A method, system and device for visual inspection of defects in a car cover base film

Abstract

The application relates to the field of computer vision, in particular to a car clothes base film flaw visual detection method, system and equipment, the method comprising the following steps: determining the abnormality coefficient of each image according to the gray value and area of the abnormal area in each image, so as to obtain an abnormality coefficient sequence; analyzing the extreme value in the air pressure sequence to obtain a shock frequency sequence and a target air pressure sequence; determining the high-density area corresponding to each abnormality coefficient based on the distribution density and space weight of each data point in the two-dimensional section corresponding to the abnormality coefficient, and extracting the target data point of the high-density area; performing curve fitting on the real-time shock frequency sequence and the real-time target air pressure sequence to obtain a running trend curve; according to the order from small to large of the abnormality coefficient, performing curve fitting on the corresponding target data points to obtain a target trend curve; and outputting the corresponding detection result according to the similarity between the target trend curve and the running trend curve. The application can improve the accuracy of base film flaw identification.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and in particular to a method, system and device for visually detecting defects in car wrap base film. Background Technology

[0002] TPU (Thermoplastic Polyurethane) base film is the core layer of paint protection film, made of thermoplastic polyurethane. It has excellent elasticity, abrasion resistance, yellowing resistance, and self-healing properties, and is mainly produced through a casting process.

[0003] In the production process of car wrap base film, the air knife is responsible for emitting a stable and uniform high-speed airflow to press the high-temperature, viscous base film material onto the surface of the cooling roller, ensuring rapid and uniform cooling and shaping. However, during operation, the air knife is also prone to blockage due to the accumulation of impurities. The airflow speed and pressure at the blockage point decrease, causing the base film material in the corresponding area to not adhere tightly to the cooling roller. This results in uneven cooling of different parts of the base film material in the transverse direction, causing transverse stripes. Moreover, the blockage of the air knife cannot heal itself unless manually removed, and the degree of blockage will only become more and more serious. This will lead to more and more serious defects in the produced base film. Therefore, it is necessary to identify defects in the early stage of blockage, when the transverse stripes are still mild, in order to avoid losses and improve product quality. However, in the early stage of air knife blockage, the defects caused to the base film are very minor and do not have obvious image features, making it difficult to identify them online. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a method, system, and device for visual inspection of defects in automotive coating base film. The specific technical solution adopted is as follows:

[0005] In a first aspect, a method for visually detecting defects in automotive coating base film is provided, the method comprising:

[0006] During the transmission of the base film, multiple frames of images are continuously acquired. Based on the gray value and area of ​​the abnormal region in each frame, the abnormality coefficient of the image is determined to obtain the abnormality coefficient sequence. The abnormal region is obtained by thresholding the image.

[0007] The extreme values ​​in the pressure sequence are analyzed to obtain the corresponding oscillation frequency sequence and target pressure sequence. The pressure sequence is obtained by continuously acquiring the pressure value of the pressure stabilizing tank during the base membrane transfer process. The oscillation frequency sequence is calculated by the reciprocal of the difference between the timestamps corresponding to adjacent extreme values. The target pressure sequence is calculated by the average of adjacent extreme values.

[0008] In three-dimensional space, based on the distribution density and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted; the three-dimensional space is constructed based on the time-aligned anomaly coefficient sequence, oscillation frequency sequence and target air pressure sequence.

[0009] The real-time oscillation frequency sequence is curve-fitted with the real-time target air pressure sequence to obtain the running trend curve. According to the order of the anomaly coefficient from small to large, the corresponding target data points are curve-fitted to obtain the target trend curve.

[0010] Based on the similarity between the target trend curve and the running trend curve, the corresponding detection results are output.

[0011] Optionally, multiple frames of images are continuously acquired during the base film transfer process. Based on the grayscale value and area of ​​the abnormal region in each frame, the anomaly coefficient of the image is determined to obtain an anomaly coefficient sequence, including:

[0012] During the transmission of the base film, multiple frames of images are continuously acquired. Based on a preset threshold, each image is segmented to obtain the abnormal region of the image.

[0013] The thickness coefficient of each anomalous region is determined based on the difference between the average pixel value of each anomalous region and the average pixel value of the corresponding image.

[0014] Based on the area of ​​the largest abnormal region selected from multiple frames of images, each abnormal region is normalized to obtain the proportional weight of the abnormal region.

[0015] The anomaly index of an anomaly region is obtained by multiplying the proportional weight of each anomaly region by the thickness coefficient of that anomaly region.

[0016] The average of the anomaly indices of multiple abnormal regions in each frame of the image is calculated to obtain the anomaly coefficient of the image.

[0017] The anomaly coefficients of multiple frames of images are sorted in chronological order to obtain an anomaly coefficient sequence.

[0018] Optionally, the extreme values ​​in the pressure sequence are analyzed to obtain the corresponding oscillation frequency sequence and the target pressure sequence, including:

[0019] The pressure value of the pressure stabilizing tank is continuously acquired during the transfer of the base membrane to obtain the pressure sequence;

[0020] The gradient of the pressure sequence is calculated, and multiple extreme values ​​of the pressure sequence are determined based on the points where the sign of the gradient changes. The multiple extreme values ​​are then sorted in chronological order to obtain an extreme value sequence; the extreme values ​​include peak values ​​and valley values.

[0021] Calculate the reciprocal of the difference between the timestamps corresponding to each adjacent extreme value in the extreme value sequence to obtain the oscillation frequency, and thus obtain the oscillation frequency sequence;

[0022] The target air pressure is obtained by calculating the mean of each adjacent peak and valley in the extreme value sequence, thus obtaining the target air pressure sequence.

[0023] Optionally, in three-dimensional space, based on the distribution density and spatial weight of each data point within the two-dimensional cross-section corresponding to each anomaly coefficient, a high-density region corresponding to that anomaly coefficient is determined, and target data points in that high-density region are extracted, including:

[0024] A three-dimensional space is constructed based on the three-dimensional point set of the time-aligned anomaly coefficient sequence, oscillation frequency sequence, and target pressure sequence;

[0025] For each anomaly coefficient corresponding to a two-dimensional cross section, a circular region is constructed in the neighborhood with each data point in the two-dimensional cross section as the center point. The number of data points contained in the circular region under different radii is obtained, and the local density coefficient of the data point under the radius is calculated based on the inverse relationship between the number of data points and the radius.

