Flexible film residue positioning and removal device based on vision algorithms

Abstract

The present application relates to a kind of flexible film residual position removal device based on visual algorithm, it is related to the technical field of cleaning equipment, including roller system, for supporting and conveying flexible film;Visual camera is arranged at one side of roller system, for scanning and obtaining image data to the surface of flexible film;Cleaning module includes servo motor, synchronous wheel, sliding block and nozzle, servo motor and synchronous wheel constitute transmission system, for driving sliding block to move, nozzle is installed on sliding block;And control system, for receiving the image data of visual camera, identify the residual position on the surface of flexible film, and control servo motor in cleaning module drive sliding block to move to corresponding position, by nozzle carry out point cleaning.The present application accurately identifies residual position by visual algorithm and realizes point cleaning in combination with servo drive system, can significantly improve flexible film cleaning efficiency and precision, while supporting double-side synchronous processing, reduces manual intervention and repeated operation.

Description

Technical Field

[0001] This invention relates to the field of cleaning equipment technology, and in particular to a flexible film residue positioning and removal device based on a visual algorithm. Background Technology

[0002] Flexible thin-film materials are widely used in industrial fields such as electronic displays, optical devices, and packaging materials. During production and processing, various residues easily adhere to the film surface, including dust particles, fibers, electrostatic adsorbents, and minute contaminants generated during the process. Traditional film cleaning methods mainly rely on manual wiping, mechanical brushing, electrostatic dust removal, and airflow purging. These methods are inefficient when dealing with large-area films and have limited cleaning precision. As the requirements for film manufacturing processes continue to increase, the standards for surface cleanliness are becoming increasingly stringent, and traditional cleaning methods can no longer meet the needs of modern industrial production.

[0003] Existing membrane cleaning equipment suffers from several technical limitations. First, it lacks precise residue identification and positioning capabilities, failing to differentiate between different types and sizes of contaminants, resulting in a blind and inefficient cleaning process. Second, cleaning operations are mostly single-sided, unable to clean both sides of the membrane simultaneously, increasing process complexity and processing time. Furthermore, existing equipment is prone to damaging or deforming flexible membranes when handling them in suspended or tensile conditions. Simultaneously, traditional cleaning methods lack intelligent control, failing to target residues based on their specific location and characteristics, often requiring repeated operations or manual intervention, impacting production efficiency and product quality consistency. Summary of the Invention

[0004] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a flexible film residue positioning and removal device based on a vision algorithm. By accurately identifying the location of residue through a vision algorithm and combining it with a servo drive system to achieve point-to-point cleaning, the device can significantly improve the cleaning efficiency and accuracy of flexible films. At the same time, it supports simultaneous processing on both sides, reducing manual intervention and repetitive operations.

[0005] To achieve the above objectives, the present invention provides a flexible film residue positioning and removal device based on a visual algorithm, which adopts the following technical solution: Roller systems are used to support and transport flexible films; A vision camera, mounted on one side of the roller system, is used to scan the surface of the flexible film and acquire image data; A cleaning module includes a servo motor, a synchronous pulley, a slider, and a nozzle. The servo motor and synchronous pulley form a transmission system for driving the slider to move. The nozzle is mounted on the slider. The control system is used to receive image data from the vision camera, identify the location of residues on the surface of the flexible film, and control the servo motor in the cleaning module to drive the slider to the corresponding position for targeted cleaning through the nozzle.

[0006] Furthermore, in the aforementioned device, the vision camera is a high-speed linear scan camera, used to scan the moving flexible film line by line.

[0007] Furthermore, in the above-mentioned device, the vision camera includes a time delay compensation module. The time delay compensation module calculates the predicted displacement of the residue in the moving direction of the film based on the real-time running speed of the flexible film and the total delay of camera exposure, image processing, and servo response, and synchronously corrects the coordinates of the residue detected by vision, so that the nozzle can still accurately align with the target position for fixed-point cleaning under the condition of continuous film movement.

[0008] Furthermore, in the aforementioned device, the control system includes a positioning algorithm processor for analyzing the image data and identifying the location and shape of the residue.

