Spandex wrap yarn core exposure defect real-time detection system based on industrial vision
Patent Information
- Application Number
- CN202510730996.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-03
- Publication Date
- 2025-09-19
AI Technical Summary
Existing yarn defect detection systems have difficulty effectively identifying exposed core defects in spandex-coated yarns, especially during the production of air-coated yarns. This is because existing image analysis algorithms are unable to accurately extract and analyze specific visual features related to exposed cores, resulting in low recognition accuracy.
A real-time detection system for exposed core defects in spandex covered yarns is constructed based on industrial vision. By identifying the image feature model layer by layer, the network nodes are first identified, and then the spandex inner core and outer covering fibers are identified. Combined with the equipment parameters, the image stretch coefficient and quality parameters are obtained to identify equipment abnormalities.
The accuracy and robustness of identifying spandex covered yarn core defects are improved, misjudgment caused by tension changes is reduced, and accurate identification of equipment abnormalities is achieved.
Smart Images
Figure CN120672679A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image analysis, and in particular to a real-time detection system for spandex coated yarn exposed core defects based on industrial vision. Background Art
[0002] In the actual production process of air-covered yarn, fluctuations in raw material quality, changes in the status of production equipment, improper process parameter settings, and the inherent complexity and instability of the airflow itself can easily lead to various defects in the yarn, which directly affect the quality of the final product.
[0003] With the continuous development and application of image data processing, particularly machine vision technology, in industrial inspection, several general-purpose yarn defect image analysis systems have been introduced into the textile industry. For example, while yarn evenness testers based on capacitance principles can detect overall yarn thickness variations, they cannot identify cosmetic defects such as "exposed cores" caused by material differences through image analysis, nor can they effectively analyze the microscopic morphology of network nodes. On the other hand, while image analysis systems based on conventional optical imaging principles may detect some more obvious defects such as thick sections, fine details, or discolored fibers, existing image processing and analysis algorithms have significant limitations in accurately identifying the unique network node structure of air-wrapped yarns, distinguishing subtle spandex exposed cores from normal yarn surface textures through image analysis, and comprehensively utilizing image information to assess the coating state of the outer fiber. These existing image analysis methods struggle to effectively extract and analyze the specific visual features associated with "exposed cores" defects.
[0004] Therefore, a real-time detection system for spandex covered yarn exposed core defects based on industrial vision is proposed. Summary of the Invention
[0005] The purpose of the present invention is to provide a real-time detection system for exposed core defects of spandex coated yarn based on industrial vision, which obtains production images and equipment parameters; identifies network nodes in the production image to obtain a first image feature; identifies the spandex inner core according to the first image feature to obtain a second image feature; identifies the outer fiber according to the first image feature and the second image feature to obtain a third image feature; obtains the production image feature according to the first image feature, the second image feature and the third image feature; obtains the image stretching coefficient according to the equipment parameters and the production image; obtains the quality parameters according to the image stretching coefficient and the production image feature; obtains the equipment fluctuation data according to the difference between the equipment parameters and the standard parameters; obtains the equipment abnormality coefficient according to the production image feature, the image stretching coefficient and the equipment fluctuation data obtained by continuous production; the present invention realizes the accurate identification of defects through image recognition and analysis.
[0006] To achieve the above object, the present invention provides the following technical solutions: A real-time detection system for spandex covered yarn core defects based on industrial vision, including: The data acquisition unit captures the production image of the air-covered yarn on the production line; based on the acquisition time of the production image, the equipment parameters are collected in the production equipment control system; The image processing unit identifies the network nodes of the air-covered yarn in the production image to obtain a first image feature; identifies the spandex inner core in the production image based on the first image feature to obtain a second image feature; identifies the outer covering fiber in the production image based on the first image feature and the second image feature to obtain a third image feature; and obtains a production image feature based on the first image feature, the second image feature, and the third image feature; The quality recognition unit recognizes the production image in combination with the equipment parameters to obtain the image stretch coefficient; the quality parameters of the air-covered yarn are obtained based on the image stretch coefficient and the production image feature recognition; The equipment early warning unit identifies and obtains equipment fluctuation data based on the difference between equipment parameters and standard parameters; and identifies and obtains equipment abnormality coefficients based on production image features, image stretching coefficients and equipment fluctuation data obtained from continuous production.
