Machine vision fused weftless tape machine process parameter self-adaptive acquisition system
By integrating machine vision into the adaptive acquisition system for process parameters of non-woven tape machines, the width, cord spacing, and surface defects of the non-woven tape can be monitored and adjusted in real time. This solves the problems of low efficiency and subjective influence of traditional inspection methods, and achieves efficient quality control and material utilization.
Patent Information
- Application Number
- CN202511454944.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-13
- Publication Date
- 2025-11-11
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In traditional non-woven tape production, process parameter control relies on manual timed sampling or offline testing, which is inefficient, easily affected by subjective factors, and lacks comprehensive perception and real-time feedback on product quality, resulting in defective products and material waste.
The adaptive acquisition system for process parameters of the non-woven tape machine, which integrates machine vision, captures images in real time through a linear scan camera and multimodal lighting equipment. Combined with data processing and control modules, it enables real-time monitoring and adaptive adjustment of the width, cord spacing, and surface defects of the non-woven tape.
It enables multi-dimensional real-time quality monitoring of the non-woven tape production process, improves the production pass rate, reduces the generation of defective products, and ensures product quality consistency and material utilization.
Smart Images

Figure CN120931639A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial automation technology, specifically to an adaptive acquisition system for process parameters of a non-woven belt conveyor that integrates machine vision. Background Technology
[0002] In the production of non-woven tires, many process parameters must be precisely controlled within extremely small tolerances. These key parameters include, but are not limited to: the total width of the sheet, the straightness of the edges, the spacing (or density) of the cords, the arrangement angle of the cords, and the quality of the sheet surface (such as the presence of defects such as bubbles, impurities, and broken rubber). Any slight deviation in these parameters may lead to serious quality problems in the final tire product, such as stress concentration, high-speed vibration, or premature failure.
[0003] Traditional quality control methods largely rely on manual, timed sampling or offline testing. This approach is not only inefficient and slow to respond, but the test results are also easily affected by the subjective factors of the operators, making it difficult to guarantee the consistency of product quality and potentially leading to a large number of defective products and serious material waste.
[0004] To overcome the drawbacks of manual inspection, some production lines employ sensors for online monitoring. For example, load cells or tension sensors are used to monitor and control sheet tension, or X-ray sensors are used to measure sheet thickness (gram weight). However, these single-parameter sensor systems cannot provide geometric information about the product and lack a comprehensive, holistic understanding of product quality.
[0005] In recent years, machine vision technology has been widely used for surface defect detection on various rolled materials, such as textiles, paper, and thin metal sheets. These systems typically employ high-speed cameras and image processing algorithms to identify surface defects. However, most of these vision inspection systems exist as independent quality inspection "checkpoints," primarily functioning to mark or reject defective products, rather than directly participating in real-time feedback control of the production process. They generally do not provide accurate geometric parameter measurements suitable for closed-loop control, or even if they do provide measurement data, they fail to form a tightly integrated, adaptive control loop with the actuators of the production equipment (such as guide rollers and tension regulators). Summary of the Invention
[0006] To address the aforementioned deficiencies in the prior art, this invention provides an adaptive acquisition system for process parameters of a non-woven belt conveyor that integrates machine vision.
[0007] The present invention provides an adaptive acquisition system for process parameters of a non-woven tape machine that integrates machine vision, comprising a data acquisition module, a data processing and parameter extraction module, and a control module. The data acquisition module is set on the production path of the non-woven tape machine and is used to capture digital images of the non-woven tape flowing through its field of view in real time. The data processing and parameter extraction module is electrically connected to the data acquisition module and is used to receive and process the digital image and extract process parameters from the digital image. The process parameters include material width, cord spacing and surface defect information. The control module is electrically connected to the data processing and parameter extraction module and the controllable execution component of the weftless belt machine, respectively. It is used to receive the process parameters, compare them with the preset target parameter values to generate a deviation signal, and generate a control command based on the deviation signal and send it to the controllable execution component of the weftless belt machine to drive it to make adjustments, so that the process parameters approach the target parameter values.
[0008] Preferably, the data acquisition module includes a linear scan camera and a multimodal illumination device.
[0009] Preferably, the multimodal lighting device includes: A backlight illumination unit is disposed below the non-woven strip and is used to generate a high-contrast contour image of the non-woven strip; A dark field illumination unit is positioned above the non-woven belt to illuminate it at a low angle, thereby highlighting the surface cord texture and surface defects.
