High-density silk screen denting control method and system based on visual identification
By acquiring reed image data through visual recognition technology, generating dynamic coding models, and constructing image enhancement and depth analysis models, the problem of automated identification and processing of minute defects in the process of high-density wire mesh reed threading was solved, achieving efficient and accurate defect cleaning and separation, and improving production efficiency and quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-15
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies struggle to automate, accurately identify, and process minute defects such as dirt, adhesion, extrusion deformation, or micro-spacing anomalies during the high-density wire mesh weaving process. Furthermore, defect cleaning and separation operations require manual intervention or are inefficient.
A high-density wire mesh reed threading control method based on vision recognition is adopted. By acquiring reed image data, a dynamic coding model is generated, an image enhancement model and a depth analysis model are constructed, and combined with a three-dimensional image control path, the automatic, accurate identification and real-time processing of defects are achieved.
It significantly improves the accuracy and automation level of defect detection and handling, achieves seamless integration of defect detection, cleaning and separation, improves the overall efficiency and quality of reed threading operations, and reduces the need for manual intervention.
Smart Images
Figure CN121661037A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of visual image analysis, recognition and processing methods, specifically relating to a high-density wire mesh reed threading control method and system based on visual recognition. Background Technology
[0002] High-density reed threading is a crucial step in the textile industry, and its quality directly affects the performance of the final product and the efficiency of subsequent processing. With the development of textile technology and the continuous improvement of product quality requirements, higher demands are placed on the precision, automation level, and efficiency of the reed threading process. Traditional reed threading methods or systems with low automation levels are difficult to meet the production needs of high density and high precision. In particular, when dealing with increasingly complex defect types, the need for automated control, accurate defect detection based on vision recognition, and efficient defect handling capabilities is becoming increasingly urgent.
[0003] Problems with existing technology: Currently, the field of high-density wire mesh reed threading mostly employs manual visual inspection or automated inspection methods based on traditional machine vision technology. These methods can usually identify relatively obvious defects, such as large-area damage or significant foreign objects. However, for complex and delicate issues that occur during the high-density wire mesh reed threading process, such as fine dirt, slight adhesion between reeds, extrusion deformation, or abnormal micro-gap, existing technologies still have limitations in terms of automation, accurate identification, and real-time processing. In addition, the cleaning and separation of defects often require manual intervention or are carried out through independent, inefficient mechanical devices, making it difficult to achieve seamless integration with the inspection process. Summary of the Invention
[0004] The purpose of this invention is to provide a high-density wire mesh reed threading control method and system based on visual recognition, which can improve the accuracy and efficiency of reed threading control and solve the problem that it is difficult to achieve automated and accurate identification of reed threading defects (such as dirt, adhesion, compression, and abnormal spacing) in the existing high-density wire mesh reed threading process.
[0005] The specific technical solution adopted by this invention is as follows: A high-density wire mesh reed threading control method based on visual recognition includes the following steps: Acquire reed image data, generate contour data information and dynamic coding model, and adjust coding parameters in real time according to the arrangement density; An image enhancement model is constructed, which generates composite image data information with fused focus through contour data information and attaches it to a dynamic coding model; Based on image focus, hierarchical generation and extraction of image data element partitions are performed; Based on the partitioning of image data elements and combined with contour data information, a depth analysis model and separation method for the corresponding focus area are generated. By using a deep analysis model to correlate healthy elements in image data feature partitions, cleanup instructions and cleanup methods are generated. By attaching cleaning instructions and cleaning methods, deep analysis models and separation methods to the dynamic coding model, a real-time 3D image based on the control path is constructed. Based on real-time 3D images and control paths, the reed threading control is completed.
[0006] According to another aspect of the present invention, the image data is acquired through a visual imaging system, the visual imaging system including a switchable backlight compensation module, which generates a composite image by combining a neural network: When the backlight is turned on, it is used to identify the position, spacing, arrangement, and connection status of the orderly arranged reeds. When the backlight is off, the boundary relationship of the adhered reed pieces is analyzed by image contrast, and the connection status of the reed pieces is determined by combining the backlight-thickness relationship model.