[0007] The equipment parameters include raw material feeding parameters, spraying and pressing processing parameters, traction molding parameters, and equipment environment parameters; The raw material feeding parameters include the spandex yarn feeding roller speed, the spandex yarn drafting multiple setting value, the spandex yarn feeding tension, the outer fiber feeding roller speed, the outer fiber overfeed rate, and the outer fiber feeding tension; The spraying processing parameters include the network nozzle air supply pressure, compressed air flow, compressed air temperature and compressed air humidity; The traction forming parameters include yarn output speed, godet speed, winding tension, and winding speed; The device environmental parameters include the temperature and humidity of the environment in which the device is located.
[0008] Constructing an image feature recognition model to recognize production images, including a first feature recognition layer, a second feature recognition layer, and a third feature recognition layer; The first feature recognition layer recognizes the network node area in the production image, extracts the distribution position of the network node in the image and the image features of the distribution area, and obtains the first image feature; the image feature of the distribution area of the network node is the image pixel data of the network node distribution area; The second feature recognition layer recognizes the distribution of the first image feature in the production image to determine the distribution path of the spandex inner core; recognizes the distribution path of the spandex inner core to obtain exposed inner core data and covered inner core data; and obtains the second image feature based on the exposed inner core data and covered inner core data; The third feature recognition layer identifies the outer fiber area of the air-covered yarn based on the distribution of the first image feature and the second image feature in the production image, and obtains the third image feature based on the identification of the outer fiber area; the third image feature includes the distribution data of the outer fiber area and the image pixel data of the distribution area.
[0009] The process of obtaining the image stretching coefficient is as follows: Identify the production image and obtain the production image features; Identifying the second image feature in the production image feature, and obtaining the stretching feature data of the spandex inner core based on the distribution path, distribution type, and distribution structure of the spandex inner core; The image stretch coefficient of the air covered yarn in the production image is identified and determined based on the stretch feature data in combination with the first image feature, the third image feature and the equipment parameters.
[0010] The quality parameter identification process of the air covered yarn includes: Constructing a covered yarn quality recognition model; identifying the image stretch coefficient and production image features through the covered yarn quality recognition model to determine quality parameters; The training process of the covered yarn quality recognition model includes: collecting an air-covered yarn training set, wherein the air-covered yarn training set includes an air-covered yarn production image and an air-covered yarn quality label; identifying the air-covered yarn production image to obtain production image features and image stretch coefficients of the air-covered yarn production image; The quality recognition model of air-covered yarn is obtained through training based on the production image features, image stretching coefficient and quality labels of air-covered yarn production images.
[0011] The process of obtaining the equipment abnormality coefficient includes: Collect production images and equipment parameters at corresponding moments according to time periods, and identify equipment fluctuation data based on the differences between the equipment parameters and standard parameters; the standard parameters are preset normal equipment parameters; Identify the production image to obtain the production image features and image stretching coefficient; According to the spatiotemporal variation characteristics of production image features, combined with image stretching coefficient and equipment fluctuation data, identification is performed to determine the equipment anomaly coefficient.
[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention constructs a layer-by-layer image feature recognition model based on the structural characteristics of air-coated yarn; first, network nodes with obvious characteristics are identified; based on the distribution of network nodes, the distribution of spandex inner cores is identified; based on the distribution of network nodes and spandex inner cores, outer fibers with various shapes and complex distributions are further identified, and the complex image analysis tasks are split, which reduces the weight of a single recognition task and improves the accuracy and robustness of recognition.
[0013] 2. This application proposes an image stretching coefficient, which is used to identify the stretching state of the air-coated yarn in the image; the image stretching coefficient and production image characteristics are identified through the coated yarn quality recognition model, and the recognition process of the production feature data is adjusted through the image stretching coefficient, so that the benchmark for standard judgment is more in line with the actual state of the current sequence, reducing misjudgments caused by tension changes and improving recognition accuracy.