[0010] Preferably, the data processing and parameter extraction module is configured as follows: Material width is extracted from the contour image generated by the backlight unit; Based on the texture image generated by the dark field illumination unit, information on the curtain spacing and surface defects is extracted.
[0011] Preferably, the extraction of material width from the contour image generated by the backlight illumination unit specifically includes: Non-maximum suppression and hysteresis thresholding are applied to the contour image to generate edge contours with a single pixel width. For each pixel row in the contour image, the horizontal coordinates of the left and right edge points are identified from the edge contour; The difference between the horizontal coordinates of the right edge point and the left edge point is calculated, and the calculation results of multiple rows are averaged to determine the width of the material.
[0012] Preferably, the data processing and parameter extraction module is further configured to run a deep learning-based defect classification model for identifying and classifying surface defects.
[0013] Preferably, the deep learning-based defect classification model is a convolutional neural network, which is trained on an image dataset containing multiple labeled defect types to classify the surface defects into preset defect types, including bubbles, scratches, and uneven coatings.
[0014] Another aspect of the present invention includes an adaptive acquisition method for process parameters of a weftless tape machine that integrates machine vision, comprising the following steps: Continuous digital images of the weftless belt are captured using a data acquisition module installed on the production path of the weftless belt machine; The digital image is analyzed by the data processing and parameter extraction module to extract process parameters, including material width, cord spacing and surface defect information. In the control module, the process parameters are compared with preset target parameter values to generate a deviation signal; Based on the deviation signal, control commands are generated and sent to the controllable actuators of the non-woven belt machine for closed-loop adjustment, thereby making the real-time process parameters approach the target parameter values.
[0015] Compared with the prior art, the advantages of this invention are: By employing multimodal illumination (combining backlighting and dark field), this invention can simultaneously capture contour images for precise dimensional measurements and texture images for analyzing cord distribution and surface defects in a single scan. This enables the system to monitor key parameters across multiple dimensions, such as width, edge straightness, cord spacing, and surface defects, in real time and synchronously, gaining a comprehensive and holistic understanding of the product's quality status.
[0016] The system compares the extracted process parameters with the target values in real time, and generates control commands in real time through control algorithms to drive the actuators of the non-woven belt machine (such as guide rollers and tension regulators) to make adaptive adjustments. This seamless integration from detection to control transforms quality control from reactive post-production screening to proactive prevention and real-time optimization during the process, significantly improving the first-pass yield of production. Attached Figure Description
[0017] Figure 1 This is a system architecture diagram of the adaptive acquisition system for process parameters of a non-woven tape machine that integrates machine vision, as proposed in this invention. Detailed Implementation
[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0019] refer to Figure 1 The adaptive acquisition system for process parameters of the non-woven tape machine that integrates machine vision includes a data acquisition module, a data processing and parameter extraction module, and a control module.
[0020] The data acquisition module is located on the production path of the non-woven tape machine and is used to capture digital images of the non-woven tape flowing through its field of view in real time. This module includes a linear scan camera and a multimodal illumination system.
[0021] In order to highlight the features that need to be detected to the greatest extent and suppress the interference of irrelevant information, this scheme constructs a specific optical path through a backlight illumination unit and a dark field illumination unit to extract information in different dimensions.
[0022] The backlight unit is a large-area, highly uniform diffused light source mounted below the non-woven belt. When light passes through the camera lens, the opaque non-woven belt blocks the light, creating a high-contrast black silhouette on the camera sensor. This illumination method is used to measure the outer contour dimensions of objects, accurately determining the left and right edge positions of the non-woven belt to calculate the total width and detect any irregularities such as burrs or gaps at the edges.
[0023] The dark field illumination unit consists of one or more strip light sources that illuminate from the sides or front of the non-woven zone at an extremely low angle (close to grazing). Under this illumination geometry, a flat, smooth rubber surface reflects light away from the camera, thus appearing as a dark background in the image. Conversely, any structure with three-dimensional texture features, such as raised filaments, surface pits, bubbles, or attached foreign matter, scatters light towards the camera, thus standing out brightly against a dark background. Dark field illumination is used to detect minute surface imperfections and enhance texture details.
[0024] By combining backlighting with dark field illumination, the system can simultaneously acquire contour information for dimensional measurement and texture information for surface quality analysis in a single scan.
[0025] The data processing and parameter extraction module is electrically connected to the data acquisition module and is used to receive and process the digital image and extract process parameters from the digital image.