[0007] According to another aspect of the present invention, the dynamic coding model includes: A basic code is generated based on the initial reed spacing and reed thickness; Real-time monitoring of reed density fluctuations and dynamic correction of encoding parameters through algorithms; The system combines deep learning models to predict potential adhesion risk areas and generates needle expansion separation priority instructions.
[0008] According to another aspect of the present invention, the separation method includes: Based on the separation path points in the 3D image, the needle expander is controlled to perform a separation operation. When a broken or tilted reed is detected, the alarm module is triggered and the automated process is paused, generating a manual maintenance instruction.
[0009] According to another aspect of the present invention, the method further includes the following steps: Control the cutting device to complete the reed threading and wire hanging operation, and record the quality inspection results during the threading process; The quality inspection results include: Record the health factor classification data for each reed insertion point; Generate a visual quality report, upload it to the production management system, and update the correction parameters of the dynamic coding model synchronously.
[0010] According to another aspect of the present invention, the three-dimensional image construction process includes: The real-time acquired separation path points and cleaning positions are mapped to a three-axis motion trajectory model, and the path point coordinates are dynamically adjusted based on the cleaning position during the reed-threading process; By optimizing the separation path priority through reinforcement learning algorithms, the device travel time can be reduced.
[0011] A high-density wire mesh reed threading control system based on vision recognition includes: The image acquisition and enhancement module is equipped with a visual imaging system with switchable backlight compensation, which is used to acquire reed image data and generate composite images with fused focus, and combined with a neural network model to realize the position recognition of ordered reeds and the analysis of adhesion boundaries. The dynamic modeling unit generates basic coding parameters based on the initial reed spacing and reed thickness, dynamically corrects the coding model by monitoring density fluctuations in real time, and integrates deep learning algorithms to predict areas of adhesion risk. The partitioning analysis engine classifies and partitions composite images according to image focus, combines contour data to generate a depth analysis model and separation method, and generates cleaning instructions by associating health elements. The three-dimensional path control module, based on the mapping of separation path points and cleaning positions to a three-axis motion trajectory model, optimizes the separation path priority through reinforcement learning algorithms, controls the needle expansion device to perform separation operations, and triggers an alarm module to pause the process when a breakage or tilt is detected. The quality feedback system records the health factor classification data and visualized quality reports of the reed placement location, and synchronously updates the correction parameters of the dynamic coding model to the production management system.
[0012] According to another aspect of the present invention, an electronic device is also provided, the electronic device including a memory and a processor; the memory is used to store a program; the processor executes the program to implement the method described in any one of the foregoing.
[0013] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a computer program that, when executed by a processor, implements the method described in any one of the preceding embodiments.
[0014] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in any one of the preceding embodiments.
[0015] The technical effects achieved by this invention are as follows: This invention achieves automated and accurate identification, cleaning, and separation of defects in high-density wire mesh reed by constructing a dynamic coding model based on visual recognition, an image enhancement model, a depth analysis model, and a three-dimensional image working path, combined with a precision motion control module and a multi-functional integrated actuator.
[0016] This invention, by combining focus and grading methods, can perform high-precision automated identification, region extraction, and targeted processing of complex and subtle defects such as reed dirt, adhesion, compression, and abnormal spacing. It significantly improves the accuracy and automation level of defect detection and processing, and further realizes the seamless integration and automated work path of defect detection, cleaning, and separation. This greatly improves the overall efficiency and quality of reed threading operations, and significantly reduces the need for and frequency of manual intervention.
[0017] This invention can correct the coding and analysis model parameters in real time based on the processing results, ensuring the long-term stability of the system and the continuous optimization of the processing effect, thereby improving the system's intelligence level and robustness.
[0018] This invention improves the automation level of cleaning and reed threading control by combining cleaning instructions and cleaning methods, deep analysis models and separation methods with dynamic coding models to generate control paths and real-time 3D images, significantly improving the overall efficiency and quality of reed threading operations, enhancing system robustness, and enabling continuous optimization. Attached Figure Description
[0019] Figure 1 This is a flowchart of the reed-threading control method of the present invention; Figure 2 This is a schematic diagram of the reed-threading control system of the present invention; Figure 3 This is a schematic diagram showing the position between the image acquisition and enhancement module and the reed in this invention. Detailed Implementation
[0020] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.