[0026] Based on the local density coefficient and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted.

[0027] Optionally, based on the local density coefficient and spatial weight of each data point in the two-dimensional cross-section corresponding to each anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted, including:

[0028] For each anomaly coefficient corresponding to a two-dimensional cross section, the weighting coefficient of the anomaly coefficient is calculated based on the total number of data points in the two-dimensional cross section, the maximum value of the oscillation frequency in the two-dimensional cross section, and the maximum value of the target air pressure in the two-dimensional cross section.

[0029] Based on the local density coefficient of each data point in the two-dimensional cross section corresponding to each anomaly coefficient and the weight coefficient of the anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted.

[0030] Optionally, based on the local density coefficient of each data point in the two-dimensional cross-section corresponding to each anomaly coefficient and the weight coefficient of the anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted, including:

[0031] For each two-dimensional cross section corresponding to anomaly coefficient, the density coefficient of the data point at that radius is determined by multiplying the local density coefficient of each data point at different radii with the weight coefficient of the anomaly coefficient.

[0032] From the two-dimensional cross-section corresponding to each anomaly coefficient, select the largest density coefficient, and determine the circular area formed by the data point corresponding to the density coefficient and its radius as the high-density area, and determine the data point as the target data point of the high-density area.

[0033] Optionally, the real-time oscillation frequency sequence is curve-fitted with the real-time target pressure sequence to obtain the operating trend curve. Then, according to the order of anomaly coefficients from smallest to largest, the corresponding target data points are curve-fitted to obtain the target trend curve, including:

[0034] Arrange the target data points corresponding to the abnormal coefficients within the preset abnormal range according to the order of abnormal coefficients from smallest to largest, and establish a two-dimensional coordinate system with the oscillation frequency corresponding to each target data point as the horizontal axis and the target air pressure as the vertical axis. Fit each target data point according to curve fitting to obtain the target trend curve.

[0035] The real-time oscillation frequency sequence and the real-time target air pressure sequence, which are collected synchronously during the current production process, are aligned by timestamp, mapped to the two-dimensional coordinate system, and fitted using curve fitting to obtain the operating trend curve.

[0036] Optionally, based on the similarity between the target trend curve and the running trend curve, the corresponding detection results are output, including:

[0037] Based on the DTW algorithm, the similarity between the target trend curve and the running trend curve is calculated;

[0038] In response to the similarity being greater than a first preset value, a first detection result is output; the first detection result is used to indicate that the device has a risk of clogging.

[0039] In response to the similarity being greater than a first preset value and less than a second preset value, a second detection result is output; the second detection result is used to indicate that the device is blocked, and the second preset value is greater than the first preset value.

[0040] Secondly, a visual inspection system for defects in automotive coating base film is provided, the system comprising:

[0041] The acquisition module is used to continuously acquire multiple frames of images during the base film transfer process. Based on the gray value and area of ​​the abnormal region in each frame of image, the abnormality coefficient of the image is determined to obtain the abnormality coefficient sequence. The abnormal region is obtained by thresholding the image.

[0042] The analysis module is used to analyze the extreme values ​​in the pressure sequence to obtain the corresponding oscillation frequency sequence and target pressure sequence. The pressure sequence is obtained by continuously acquiring the pressure value of the pressure stabilizing tank during the base membrane transfer process. The oscillation frequency sequence is calculated by the reciprocal of the difference between the timestamps corresponding to adjacent extreme values. The target pressure sequence is calculated by the average of adjacent extreme values.

[0043] The determination module is used to determine the high-density region corresponding to each anomaly coefficient in three-dimensional space based on the distribution density and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, and to extract the target data points in the high-density region; the three-dimensional space is constructed based on the time-aligned anomaly coefficient sequence, oscillation frequency sequence and target air pressure sequence;

[0044] The fitting module is used to perform curve fitting between the real-time oscillation frequency sequence and the real-time target air pressure sequence to obtain the running trend curve. According to the order of the anomaly coefficient from small to large, the corresponding target data points are curve fitted to obtain the target trend curve.

[0045] The output module is used to output the corresponding detection results based on the similarity between the target trend curve and the running trend curve.

[0046] Thirdly, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein the processor executes the computer program to implement the method of the first aspect.

[0047] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this application.

[0048] This application offers the following advantages: it deeply correlates the visual features of the base film transmission image with the dynamic air pressure signal of the pressure stabilizing tank in the process equipment. By learning from historical data, a target trend curve representing the stable correlation between "defect degree - air pressure characteristics" is constructed in a three-dimensional feature space. In real-time detection, the oscillation frequency sequence and the target air pressure sequence are analyzed simultaneously to generate an operating trend curve. By calculating its similarity to the target trend curve, a quantitative assessment and early warning of the degree of air knife blockage is achieved. This overcomes the limitations of traditional pure visual inspection, which is highly dependent on image contrast and insensitive to weak early defects. It also overcomes the insufficiency of simply monitoring air pressure signals, which cannot directly correlate with product quality. It can issue early warnings before defects form obvious visual features, enabling proactive intervention. Through a data-driven correlation model, it effectively eliminates random interference, significantly reducing the false alarm rate. Furthermore, it provides a graded alarm mechanism, offering operators complete data support from early warning and auxiliary judgment to maintenance decisions, thereby significantly improving the intelligence level, product yield, and operational stability of the invisible car wrap base film production line. Attached Figure Description

[0049] To more clearly illustrate the technical solutions and advantages in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a method for visually detecting defects in a car wrap base film, as described in one embodiment.

[0051] Figure 2 This is a schematic diagram of the structure of a visual inspection system for defects in a car wrap base film in one embodiment;

[0052] Figure 3 This is a schematic diagram of the structure of an electronic device in one embodiment. Detailed Implementation

[0053] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a visual inspection method, system, and device for defects in car wrap base film proposed in this application. 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.

