Film gluing quality detection system and detection method based on visual detection

Abstract

The invention relates to the technical field of film production, and particularly discloses a film gluing quality detection system and detection method based on visual detection. A light source module; an image preprocessing module; an image splicing module; a defect detection and identification module; a motion control module; a data management module; the man-machine interaction module is electrically connected with the image acquisition module, the light source module, the image preprocessing module, the defect detection and recognition module, the image splicing module, the motion control module and the data management module; through cooperation of the image acquisition module, the light source module, the image preprocessing module, the image splicing module, the defect detection and identification module, the motion control module, the data management module and the man-machine interaction module, efficient and high-precision detection of the film gluing quality is realized, the possibility of defect omission is reduced, and the detection efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of film production, and particularly relates to a film laminating quality detection system and a detection method based on visual detection. Background Art

[0002] A film, also known as a film or a soft film, is a traditional image recording material based on silver salt photosensitive technology. In film production, in order to improve the durability and visual effect of the film, the film needs to be laminated. By covering the surface of the film with glue or a film, it can prevent the film from physical damage, such as scratches and abrasions, and at the same time, it can also resist the erosion of external factors such as dust and water vapor, and extend the service life of the film.

[0003] Currently, the methods for detecting the quality of films usually use optical instruments, such as inspection lights, microscopes, projectors, etc., to observe and analyze the microscopic structure and appearance of the film laminating. By observing the microscopic morphology, thickness distribution, etc. of the adhesive layer, the laminating quality is judged. For example, the patent document with the application number 202410824170.5 discloses a medical film defect detection device, including a base, an inspection light and a cross plate. The inspection light is fixedly connected to the top of the base. A fixing frame is fixedly connected to the top of the base. The top of the fixing frame is movably connected with a fixing frame. The bottom of the fixing frame is fixedly connected with a connecting rod, and the connecting rod penetrates through the fixing frame. The bottom of the fixing frame is symmetrically fixedly connected with rubber plates, and the number of rubber plates is two. A fixing mechanism is arranged at the bottom of the connecting rod. A vertical rod is fixedly connected to the top of the base. The cross plate is a hollow structure, and the inner wall of the cross plate is slidably connected with the outer wall of the vertical rod. This invention controls the movement of the fixing frame to fix the film respectively, so that the film remains stable during detection. By rotating the rotating rod, the cross plate is fixed or unlocked, so as to adjust the height of the detection plate, and the effect of detecting films of different sizes respectively is realized, which has good practicability.

[0004] However, when using the above optical instruments to detect the film laminating quality, it is usually necessary to observe and analyze different parts of the film one by one, and the field of view of the optical instrument is relatively small, and only a small part of the pulled film can be observed at a time, thus reducing the efficiency of detecting the film laminating quality. Therefore, we need to propose a film laminating quality detection system and a detection method based on visual detection to solve the above existing problems, so that it can use a high-speed camera and an image processing algorithm to quickly obtain the image information on the film surface, and use multiple cameras or stitching technology to realize the detection of large-area films, reduce the possibility of missing defects, and improve the efficiency of detecting the film laminating quality. Summary of the Invention

[0005] The object of the present invention is to provide a visual inspection-based film laminating quality inspection system and inspection method, which can utilize a high-speed camera and image processing algorithms to quickly acquire the image information of the film surface, and use multiple cameras or stitching techniques to achieve the inspection of large-area films, reducing the possibility of defect omission and improving the efficiency of film laminating quality inspection, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A visual inspection-based film laminating quality inspection system includes: an image acquisition module for shooting film images at a frame rate synchronized with the film transmission speed;

[0008] A light source module for providing illumination conditions for the image acquisition module by selecting different light source types;

[0009] An image preprocessing module for preprocessing the film images captured by each camera. The image acquisition module is electrically connected to the light source module and the image preprocessing module respectively;

[0010] An image stitching module for stitching the images captured by different cameras to form a complete film image;

[0011] A defect detection and recognition module for analyzing the pre-stitched image using image processing algorithms to identify various defects in the large-area film image. The image stitching module is electrically connected to the image preprocessing module and the defect detection and recognition module respectively;

[0012] A motion control module for precisely controlling the film transmission speed and position through a PID control algorithm. The motion control module is electrically connected to the image acquisition module;

[0013] A human-machine interaction module, which is electrically connected to the image acquisition module, the light source module, the image preprocessing module, the defect detection and recognition module, the image stitching module and the motion control module respectively.