[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0022] According to embodiments of the present invention, a method embodiment of a high-density wire mesh reed threading control method based on visual recognition is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.
[0023] like Figure 1 As shown, the high-density wire mesh reed threading control method based on visual recognition includes the following steps: S1. Acquire reed image data, generate contour data information and dynamic coding model, and adjust coding parameters in real time according to the arrangement density; S2. Construct an image enhancement model, generate composite image data information with fused focus through contour data information, and attach it to the dynamic coding model; S3. Based on the image focus, generate and extract image data element partitions in a hierarchical manner; S4. Based on the image data element partitioning and combined with contour data information, generate a depth analysis model and separation method for the corresponding focus area; S5. By using a deep analysis model to associate healthy elements in image data feature partitions, cleanup instructions and cleanup methods are generated: S6. By attaching cleaning instructions and cleaning methods, depth analysis models and separation methods to the dynamic coding model, a real-time 3D image based on the control path is constructed. S7. Based on the real-time 3D image and control path, complete the reed threading control.
[0024] In step S1 above, the image data refers to the data of the reeds and the gaps between them. The acquisition methods include, but are not limited to, taking pictures, recording videos, and laser scanning. The contour data information is the arrangement or method of the reeds, and the contour data information is obtained by analyzing the image data.
[0025] Furthermore, image data is acquired through a visual imaging system, which includes a switchable backlight compensation module and generates composite images by combining them with a neural network. When the backlight is turned on, it is used to identify the position, spacing, arrangement, and connection status of the orderly arranged reeds. When the backlight is off, the boundary relationship of the adhered reeds is analyzed by image contrast, and the connection status of the reeds is determined by combining the backlight-thickness relationship model. When the backlight is off, glare and diffuse reflection are avoided from affecting the image acquisition results of the image data acquisition device.
[0026] The hardware configuration of the visual imaging system is as follows: It includes a switchable backlight compensation module (such as an LED backlight panel, light strip, etc.) and a high-resolution industrial camera. The backlight compensation module can switch on / off states via a controller (such as a PLC) to ensure optimal lighting conditions in different scenarios. For example, when the backlight is on, the light source illuminates the reeds evenly, highlighting the orderly arrangement of the reed positions, spacing, and connection status; when the backlight is off, the image contrast is enhanced by ambient light or side light, highlighting the boundary details of the adhesion areas.
[0027] The backlight compensation module operates in the following ways: When the backlight is on, the outline of the reeds in the image is clearer, making it easier to identify the position, spacing and arrangement of the orderly arranged reeds (such as linear arrangement and staggered arrangement). The center point of the reeds is quickly located by template matching algorithm (such as normalized cross-correlation NCC) and the adjacent spacing is calculated. In addition, the connection status of the reeds (such as whether the welding point is complete) is detected by morphological operations (such as opening operation).
[0028] When the backlight is off, the boundary relationship of the stuck reeds is analyzed by relying on image contrast. For example, the local contrast of the image is enhanced by histogram equalization, and the boundary of the stuck area is extracted by edge detection algorithm. At the same time, the backlight-thickness relationship model (a mathematical model established based on experimental data) is introduced. The thickness of the reeds is inferred from the gray value of the image when the backlight is off, and the sticking status is further determined (such as whether the sticking is caused by the change in thickness).
[0029] Furthermore, the generation of composite images relies on deep learning models (such as generative adversarial networks, GANs). For example, a clear outline image with backlighting on can be used as input to train a generator network to simulate the adhesion boundary features when the backlighting is off. The final output is a composite image that fuses the two lighting conditions, which improves the model's adaptability to complex scenes, reduces recognition errors under single lighting conditions, and enhances the robustness and accuracy of image data.
[0030] It should be further explained that the dynamic coding model is based on the sequential coding of contour data information and the corresponding spacing between reeds. The coding includes at least one set of codes corresponding to reeds, and may also include codes generated by the spacing between reeds.