[0014] 3. The present invention continuously monitors the air-covered yarn production line, collects production images and equipment parameters at corresponding moments according to a time period; obtains production image features and image stretching coefficients; identifies equipment fluctuation data based on the differences between equipment parameters and standard parameters; identifies equipment abnormality coefficients based on the spatiotemporal variation characteristics of production image features in combination with image stretching coefficients and equipment fluctuation data, thereby accurately identifying equipment abnormalities through production defects of air-covered yarn. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 Schematic diagram of the structure of a real-time detection system for spandex covered yarn exposed core defects based on industrial vision of the present invention; Figure 2 Schematic diagram of the structure of the image feature recognition model of the present invention; Figure 3 Schematic diagram of the process for obtaining the quality parameters of the air-covered yarn of the present invention. DETAILED DESCRIPTION
[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0017] The production of air-covered yarn primarily relies on a web nozzle and high-speed compressed air. The spandex core and outer covering fibers are fed into the web nozzle's channel at specific speeds. As the spandex core and outer covering fibers enter the web nozzle, high-speed, high-pressure compressed air flows through a specialized cavity structure within the nozzle, generating a strong and complex forced airflow. The impact and disturbance of the airflow blow the outer covering fibers apart and deform them. These deformed outer covering fibers entwine and wrap around the taut spandex core in an irregular pattern, forming the air-covered yarn.
[0018] Among them, under the action of high-speed airflow, the outer fibers form local, intermittent, and relatively tight entanglement points around the spandex inner core, namely network nodes; network nodes are the key to the structural stability of the air-covered yarn and are the main bonding points of the outer fibers wrapped around the spandex inner core. They usually appear as small bumps or dividing line areas at intervals; their shape, size, structure, density and distribution uniformity are important indicators of sparse quality.
[0019] The exposed core defect refers to the phenomenon that the spandex core is not completely and effectively covered by the outer fiber, resulting in part or all of the spandex filaments being exposed on the surface. This is one of the most common defects of air-covered yarn and one of the most quality-affecting defects. In order to ensure the production quality of air-covered yarn, it is necessary to identify the exposed core phenomenon.
[0020] Example 1:
[0021] The present invention provides a real-time detection system for spandex coated yarn core defects based on industrial vision, the structure of which is as follows: Figure 1 As shown, it includes: a data acquisition unit, an image processing unit, a quality recognition unit and an equipment early warning unit.
[0022] The data acquisition unit captures the production image of the air-covered yarn on the production line; and collects equipment parameters in the production equipment control system according to the acquisition time of the production image.
[0023] The equipment parameters include raw material feeding parameters, spraying and pressing processing parameters, traction molding parameters, and equipment environment parameters; The raw material feeding parameters include the spandex yarn feeding roller speed, the spandex yarn drafting multiple setting value, the spandex yarn feeding tension, the outer fiber feeding roller speed, the outer fiber overfeed rate, and the outer fiber feeding tension; The spandex yarn feeding roller speed is used to control the speed at which the spandex yarn enters the processing area; the spandex yarn drafting multiple setting value is the multiple by which the spandex yarn is stretched in the processing area, which is calculated based on the spandex feeding speed and the reference speed or directly set; the spandex yarn feeding tension is the tension of the spandex yarn when it enters the processing area; The outer fiber feeding roller speed is used to control the speed at which the outer fiber enters the processing area; the outer fiber overfeed rate is the excess percentage of the yarn feeding speed relative to the reference speed; and the outer fiber feeding tension is the tension of the outer fiber when it enters the processing area.
[0024] The spraying processing parameters include the network nozzle air supply pressure, compressed air flow, compressed air temperature and compressed air humidity; The network nozzle air supply pressure is the compressed air pressure supplied to the network nozzle; the compressed air flow is the compressed air flow through the network nozzle; the compressed air temperature is the temperature of the compressed air before entering the nozzle; and the compressed air humidity is the humidity of the compressed air.