[0014] Preferably, the image acquisition module mainly consists of a high-speed camera and a lens. The high-speed camera is used to quickly shoot the surface image of the moving film to ensure that each frame can be clearly captured during the rapid transmission of the film. The lens selects a suitable focal length and aperture according to the detection requirements to obtain high-quality images.

[0015] Preferably, the light source module selects different light source types according to the characteristics of the film and the detection requirements to provide uniform and stable illumination conditions for the image acquisition module. The light source types include parallel light, backlight and ring light.

[0016] Preferably, the image preprocessing module preprocesses the acquired original image, and the operation process of the preprocessing includes denoising, grayscale conversion, contrast enhancement, and normalization;

[0017] The process of the image preprocessing module for denoising is as follows:

[0018] A1. Calculate the weight function of the image according to the original image, and the calculation formula of the weight function is:

[0019] where k σ (i,j) is the value of the weight function at the image pixel coordinate (i,j), x and y are the coordinates of the central pixel respectively, and σ is the standard deviation for controlling the smoothing intensity of the filter;

[0020] A2. Filter the image according to the weight function, and the filtering formula is:

[0021] f(m,o) = ∑ (i,j)∈n k σ (i,j)*I(i,j), where f(m,o) is the pixel value of the denoised image at the coordinate (m,o), n is the number of neighborhood images centered on the coordinate (m,o), and I(i,j) is the pixel value of the image to be filtered at the coordinate (i,j).

[0022] Preferably, the process of the image stitching module for stitching images taken by different cameras is as follows:

[0023] B1. Extract feature points from each preprocessed image using the scale-invariant feature algorithm;

[0024] B2. Calculate the distances between the feature points and find the pairs of matching feature points between different images;

[0025] The distance between feature points can be calculated using the Euclidean distance between feature points;

[0026] B3. According to the pairs of matching feature points, use the random sample consensus algorithm to evaluate the transformation matrix between the images;

[0027] The random sample consensus algorithm satisfies the following relationship:

[0028] where x1 and y1 are the coordinates of the feature points in the image to be transformed respectively, x2 and y2 are the coordinates of the matching feature points in the transformed image respectively, and H is a 3*3 transformation matrix;

[0029] B4. Fuse the transformed images to generate a complete stitched image.

[0030] Preferably, in step B1, the method for extracting feature points is as follows: construct the scale space of the image through Gaussian convolution, detect the extreme points in the scale space, accurately locate the detected extreme points, remove the edge responses and points with low contrast, then assign a direction to each key point, and finally generate feature points containing the gradient information of the area around the key point. The gradient information includes the gradient magnitude and the gradient direction. Among them, the calculation formula for the gradient magnitude is:

[0031]

[0032] The calculation formula for the gradient direction is:

[0033]

[0034] In the formula, F(x, y) is the gradient magnitude of the image at the pixel point (x, y), g(x + 1, y) and g(x - 1, y) are the gray values of the two adjacent horizontal neighborhood pixel points of the expected processed image at (x, y) respectively, g(x, y + 1) and g(x, y - 1) are the gray values of the two adjacent vertical neighborhood pixel points of the expected processed image at (x, y) respectively, and θ(x, y) is the gradient direction of the image at the pixel point (x, y).

[0035] Preferably, in step B4, the weighted average fusion algorithm is used for fusion during fusion. The formula of the weighted average fusion algorithm is as follows:

[0036] G(x1, y1) = w1 * I1(x1, y1) + w2 * I2(x1, y1), where G(x1, y1) is the pixel value of the fused image, I1(x1, y1) and I2(x1, y1) are the pixel values of the two images at the corresponding position (x1, y1) respectively, w1 and w2 are the weight coefficients of the two images at the corresponding position (x1, y1) respectively, and w1 + w2 = 1.

[0037] Preferably, the process of the defect detection and recognition module analyzing and recognizing various defects in the spliced image is as follows:

[0038] C1. Use the Canny operator edge detection algorithm to extract the edge information of the object in the spliced image;

[0039] C2. Find the corner points in the image through the Harris corner detection algorithm. Each corner point corresponds to a prominent feature point in the image;

[0040] C3. Set a threshold that can separate the normal area from the defect area according to the statistical features of the image edge information and corner points;

[0041] C4. Use the k-means clustering algorithm to segment the pixels in the image into different regions according to their pixel characteristics;

[0042] C5. Perform dilation and erosion operations on the segmented image to make the boundaries of the defect regions clearer;

[0043] C6. Match the extracted image features with the defect features in the pre-established defect feature library to determine the defect type;

[0044] C7. Perform pixel statistics on the defect positions and calculate the area size of the defects to evaluate the severity of the defects.