[0031] Furthermore, by combining a visual imaging system with switchable backlight compensation, it ensures that the clear reed arrangement can be captured under different lighting conditions (such as backlight on / off). Contour data information is extracted through edge detection algorithms (such as the Canny operator) to characterize the geometry, arrangement order, and spacing of the reeds. The dynamic coding model generates basic codes (e.g., binary or grayscale codes) based on the initial reed spacing and reed thickness, and dynamically corrects the coding parameters by monitoring density fluctuations in real time (such as small changes in the spacing between adjacent reeds, changes in the vertical distance between the two ends of the reeds, etc.) to ensure that the model adapts to dynamic changes in the production process. This process includes real-time acquisition, or pre-scanning to generate complete reed image data to achieve advance path planning.
[0032] According to step S2, the image enhancement model is constructed by combining contour data information with neural networks (such as convolutional neural networks CNN) to generate composite images. This model uses an attention mechanism to weight key regions in the image (such as adhesion boundaries, abnormal spacing, etc.) to generate composite images with "focus". For example, areas with dirt or adhesion are given higher attention, so they are prioritized or focused on in subsequent analysis. This process relies on parameters in the dynamic coding model (such as the current reed density) as input to ensure that the enhancement model is synchronized with the actual production status.
[0033] Furthermore, the corresponding elements of the image enhancement model include at least the arrangement of reeds and their corresponding codes.
[0034] According to step S3, it should be noted that the partitioning of image data elements is based on the grading results of focus level. For example, the focus level is divided into three levels: high (adhesion risk area), medium (slight dirt), and low (normal area). The pixels in the image are classified by clustering algorithm (such as K-means) to extract areas of different levels. The key to this step is to achieve efficient data segmentation, reduce redundant calculations, ensure the integrity of key areas, and implement corresponding cleaning speeds or materials for different focus areas. For example, when there is dirt such as sticky thread ends, a device such as a vacuum cleaner head is used for cleaning. If it is classified as sticky or clamped between adjacent reeds, clamps are used for clamping. The cleaning device may also include a micro robotic arm with replaceable head tools, a micro five-axis clamping mechanism, etc.
[0035] Furthermore, focus can be graded using the following methods: reed arrangement (low), dirt on the reed surface or between reeds (high or medium level depending on the amount or type of dirt), reed adhesion (high level; adhesion usually refers to local adhesion or connection between adjacent reeds, reflecting reed deformation, etc.), reed compression (high level; compression usually refers to the overall or more than two-thirds of adjacent reeds being in contact, possibly indicating overall deformation between reeds or reed damage, etc.), and abnormal spacing (high level; usually refers to large spacing between reeds, or different spacing between the ends of adjacent reeds, such as tilting or breakage). Specifically, focus is assigned different levels based on the different conditions mentioned above (this process is also called grading). The function of focus is to further magnify the local area during the reed threading control operation, used to generate maintenance instructions, and through magnification, to improve recognition accuracy and efficiency.
[0036] Furthermore, image data element partitioning refers to the extraction of regions from various reed data for hierarchical or focus-based recognition.
[0037] According to step S4, for image data feature partitioning, the deep analysis model further magnifies key areas (such as adhesion boundaries), and generates separation strategies by combining contour data information. For example, for high-focus areas, a local magnification algorithm (such as bilinear interpolation) is used to improve resolution, and the boundaries are accurately extracted through morphological operations (such as erosion and dilation). The separation method is based on a deep learning model (such as U-Net) to predict the optimal separation path and generate a needle expansion separation priority instruction.
[0038] According to step S5, the classification of health elements is based on the results of image data element partitioning. For example, areas of adhesion, compression, and dirt are marked as unhealthy elements, while normal areas are marked as healthy elements. The generation of cleaning instructions needs to be combined with the type of health element (such as dirt or breakage) and the corresponding cleaning method (such as vacuum cleaning or manual repair) is selected. This step relies on parameters in the dynamic coding model (such as the current reed position) for real-time positioning or pre-scanning acquisition.
[0039] According to step S6, the cleaning instructions and cleaning methods, deep analysis models and separation methods are attached to the dynamic coding model, as well as the coding in the dynamic model, to construct a working path that enables the machine to perform cleaning and separation simultaneously. For example, the separation path points and cleaning positions are mapped by a three-axis motion trajectory model (X / Y / Z axes). Another example is to generate three-dimensional path point coordinates for the adhesion area and combine reinforcement learning algorithms to optimize path priority, reduce the device's idle travel time, and improve the cleaning optimization speed during the reed threading process.