[0025] The traction forming parameters include yarn output speed, godet speed, winding tension, and winding speed; The yarn output speed is the speed at which the finished covered yarn leaves the processing area and is pulled; the guide wheel speed is the connection of the guide mechanism, which is usually matched with the line speed; the winding tension is the tension of the yarn during winding and forming; the winding speed is the speed of the winding bobbin.
[0026] The equipment environmental parameters include the temperature and humidity of the environment in which the equipment is located; specifically, the ambient temperature and humidity of the production workshop.
[0027] The influence of the above parameters on the quality of air covered yarn products is as follows: Spandex yarn feeding parameters: When the spandex draft ratio is too high or the tension is inappropriate, the spandex yarn can easily become too thin, too long, or break, causing core exposure. The matching of feed speed and other speeds is also critical. Outer fiber feeding parameters: When the outer fiber overfeed rate and tension are insufficient, the coating will become loose and uneven, which is prone to core exposure.
[0028] Spraying processing parameters: The pressure, flow, temperature and humidity of the network nozzle directly affect the quality of network node formation. When the network points are poorly connected or uneven, the spandex yarn can easily be exposed from the network gap.
[0029] Traction forming parameters: The matching of line output speed and each feeding speed, as well as the size of winding tension, will affect the final shape of the line and the coating state of the inner core. Excessive tension may also cause core exposure.
[0030] Equipment environmental parameters: Temperature and humidity will affect the mechanical properties and static electricity of the fiber, indirectly affecting the stability of the processing process, which may affect the coating effect.
[0031] The present invention photographs the air-covered yarn on the production line to obtain a production image; and obtains the equipment parameters at the corresponding moment of the production image, including raw material feeding parameters, spray processing parameters, traction molding parameters, and equipment environment parameters; and correlates the equipment parameters with the bare core defects identified in the image for analysis, providing data support for subsequent abnormal identification of the air-covered yarn.
[0032] The image processing unit identifies the network nodes of the air-covered yarn in the production image to obtain a first image feature; identifies the spandex inner core in the production image based on the first image feature to obtain a second image feature; identifies the outer fiber in the production image based on the first image feature and the second image feature to obtain a third image feature; and obtains the production image feature based on the first image feature, the second image feature, and the third image feature.
[0033] Construct an image feature recognition model to identify production images. The model is built based on a deep learning network model. Its structure is as follows Figure 2 As shown, it includes a first feature recognition layer, a second feature recognition layer and a third feature recognition layer; The first feature recognition layer identifies the network node area in the production image, extracts the distribution position and distribution area image features of the network nodes in the image, and obtains the first image feature; the distribution area image feature of the network node is the image pixel data of the network node distribution area; wherein, based on the distribution area image feature of the network node, the distribution area and shape of the network node area can be identified; The second feature recognition layer identifies the distribution of the first image feature in the production image to determine the distribution path of the spandex core; identifies the distribution path of the spandex core to obtain exposed core data and covered core data; obtains the second image feature based on the exposed core data and covered core data; the exposed core data is the image pixels of the exposed area of the spandex core, and the covered core data is the image pixels of the unexposed area of the spandex core.
[0034] A core-exposed defect occurs when the spandex core is not fully and effectively covered by the wrapped fibers, resulting in partial or complete exposure of the spandex filaments on multiple surfaces. This is one of the most common defects in air-covered yarns and one of the most detrimental to quality. Core-exposed defects can occur randomly, influenced by factors such as raw material rotation and transient pulse disturbances. However, they can also occur periodically or be concentrated in specific locations due to equipment issues. The distribution of core-exposed defects includes: Near the network node: If the network node forms a poor structure, such as a loose structure, loose entanglement, too small size, or is idle during the formation process, it may also cause the network node itself to be unable to effectively bind all peripheral fibers.
[0035] Between network nodes: This is a common area where core dew occurs. Because network nodes are sparsely distributed, the area between two network nodes has a relatively weaker fiber binding force. If the network node density is insufficient, or the number of sparse fibers is insufficient, or the fluffiness is insufficient, core dew is likely to occur in the middle part between the nodes.
[0036] Therefore, by combining the distribution positions of network nodes and the image features of the distribution area, the exposed core area can be identified more accurately; thereby accurately identifying the exposed inner core data and the covered inner core data.
