therefore, in the target row, for any pixel, if the comprehensive defect score of the pixel is greater than the preset crystal point alarm threshold, it is determined that there is a crystal point defect at the pixel.

[0084] If the overall defect score of a pixel is less than or equal to the preset crystal point alarm threshold, it is necessary to further determine the possibility that the pixel is dust. Since the larger the data value in the transmittance map, the better the light transmittance, if the data value of the pixel in the transmittance map is less than the preset light transmittance threshold, it is determined that there is a dust defect at the pixel.

[0085] It should be noted that in this embodiment of the present invention, the preset crystal point alarm threshold is 0.65 and the preset light transmission threshold is 0.5. The specific values ​​can be adjusted according to the implementation scenario and are not limited here.

[0086] In summary, by acquiring fringe images of multiple rows of pixels in a polypropylene film sequentially and storing them in a row-level buffer, the continuity and integrity of the image data can be guaranteed. Simultaneously, frequency domain processing and signal analysis are performed on the fringe images of each row of pixels to obtain a refraction phase map, fringe density map, and transmittance map. This multi-dimensional information acquisition method can more accurately capture various features of the film. Crystal point defects, due to their tiny spherical structures, form optical microlens effects, manifesting as high-frequency local frequency abrupt changes in the fringe image. Therefore, filtering the fringe density map to obtain a low-frequency density map and comparing the differences between the two to obtain a microlens distortion map makes this microlens distortion feature in the film more obvious, thereby improving the subsequent detection sensitivity for such defects. Furthermore, in a transmission-type fringe deflection imaging system, there is a significant engineering principle: when the film is displaced due to jitter, the fringe frequency in the imaging field of view changes, and the sensitivity of phase measurement also changes accordingly. Therefore, by utilizing the dynamic response characteristics between the data values ​​in the low-frequency density map and the data values ​​in the refraction phase map, a dynamic gain coefficient is calculated, and the refraction phase map is adaptively corrected to obtain an anti-jitter phase map. This step effectively eliminates phase errors caused by equipment vibration, film jitter, and other factors, improving the accuracy of phase information. Next, the differences in the anti-jitter phase maps of different rows are analyzed, and the phase curvature map of the target row is calculated, revealing the degree and trend of film surface curvature from the perspective of phase change. Finally, the transmittance map, phase curvature map, and microlens distortion map of the target row are fused to calculate the comprehensive defect score for each pixel in the target row. Defect detection is then performed based on the comprehensive defect score and the numerical characteristics of the transmittance map. By fully considering various film characteristics, this method can more comprehensively and accurately determine whether each pixel has defects, as well as the type and severity of those defects, greatly improving the accuracy and comprehensiveness of defect detection.

[0087] This invention also provides a polypropylene film defect detection system; please refer to [link / reference]. Figure 2The diagram shows a system block diagram, including a data acquisition module 201 for implementing step S1 in the above method embodiment; a feature analysis module 202 for implementing step S2 in the above method embodiment; and a defect detection module 203 for implementing step S3 in the above method embodiment.

[0088] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the polypropylene film defect detection system and the polypropylene film defect detection method embodiment provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiment, which will not be repeated here.

[0089] This invention also provides a polypropylene film defect detection device; please refer to [link to relevant documentation]. Figure 3 The diagram shows a schematic of the device structure, including a processor 300, a memory 301, a bus 302, and a communication interface 303. The processor 300, the communication interface 303, and the memory 301 are connected via the bus 302. The memory 301 may contain high-speed random access memory, and the bus 302 may be an ISA bus, a PCI bus, or an EISA bus, etc. The processor 300 may be an integrated circuit chip with signal processing capabilities. The memory 301 stores at least one instruction, at least one program, a code set, or an instruction set. When the processor loads and executes the at least one instruction, at least one program, a code set, or an instruction set, it implements the steps in a polypropylene film defect detection method.

[0090] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0091] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method of detecting defects in a polypropylene film, characterized by, The method comprises: acquiring and storing the fringe images of the polypropylene film in a time sequence, and performing frequency domain processing and signal analysis on the fringe images of each row of pixels to obtain the refractive phase image, the fringe density image and the light transmittance image of each row of pixels; filtering the fringe density image to obtain a low-frequency density image, comparing the fringe density image and the low-frequency density image to obtain the microlens distortion image, and calculating the dynamic gain coefficient based on the change response characteristics between the data values in the low-frequency density image and the data values in the refractive phase image to adaptively correct the refractive phase image and obtain the anti-shake phase image; analyzing the change differences of the anti-shake phase images of different rows, calculating the phase curvature image of the target row, and fusing the light transmittance image, the phase curvature image and the microlens distortion image of the target row to calculate the comprehensive defect score of each pixel point on the target row and perform defect detection on the target row on the film based on the numerical characteristics of the comprehensive defect score and the light transmittance image.