[0054] 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 application pertains.

[0055] The following, with reference to the accompanying drawings, details a specific scheme for a visual inspection method for defects in car wrap base film provided in this application. For example... Figure 1 As shown, the method includes:

[0056] S11. During the transmission of the base film, multiple frames of images are continuously acquired. Based on the gray value and area of ​​the abnormal region in each frame of image, the abnormality coefficient of the image is determined to obtain the abnormality coefficient sequence.

[0057] The abnormal regions are obtained by thresholding the image.

[0058] The steps for collecting and processing data from the cooling rollers are as follows:

[0059] (1) After the base film material has completely left the cooling roller, the equipment is arranged between the conveying area and the first guide roller: (a) A high-brightness, uniformly illuminated LED light source is arranged below the conveying area. The width of the light source should be greater than the width of the base film material to ensure consistent illumination in the detection area; (b) A linear array camera with a linear speed greater than 200 meters per minute is arranged above the conveying area to acquire the transmission image of the cooled base film material.

[0060] (2) Couple the rotary encoder to the main drive roller of the production line to ensure that the camera capture frame is strictly synchronized with the film displacement, thus eliminating image ghosting or stretching.

[0061] The pressure stabilizing tank connected to the air knife nozzle is responsible for controlling the flow rate and air pressure of the blown air knife. When the air knife nozzle is blocked, the sensor at the blockage point detects a pressure drop and increases the back pressure inside the tank to increase the pressure at the air knife nozzle. Since the defects generated on the base film are related to the blockage of the air knife nozzle, the defects are related to the pressure change inside the pressure stabilizing tank. By detecting and analyzing the pressure change inside the pressure stabilizing tank, the defects can be detected.

[0062] The uneven horizontal stripes on the base film surface are caused by the different air pressures at different locations of the air knife. The greater the degree of blockage, the greater the difference in air pressure, and the more severe the defects. The pressure inside the pressure stabilizing tank can be directly quantified. Therefore, in order to further analyze the relationship between the two, it is necessary to quantify the severity of the defects.

[0063] When horizontal stripes appear on the base film, the thickness at that location will be greater than that of the normal area. When light shines on the area where horizontal stripes appear, the thicker area will appear darker in the image. Therefore, in the transmission image, the severity of the defect is related to the pixel grayscale of the image. The smaller the grayscale, the deeper the defect. When the equipment is working normally, the thickness of the base film produced is uniform, and it appears as a uniform grayscale in the transmission image.

[0064] Therefore, in one embodiment, multiple frames of images are continuously acquired during the transfer of the base film. Based on the grayscale value and area of ​​the abnormal region in each frame, the abnormality coefficient of the image is determined to obtain an abnormality coefficient sequence, including:

[0065] During the transmission of the base film, multiple frames of images are continuously acquired. Based on a preset threshold, each image is segmented to obtain the abnormal region of the image.

[0066] The thickness coefficient of each anomalous region is determined based on the difference between the average pixel value of each anomalous region and the average pixel value of the corresponding image.

[0067] Based on the area of ​​the largest abnormal region selected from multiple frames of images, each abnormal region is normalized to obtain the proportional weight of the abnormal region.

[0068] The anomaly index of an anomaly region is obtained by multiplying the proportional weight of each anomaly region by the thickness coefficient of that anomaly region.

[0069] The average of the anomaly indices of multiple abnormal regions in each frame of the image is calculated to obtain the anomaly coefficient of the image.

[0070] The anomaly coefficients of multiple frames of images are sorted in chronological order to obtain an anomaly coefficient sequence.

[0071] It should be noted that, except for data explicitly specified as being collected in real time, the abnormal coefficient sequence, oscillation frequency sequence, and target air pressure sequence used in this application to construct the target trend curve are all based on data recorded during historical production processes.

[0072] The average gray level generated by the base film is calculated from grayscale images captured when the historical equipment was working normally. Meanwhile, the maximum value in the graph under normal conditions As a preset threshold for segmentation, specifically, during the system initialization or calibration phase, under stable production conditions where the equipment is flawless and the air knife is working normally, multiple frames of base film transmission images are acquired, and the maximum pixel grayscale value in all normal images is counted. This is used as a preset threshold. In historical images that produced anomalies, the preset threshold is used for segmentation, where pixel values ​​are less than [a certain threshold value]. The region is designated as an abnormal region (defective region). Calculate the average pixel value of a given abnormal region. The greater the deviation of the average pixel grayscale of the abnormal area from that of the normal area, the more severe the defect (abnormality). Therefore, the thickness coefficient of the abnormal area... The calculation formula is: , This represents the average pixel value of the abnormal region. This represents the average pixel value of the image to which the abnormal region belongs.

[0073] The severity of defects is related not only to thickness but also to the size of the affected area. All areas of abnormal thickness in historical images are segmented as defective regions. The size of the defect is represented by the pixel area occupied by the defective region in the image. The pixel area of ​​the largest defective region across multiple frames is then selected. , To determine the maximum statistical value of the area of ​​all abnormal regions within a preset historical statistical period (e.g., the past 24 hours or the previous production batch), the proportional weight of the abnormal regions is... The calculation formula is: ;in, The proportional weight for each abnormal region, The area of ​​the abnormal region is divided by the maximum area value to normalize the thickness coefficient.

[0074] For two defective areas of unequal size, the one with the larger area tends to have a greater thickness. This is because the thickness of the base film is caused by the varying degrees of adhesion to the cooling roller. The larger the area of ​​incomplete adhesion, the greater the thickness caused by necking. These two factors are correlated. Therefore, the degree of abnormality of a defect is represented by the product of two coefficients. The abnormality index for each abnormal area... The calculation formula is: ;in, Anomaly index for each anomalous region, Here, represents the thickness coefficient of the anomalous region, and represents the proportional weight of the anomalous region. and proportional weight After normalization, the two coefficients take values ​​in the range [0,1], therefore the anomaly index... The region range is [0,1], and both the thickness coefficient and the proportional weight are positively correlated with the severity of the defect. The closer it is to 1, the more serious the defect it represents.