[0040] According to step S7, the final control is based on real-time 3D images and path planning to drive the needle expansion device to perform separation and cleaning operations. If a broken or tilted reed is detected, an alarm module is triggered and the process is paused to ensure production safety.
[0041] Based on the above steps, through multi-stage data processing and model collaboration, real-time monitoring and dynamic adjustment of the high-density wire mesh reed threading process are achieved, which significantly improves production efficiency and product quality while reducing manual intervention.
[0042] In practical applications, the real-time dynamic positioning scan and advance scan methods are preferably performed periodically. For example, if the inspection is correct or there is no quality problem with the reed thread, the advance scan method is used to check for defects. In actual operation, the application of computing power is reduced. After three cycles of actual operation, the real-time dynamic positioning scan is started to check for potential problems.
[0043] As an optional embodiment, the dynamic coding model includes: A basic code is generated based on the initial reed spacing and reed thickness; Real-time monitoring of reed density fluctuations and dynamic correction of encoding parameters through algorithms; The system combines deep learning models to predict potential adhesion risk areas and generates needle expansion separation priority instructions.
[0044] Based on the above, the basic coding is generated based on the initial reed spacing (such as the design value) and reed thickness (such as the nominal value), and adopts binary or grayscale coding form. For example, the reed spacing is discretized into a fixed-length coding sequence, and each code corresponds to the state (normal / abnormal) of a reed position. The initial parameters are optimized by combining historical production data to ensure the universality of the coding model.
[0045] Furthermore, the density fluctuations of the reeds are monitored in real time (e.g., by measuring the distance between adjacent reeds using a laser sensor), and the density change trend is analyzed using a sliding window algorithm (e.g., moving average). If a sudden density change is detected (e.g., the local spacing shrinks beyond a threshold), the encoding parameters are dynamically adjusted (e.g., the attention weight of the corresponding area is increased) to ensure that the model is synchronized with the actual production status.
[0046] Furthermore, deep learning predicts adhesion risk areas. Deep learning models (such as LSTM or Transformer) are used to analyze historical production data (such as image features and encoding parameters when adhesion occurs) to predict potential adhesion risk areas. For example, the current encoding parameters and image data are input, and a risk probability distribution map is output, and a needle separation priority instruction is generated (such as prioritizing high-risk areas).
[0047] Based on the above, the dynamic coding model achieves adaptive adjustment to complex production environments through real-time correction and deep learning prediction, significantly reducing production failures caused by adhesion.
[0048] As an optional embodiment, the separation method includes: Based on the separation path points in the 3D image, the needle expander is controlled to perform a separation operation. When a broken or tilted reed is detected, the alarm module is triggered and the automated process is paused, generating a manual maintenance instruction.
[0049] The separation force is dynamically adjusted or pre-adjusted according to the material characteristics of the reed. When the pressure sensed by the needle expander exceeds the threshold during the separation of the reed, it indicates that the needle expander is in the wrong position. The separation is then re-identified by running the reed-piercing control method to avoid breakage of the needle expander or physical damage to the reed.
[0050] Furthermore, the separation path planning based on 3D images involves generating separation path points based on the cleaning positions and risk areas in real-time 3D images, and mapping them to a three-axis motion trajectory model (X / Y / Z axes). For example, a spiral separation path is generated for high-risk areas to ensure that the expanding needle fully contacts the adhesion boundary. The coordinates of the path points need to be dynamically adjusted to take into account possible mechanical vibrations or deformations during the reed threading process.
[0051] Furthermore, the control logic of the needle expander is as follows: the needle expander is driven by a motor and needs to provide real-time feedback of position and force sensor data when performing the separation operation. For example, the needle expander pressure is adjusted through a PID control algorithm to avoid excessive force that could damage the reed. At the same time, the path priority is optimized by combining reinforcement learning algorithms (such as prioritizing high-risk areas) to reduce the device's idle travel time.
[0052] Furthermore, the anomaly detection and alarm mechanism includes: real-time monitoring of the reed's condition (such as breakage or tilting) through a vision system; if an anomaly is detected (such as the reed's center offset exceeding a threshold), an alarm module (such as an audible and visual alarm) is triggered and the automated process is paused. At this time, a manual maintenance instruction is generated and pushed to the operation terminal to ensure timely handling of the fault.