[0009] Furthermore, the aforementioned device also includes a second cleaning module disposed on the other side of the roller system for cleaning the other side of the flexible film.

[0010] Furthermore, in the above-mentioned device, the second cleaning module is arranged in a central mirror image with the cleaning module, which can clean both sides of the flexible film simultaneously.

[0011] Furthermore, in the above-described device, the nozzle is positioned relatively perpendicular to the centerline of the roller system to ensure that the airflow does not adversely affect the film.

[0012] Furthermore, the aforementioned device includes dynamic adjustment of the injection pressure, pulse width, and injection frequency. The adaptive adjustment determines the target injection parameter range based on the size, adhesion strength, and morphological characteristics of the residue, and sets an upper limit for the error of the injection parameters. The injection parameters are optimized and corrected by evaluating the cleaning effect after the spraying through real-time feedback.

[0013] Furthermore, in the aforementioned device, the control system can automatically adjust the spray pressure and purging time of the nozzle according to the type of residue. The automatic adjustment is based on the size, adhesion degree, area and morphological characteristics of the residue to classify the residue, and matches the corresponding spray pressure range and purging time range for different types of residue, so as to achieve differentiated cleaning of light, heavy and stubborn residues.

[0014] Furthermore, the aforementioned device also includes a CLS scanner to assist the vision camera in performing thin film surface inspection.

[0015] In summary, compared with the prior art, the present invention has at least one of the following beneficial technical effects: The flexible film residue positioning and removal device based on vision algorithms of the present invention integrates a vision camera, a servo-driven cleaning module, and an intelligent control system to achieve precise identification and targeted removal of residues on the surface of flexible films. The device acquires image data in real time during film transport, automatically identifies the location and characteristics of residues through vision algorithms, and drives the cleaning module to precisely move to the target location for targeted blowing cleaning. Compared with traditional blind, comprehensive cleaning methods, this significantly improves cleaning efficiency and accuracy, reduces unnecessary cleaning operations, and avoids over-processing or damage to the film. It enhances production efficiency and product consistency while ensuring cleaning quality. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0017] Figure 1 An isometric view of a flexible film residue positioning and removal device based on a vision algorithm is shown.

[0018] Figure 2 An exploded view of a flexible film residue positioning and removal device based on a vision algorithm is shown.

[0019] Figure 3 A top view of the cleaning module components is shown.

[0020] Figure 4 A cross-sectional view of the cleaning module assembly is shown.

[0021] Figure 5 A flowchart illustrating the method for residue classification and cleaning parameter setting is provided. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Furthermore, it should be understood that the specific embodiments described herein are only for illustration and explanation of this application and are not intended to limit this application.

[0023] It should be noted that the order of description of the following embodiments is not intended to limit the preferred order of the embodiments of this application. Furthermore, the descriptions of each embodiment in the following embodiments have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0024] The method steps described in this embodiment of the invention can be executed in the order described in the specific implementation, or the execution order of each step can be adjusted according to actual needs, provided that the technical problem can be solved. These are not listed one by one here.

[0025] The present invention will be further described in detail below with reference to the accompanying drawings.

[0026] Reference Figure 1 The flexible film residue positioning and removal device based on vision algorithms includes major components such as a roller system, a vision camera, a cleaning module, and a control system. Through a coordinated workflow involving a vision subsystem, AI positioning, and a pneumatic gun mechanical actuator, the device achieves an overall cleaning efficiency of over 99%.

[0027] like Figure 1 As shown, the first roller system 2 and the second roller system 22 are configured as cylindrical assemblies for supporting and conveying the flexible film 7. The first roller system 2 and the second roller system 22 are spaced apart to maintain the tension and positioning of the flexible film 7 as it passes through the cleaning process. The flexible film 7 moves along the path defined by the first roller system 2 and the second roller system 22, passing through the detection and cleaning area.

[0028] A first vision camera 1 is mounted on a side plate 5, positioned on one side of the first roller system 2, and is used to scan the surface of the flexible film 7 and acquire image data. The first vision camera 1 is aligned with the first roller system 2 to scan the film surface. In some embodiments, the first vision camera 1 covers a continuously dynamically moving area of ​​the flexible film 7, performing real-time image acquisition during production line operation.