[0075] Calculate the mean of the anomaly indices of multiple anomalous regions in the a-th frame image to obtain the anomaly coefficient of the a-th frame image. Let k be the total number of frames. The anomaly coefficients of the frames are sorted chronologically to obtain the anomaly coefficient sequence. The expression is .

[0076] S12. Analyze the extreme values ​​in the pressure sequence to obtain the corresponding oscillation frequency sequence and target pressure sequence.

[0077] Among them, the pressure sequence is obtained by continuously acquiring the pressure value of the pressure stabilizing tank during the base membrane transmission process, the oscillation frequency sequence is calculated by the reciprocal of the difference between the timestamps corresponding to adjacent extreme values, and the target pressure sequence is calculated by the average of adjacent extreme values.

[0078] Pressure stabilizing tanks typically have a sensitive pressure feedback system. When the air pressure at the blade edge weakens, the system increases the exhaust to compensate. However, the pressure in the unblocked area will rise sharply. The PID algorithm (Proportional-Integral-Derivative Control) will then control the reduction of air pressure. Because the blockage persists, the air pressure will oscillate back and forth. This oscillation will change according to the degree of blockage at the blade edge, and the degree of blockage at the blade edge is related to the severity of the base film at the production site during this period.

[0079] When the air inlet is blocked, causing the air pressure inside the pressure stabilizing tank to change, the pointer of the internal pressure gauge in the pressure stabilizing tank will swing back and forth between the optimal air pressure value. For the same equipment, its mechanical state and the PID algorithm embedded in the equipment are the same. Therefore, under the same degree of blockage of the air inlet, the swing amplitude and frequency of the pressure gauge pointer are the same. So, the change in the value of the pressure gauge in the pressure stabilizing tank can be observed to determine its blockage status with the air inlet, thereby indirectly determining the degree of defect caused by the blockage.

[0080] Because industrial equipment collects discrete data in the form of interval sampling, and pressure stabilizing tanks are key process equipment, they use high frequency for data recording, with a typical sampling interval of one hundred milliseconds. Therefore, the air pressure change data obtained from the database is discrete data collected at 0.1-second intervals.

[0081] Because there is a physical distance between the air knife edge (the source of abnormal air pressure) and the linear scan camera (the defect detection point), the speed of base film delivery needs to be considered. and distance Calculate the delay time The moment. Acquired image data and time The collected air pressure data were aligned.

[0082] In one embodiment, the extreme values ​​in the pressure sequence are analyzed to obtain the corresponding oscillation frequency sequence and the target pressure sequence, including:

[0083] The pressure value of the pressure stabilizing tank is continuously acquired during the transfer of the base membrane to obtain the pressure sequence;

[0084] The gradient of the pressure sequence is calculated, and multiple extreme values ​​of the pressure sequence are determined based on the points where the sign of the gradient changes. The multiple extreme values ​​are then sorted in chronological order to obtain an extreme value sequence; the extreme values ​​include peak values ​​and valley values.

[0085] Calculate the reciprocal of the difference between the timestamps corresponding to each adjacent extreme value in the extreme value sequence to obtain the oscillation frequency, and thus obtain the oscillation frequency sequence;

[0086] The target air pressure is obtained by calculating the mean of each adjacent peak and valley in the extreme value sequence, thus obtaining the target air pressure sequence.

[0087] Search the equipment operation database for changes in the pressure tank pressure during the period from the appearance of defects in the base membrane to the shutdown of the equipment, and record the pressure values. If the changes are represented as a sequence, then the pressure sequence is... for , For the first The air pressure value of the second data collection. This refers to the number of times data is collected from the occurrence of a defect to the time the equipment is shut down.

[0088] When the gas pressure inside the pressure vessel oscillates, its numerical change presents a wave with an unstable amplitude at a certain frequency. Therefore, there are several maxima and minima in the gas pressure sequence, which are the positions of wave peaks and troughs. The time interval between adjacent wave peaks can be used to represent the frequency of gas pressure oscillation.

[0089] The gradient of the barometric pressure sequence is calculated using the one-dimensional Canny algorithm. The points where the gradient values ​​change are the extreme values ​​in the sequence. The elements containing the extreme values ​​are then stored in the extreme value sequence in a table. , , For sequence The Middle The extreme values ​​are at The position index in the middle, This represents the number of times a peak and a trough occur.

[0090] Because the pressure changes in the pressure tank exhibit irregular and non-periodic waves, and to perform more detailed statistical analysis of the data, the frequency is calculated using the interval between adjacent peaks and troughs. Since wave peaks and troughs alternate, the calculation sequence... The difference sequence is the time interval between adjacent peaks and troughs, and the reciprocal of the time interval is the frequency of one change.

[0091] Calculate the difference between the timestamps corresponding to each adjacent extreme value to obtain a difference sequence. Use the reciprocal of each element in the difference sequence as the frequency of air pressure change, and denote this as an oscillation frequency sequence. , , For the first The peak represented by the element and the first element Each element represents the frequency of change between troughs.

[0092] The pressure fluctuations inside the pressure tank are caused by the lag in the pressure compensation system. Simply put, the pressure sensor at the air inlet, due to blockage, sends a reduced pressure signal back to the control system. At this point, a PID algorithm adjusts the tank pressure to compensate for the pressure at the air inlet. However, because the machine's operation is not as sensitive as the algorithm's calculation, it exhibits a certain lag. When increasing the tank pressure to compensate for the air pressure at the air inlet, once the correct pressure value is reached, the algorithm issues a stop command. Upon receiving the command, the machine closes the control valve, stopping the air pump. During the closing process, the machine continues to operate for a period, causing the tank pressure to rise further, exceeding the algorithm's calculated pressure value. At this point, the algorithm again controls the reduction of the tank pressure, thus creating a cycle. Furthermore, the degree of blockage at the air inlet changes, causing the algorithm's calculated optimal tank pressure value to change accordingly, resulting in the tank pressure constantly fluctuating around the optimal pressure. Therefore, it can be considered that... The midpoint between adjacent peaks and troughs is close to the optimal air pressure value calculated by the algorithm at that time. The optimal air pressure value calculated by the algorithm is highly correlated with the degree of blockage at that time. Since the degree of blockage is related to the degree of defect, the optimal air pressure value can be used as the standard for judging the air pressure change and the degree of defect.