[0026] Specifically, it includes: First, continuous line data is received from a line scan camera. Simultaneously, real-time speed and position information of the production line is acquired via an encoder connected to the drive roller of the non-woven belt conveyor to ensure the correct aspect ratio between rows of images in the direction of motion.
[0027] It is understandable that the original image will inevitably be affected by factors such as electromagnetic interference, dust, and vibration in the industrial environment, resulting in noise. Preprocessing is necessary to improve the accuracy of subsequent processing. This embodiment uses a Gaussian filter or a median filter to smooth the image. These filters can effectively suppress random noise while better preserving edge and detail information in the image, avoiding excessive blurring.
[0028] Next, based on the contour image generated by the backlight illumination unit, the material width is extracted, specifically: Non-maximum suppression and hysteresis thresholding are applied to the contour image to generate a single-pixel-width edge contour.
[0029] For each pixel row in the contour image, the horizontal coordinates of the left and right edge points are identified from the edge contour.
[0030] The difference between the horizontal coordinates of the right edge point and the left edge point is calculated, and the calculation results of multiple rows are averaged to determine the width of the material.
[0031] Next, based on the texture image generated by the dark field illumination unit, the curtain spacing information is extracted, specifically: Select one or more representative regions of interest from the texture image, and apply a two-dimensional fast Fourier transform to each region of interest to convert it from the spatial domain to the frequency domain.
[0032] In the generated spectrum, the periodic parallel structure of the cord will appear as a pair of bright energy concentration points (peaks) that are symmetrical about the center of the spectrum. The positions of these peaks directly correspond to the dominant frequency and direction of the cord structure.
[0033] These energy peaks are automatically located using a peak detection algorithm. The distance from the peak to the center of the spectrum is proportional to the spatial frequency of the curtain. The average curtain spacing in the region of interest can be calculated using this distance. Meanwhile, the direction of the straight line connecting the peak and the center reflects the arrangement angle of the curtain.
[0034] Then, based on the contour image generated by the backlight illumination unit, a deep learning-based defect classification model is used to identify and classify surface defects.
[0035] This approach trains the model using a large dataset containing numerous labeled images. In each image, various defects (such as bubbles, scratches, and uneven coatings) are manually and precisely labeled with their location and category. This dataset is then used to train a convolutional neural network (e.g., an efficient object detection network like YOLOv8). During training, the network automatically learns and masters deep, abstract visual features that distinguish different defect types. After training, the model enters an online real-time inference phase, where it is deployed to the data processing and parameter extraction module. The system inputs real-time captured texture images into the model, and the model outputs the location (usually represented by bounding boxes) of any potential defects in the image and their corresponding category.
[0036] Finally, all extracted parameters (such as width value, average spacing, defect type, and coordinates) are packaged into a structured data frame and sent to the control module via industrial Ethernet.
[0037] The control module is electrically connected to both the data processing and parameter extraction module and the controllable actuator of the weftless belt machine. It receives the real-time process parameters, compares them with preset target parameter values to generate a deviation signal, and generates a control command based on the deviation signal, sending it to the controllable actuator of the weftless belt machine to drive it to adjust, thereby bringing the real-time process parameters closer to the target parameter values. Specifically, it includes: First, the system receives data frames transmitted from the data processing and parameter extraction module and parses out various real-time process parameters.
[0038] Next, the target parameter values and allowable tolerance ranges stored internally by the operator are compared with the received real-time parameter values.
[0039] If the absolute value of the deviation exceeds the preset tolerance range, for the control of continuous variables (such as width and spacing), this module uses a proportional-integral-derivative (PID) control algorithm to calculate an optimal control output.
[0040] Ultimately, this module translates the calculated control output into specific operational instructions for the connected actuators. For example: Width deviation adjustment: If the width is too small, an analog voltage signal is output to drive the servo motor to fine-tune the position of the edge guide roller, so that it expands outward; or the opening of the extruder die head is adjusted.
[0041] Cord spacing deviation adjustment: If the cord spacing is too large, the output signal instructs a certain tension controller on the cord frame to increase the tension, or to finely adjust the angle of the cord combing guide device.
[0042] Defect Handling: If a serious defect exists, an alarm system (audio-visual alarm) is triggered, and an online marking device (such as inkjet or labeling) is activated to mark the corresponding position on the non-woven tape for subsequent processes to identify and reject. Simultaneously, information such as the type, location, and time of the defect is recorded in the production log.