[0029] The first CLS scanner 11 is located near the first vision camera 1 and provides auxiliary detection functions for thin film surface inspection. The first CLS scanner 11 works in conjunction with the first vision camera 1 to enhance the ability to identify residues on the surface of the flexible thin film 7.

[0030] The first cleaning module 3 and the second cleaning module 4 are located on opposite sides of the path of the flexible film 7, forming a mirror configuration relative to the centerline of the film. The first cleaning module 3 is located on the upper side of the film path, and the second cleaning module 4 is located on the lower side of the film path. This arrangement enables the device to handle residues on both surfaces of the flexible film 7.

[0031] Side plate 5 provides structural support for mounting the first vision camera 1 and cleaning modules 3 and 4. Side plate 5 maintains the spatial relationship between the various components and ensures stable positioning relative to the first roller system 2 and the second roller system 22.

[0032] The control system receives image data from the first vision camera 1 and identifies the location of residues on the surface of the flexible film 7. The control system analyzes the image data, determines the coordinates of the residues in the workspace, and generates corresponding control commands. In some embodiments, the control system processes continuous high-resolution images using AI positioning algorithms to accurately identify the location and shape of residues on the film surface.

[0033] Reference Figure 2 The detailed structure of the cleaning module includes a transmission system and a fixed-point cleaning mechanism. The first cleaning module 3 and the second cleaning module 4 each include a servo motor, a synchronous wheel, a slider 31, and a nozzle 32. The servo motor 8 and the synchronous wheel 81 form the transmission system, used to drive the slider 31 to move.

[0034] like Figure 2 As shown, the servo motor 8 is located inside the cleaning module and is coupled to the synchronous pulley 81 via a mechanical connection. The synchronous pulley 81 forms part of the transmission system, converting the rotational motion of the servo motor 8 into the linear motion of the slider 31. The transmission system achieves precise positioning of the slider 31 in the horizontal and vertical directions through closed-loop servo control.

[0035] The slider 31 is configured to move along a straight path within the cleaning module, carrying the nozzle 32 to a precise position for directional cleaning. The nozzle 32 is mounted on the slider 31 and moves accordingly as the slider 31 moves. The nozzle 32 is mounted on a movable support structure, achieving point-to-point cleaning through precise positioning in the X / Y directions.

[0036] Reference Figure 2 A second vision camera 10 is positioned on the other side of the roller system, forming a dual-camera configuration with the first vision camera 1. The second vision camera 10 is used to scan and monitor the other side of the flexible film. A second CLS scanner 101 is located near the second vision camera 10, providing auxiliary functions for film surface inspection.

[0037] The control system calculates the optimal blowing path and timing based on the coordinates of the residue identified by the vision camera. The control system then controls the servo motor 8 in the cleaning module to move the slider 31 to the corresponding position, aligning the nozzle 32 with the target location on the flexible film surface. After positioning, the control system triggers the nozzle 32 to perform targeted cleaning, using high-pressure airflow to blow the residue away from the film surface.

[0038] In some embodiments, the jet pressure and purging time of nozzle 32 are automatically adjusted according to the type of residue to ensure thorough cleaning without damaging the flexible membrane. The drive system achieves rapid response and precise positioning through servo control, enabling nozzle 32 to quickly move to the target location to perform the cleaning operation after detecting residue.

[0039] Reference Figure 3 The first vision camera 1 and the second vision camera 10 are high-speed line scan cameras used to scan the moving flexible film line by line. The high-speed line scan cameras scan line by line along the moving direction of the flexible film to acquire high-resolution images in real time. In some embodiments, the high-speed line scan cameras combine AI positioning algorithms and high-speed image processing to detect minute defects under high-speed movement on the production line.

[0040] like Figure 3 As shown, the servo motor 8 is located near the first roller system 2 and the second roller system 22, and is mechanically coupled to the synchronous pulley 81. The servo motor 8 provides driving force to the positioning mechanism, and the synchronous pulley 81 forms part of the transmission system, converting the rotational motion of the servo motor 8 into the linear motion of the slider assembly. The synchronous pulley 81 is configured to mesh with the drive mechanism to achieve precise control of the nozzle along the cleaning path.