[0037] The third feature recognition layer identifies the outer fiber area of the air-covered yarn based on the distribution of the first image feature and the second image feature in the production image, and obtains the third image feature based on the identification of the outer fiber area; the third image feature includes the distribution data of the outer fiber area and the image pixel data of the distribution area.
[0038] The present invention constructs a layer-by-layer image feature recognition model based on the structural characteristics of air-coated yarns. First, network nodes with obvious characteristics are identified. Based on the distribution of network nodes, the distribution of spandex inner cores is identified. Based on the distribution of network nodes and spandex inner cores, outer fibers with various shapes and complex distributions are further identified. The complex image analysis tasks are split, the weight of a single recognition task is reduced, and the accuracy and robustness of recognition are improved.
[0039] The quality recognition unit recognizes the production image in combination with the equipment parameters to obtain the image stretch coefficient; and obtains the quality parameters of the air-covered yarn based on the image stretch coefficient and the production image feature recognition.
[0040] During the production process of air-covered yarn, tension fluctuates, resulting in different apparent elongation or relaxation of the lines in the image. This can easily lead to errors in quality identification based directly on production image features. Therefore, this application proposes an image stretch coefficient.
[0041] The idea for calculating the image stretch coefficient is: when the air-coated yarn is in different stretching states, there is a difference between the actual state data and the initial state data; the initial state data is the length data and structural data of the spandex inner core when it is not subject to external force; the actual state data is the length data and structural data of the spandex inner core when it is subject to external force; the above-mentioned length data is the length of the spandex inner core, and the structural data is the spatial structural state of the spandex inner core.
[0042] According to the structural data of the actual state and the initial state, the length data of the actual state is corrected to obtain the corrected length data; according to the ratio of the corrected length data to the length data of the initial state, the image stretching coefficient is obtained; when the image stretching coefficient is 1, it indicates that it is not affected by external force; when the image stretching coefficient is greater than 1, it indicates that it is affected by tension and the length becomes longer; when the image stretching coefficient is less than 1, it indicates that the length becomes shorter.
[0043] In this embodiment, the process of obtaining the image stretching coefficient is as follows: Identify the production image and obtain the production image features; The second image feature in the production image feature is identified, and the stretching feature data of the spandex core is obtained based on the distribution path, distribution type and distribution structure of the spandex core; the stretching feature data includes parameters such as the diameter of the spandex core at different stages, the degree of bending, the distribution of the exposed core area, and the spatial morphology of the spandex core.
[0044] The image stretch coefficient of the air covered yarn in the production image is identified and determined based on the stretch feature data in combination with the first image feature, the third image feature and the equipment parameters.
[0045] Specifically, the stretch feature data is identified and corrected through the first image feature, the third image feature and the equipment parameters to obtain the actual state data, and then the actual state data is identified with the preset air-covered yarn initial state data to obtain the image stretch coefficient.
[0046] This application proposes an image stretching coefficient, which is used to identify the stretching state of the air-covered yarn in the image, and to adjust the recognition of production image features, so that the benchmark for standard judgment is more in line with the actual state of the current sequence, reducing misjudgments caused by tension changes and improving recognition accuracy.
[0047] The quality parameter identification process of the air covered yarn is as follows: Figure 3 Shown, including: Based on the neural network model, a covered yarn quality recognition model is constructed; the image stretch coefficient and production image features are identified by the covered yarn quality recognition model to determine the quality parameters; The training process of the covered yarn quality recognition model includes: collecting an air-covered yarn training set, wherein the air-covered yarn training set includes an air-covered yarn production image and an air-covered yarn quality label; identifying the air-covered yarn production image to obtain production image features and image stretch coefficients of the air-covered yarn production image; The quality recognition model of air-covered yarn is obtained through training based on the production image features, image stretching coefficient and quality labels of air-covered yarn production images.
[0048] The present invention constructs a covered yarn quality recognition model; the image stretching coefficient and production image characteristics are identified by the covered yarn quality recognition model, and the recognition process of production feature data is adjusted by the image stretching coefficient to accurately identify the quality parameters of the air-covered yarn in the image.