[0045] Preferably, the control process of the PID control algorithm is as follows:

[0046] D1. Measure the actual transmission speed and position of the film in real time through a speed sensor and a position sensor;

[0047] D2. Calculate the speed transmission error and the position error according to the current transmission speed, the current position of the film, as well as the target speed and the target position;

[0048] D3. Calculate the speed control quantity and the position control quantity for adjustment respectively by using a PID controller according to the speed transmission error and the position error;

[0049] D4. Convert the calculated speed control quantity and position control quantity into corresponding control signals, output the control signals to the actuator of the transmission device, and control the transmission speed and position of the film by adjusting the rotation speed and direction of the motor;

[0050] D5. Repeat steps D1 - D4, monitor and adjust the transmission speed and position of the film in real time, and form a closed-loop feedback control system to ensure that the film maintains a stable motion state throughout the detection process.

[0051] Based on the above-described film laminating quality detection system based on vision detection, the present invention also proposes a film laminating quality detection method based on vision detection, including the following steps:

[0052] S1. Set detection parameters through the human-machine interaction module;

[0053] S2. Connect the motion control module to the production line transmission device and set the transmission rate and position parameters of the film;

[0054] S3. The light source module provides uniform and stable illumination for the film according to the set illumination parameters. The image acquisition module, under the synchronous control of the motion control module, quickly captures the image of the film surface during motion at the set shooting frame rate to obtain a high-quality film surface image;

[0055] S4. The image preprocessing module preprocesses the film surface images collected by each camera to make the detailed information in the preprocessed images prominent;

[0056] S5. The image stitching module stitches the film surface images collected by different cameras to obtain a complete film image;

[0057] S6. The defect detection and recognition module uses image processing algorithms to analyze and judge the complete film image, identify various defects on the film surface, determine the types of defects, and quantitatively evaluate the defects;

[0058] S7. The human-computer interaction module displays the detection results and gives an alarm prompt when serious defects or abnormal conditions are detected on the film;

[0059] S8. Continuously perform lamination quality detection on the subsequent transmitted films and monitor the lamination quality of the films in real time.

[0060] A film lamination quality detection system and detection method based on visual detection proposed by the present invention have the following advantages compared with the prior art:

[0061] 1. Through the coordinated cooperation of the image acquisition module, light source module, image preprocessing module, image stitching module, defect detection and recognition module, motion control module, and human-computer interaction module in the present invention, the image acquisition module, with the cooperation of the motion control module, can capture film images at a frame rate synchronized with the film transmission speed. The light source module provides good lighting conditions, and the image acquisition module obtains high-quality image information. Then, the image preprocessing module optimizes the original image, and the image stitching module stitches the preprocessed images to form a complete large-area image. Then, the defect detection and recognition module accurately identifies the defect information in the large-area image, and the human-computer interaction module facilitates the users to monitor and operate. Through the above steps, efficient and high-precision detection of the film lamination quality is achieved, the possibility of defect omission is reduced, and the detection efficiency is improved.

[0062] 2. By the method of first stitching the images taken by multiple cameras after preprocessing and then performing defect detection and recognition in the present invention, compared with the method of performing defect detection on each camera image separately and then performing integrated analysis, unified detection of the stitched images can reduce the repeated calculation and processing processes, improve the detection efficiency, and save time and computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 Shows a system block diagram according to an embodiment of the present invention;

[0064] Figure 2Shows a flowchart of the image stitching module for stitching images captured by different cameras according to an embodiment of the present invention;

[0065] Figure 3 Shows a flowchart of the defect detection and recognition module for analyzing and recognizing various defects in the stitched image according to an embodiment of the present invention;

[0066] Figure 4 Shows a flowchart of the control process of the PID control algorithm according to an embodiment of the present invention;

[0067] Figure 5 Shows a flowchart of the film laminating quality detection method according to an embodiment of the present invention. Detailed implementation manners

[0068] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. The specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0069] The present invention provides a Figures 1-4 film laminating quality detection system based on vision detection as shown, including an image acquisition module, a light source module, an image preprocessing module, an image stitching module, a defect detection and recognition module, a motion control module, and a human-machine interaction module. The human-machine interaction module is electrically connected to the image acquisition module, the light source module, the image preprocessing module, the defect detection and recognition module, the image stitching module, and the motion control module respectively. The image acquisition module is electrically connected to the light source module and the image preprocessing module respectively. The image stitching module is electrically connected to the image preprocessing module and the defect detection and recognition module respectively. The motion control module is electrically connected to the image acquisition module.