[0041] The control system includes a positioning algorithm processor for analyzing image data and identifying the location and shape of residues. The positioning algorithm processor utilizes a pre-trained AI model to analyze continuous high-resolution images acquired by a high-speed linear scan camera. In some embodiments, the positioning algorithm processor accurately identifies the location and shape of residues on the flexible film surface and calculates the coordinates of the residues in the workspace.

[0042] The pre-trained AI model, trained using machine learning algorithms, is capable of identifying different types and shapes of residues. The positioning algorithm processor receives image data from a high-speed linear scan camera, analyzes and processes it through the AI ​​model, and generates location information and morphological characteristics of the residues. The processor then transmits the analysis results to other components of the control system for subsequent cleaning operation control.

[0043] In some implementations, the combined configuration of a vision camera and a positioning algorithm processor enables comprehensive coverage of a continuously moving thin film region. The line-by-line scanning capability of a high-speed linear scan camera, combined with the real-time processing power of the AI ​​positioning algorithm, ensures accurate identification and positioning of surface residues during the high-speed movement of the thin film.

[0044] In some implementations, a high-speed linear scan camera combined with a dynamic calibration system achieves a precise spatial correspondence between the camera and the nozzle. The dynamic calibration system establishes a mapping between the image coordinate system and the physical coordinate system, ensuring that the location of the residue detected visually can be accurately converted into the target cleaning location of the nozzle. The dynamic calibration system includes a calibration algorithm module that updates the coordinate transformation matrix in real time based on the motion state of the flexible film and the system's geometric parameters, compensating for positional deviations caused by film movement, mechanical vibration, or temperature changes.

[0045] Specifically, in this embodiment, a spatial mapping relationship is established between the high-speed linear array camera and the nozzle through a dynamic calibration system. The dynamic calibration uses a two-dimensional standard calibration board with a unit size of 5 mm × 5 mm. By acquiring no fewer than 200 sets of calibration images under motion conditions, the transformation matrix from the image coordinate system to the nozzle coordinate system is calculated, with a calibration error not exceeding 0.15 mm. To compensate for the time delay during the continuous movement of the thin film, a time delay compensation module is included to correct the total system delay in real time throughout the entire process of "camera exposure—image transmission—algorithm processing—servo response." Actual tests show that the system delay is approximately 38–55 ms, including approximately 8 ms for camera exposure, approximately 6 ms for image transmission, approximately 12–18 ms for algorithm inference, and approximately 12–20 ms for servo response. The time delay compensation module, based on the real-time detected film linear velocity (e.g., 0.5–1.2 m / s), calculates the predicted displacement of the residue within the aforementioned total delay time (typically ranging from 18–66 mm), and superimposes this predicted displacement onto the initial coordinates of the residue obtained through visual recognition. This achieves dynamic correction of the nozzle target position, keeping the nozzle positioning error within ±0.5 mm. Furthermore, the system updates the film velocity and delay estimate every 20 ms, allowing the compensation results to change in real-time with equipment vibration or speed fluctuations.

[0046] Continue to refer to Figure 1 and Figure 2 The second cleaning module 4 is located on the other side of the roller system, forming a symmetrical configuration with the first cleaning module 3. The second cleaning module 4 is used to clean the other side of the flexible film 7, realizing a double-sided cleaning function. The second cleaning module 4 includes the same component structure as the first cleaning module 3, including components such as a servo motor, a synchronous wheel, a slider, and a nozzle.

[0047] The second cleaning module 4 is centrally mirrored with the first cleaning module 3, forming a symmetrical layout relative to the centerline of the flexible film 7. This mirror configuration allows the two cleaning modules to clean both sides of the flexible film 7 simultaneously. In some embodiments, the nozzles of the second cleaning module 4 correspond spatially to the nozzles of the first cleaning module 3, ensuring coordination of the double-sided cleaning operation.