[0053] Based on the above, efficient and safe reed threading operations can be achieved through three-dimensional path planning and intelligent control, while alarm mechanisms ensure production continuity.
[0054] As an optional embodiment, the reed threading control method further includes the following steps: The device is controlled to complete the reed threading and wire hanging operation, and the quality inspection results during the threading process are recorded. When the device is controlled to complete the reed threading and wire hanging operation, it is necessary to accurately position the device by combining the path point coordinates in the three-dimensional image. For example, the reed position can be corrected in real time by using a vision servo system to ensure that the wire hanging action is synchronized with the reed arrangement. Key parameters (such as reed speed and pressure) are recorded during the operation for subsequent analysis. The quality inspection results include: Record the health factor classification data for each reed insertion point; Generate a visual quality report, upload it to the production management system, and update the correction parameters of the dynamic coding model synchronously.
[0055] The quality inspection includes health element classification (such as normal, dirt, adhesion, and breakage) and the generation of visual quality reports. For example, the health status of each reed position is marked by the image data element partitioning results, and a heat map is generated to show the distribution of risk areas. The report includes statistical indicators (such as pass rate and failure rate) and recommended measures (such as cleaning cycle optimization).
[0056] Furthermore, the quality inspection results are synchronized to the production management system (such as MES) and used to correct the dynamic coding model. For example, if a batch has a high frequency of sticking failures, the coding parameters are adjusted (such as increasing the attention weight of the corresponding area) to improve the stability of subsequent production.
[0057] Based on the above, closed-loop control enables continuous optimization of the production process, significantly improving product quality and equipment operating efficiency.
[0058] As an optional embodiment, the 3D image construction process includes: The real-time acquired separation path points and cleaning positions are mapped to a three-axis motion trajectory model, and the path point coordinates are dynamically adjusted based on the cleaning position during the reed-threading process; By optimizing the separation path priority through reinforcement learning algorithms, the device travel time can be reduced.
[0059] Among them, the real-time acquired separation path points (such as the coordinates of the adhesion area) and cleaning positions (such as the coordinates of the dirt area) need to be uniformly mapped to the three-axis motion trajectory model. For example, a coordinate transformation algorithm (such as homogeneous transformation) is used to convert the image coordinate system into the mechanical coordinate system, and the path point coordinates are dynamically adjusted to compensate for mechanical errors.
[0060] Furthermore, reinforcement learning algorithms (such as Q-learning) optimize path priorities by simulating the execution effects of different path strategies (such as idle travel time and separation success rate). For example, training an agent to learn to prioritize high-risk areas reduces the number of device round trips and improves overall efficiency.
[0061] Based on the above, path optimization reduces idle time while ensuring that separation and cleanup operations run simultaneously and improving operational accuracy.
[0062] According to another aspect of the present invention, an electronic device is also provided, the electronic device including a memory and a processor; the memory is used to store a program; the processor executes the program to implement the method of any of the foregoing.
[0063] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a computer program that, when executed by a processor, implements the method of any of the foregoing.
[0064] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in any of the foregoing.
[0065] Please refer to Figure 2 and Figure 3 A high-density wire mesh reed threading control system based on vision recognition includes: The image acquisition and enhancement module is equipped with a visual imaging system with switchable backlight compensation, which is used to acquire reed image data and generate composite images with fused focus, and combined with a neural network model to realize the position recognition of ordered reeds and the analysis of adhesion boundaries. The dynamic modeling unit generates basic coding parameters based on the initial reed spacing and reed thickness, dynamically corrects the coding model by monitoring density fluctuations in real time, and integrates deep learning algorithms to predict areas of adhesion risk. The partitioning analysis engine classifies and partitions composite images according to image focus, combines contour data to generate a depth analysis model and separation method, and generates cleaning instructions by associating health elements. The three-dimensional path control module, based on the mapping of separation path points and cleaning positions to a three-axis motion trajectory model, optimizes the separation path priority through reinforcement learning algorithms, controls the needle expansion device to perform separation operations, and triggers an alarm module to pause the process when a breakage or tilt is detected. The quality feedback system records the health factor classification data and visualized quality reports of the reed placement location, and synchronously updates the correction parameters of the dynamic coding model to the production management system.