[0070] Through the coordinated cooperation of the image acquisition module, light source module, image preprocessing module, image stitching module, defect detection and recognition module, motion control module, and human-computer interaction module, it is possible to automatically, quickly, and accurately detect the quality of film laminating. The image acquisition module, in cooperation with the motion control module, can capture film images at a frame rate synchronized with the film transmission speed. The light source module provides good lighting conditions, and the image acquisition module obtains high-quality image information. Then, the image preprocessing module is used to optimize the original image. The image stitching module stitches the preprocessed images together to form a complete large-area image. Then, the defect detection and recognition module accurately identifies the defect information in the large-area image. The human-computer interaction module facilitates the operator to monitor and operate. Through the above steps, the efficient and high-precision detection of the quality of film laminating is achieved, the possibility of defect omission is reduced, the detection efficiency is improved, and the problem of low detection efficiency existing in traditional optical instrument detection is effectively solved.

[0071] The image acquisition module mainly consists of a high-speed camera and a lens. The high-speed camera is used to quickly capture the surface image of the moving film to ensure that each frame can be clearly captured during the rapid transmission of the film. The lens selects a suitable focal length and aperture according to the detection requirements to obtain high-quality images and ensure that the clarity and resolution of the images meet the requirements of subsequent processing. Using the fast capture ability of the high-speed camera, a large amount of image information on the film surface can be obtained in a short time, effectively solving the problem of slow detection speed of optical instruments.

[0072] When setting the parameters of the high-speed camera, the shooting frame rate is determined according to the film transmission speed and the minimum defect size. The shooting frame rate needs to meet where f is the shooting frame rate, v is the film transmission speed, and d is the minimum defect size; the resolution and exposure time of the camera are configured according to the detection accuracy requirements and storage capacity;

[0073] When setting the parameters of the lens, a suitable aperture size and focal length are selected according to the shooting environment and film characteristics. When the focal length is short, the field of view is large, but the objects in the image are relatively small. When the focal length is long, the field of view is small, but the magnification of the objects is large. A large aperture has a small depth of field and a large light input; a small aperture has a large depth of field and a small light input.

[0074] The light source module selects different light source types according to the characteristics of the film and the detection requirements to provide uniform and stable lighting conditions for the image acquisition module. The light source types include parallel light, backlight, and ring light. Different light source types can highlight different features on the film surface, enhance the contrast of the image, make defects easier to be detected. By optimizing the lighting conditions, the quality and recognizability of the image are improved, facilitating more accurate identification of defects on the film surface and reducing false positives and missed detections caused by uneven illumination.

[0075] When selecting the light source type, first analyze the material, surface characteristics of the film, and the type of detected defects, and then conduct a light source test. Take pictures of the film under different light source types, observe the contrast and clarity of the defects in the pictures, and select the light source type that can best highlight the defects.

[0076] The image preprocessing module preprocesses the film images captured by each camera. The operation process of preprocessing includes denoising, grayscale conversion, contrast enhancement, and normalization. The quality of the images after preprocessing can be significantly improved, enhancing the accuracy and reliability of defect detection.

[0077] The process of the image preprocessing module for denoising is as follows:

[0078] A1. Calculate the weight function of the image based on the original image. The calculation formula of the weight function is:

[0079] where k σ (i,j) is the value of the weight function at the image pixel coordinate (i,j), x and y are the coordinates of the central pixel respectively, and σ is the standard deviation controlling the smoothing intensity of the filter;

[0080] A2. Filter the image according to the weight function. The filtering formula is:

[0081] f(m,o) = ∑ (i,j)∈n k σ (i,j) * I(i,j), where f(m,o) is the pixel value of the denoised image at the coordinate (m,o), n is the number of neighborhood images centered on the coordinate (m,o), and I(i,j) is the pixel value of the image to be filtered at the coordinate (i,j);

[0082] The image preprocessing module performs grayscale conversion on the denoised image through a grayscale model. The formula of the grayscale model is:

[0083] where Gray is the grayscale value of the image after grayscale processing, R(m,o), G(m,o), and B(m,o) are the component values of the red, green, and blue channels at the coordinate (m,o) respectively, G(m,o) << 2 means shifting the component value of the green channel two bits to the left, << is the bit operation symbol, and the calculation efficiency can be improved through bit operations;

[0084] The image preprocessing module enhances the contrast of the grayscale processed image through a grayscale transformation function. The formula of the grayscale transformation function is:

[0085] where T r is the probability density of the gray level r, and L riis the number of occurrences of the gray level r, and N is the total number of gray levels; through gray-scale transformation, the gray-scale range of the entire image is expanded, the originally narrow gray-scale distribution becomes wider, the dark part becomes darker, the bright part becomes brighter, and the details and layering of the image become more obvious, achieving the effect of enhancing the image contrast;

[0086] The image preprocessing module normalizes the image after contrast enhancement, and the formula for normalization is:

[0087] where g(x, y) is the value of the pixel coordinate (x, y) after normalization, T r (x, y) is the value of the pixel (x, y) in the original image, f min and f max are respectively the minimum pixel value and the maximum pixel value in the original image, and L is the maximum value of the target range;

[0088] The image stitching module stitches the images taken by different cameras to form a complete film image, achieving the complete detection of a large-area film, expanding the detection field of view, and avoiding the problem of missing defects caused by the limited field of view of the camera.

[0089] The process of the image stitching module stitching the images taken by different cameras is as follows:

[0090] B1. Use the scale-invariant feature algorithm to extract feature points from each preprocessed image;

[0091] The method of extracting feature points: construct the scale space of the image through Gaussian convolution, detect the extreme points in the scale space, accurately locate the detected extreme points, remove the edge responses and low-contrast points, then assign a direction to each key point, and finally generate feature points containing the gradient information of the area around the key point. The gradient information includes the gradient magnitude and the gradient direction. Among them, the formula for the gradient magnitude is:

[0092]

[0093] The formula for the gradient direction is:

[0094]

[0095] In the formula, F(x, y) is the gradient magnitude of the image at the pixel point (x, y), g(x + 1, y) and g(x - 1, y) are respectively the gray values of the adjacent two horizontal direction neighborhood pixel points at (x, y) in the expected processed image, g(x, y + 1) and g(x, y - 1) are respectively the gray values of the adjacent two vertical direction neighborhood pixel points at (x, y) in the expected processed image, and θ(x, y) is the gradient direction of the image at the pixel point (x, y);

[0096] B2. Calculate the distances between the feature points and find the pairs of matching feature points between different images;

[0097] The distances between the feature points can be calculated using the Euclidean distance between the feature points;

[0098] B3. According to the pairs of matching feature points, use the Random Sample Consensus (RANSAC) algorithm to evaluate the transformation matrix between the images;

[0099] The Random Sample Consensus (RANSAC) algorithm satisfies the following relationship:

[0100] where x1 and y1 are the coordinates of the feature points in the image to be transformed, x2 and y2 are the coordinates of the matching feature points in the transformed image, H is a 3×3 transformation matrix, and in actual calculation, at least 4 pairs of matching feature points are required to solve for H;

[0101] B4. Fuse the transformed images to generate a complete stitched image;

[0102] When fusing, the weighted average fusion algorithm is used for fusion. The formula of the weighted average fusion algorithm is as follows:

[0103] G(x1,y1) = w1 * I1(x1,y1) + w2 * I2(x1,y1), where G(x1,y1) is the pixel value of the fused image, I1(x1,y1) and I2(x1,y1) are the pixel values of the two images at the corresponding position (x1,y1) respectively, w1 and w2 are the weight coefficients of the two images at the corresponding position (x1,y1), and w1 + w2 = 1; the weight coefficients can be determined according to the factor of the distance from the pixel point to the stitching seam to achieve smooth transition. For the pixel points closer to the stitching seam in the image, the value of the weight coefficient w1 will be larger and the value of the weight coefficient w2 will be smaller.