[0048] The dual-sided cleaning system achieves synchronous or alternating operation of the upper and lower cleaning modules through a timing coordination mechanism. The control system coordinates the working timing of the first cleaning module 3 and the second cleaning module 4 based on the detection results of the first vision camera 1 and the second vision camera 10. In some embodiments, when the flexible film 7 is fed, the upper and lower cameras simultaneously scan and locate the residue on the upper and lower surfaces, and the two servo-controlled cleaning modules synchronously or alternately blow air onto both sides.

[0049] The anti-collision protection mechanism prevents the upper and lower nozzles from colliding during the cleaning process by monitoring the position status of the two cleaning modules. The control system tracks the position information of the sliders in the first cleaning module 3 and the second cleaning module 4 in real time to establish a safe working area limit. In some embodiments, the anti-collision protection mechanism includes position sensors and safety algorithms. When the system detects that the nozzles of the two cleaning modules are close to a preset safe distance, it automatically adjusts the cleaning sequence or suspends the operation of the corresponding module.

[0050] The fusion processing algorithm for double-sided residue detection integrates image data from the first visual camera 1 and the second visual camera 10. The fusion processing algorithm analyzes the residue distribution on both sides of the flexible film 7 to generate a comprehensive cleaning strategy. In some embodiments, the fusion processing algorithm optimizes the working sequence and cleaning parameters of the cleaning module based on the location, size, and type of the residue, avoiding duplicate cleaning and missed cleaning.

[0051] The dual-sided synchronous cleaning configuration significantly improves overall cleaning efficiency. The first cleaning module 3 and the second cleaning module 4 can process both sides of the flexible film 7 simultaneously or alternately, reducing the number of times the film passes through the cleaning area. In some embodiments, the dual-sided parallel working mode more than doubles the cleaning efficiency compared to single-sided cleaning.

[0052] The cleaning system performs a non-contact purging operation to prevent damage to the flexible film 7. Nozzles use high-pressure airflow to blow residues away from the film surface, avoiding scratches or deformation that could result from mechanical contact. In some embodiments, non-contact purging ensures high precision and rapid purging speed while maintaining the integrity and surface quality of the flexible film 7.

[0053] Reference Figure 4 The nozzles are positioned relatively perpendicular to the centerline of the roller system to ensure that the airflow does not adversely affect the film. Figure 4 Two cross-sectional views of the cleaning module assembly are shown, illustrating the internal structure and component arrangement of the cleaning mechanism under different operating conditions.

[0054] like Figure 4As shown in the top view, in the first configuration state, the cleaning module has the slider mechanism located at the top of the assembly. The nozzle assembly extends from the slider mechanism and is oriented towards the film path. The housing has an opening on the left side and includes an angled guide vane surface at the bottom for guiding airflow or debris. The internal cavity of the housing provides space for assembly movement and debris collection.

[0055] Figure 4 The lower view shows the cleaning module in its second configuration state, demonstrating the repositioning of the movable components. The slider mechanism has moved to different positions within the housing, and the nozzle assembly has changed its position accordingly. The angled guide vane surface remains at the bottom of the housing. The drive mechanism assembly, located at the bottom of the assembly, is connected to the slider mechanism to enable its movement within the housing.

[0056] Two views illustrate the range of motion of the cleaning component within the housing structure. The cross-sectional view reveals the internal channels and spatial relationships between the slider mechanism, nozzle assembly, and housing wall. The lower guide vane surface creates channels to guide material to the collection area. The component design allows the nozzle to traverse different positions while maintaining proper orientation relative to the film processing path.

[0057] The vertical arrangement of the nozzles relative to the centerline of the roller system ensures precise control of the airflow direction. In some embodiments, this configuration prevents the airflow from exerting lateral forces or disturbances on the flexible film, avoiding film misalignment or wrinkling during cleaning. The airflow generated by the vertically positioned nozzles acts directly on residues on the film surface without affecting the overall stability of the film.

[0058] An angled surface formed by a baffle at the bottom of the housing guides the blown-off residue and airflow toward the collection area. The angled design of the baffle optimizes the airflow path and prevents residue from re-adhering to the membrane surface during cleaning. In some embodiments, the angle and shape of the baffle surface are optimized according to the characteristics of different types of residue to improve collection efficiency.