2. The method for detecting defects in a polypropylene film according to claim 1, characterized in that, The method for obtaining the microlens distortion image comprises: in the low-frequency density image and the fringe density image corresponding to each row of pixels of the polypropylene film, taking the absolute value of the difference between the data value of the pixel point at the same position in the fringe density image and the data value in the low-frequency density image as the density deviation factor; composing the microlens distortion image corresponding to each row of pixels according to the density deviation factors corresponding to each row of pixels of the polypropylene film.

3. The method for detecting defects in a polypropylene film according to claim 1, characterized in that, The method for obtaining the dynamic gain coefficient comprises: in the low-frequency density image of each row of pixels of the polypropylene film, taking the median value in a preset window as a reference value; in each row of pixels of the polypropylene film, taking the sum of the data value of each pixel point in the low-frequency density image and a preset constant as the denominator, and taking the reference value as the numerator as the dynamic gain coefficient corresponding to each pixel point.

4. The method of claim 1, wherein the step of detecting the defect of the polypropylene film is performed by using a light source having a wavelength of 660 nm. The method for obtaining the anti-shake phase image comprises: performing mean filtering on the refractive phase image of each row of pixels of the polypropylene film to obtain a low-frequency phase image, and calculating the difference between the data value of the pixel point at the same position in the refractive phase image and the data value in the low-frequency phase image in the low-frequency phase image and the refractive phase image of each row of pixels to obtain the local phase image of each row of pixels; in each row of pixels of the polypropylene film, taking the product of the data value of each pixel point in the local phase image and the corresponding dynamic gain coefficient as the anti-shake phase of each pixel point; composing the anti-shake phase image of the polypropylene film by the anti-shake phase of each row of pixels.

5. The method of claim 1, wherein the step of detecting the defect of the polypropylene film is performed by using a machine vision system. The method for obtaining the phase curvature image comprises: in the row-level buffer, the rows with a preset step length interval after the acquisition time of the target row are taken as the latest rows, and the rows with a preset step length interval before the acquisition time of the target row are taken as the historical comparison rows; in the anti-shake phase images of the target row, the latest row and the historical comparison row, for any one pixel point at the same position in each row, the sum of the data values of the pixel point in the latest row and the historical comparison row is subtracted by twice the data value in the target row, and the absolute value of the obtained difference is taken as the phase curvature of the pixel point in the target row; the target curvatures of all pixel points on the target row are composed to obtain the phase curvature image of the target row.

6. The method of claim 1, wherein the step of detecting the defect of the polypropylene film is performed by using a machine vision system. The method for obtaining the comprehensive defect score comprises: For any one pixel point in the target row, the data value of the pixel point in the microlens distortion map is multiplied by a preset balance factor, and then added to the data value of the pixel point in the phase curvature map, as the superimposed topographic feature value of the pixel point; The superimposed topographic feature value of the pixel point is multiplied by the data value of the pixel point in the transmittance map and normalized to obtain the comprehensive defect score of the pixel point.

7. The method of claim 1, wherein the step of detecting the defect of the polypropylene film is performed by using a machine vision system. The defect detection on the target row on the film according to the comprehensive defect score and the numerical characteristics of the transmittance map comprises: On the target row, for any one pixel point, if the comprehensive defect score of the pixel point is greater than a preset crystal point alarm threshold, it is determined that a crystal point defect exists at the pixel point; If the comprehensive defect score of the pixel point is less than or equal to the preset crystal point alarm threshold, if the data value of the pixel point in the transmittance map is less than a preset transmittance threshold, it is determined that a dust defect exists at the pixel point.

8. The method of claim 1, wherein the polypropylene film is a biaxially oriented film. The method for obtaining the low-frequency density map comprises: After median filtering the fringe density map of each row, the low-frequency density map of each row is obtained.

9. A polypropylene film defect detection system characterized by, The system comprises: A data acquisition module is configured to acquire fringe images of a plurality of rows of pixel points of a polypropylene film in a time sequence and store the fringe images in a row-level buffer, and perform frequency domain processing and signal analysis on the fringe images of each row of pixel points to obtain a refractive phase map, a fringe density map and a transmittance map of each row of pixel points; A feature analysis module is configured to filter the fringe density map to obtain a low-frequency density map, compare difference characteristics of the fringe density map and the low-frequency density map to obtain a microlens distortion map, calculate a dynamic gain coefficient by using a change response characteristic between data values in the low-frequency density map and data values in the refractive phase map, and perform adaptive correction on the refractive phase map to obtain an anti-shake phase map, and analyze change difference conditions of the anti-shake phase maps of different rows to calculate a phase curvature map of a target row; A defect detection module is configured to fuse the transmittance map, the phase curvature map and the microlens distortion map of the target row, calculate a comprehensive defect score of each pixel point on the target row, and perform defect detection on the target row on the film according to the comprehensive defect score and the numerical characteristics in the transmittance map.

10. A polypropylene film defect detection apparatus characterized by comprising: The system comprises a processor and a memory, the memory stores at least one instruction, at least one program, a code set or an instruction set, and the processor loads and executes the at least one instruction, the at least one program, the code set or the instruction set to implement the steps of the polypropylene film defect detection method according to any one of claims 1-8.

Images