[0104] The defect detection and recognition module uses image processing algorithms to analyze the pre-stitched image, identify various defects in the large-area film image, and realize the quantitative evaluation of the film laminating quality, overcoming the subjectivity and inaccuracy of manual visual inspection and greatly improving the detection accuracy;

[0105] The process of the defect detection and recognition module analyzing and identifying various defects in the stitched image is as follows:

[0106] C1. Use the Canny operator edge detection algorithm to extract the edge information of the objects in the stitched image;

[0107] C2. Find the corner points in the image through the Harris corner detection algorithm, and each corner point corresponds to a prominent feature point in the image;

[0108] C3. Set a threshold that can separate the normal area from the defective area based on the edge information of the image and the statistical characteristics of the corner points. If it is found that the edge length of the defective area is significantly shorter, a certain proportion of the average edge length can be used as the threshold, and the edge area smaller than this threshold may be considered as the defective area. For the number of corner points, a threshold can be set to distinguish the normal area and the defective area according to the difference in the number of corner points between the normal area and the defective area;

[0109] C4. Use the k-means clustering algorithm to segment different areas of the pixel points in the image according to their pixel point characteristics;

[0110] C5. Perform dilation and erosion operations on the segmented image to make the boundary of the defective area clearer. The dilation operation can expand the foreground area and connect some disconnected defective areas, and the erosion operation can remove some isolated noise points and small interference areas to make the boundary clearer;

[0111] C6. Match the extracted image features with the defective features in the pre-established defective feature library to determine the type of defect. The defective feature library contains the feature descriptions of various known defects. Find the defective feature with a high similarity to the extracted image features, and the type corresponding to this defective feature is the determined type of defect;

[0112] C7. Perform pixel statistics on the defective position to calculate the area size of the defect to evaluate the severity of the defect;

[0113] The motion control module is used to cooperate with the transmission device of the production line, and precisely control the transmission speed and position of the film through the PID control algorithm to ensure that the film maintains a stable motion state during the detection process. At the same time, it is synchronized with the image acquisition module to ensure that the image acquisition module can capture clear images.

[0114] The control flow of the PID control algorithm is as follows:

[0115] D1. Measure the actual transmission speed and position of the film in real time through the speed sensor and the position sensor;

[0116] D2. Calculate the speed transmission error and the position error according to the current transmission speed, current position of the film, as well as the target speed and target position;

[0117] D3. Calculate the speed control amount and the position control amount for adjustment respectively by using the PID controller according to the speed transmission error and the position error;

[0118] D4. Convert the calculated speed control amount and position control amount into corresponding control signals, output the control signals to the actuator of the transmission device, and control the transmission speed and position of the film by adjusting the rotation speed and direction of the motor;

[0119] D5. Repeat steps D1 - D4, and monitor and adjust the transmission speed and position of the film in real time to form a closed - loop feedback control system to ensure that the film maintains a stable motion state throughout the detection process;

[0120] When the motion control module is synchronized with the image acquisition module, by setting a trigger signal output port in the motion control module, when the film reaches the specified position or meets the target speed condition, the motion control module sends a trigger signal to the image acquisition module. After receiving the trigger signal, the image acquisition module immediately performs image acquisition, ensuring the precise synchronization of image acquisition and film movement, and ensuring the consistency and accuracy of the acquired images.

[0121] The human - machine interaction module is used to provide an intuitive operation interface for the operator, display the detection results, and give an alarm prompt when serious defects or abnormal conditions are detected in the film, enabling the operator to understand the operating state of the detection system in real time through the operation interface and handle abnormal conditions in a timely manner.

[0122] For the convenience of data storage and management, a data management module is also included. The defect detection and recognition module is electrically connected to the data management module. The data management module stores and manages the acquired image data, detection results, and production information, facilitating the query, statistics, and analysis of historical data, and promptly discovering potential problems in the production process.

[0123] Based on the above - described film laminating quality detection system based on visual inspection, the present invention also provides a film laminating quality detection method based on visual inspection. As Figure 5 shown, it includes the following steps:

[0124] S1. Set detection parameters through the human - machine interaction module. The detection parameters include the shooting frame rate of the high - speed camera, exposure time, brightness and color of the light source, and relevant thresholds of the image - processing algorithm;

[0125] S2. Connect the motion control module to the production line transmission device and set the transmission rate and position parameters of the film;

[0126] S3. The light source module provides uniform and stable illumination for the film according to the set lighting parameters. Under the synchronous control of the motion control module, the image acquisition module quickly shoots the surface image of the moving film at the set shooting frame rate to obtain a high - quality surface image of the film;

[0127] S4. The image pre - processing module pre - processes the surface image of the film collected by each camera to make the detailed information in the pre - processed image prominent;