[0059] In some implementations, the nozzle's jet parameters are adaptively adjustable, achieving dynamic optimization of pressure, pulse width, and frequency through intelligent control algorithms. The adaptive adjustment system automatically calculates and sets the corresponding jet parameters based on residue characteristic information identified by the positioning algorithm processor, including residue size, shape, adhesion strength, and material type. The pressure adjustment range covers the operating range of 0.1-1.0 MPa, the pulse width control accuracy reaches the 1-millisecond level, and the frequency adjustment range is 1-100 Hz. The system establishes a database mapping residue characteristics to jet parameters and continuously optimizes the parameter selection strategy through machine learning algorithms, ensuring that the most suitable jet configuration is used for each cleaning operation. An error upper limit control mechanism monitors the deviation between the actual jet parameters and the set values; when the deviation exceeds a preset threshold, the system automatically corrects the parameters or issues an alarm.

[0060] Specifically, in this embodiment, the nozzle's jet parameters employ an adaptive adjustment strategy, automatically determining the optimal jet pressure, pulse width, and jet frequency based on the residue characteristics identified by the positioning algorithm processor. In practical applications, the system adjusts the jet pressure based on the residue area (0.2–8 mm). 2 The parameters are mapped based on features such as adhesion strength level (divided into L1–L5) and edge morphology (circular, irregular, polygonal debris). For example, when an area smaller than 1 mm is detected... 2 For light residues with an adhesion strength of L1, the spray pressure is set to 0.15–0.25 MPa, the pulse width to 3–6 ms, and the frequency to 10–20 Hz; for areas of 1–4 mm², the spray pressure is set to 0.15–0.25 MPa, the pulse width to 3–6 ms, and the frequency to 10–20 Hz. 2 For moderate residues with an adhesion strength of L2–L3, the spray pressure should be adjusted to 0.28–0.45 MPa, pulse width 6–12 ms, and frequency 20–40 Hz; while for areas larger than 4 mm², the spray pressure should be adjusted to 0.28–0.45 MPa, pulse width 6–12 ms, and frequency 20–40 Hz. 2 For stubborn residues with adhesion strength reaching L4–L5, the spray pressure is increased to 0.50–0.85 MPa, pulse width 12–20 ms, and frequency 40–80 Hz. The system sets the upper limit of error for the spray parameters to ±5%, and monitors the cleaning effect in real time after spraying using an airflow feedback sensor installed near the cleaning module. When the residue removal rate is lower than 95% or the feedback signal deviation exceeds a threshold (e.g., 8%), the control system automatically performs secondary optimization of the spray parameters to ensure that the spraying action thoroughly cleans the residue without damaging the flexible film.

[0061] Furthermore, the control system automatically adjusts the nozzle's spray pressure and purging time based on the type of residue. In some embodiments, the control system establishes a correspondence between residue type and cleaning parameters by analyzing the morphological characteristics of the residue identified by the positioning algorithm processor. The control system includes a parameter adjustment module that automatically calculates the most suitable spray pressure range and purging duration based on the size, adhesion, and material properties of the residue. The parameter adjustment module adjusts the nozzle's operating parameters in real time using a preset parameter database and adaptive algorithms, ensuring thorough cleaning without damaging the flexible membrane. The spray pressure adjustment range covers multiple levels from low-pressure fine cleaning to high-pressure powerful removal, and the purging time control accuracy reaches the millisecond level, achieving precise cleaning of different types of residue.

[0062] Specifically, refer to Figure 5 First, feature extraction of the residue image is performed. After the image acquired by the high-speed linear scan camera is analyzed by the positioning algorithm processor, the key feature parameters of the residue are extracted, including the area A (typically ranging from 0.2 to 10 mm). 2 The parameters are: 1) Roundness C (0.3–1), 2) Adhesion strength index F (based on texture gradient and pixel grayscale difference, taking levels 1–5), and 3) Edge roughness R (0–1). The adhesion strength index F is determined by the average value of F = k·(∂I / ∂x + ∂I / ∂y), where k is a normalization constant.