[0093] The average of the pressure values ​​of adjacent peaks and troughs is taken as the optimal pressure value (target pressure) calculated by the algorithm between the two, and this value is recorded as the target pressure sequence. , , For the first The value of the first peak and the first The optimal air pressure value between the two is calculated by averaging the values ​​of the troughs.

[0094] The above method calculates the frequency of air pressure oscillation and the optimal air pressure value during the process from the moment the defect appears to the machine shutdown. However, the changes in these two values ​​are not directly related to the degree of defect. It is necessary to analyze the data changes in multiple processes. If the air pressure oscillation frequency and the optimal air pressure value show the same value at the same degree of defect in different processes, it indicates that the air pressure oscillation frequency and the optimal air pressure value are highly correlated and can be used to represent the degree of defect.

[0095] S13. In three-dimensional space, based on the distribution density and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, determine the high-density region corresponding to the anomaly coefficient, and extract the target data points in the high-density region.

[0096] The three-dimensional space is constructed based on time-aligned anomaly coefficient sequences, oscillation frequency sequences, and target pressure sequences.

[0097] During the production of TPU base film, the base film is conveyed on the guide rollers at a speed of 40 meters per minute. This speed is far lower than the data acquisition frequency of the equipment (0.1 seconds), so in terms of degree sequence... Every time a flaw occurs in the sequence and The corresponding air pressure frequency value and target air pressure can be found in the data. Therefore, each time a defect is detected, a three-dimensional point coordinate system can be formed by the anomaly coefficient, oscillation frequency and target air pressure. By projecting the three-dimensional point coordinates corresponding to the defect time in the whole process into the three-dimensional space, a point set can be obtained. By projecting the three-dimensional point sets of defect times in different processes into the three-dimensional space, the changes in oscillation frequency and target air pressure corresponding to different defect degrees can be further obtained.

[0098] In different processes, the anomaly coefficients and corresponding oscillation frequencies and target air pressures at the moments when defects occur are calculated. The resulting three-dimensional point sets are then projected into a three-dimensional space with the anomaly coefficients, oscillation frequencies, and target air pressures as dimensions. In other words, a three-dimensional space is constructed based on the three-dimensional point sets of the time-aligned anomaly coefficient sequence, oscillation frequency sequence, and target air pressure sequence.

[0099] In one embodiment, in three-dimensional space, based on the distribution density and spatial weight of each data point within a two-dimensional cross-section corresponding to each anomaly coefficient, a high-density region corresponding to that anomaly coefficient is determined, and target data points in that high-density region are extracted, including:

[0100] A three-dimensional space is constructed based on the three-dimensional point set of the time-aligned anomaly coefficient sequence, oscillation frequency sequence, and target pressure sequence;

[0101] For each anomaly coefficient corresponding to a two-dimensional cross section, a circular region is constructed in the neighborhood with each data point in the two-dimensional cross section as the center point. The number of data points contained in the circular region under different radii is obtained, and the local density coefficient of the data point under the radius is calculated based on the inverse relationship between the number of data points and the radius.

[0102] Based on the local density coefficient and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted.

[0103] Specifically, based on the local density coefficient and spatial weight of each data point in the two-dimensional cross-section corresponding to each anomaly coefficient, the high-density region corresponding to that anomaly coefficient is determined, and the target data points in that high-density region are extracted, including:

[0104] For each anomaly coefficient corresponding to a two-dimensional cross section, the weighting coefficient of the anomaly coefficient is calculated based on the total number of data points in the two-dimensional cross section, the maximum value of the oscillation frequency in the two-dimensional cross section, and the maximum value of the target air pressure in the two-dimensional cross section.

[0105] Based on the local density coefficient of each data point in the two-dimensional cross section corresponding to each anomaly coefficient and the weight coefficient of the anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted.

[0106] Specifically, based on the local density coefficient of each data point in the two-dimensional cross-section corresponding to each anomaly coefficient and the weight coefficient of that anomaly coefficient, the high-density region corresponding to that anomaly coefficient is determined, and the target data points in that high-density region are extracted, including:

[0107] For each two-dimensional cross section corresponding to anomaly coefficient, the density coefficient of the data point at that radius is determined by multiplying the local density coefficient of each data point at different radii with the weight coefficient of the anomaly coefficient.

[0108] From the two-dimensional cross-section corresponding to each anomaly coefficient, select the largest density coefficient, and determine the circular area formed by the data point corresponding to the density coefficient and its radius as the high-density area, and determine the data point as the target data point of the high-density area.

[0109] Under the same anomaly coefficient (defect degree), the oscillation frequency and target air pressure will be distributed in a two-dimensional plane. Because the equipment environment is different at the time of defect generation, the points in the plane may not be concentrated. At this time, it is necessary to judge the distribution density of spatial points. Points in a high-density area indicate that the oscillation frequency and target air pressure are more likely to be distributed in this area under the same defect degree in different equipment environments. Furthermore, the correlation between the two air pressure change coefficients and the generated anomaly coefficient is relatively large in a high-density area. Therefore, in three-dimensional space, the denser the distribution of spatial points in the two-dimensional cross-section obtained by dimensionality reduction of different anomaly coefficients, the more likely the data is to be related to the defect degree.

[0110] In a two-dimensional cross-section with a fixed anomaly coefficient, the point density of the region can be calculated by gradually expanding the radius of a circle with one spatial point as the center. As the circular region increases, the increase in radius is inversely proportional to the density of the region, while the number of points contained in the region is directly proportional to the density. The density of the region can be represented by the product of the radius and the number of points contained. This coefficient has only one peak value, and the peak value is the critical value of the radius expansion. The circle formed by the radius and the center is the high-density region.