[0128] S5. The film surface images collected by different cameras are stitched by the image stitching module to obtain a complete film image;

[0129] S6. The complete film image is analyzed and judged by the defect detection and recognition module using image processing algorithms to identify various defects on the film surface, determine the type of the defects, and quantitatively evaluate the defects;

[0130] S7. The detection results are displayed by the human-computer interaction module, and an alarm prompt is given when serious defects or abnormal conditions are detected on the film;

[0131] S8. Continuously detect the laminating quality of the subsequent transmitted films and monitor the laminating quality of the films in real time;

[0132] Multiple cameras are used to simultaneously capture different parts of the film from different angles or positions, and then the images of these parts are stitched into a complete large-area image by image stitching technology. In this way, the fields of view of multiple cameras can cover a large area of the film, and they are combined through stitching technology to achieve a comprehensive detection of the entire film surface, thereby reducing the possibility of missing defects and improving the detection efficiency.

[0133] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A film laminating quality detection system based on visual detection, characterized in that: Including: An image acquisition module, which is used to capture film images at a frame rate synchronized with the film transmission speed; A light source module, which provides illumination conditions for the image acquisition module by selecting different light source types; An image preprocessing module, which is used to preprocess the film images captured by each camera. The image acquisition module is electrically connected to the light source module and the image preprocessing module respectively; An image stitching module, which stitches the images captured by different cameras to form a complete film image; A defect detection and recognition module, which uses image processing algorithms to analyze the pre-stitched images and identify various defects in the large-area film images. The image stitching module is electrically connected to the image preprocessing module and the defect detection and recognition module respectively; A motion control module, which precisely controls the transmission speed and position of the film through a PID control algorithm. The motion control module is electrically connected to the image acquisition module; A human-machine interaction module, which is electrically connected to the image acquisition module, the light source module, the image preprocessing module, the defect detection and recognition module, the image stitching module and the motion control module respectively.

2. The film laminating quality detection system based on visual detection according to claim 1, characterized in that: The image acquisition module mainly consists of a high-speed camera and a lens. The high-speed camera is used to quickly capture the surface image of the moving film to ensure that each frame can be clearly captured during the rapid transmission of the film. The lens selects a suitable focal length and aperture according to the detection requirements to obtain high-quality images.

3. The film laminating quality detection system based on visual detection according to claim 1, wherein: The light source module selects different light source types according to the characteristics of the film and the detection requirements to provide uniform and stable illumination conditions for the image acquisition module. The light source types include parallel light, backlight and ring light.

4. A film laminating quality detection system based on visual detection according to claim 1, characterized in that: The image preprocessing module preprocesses the acquired original images. The operation process of the preprocessing includes denoising, grayscale conversion, contrast enhancement and normalization; The process of the image preprocessing module for denoising is as follows: A1. Calculate the weight function of the image according to the original image. The calculation formula of the weight function is: where k σ (i, j) is the value of the weight function at the image pixel coordinates (i, j), x and y are the coordinates of the central pixel respectively, and σ is the standard deviation that controls the smoothing intensity of the filter; A2. Filter the image according to the weight function. The filtering formula is: f(m,o) = ∑ (i,j)∈n k σ (i,j)*I(i,j), where f(m,o) is the pixel value of the denoised image at the coordinate (m,o), n is the number of neighborhood images centered at the coordinate (m,o), and I(i,j) is the pixel value of the image to be filtered at the coordinate (i,j).

5. A film laminating quality detection system based on visual detection according to claim 1, characterized in that: The process of the image stitching module for stitching the images captured by different cameras is as follows: B1. Use the scale-invariant feature algorithm to extract feature points from each preprocessed image; B2. Calculate the distances between the feature points and find the matching feature point pairs between different images; The distance between the feature points can be calculated using the Euclidean distance between the feature points; B3. According to the matching feature point pairs, use the random sample consensus algorithm to evaluate the transformation matrix between the images; The random sample consensus algorithm satisfies the following relationship: Wherein, x1 and y1 are respectively the coordinates of the feature points in the image to be transformed, x2 and y2 are respectively the coordinates of the matching feature points in the transformed image, and H is a 3×3 transformation matrix; B4. Fuse the transformed images to generate a complete stitched image.