[0063] Next, residue classification is performed. Residues are graded according to preset residue classification rules. For example: light residues (L1 category): A < 1 mm 2 And F ≤ 2; Moderate residue (L2–L3 class): 1 mm 2 ≤ A≤ 4 mm 2 Or F = 3; Refractory residues (L4–L5): A > 4 mm 2 If F ≥ 4 and R > 0.5, it will be automatically upgraded to class L5. The classification result will serve as the basis for subsequent cleaning parameter selection.

[0064] Next, a matching strategy for injection pressure and purging time is implemented. In some implementations, to avoid insufficient cleaning or film damage caused by misclassification, an upper limit constraint of ±5% is imposed on the selected parameters. For example, when there are fluctuations in the estimated residue adhesion strength F, the final injection parameters are ensured to be within the safe boundary by using P_actual = clamp(P_estimated ± ΔP, ±5%) and T_actual = clamp(T_estimated ± ΔT, ±5%).

[0065] Finally, real-time evaluation and secondary adjustments are performed on the cleaning effect. Optical micro-sensors or airflow feedback sensors are installed near the cleaning module to determine whether the residue removal rate after blowing reaches a threshold (≥95%). The evaluation steps include: If the residual pixel value is less than 5% after cleaning, the cleaning is considered successful. If the residual pixel value is ≥ 5% or the feedback wind pressure deviation is > 8%, the control system will trigger a secondary adjustment. The secondary adjustment can be set as follows: increase pressure by 0.05–0.15 MPa; increase purging time by 3–8 ms; and repeat the purging process. If the point is still not removed after three purging cycles, it will be marked as a "high adhesion point" and handled manually or by a dedicated module at the back end.

[0066] Furthermore, the device described in this embodiment of the invention also includes a CLS scanner, which serves as an auxiliary detection component, working in conjunction with a vision camera to enhance the ability to identify and locate residues on the surface of flexible thin films. (See also...) Figure 2 The first CLS scanner 11 and the second CLS scanner 101 are located adjacent to the first vision camera 1 and the second vision camera 10, respectively, to provide supplementary detection functions for thin film surface detection.

[0067] In some implementations, a CLS scanner performs three-dimensional contour detection on the surface of a flexible thin film using laser scanning technology. The CLS scanner emits a laser beam and receives the reflected signal, generating height information and texture data of the thin film surface. This three-dimensional detection capability complements the two-dimensional image information from the vision camera, providing thickness, height, and three-dimensional shape characteristics of the residue.

[0068] The auxiliary detection function of the CLS scanner integrates image data from a vision camera using a multi-sensor data fusion algorithm. The data fusion algorithm combines the 3D contour information acquired by the CLS scanner with the high-resolution image data from the vision camera to generate a more complete and accurate description of residue features. In some implementations, this multi-sensor fusion method significantly improves the detection accuracy of minute residues and low-contrast contaminants.

[0069] The configuration of the first CLS scanner 11 and the second CLS scanner 101 enables comprehensive auxiliary inspection of both sides of the flexible film. The two CLS scanners, together with their corresponding vision cameras, form a detection unit, responsible for auxiliary scanning of the upper and lower surfaces of the film, respectively. This dual-sided auxiliary inspection configuration ensures the completeness and reliability of residue identification.

[0070] The detection data from the CLS scanner is transmitted to the positioning algorithm processor of the control system via a real-time processing algorithm. The positioning algorithm processor performs correlation analysis between the 3D data from the CLS scanner and the image data from the vision camera to extract comprehensive feature information of the residue. In some implementations, this multi-dimensional feature extraction method enables the system to identify minute residues that are difficult to detect using traditional single detection methods.

[0071] The device forms a complete closed-loop control process through identification, positioning, blowing, and collection. After the vision camera and CLS scanner complete the residue identification, the control system performs precise positioning calculations, driving the cleaning module to perform targeted blowing operations. The collection module quickly recovers the blown-off residue, preventing it from re-adhering to the film surface or contaminating the working environment.

[0072] The collection module is located below and to the side of the cleaning area, and collects the residue removed by the blown airflow using a negative pressure suction system. In some embodiments, the collection module includes a suction device and a filtration system. The suction device generates a negative pressure airflow to draw the blown-away residue into the collection chamber, and the filtration system separates and processes the collected residue.