[0111] Specifically, in a two-dimensional cross-section with a fixed degree of defect, a data point is randomly selected, and a circle is drawn with that data point as the center and a radius of... The region is formed into a circular area, and the number of data points contained in the region is... Then the local density coefficient of the circular region can be obtained. Local density coefficient The calculation formula is: .

[0112] As the radius increases, the density coefficient will reach a peak. When this peak is the maximum value and the coefficient changes by first increasing and then decreasing (to avoid the situation where the function keeps decreasing and the initial point is the maximum value, rendering the function ineffective), the position of this peak can represent the critical value of radius expansion. Therefore, the contribution of the coefficient of the number of points included in the expansion of the region area is greater than that of the radius coefficient. Thus, it is necessary to add weights to the spatial points, and the weight value should be related to the number of points in the two-dimensional space. When the number of points in the two-dimensional space is large, the contribution of each point is small, and when the number of points in the two-dimensional space is small, the contribution of each point is large.

[0113] In a two-dimensional cross section with a fixed anomaly coefficient, the total number of data points contained in that cross section is counted. At the same time, the maximum value of the oscillation frequency in this cross section is calculated. and the maximum value of the target air pressure. Calculate the weighting coefficients of the data points in this cross section. Weighting coefficient The calculation formula is: .

[0114] Calculate the weighting coefficients With local density coefficient The product of these two factors yields the density coefficient of a circular region centered at any data point in the two-dimensional cross-section. Density coefficient The calculation formula is: in, Density coefficients for each data point at each radius. This represents the local density coefficient of the data point within this radius. This is the weighting coefficient of the outlier coefficient corresponding to this data point.

[0115] S14. Perform curve fitting between the real-time oscillation frequency sequence and the real-time target air pressure sequence to obtain the running trend curve. According to the order of the anomaly coefficient from small to large, perform curve fitting on the corresponding target data points to obtain the target trend curve.

[0116] Anomaly coefficients within an anomaly interval are selected for analysis in three-dimensional space; the anomaly interval is... Within the specified interval, in ascending order of anomaly coefficient, the high-density regions of each data point in the two-dimensional cross-section under each anomaly coefficient are calculated. Specifically:

[0117] (a) Traverse all data points in the first two-dimensional cross-section with the smallest anomaly coefficient, and calculate the density coefficient of the circular region centered at that data point. The center point of the circular region with the highest density coefficient in the two-dimensional cross section is taken as the target data point of the plane. This point is a two-dimensional point (excluding outlier coefficients). In addition to the calculation method based on the density of the circular region, clustering algorithms such as DBSCAN clustering and Gaussian mixture model (GMM) can also be used to extract the center point of the high-density region as the target data point.

[0118] (b) When calculating the target data points of the two-dimensional cross section corresponding to the second anomaly coefficient and subsequent two-dimensional cross sections, start traversing from the point adjacent to the previous target data point until all target data points in the interval are calculated.

[0119] (c) Store the target data points of each two-dimensional cross-section into an ordered point set according to the calculation order. Establish a two-dimensional coordinate system with the oscillation frequency corresponding to each target data point as the horizontal axis and the target air pressure as the vertical axis. Perform curve fitting on the target data points in the point set to obtain a two-dimensional curve, that is, the target trend curve of air pressure oscillation frequency and target air pressure when defects occur. The fitting method can be cubic spline interpolation. If the horizontal axis (abnormal coefficient) is not monotonic during fitting, piecewise linear interpolation or table lookup can be used.

[0120] Besides the blockage of the blade, the instantaneous action of electrical components and the residual liquid in the pipeline can also cause the pressure of the pressure stabilizing tank to change and cause pressure oscillations. Therefore, long-term observation is required. When the change of the pressure stabilizer shows a high similarity to the target trend curve, it is considered that the base film has defects caused by the air blade.

[0121] The length of the trend curve is fixed, but when observing the equipment, in the initial stage, the length of the curve fitted from the equipment operating data is constantly increasing. The two lengths are not aligned, so a time warping algorithm is needed to calculate the similarity between the equipment operating data and the trend curve, thereby determining whether there are defects in the base film at the current moment.

[0122] When the pressure in the pressure tank begins to change, real-time data is recorded. The real-time oscillation frequency sequence and the real-time target pressure sequence are aligned by timestamps, mapped to a two-dimensional coordinate system, and curve fitting is used to obtain an operating trend curve. .

[0123] Running trend curve With the target trend curve Both are mapped to the same two-dimensional coordinate system (horizontal axis is oscillation frequency, vertical axis is target air pressure) to ensure comparability in the same physical space; among them, the target trend curve Based on feature points extracted chronologically from typical historical production processes, a typical evolutionary path of air knife tip blockage is constructed. The temporal information is only used to determine the order of feature points and does not participate in real-time matching; while the operational trend curve... The time series is obtained by strictly aligning the real-time oscillation frequency and the real-time target air pressure value collected synchronously during the current production process with the timestamps, forming a point series, and then by curve fitting. Its time series completely preserves the dynamic response process of the current air pressure system. Although the two curves have different sources (one is an offline reference and the other is an online trajectory), they share a unified coordinate system and are both generated according to their respective time evolution order. Therefore, the consistency of the shape can be effectively compared by similarity algorithms (such as DTW) without relying on absolute time alignment or consistent sampling intervals.

[0124] S15. Output the corresponding detection results based on the similarity between the target trend curve and the running trend curve.

[0125] In one embodiment, based on the similarity between the target trend curve and the running trend curve, the corresponding detection result is output, including:

[0126] Based on the DTW algorithm, the similarity between the target trend curve and the running trend curve is calculated;

[0127] In response to the similarity being greater than a first preset value, a first detection result is output; the first detection result is used to indicate that the device has a risk of clogging.

[0128] In response to the similarity being greater than a first preset value and less than a second preset value, a second detection result is output; the second detection result is used to indicate that the device is blocked, and the second preset value is greater than the first preset value.

[0129] The first and second preset values ​​can be set according to the actual situation. For example, the first preset value is 0.5 and the second preset value is 0.7.