6. The film laminating quality detection system based on visual detection according to claim 5, wherein: In step B1, the method of extracting feature points: construct the scale space of the image through Gaussian convolution, detect the extreme points in the scale space, accurately locate the detected extreme points, remove the edge responses and low-contrast points, then assign a direction to each key point, and finally generate feature points containing the gradient information of the area around the key point. The gradient information includes the gradient amplitude and the gradient direction. Among them, the calculation formula of the gradient amplitude is: The calculation formula of the gradient direction is: In the formula, F(x, y) is the gradient magnitude of the image at the pixel point (x, y), g(x + 1, y) and g(x - 1, y) are the gray values of the adjacent two horizontal neighborhood pixel points of the expected processed image at (x, y) respectively, g(x, y + 1) and g(x, y - 1) are the gray values of the adjacent two vertical neighborhood pixel points of the expected processed image at (x, y) respectively, and θ(x, y) is the gradient direction of the image at the pixel point (x, y).

7. The film laminating quality detection system based on visual detection according to claim 6, wherein: In step B4, the weighted average fusion algorithm is used for fusion during fusion, and the formula of the weighted average fusion algorithm is as follows: G(x1, y1) = w1 * I1(x1, y1) + w2 * I2(x1, y1), where G(x1, y1) is the pixel value of the fused image, I1(x1, y1) and I2(x1, y1) are the pixel values of the two images at the corresponding position (x1, y1) respectively, w1 and w2 are the weight coefficients of the two images at the corresponding position (x1, y1), and w1 + w2 = 1.

8. A film laminating quality detection system based on visual detection according to claim 1, characterized in that: The process of the defect detection and recognition module analyzing and recognizing various defects in the spliced image is as follows: C1. Use the Canny operator edge detection algorithm to extract the edge information of the object in the spliced image; C2. Find the corner points in the image through the Harris corner detection algorithm, and each corner point corresponds to a prominent feature point in the image; C3. Set a threshold that can separate the normal area from the defect area according to the statistical features of the image edge information and corner points; C4. Use the k-means clustering algorithm to segment different areas of the pixel points in the image according to their pixel point features; C5. Perform dilation and erosion operations on the segmented image to make the boundary of the defect area clearer; C6. Match the extracted image features with the defect features in the pre-established defect feature library to determine the defect type; C7. Perform pixel statistics on the defect position and calculate the area size of the defect to evaluate the severity of the defect.

9. A visual inspection-based film laminating quality inspection system according to claim 1, characterized in that: The control process of the PID control algorithm is as follows: D1. Measure the actual transmission speed and position of the film in real time through the speed sensor and position sensor; D2. Calculate the speed transmission error and position error according to the current transmission speed, current position of the film, as well as the target speed and target position; D3. Calculate the speed control amount and position control amount for adjustment respectively by using the PID controller according to the speed transmission error and position error; D4. Convert the calculated speed control amount and position control amount into corresponding control signals, output the control signals to the actuator of the transmission device, and control the transmission speed and position of the film by adjusting the rotation speed and steering of the motor; D5. Repeat steps D1 - D4 to monitor and adjust the transmission speed and position of the film in real time, forming a closed-loop feedback control system to ensure that the film maintains a stable motion state during the entire detection process.

10. A method for detecting the quality of film lamination based on visual inspection, based on a system for detecting the quality of film lamination based on visual inspection according to any one of claims 1-9, characterized in that: It includes the following steps: S1. Set detection parameters through the human-computer interaction module. The detection parameters include the shooting frame rate of the high-speed camera, exposure time, brightness and color of the light source, and relevant thresholds of the image processing algorithm; S2. Connect the motion control module to the production line transmission device and set the transmission rate and position parameters of the film. S3. The light source module provides uniform and stable illumination for the film according to the set lighting parameters. Under the synchronous control of the motion control module, the image acquisition module quickly captures the surface image of the moving film at the set shooting frame rate to obtain a high-quality surface image of the film. S4. The image preprocessing module preprocesses the surface image of the film collected by each camera to make the detailed information in the preprocessed image prominent. S5. The image stitching module stitches the surface images of the film collected by different cameras to obtain a complete film image. S6. The defect detection and recognition module uses image processing algorithms to analyze and judge the complete film image, identify various defects on the film surface, determine the type of defects, and quantitatively evaluate the defects. S7. The human-machine interaction module displays the detection results and gives an alarm prompt when serious defects or abnormal situations are detected on the film. S8. Continuously conduct laminating quality detection on the subsequently transmitted films and monitor the laminating quality of the films in real time.

Images