[0073] The closed-loop control process ensures the integrity of the cleaning operation through a real-time feedback mechanism. The control system monitors the execution status of each stage—identification, positioning, blowing, and collection—and coordinates the working sequence of each component through status feedback signals. In some implementations, the system will only proceed to the next detection target after the collection module confirms that the residue has been successfully recovered, ensuring the continuity and reliability of the cleaning process.

[0074] The entire closed-loop control system achieves a cleaning efficiency of over 99% through precise timing control and status monitoring. The CLS scanner's auxiliary detection function improves the accuracy of residue identification, reducing missed and false detections. In some implementations, the combination of multi-sensor collaborative detection and closed-loop control processes enables the device to handle various types and sizes of residues, maintaining stable and highly efficient cleaning performance.

[0075] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flexible film residue positioning and removal device based on a visual algorithm, characterized in that, include: Roller systems are used to support and transport flexible films; A vision camera, mounted on one side of the roller system, is used to scan the surface of the flexible film and acquire image data; A cleaning module, located on one side of the roller system, includes a servo motor, a synchronous wheel, a slider, and a nozzle. The servo motor and the synchronous wheel form a transmission system for driving the slider to move, and the nozzle is mounted on the slider. as well as The control system is used to receive image data from the vision camera, identify the location of residues on the surface of the flexible film, and control the servo motor in the cleaning module to drive the slider to the corresponding position for targeted cleaning through the nozzle. The vision camera is a high-speed linear scan camera, used to scan the moving flexible film line by line; The nozzles are positioned relatively perpendicular to the centerline of the roller system to ensure that the airflow does not adversely affect the flexible film; The high-speed linear array camera and the nozzle establish a spatial mapping relationship through a dynamic calibration system. The dynamic calibration system uses a two-dimensional standard calibration plate and calculates the transformation matrix from the image coordinate system to the nozzle coordinate system by acquiring calibration images in motion. The vision camera includes a time delay compensation module, which is used to determine the total system delay including camera exposure delay, image transmission delay, algorithm processing delay and servo response delay. Based on the real-time linear velocity of the flexible film, the module calculates the predicted displacement of the residue along the moving direction of the flexible film within the total system delay time. The predicted displacement is then superimposed on the initial coordinates of the residue obtained by visual recognition to dynamically correct the target position of the nozzle. The control system includes a positioning algorithm processor and a parameter adjustment module. The positioning algorithm processor analyzes the images acquired by the high-speed linear array camera and extracts the area A, roundness C, adhesion strength index F, and edge roughness R of the residue. The control system classifies the residues according to the area A, adhesion strength index F, and edge roughness R, following a preset residue classification rule. The parameter adjustment module determines the nozzle's injection pressure, pulse width, and injection frequency from a pre-established database of mapping relationships between residue characteristics and jet parameters, based on the residue classification results. The parameter adjustment module sets an error upper limit constraint of ±5% for the determined injection pressure and purging time, so that the actual jet parameters of the nozzle are within the corresponding safety boundary. An optical micro-sensor or airflow feedback sensor is installed near the cleaning module. The control system evaluates the cleaning effect after blowing in real time through the optical micro-sensor or airflow feedback sensor. When the residual pixel value after cleaning is less than 5%, the cleaning is considered successful. When the residual pixel value after cleaning is not less than 5% or the feedback air pressure deviation is greater than 8%, the control system performs a secondary adjustment, increasing the spray pressure by 0.05 MPa to 0.15 MPa, increasing the blowing time by 3 ms to 8 ms, and controlling the nozzle to perform another blowing. When the same residue is still not removed after three blowing cycles, the location of the residue is marked as a high adhesion point and handed over to the backend manual or dedicated module for processing.

2. The apparatus according to claim 1, characterized in that, It also includes a second cleaning module located on the other side of the roller system for cleaning the other side of the flexible film.

3. The apparatus according to claim 2, characterized in that, The second cleaning module is centrally mirrored with the first cleaning module, enabling it to clean both sides of the flexible film simultaneously.

4. The apparatus according to claim 1, characterized in that, It also includes a CLS scanner to assist the vision camera in the detection of flexible film surfaces.

Images