[0130] The above steps quantified the severity of the defect of transverse stripes on the base film. Further analysis of the correlation between the change in pressure in the pressure stabilizing tank and the degree of defect during defect formation yielded a trend curve. The similarity between the target trend curve and the operating trend curve during equipment operation was used to determine whether the currently produced base film had defects. The DTW algorithm was used to calculate the operating trend curve. With the target trend curve similarity between The process of feeding back the test results to the operators for processing is as follows:

[0131] ①. When calculated If the air knife nozzle is suspected of being blocked, the equipment needs to be inspected by staff. The first test result is output, and the data is returned to the operation platform, alerting the staff. Staff then monitor subsequent data changes and inspect the equipment. Near the air knife nozzle, staff analyze whether the sound emitted during airflow differs from usual, and simultaneously use a flashlight to illuminate the base film surface to observe whether the surface reflection is smooth. This is a non-contact inspection of the machine.

[0132] ②. When When staff observe an abnormality in the equipment, they assume that the air knife nozzle is blocked and the production base film is defective. They then output a second test result, shut down the equipment, clear the air knife nozzle, and record the data of the abnormality stage.

[0133] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0134] This application also provides a visual inspection system for defects in car wrap base film, such as Figure 2 As shown, the system includes:

[0135] The acquisition module 21 is used to continuously acquire multiple frames of images during the transmission of the base film. Based on the gray value and area of ​​the abnormal region in each frame of the image, the abnormal coefficient of the image is determined to obtain the abnormal coefficient sequence. The abnormal region is obtained by thresholding the image.

[0136] Analysis module 22 is used to analyze the extreme values ​​in the pressure sequence to obtain the corresponding oscillation frequency sequence and target pressure sequence. The pressure sequence is obtained by continuously acquiring the pressure value of the pressure stabilizing tank during the base membrane transfer process. The oscillation frequency sequence is calculated by the reciprocal of the difference between the timestamps corresponding to adjacent extreme values. The target pressure sequence is calculated by the average of adjacent extreme values.

[0137] The determination module 23 is used to determine the high-density region corresponding to each anomaly coefficient in three-dimensional space based on the distribution density and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, and to extract the target data points in the high-density region; the three-dimensional space is constructed based on the time-aligned anomaly coefficient sequence, oscillation frequency sequence and target air pressure sequence.

[0138] The fitting module 24 is used to perform curve fitting between the real-time oscillation frequency sequence and the real-time target air pressure sequence to obtain the running trend curve. According to the order of the anomaly coefficient from small to large, the corresponding target data points are curve fitted to obtain the target trend curve.

[0139] Output module 25 is used to output the corresponding detection results based on the similarity between the target trend curve and the running trend curve.

[0140] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs.

[0141] Figure 3 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the method described in any of the above embodiments. Figure 3 The electronic device 30 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0142] like Figure 3 As shown, the electronic device 30 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0143] Bus 33 includes a data bus, an address bus, and a control bus.

[0144] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.

[0145] The memory 32 may also include a program tool 325 (or utility) having a set (at least one) program module 324, such program module 324 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0146] The processor 31 executes various functional applications and data processing, such as the methods provided in any of the above embodiments, by running computer programs stored in the memory 32.

[0147] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed via input / output (I / O) interface 35. Furthermore, electronic device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 36. As shown, network adapter 36 communicates with other modules of electronic device 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0148] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0149] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in any of the above embodiments.

[0150] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0151] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0152] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the method described in any of the above embodiments.

[0153] The program code for executing the computer program product of this application can be written in any combination of one or more programming languages. The program code can be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0154] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0155] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

[0156] 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 for visually inspecting defects in car wrap base film, characterized in that, The method includes: During the transmission of the base film, multiple frames of images are continuously acquired. Based on the gray value and area of ​​the abnormal region in each frame, the abnormality coefficient of the image is determined to obtain the abnormality coefficient sequence. The abnormal region is obtained by thresholding the image. The extreme values ​​in the pressure sequence are analyzed to obtain the corresponding oscillation frequency sequence and target pressure sequence. The pressure sequence is obtained by continuously acquiring the pressure value of the pressure stabilizing tank during the base membrane transfer process. The oscillation frequency sequence is calculated by the reciprocal of the difference between the timestamps corresponding to adjacent extreme values. The target pressure sequence is calculated by the average of adjacent extreme values. In three-dimensional space, based on the distribution density and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted; the three-dimensional space is constructed based on the time-aligned anomaly coefficient sequence, oscillation frequency sequence and target air pressure sequence. The real-time oscillation frequency sequence is curve-fitted with the real-time target air pressure sequence to obtain the running trend curve. According to the order of the anomaly coefficient from small to large, the corresponding target data points are curve-fitted to obtain the target trend curve. Based on the similarity between the target trend curve and the running trend curve, the corresponding detection results are output.

2. The method for visually inspecting defects in a car wrap base film as described in claim 1, characterized in that, The process of continuously acquiring multiple frames of images during the base film transfer, and determining the anomaly coefficient of each image based on the grayscale value and area of ​​the abnormal region in each frame to obtain an anomaly coefficient sequence, includes: During the transmission of the base film, multiple frames of images are continuously acquired. Based on a preset threshold, each image is segmented to obtain the abnormal region of the image. The thickness coefficient of each anomalous region is determined based on the difference between the average pixel value of each anomalous region and the average pixel value of the corresponding image. Based on the area of ​​the largest abnormal region selected from multiple frames of images, each abnormal region is normalized to obtain the proportional weight of the abnormal region. The anomaly index of an anomaly region is obtained by multiplying the proportional weight of each anomaly region by the thickness coefficient of that anomaly region. The average of the anomaly indices of multiple abnormal regions in each frame of the image is calculated to obtain the anomaly coefficient of the image. The anomaly coefficients of multiple frames of images are sorted in chronological order to obtain an anomaly coefficient sequence.