[0066] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.
[0067] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.
Claims
1. A high-density wire mesh reed threading control method based on visual recognition, characterized in that, Includes the following steps: Acquire reed image data, generate contour data information and dynamic coding model, and adjust coding parameters in real time according to the arrangement density; An image enhancement model is constructed, which generates composite image data information with fused focus through contour data information and attaches it to a dynamic coding model; Based on image focus, hierarchical generation and extraction of image data element partitions are performed; Based on the partitioning of image data elements and combined with contour data information, a depth analysis model and separation method for the corresponding focus area are generated. By using a deep analysis model to correlate healthy elements in image data feature partitions, cleanup instructions and cleanup methods are generated. By attaching cleaning instructions and cleaning methods, deep analysis models and separation methods to the dynamic coding model, a real-time 3D image based on the control path is constructed. Based on real-time 3D images and control paths, the reed threading control is completed.
2. The high-density wire mesh reed threading control method based on visual recognition according to claim 1, characterized in that, The image data is acquired through a visual imaging system, which includes a switchable backlight compensation module and generates composite images by combining them with a neural network. When the backlight is turned on, it is used to identify the position, spacing, arrangement, and connection status of the orderly arranged reeds. When the backlight is off, the boundary relationship of the adhered reed pieces is analyzed by image contrast, and the connection status of the reed pieces is determined by combining the backlight-thickness relationship model.
3. The high-density wire mesh reed threading control method based on visual recognition according to claim 1, characterized in that, The dynamic coding model includes: A basic code is generated based on the initial reed spacing and reed thickness; Real-time monitoring of reed density fluctuations and dynamic correction of encoding parameters through algorithms; The system combines deep learning models to predict potential adhesion risk areas and generates needle expansion separation priority instructions.
4. The high-density wire mesh reed threading control method based on visual recognition according to claim 1, characterized in that, The separation method includes: Based on the separation path points in the 3D image, the needle expander is controlled to perform a separation operation. When a broken or tilted reed is detected, the alarm module is triggered and the automated process is paused, generating a manual maintenance instruction.
5. The high-density wire mesh reed threading control method based on visual recognition according to claim 1, characterized in that, It also includes the following steps: Control the cutting device to complete the reed threading and wire hanging operation, and record the quality inspection results during the threading process; The quality inspection results include: Record the health factor classification data for each reed insertion point; Generate a visual quality report, upload it to the production management system, and update the correction parameters of the dynamic coding model synchronously.
6. The high-density wire mesh reed threading control method based on visual recognition according to claim 1, characterized in that, The three-dimensional image construction process includes: The real-time acquired separation path points and cleaning positions are mapped to a three-axis motion trajectory model, and the path point coordinates are dynamically adjusted based on the cleaning position during the reed-threading process; By optimizing the separation path priority through reinforcement learning algorithms, the device travel time can be reduced.
7. A high-density wire mesh reed threading control system based on visual recognition, operating the method described in any one of claims 1-6, characterized in that, include: The image acquisition and enhancement module is equipped with a visual imaging system with switchable backlight compensation, which is used to acquire reed image data and generate composite images with fused focus, and combined with a neural network model to realize the position recognition of ordered reeds and the analysis of adhesion boundaries. The dynamic modeling unit generates basic coding parameters based on the initial reed spacing and reed thickness, dynamically corrects the coding model by monitoring density fluctuations in real time, and integrates deep learning algorithms to predict areas of adhesion risk. The partitioning analysis engine classifies and partitions composite images according to image focus, combines contour data to generate a depth analysis model and separation method, and generates cleaning instructions by associating health elements. The three-dimensional path control module, based on the mapping of separation path points and cleaning positions to a three-axis motion trajectory model, optimizes the separation path priority through reinforcement learning algorithm, controls the needle expansion device to perform separation operations, and triggers an alarm module to pause the process when a breakage or tilt is detected. The quality feedback system records the health factor classification data and visualized quality reports of the reed placement location, and synchronously updates the correction parameters of the dynamic coding model to the production management system.
8. An electronic device, characterized in that, The electronic device includes a memory and a processor; The memory is used to store programs; The processor executes the program to implement the method of any one of claims 1 to 6.
9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.