[0043] Another aspect of the present invention includes an adaptive acquisition method for process parameters of a weftless tape machine that integrates machine vision, comprising the following steps: Continuous digital images of the weftless belt are captured using a data acquisition module installed on the production path of the weftless belt machine; The digital image is analyzed by the data processing and parameter extraction module to extract process parameters, including material width, cord spacing and surface defect information. In the control module, the process parameters are compared with preset target parameter values to generate a deviation signal; Based on the deviation signal, control commands are generated and sent to the controllable actuators of the non-woven belt machine for closed-loop adjustment, thereby making the real-time process parameters approach the target parameter values.
[0044] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0045] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims
1. An adaptive acquisition system for process parameters of a non-woven belt conveyor integrating machine vision, characterized in that, It includes a data acquisition module, a data processing and parameter extraction module, and a control module; The data acquisition module is set on the production path of the non-woven tape machine and is used to capture digital images of the non-woven tape flowing through its field of view in real time. The data processing and parameter extraction module is electrically connected to the data acquisition module and is used to receive and process the digital image and extract process parameters from the digital image. The process parameters include material width, cord spacing and surface defect information. The control module is electrically connected to the data processing and parameter extraction module and the controllable execution component of the weftless belt machine, respectively. It is used to receive the process parameters, compare them with the preset target parameter values to generate a deviation signal, and generate a control command based on the deviation signal and send it to the controllable execution component of the weftless belt machine to drive it to make adjustments, so that the process parameters approach the target parameter values.
2. The adaptive acquisition system for process parameters of a non-woven belt conveyor assemblies fused with machine vision according to claim 1, characterized in that, The data acquisition module includes a linear scan camera and a multimodal illumination device.
3. The adaptive acquisition system for process parameters of a non-woven belt conveyor assemblies incorporating machine vision as described in claim 2, characterized in that, The multimodal lighting device includes: A backlight illumination unit is disposed below the non-woven strip and is used to generate a high-contrast contour image of the non-woven strip; A dark field illumination unit is positioned above the non-woven belt to illuminate it at a low angle, thereby highlighting the surface cord texture and surface defects.
4. The adaptive acquisition system for process parameters of a non-woven belt conveyor assemblies fused with machine vision according to claim 3, characterized in that, The data processing and parameter extraction module is configured as follows: Material width is extracted from the contour image generated by the backlight unit; Based on the texture image generated by the dark field illumination unit, information on the curtain spacing and surface defects is extracted.
5. The adaptive acquisition system for process parameters of a non-woven belt conveyor fusion machine according to claim 4, characterized in that, The contour image generated based on the backlight illumination unit is used to extract the material width, specifically including: Non-maximum suppression and hysteresis thresholding are applied to the contour image to generate edge contours with a single pixel width. For each pixel row in the contour image, the horizontal coordinates of the left and right edge points are identified from the edge contour; The difference between the horizontal coordinates of the right edge point and the left edge point is calculated, and the calculation results of multiple rows are averaged to determine the width of the material.
6. The adaptive acquisition system for process parameters of a non-woven belt conveyor fusion machine according to claim 4, characterized in that, The data processing and parameter extraction module is also configured to run a deep learning-based defect classification model for identifying and classifying surface defects.
7. The adaptive acquisition system for process parameters of a non-woven belt conveyor assemblies fused with machine vision according to claim 6, characterized in that, The deep learning-based defect classification model is a convolutional neural network, which is trained on an image dataset containing multiple labeled defect types to classify the surface defects into preset defect types, including bubbles, scratches, and uneven coatings.
8. A method for adaptive acquisition of process parameters of a non-woven belt conveyor integrating machine vision, characterized in that, Includes the following steps: Continuous digital images of the weftless belt are captured using a data acquisition module installed on the production path of the weftless belt machine; The digital image is analyzed by the data processing and parameter extraction module to extract process parameters, including material width, cord spacing and surface defect information. In the control module, the process parameters are compared with preset target parameter values to generate a deviation signal; Based on the deviation signal, control commands are generated and sent to the controllable actuators of the non-woven belt machine for closed-loop adjustment, thereby making the real-time process parameters approach the target parameter values.
Citation Information
Patent Citations
Cloth flaw detection method and system based on machine vision
CN119198751A
Circuit board defect detection system based on machine learning
CN119722588A
Artificial board surface defect intelligent detection method and system based on machine vision
CN120355688A
Automatic inspection of textile webs
DE3639636A1
Illuminating unit for an article-sensing camera
US20020005892A1
Cited By
Vision-based method for evaluating gluing uniformity of water-based casein glue for labeling glass bottles
CN121921322A
A grating roll-to-roll manufacturing apparatus control system
CN122194890B