3. The method for visually inspecting defects in a car wrap base film as described in claim 1, characterized in that, The analysis of extreme values ​​in the pressure sequence to obtain the corresponding oscillation frequency sequence and target pressure sequence includes: The pressure value of the pressure stabilizing tank is continuously acquired during the transfer of the base membrane to obtain the pressure sequence; The gradient of the pressure sequence is calculated, and multiple extreme values ​​of the pressure sequence are determined based on the points where the sign of the gradient changes. The multiple extreme values ​​are then sorted in chronological order to obtain an extreme value sequence; the extreme values ​​include peak values ​​and valley values. Calculate the reciprocal of the difference between the timestamps corresponding to each adjacent extreme value in the extreme value sequence to obtain the oscillation frequency, and thus obtain the oscillation frequency sequence; The target air pressure is obtained by calculating the mean of each adjacent peak and valley in the extreme value sequence, thus obtaining the target air pressure sequence.

4. The method for visually inspecting defects in a car wrap base film as described in claim 1, characterized in that, In three-dimensional space, based on the distribution density and spatial weight of each data point within a two-dimensional cross-section corresponding to each anomaly coefficient, the high-density region corresponding to that anomaly coefficient is determined, and the target data points of that high-density region are extracted, including: A three-dimensional space is constructed based on the three-dimensional point set of the time-aligned anomaly coefficient sequence, oscillation frequency sequence, and target pressure sequence; For each anomaly coefficient corresponding to a two-dimensional cross section, a circular region is constructed in the neighborhood with each data point in the two-dimensional cross section as the center point. The number of data points contained in the circular region under different radii is obtained, and the local density coefficient of the data point under the radius is calculated based on the inverse relationship between the number of data points and the radius. Based on the local density coefficient and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted.

5. The method for visually inspecting defects in a car wrap base film as described in claim 4, characterized in that, The step of determining the high-density region corresponding to each anomaly coefficient based on the local density coefficient and spatial weight of each data point in the two-dimensional cross-section corresponding to each anomaly coefficient, and extracting the target data points in the high-density region, includes: For each anomaly coefficient corresponding to a two-dimensional cross section, the weighting coefficient of the anomaly coefficient is calculated based on the total number of data points in the two-dimensional cross section, the maximum value of the oscillation frequency in the two-dimensional cross section, and the maximum value of the target air pressure in the two-dimensional cross section. Based on the local density coefficient of each data point in the two-dimensional cross section corresponding to each anomaly coefficient and the weight coefficient of the anomaly coefficient, the high-density region corresponding to the anomaly coefficient is determined, and the target data points in the high-density region are extracted.

6. The method for visually inspecting defects in a car wrap base film as described in claim 5, characterized in that, The step of determining the high-density region corresponding to each anomaly coefficient based on the local density coefficient of each data point in the two-dimensional cross-section corresponding to each anomaly coefficient and the weight coefficient of the anomaly coefficient, and extracting the target data points in the high-density region, includes: For each two-dimensional cross section corresponding to anomaly coefficient, the density coefficient of the data point at that radius is determined by multiplying the local density coefficient of each data point at different radii with the weight coefficient of the anomaly coefficient. From the two-dimensional cross-section corresponding to each anomaly coefficient, select the largest density coefficient, and determine the circular area formed by the data point corresponding to the density coefficient and its radius as the high-density area, and determine the data point as the target data point of the high-density area.

7. The method for visually inspecting defects in a car wrap base film as described in claim 1, characterized in that, The process of curve fitting the real-time oscillation frequency sequence with the real-time target air pressure sequence to obtain the operating trend curve, and curve fitting the corresponding target data points according to the order of anomaly coefficient from smallest to largest to obtain the target trend curve, includes: Arrange the target data points corresponding to the abnormal coefficients within the preset abnormal range according to the order of abnormal coefficients from smallest to largest, and establish a two-dimensional coordinate system with the oscillation frequency corresponding to each target data point as the horizontal axis and the target air pressure as the vertical axis. Fit each target data point according to curve fitting to obtain the target trend curve. The real-time oscillation frequency sequence and the real-time target air pressure sequence, which are collected synchronously during the current production process, are aligned by timestamp, mapped to the two-dimensional coordinate system, and fitted using curve fitting to obtain the operating trend curve.

8. The method for visually inspecting defects in a car wrap base film as described in claim 1, characterized in that, The step of outputting corresponding detection results based on the similarity between the target trend curve and the running trend curve includes: Based on the DTW algorithm, the similarity between the target trend curve and the running trend curve is calculated; In response to the similarity being greater than a first preset value, a first detection result is output; the first detection result is used to indicate that the device has a risk of clogging. In response to the similarity being greater than a first preset value and less than a second preset value, a second detection result is output; the second detection result is used to indicate that the device is blocked, and the second preset value is greater than the first preset value.

9. A visual inspection system for defects in car wrap base film, characterized in that, The system includes: The acquisition module is used to continuously acquire multiple frames of images during the base film transfer process. Based on the gray value and area of ​​the abnormal region in each frame of image, the abnormality coefficient of the image is determined to obtain the abnormality coefficient sequence. The abnormal region is obtained by thresholding the image. The analysis module is used to analyze the extreme values ​​in the pressure sequence to obtain the corresponding oscillation frequency sequence and target pressure sequence. The pressure sequence is obtained by continuously acquiring the pressure value of the pressure stabilizing tank during the base membrane transfer process. The oscillation frequency sequence is calculated by the reciprocal of the difference between the timestamps corresponding to adjacent extreme values. The target pressure sequence is calculated by the average of adjacent extreme values. The determination module is used to determine the high-density region corresponding to each anomaly coefficient in three-dimensional space based on the distribution density and spatial weight of each data point in the two-dimensional cross section corresponding to each anomaly coefficient, and to extract the target data points in the high-density region; the three-dimensional space is constructed based on the time-aligned anomaly coefficient sequence, oscillation frequency sequence and target air pressure sequence; The fitting module is used to perform curve fitting between the real-time oscillation frequency sequence and the real-time target air pressure sequence to obtain the running trend curve. According to the order of the anomaly coefficient from small to large, the corresponding target data points are curve fitted to obtain the target trend curve. The output module is used to output the corresponding detection results based on the similarity between the target trend curve and the running trend curve.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 8.

Images