[0049] The equipment early warning unit identifies and obtains equipment fluctuation data based on the difference between equipment parameters and standard parameters; and identifies and obtains equipment abnormality coefficients based on production image features, image stretching coefficients and equipment fluctuation data obtained from continuous production.
[0050] The process of obtaining the equipment abnormality coefficient includes: Collect production images and equipment parameters at corresponding moments according to time periods, and identify equipment fluctuation data based on the differences between the equipment parameters and standard parameters; the standard parameters are preset normal equipment parameters; Identify the production image to obtain the production image features and image stretching coefficient; According to the spatiotemporal variation characteristics of production image features, combined with image stretching coefficient and equipment fluctuation data, identification is performed to determine the equipment anomaly coefficient.
[0051] The spatiotemporal variation characteristics of the production image features include spatial variations among the first image features, the second image features, and the third image features, and temporal variations among the image features of consecutive production images.
[0052] The present invention continuously monitors the air-covered yarn production line, collects production images and equipment parameters at corresponding moments according to a time period; obtains production image features and image stretch coefficients; identifies equipment fluctuation data based on the differences between equipment parameters and standard parameters; identifies the equipment abnormality coefficient based on the spatiotemporal variation characteristics of the production image features in combination with the image stretch coefficient and equipment fluctuation data, thereby accurately identifying equipment abnormalities through production defects of the air-covered yarn.
[0053] Example 2:
[0054] The present invention provides a real-time detection system for spandex coated yarn core defects based on industrial vision, the structure of which is as follows: Figure 1 As shown, it includes: a data acquisition unit, an image processing unit, a quality recognition unit and an equipment early warning unit.
[0055] The data acquisition unit captures the production image of the air-covered yarn on the production line; and collects equipment parameters in the production equipment control system according to the acquisition time of the production image.
[0056] The equipment parameters include raw material feeding parameters, spraying and pressing processing parameters, traction molding parameters, and equipment environment parameters; The raw material feeding parameters include the spandex yarn feeding roller speed, the spandex yarn drafting multiple setting value, the spandex yarn feeding tension, the outer fiber feeding roller speed, the outer fiber overfeed rate, and the outer fiber feeding tension; The spraying processing parameters include the network nozzle air supply pressure, compressed air flow, compressed air temperature and compressed air humidity; The traction forming parameters include yarn output speed, godet speed, winding tension, and winding speed; The device environmental parameters include the temperature and humidity of the environment in which the device is located.
[0057] The present invention photographs the air-covered yarn on the production line to obtain a production image; and obtains the equipment parameters at the corresponding moment of the production image, including raw material feeding parameters, spray processing parameters, traction molding parameters, and equipment environment parameters; and correlates the equipment parameters with the bare core defects identified in the image for analysis, providing data support for subsequent abnormal identification of the air-covered yarn.
[0058] The image processing unit identifies the network nodes of the air-covered yarn in the production image to obtain a first image feature; identifies the spandex inner core in the production image based on the first image feature to obtain a second image feature; identifies the outer fiber in the production image based on the first image feature and the second image feature to obtain a third image feature; and obtains the production image feature based on the first image feature, the second image feature, and the third image feature.
[0059] Constructing an image feature recognition model to recognize production images, including a first feature recognition layer, a second feature recognition layer, and a third feature recognition layer; The first feature recognition layer identifies the network node area in the production image, extracts the distribution position and distribution area image features of the network nodes in the image, and obtains the first image feature; the distribution area image feature of the network node is the image pixel data of the network node distribution area; wherein, based on the distribution area image feature of the network node, the distribution area and shape of the network node area can be identified; The second feature recognition layer identifies the distribution of the first image feature in the production image to determine the distribution path of the spandex core; identifies the distribution path of the spandex core to obtain exposed core data and covered core data; obtains the second image feature based on the exposed core data and covered core data; the exposed core data is the image pixels of the exposed area of the spandex core, and the covered core data is the image pixels of the exposed area of the spandex core.
[0060] The third feature recognition layer identifies the outer fiber area of the air-covered yarn based on the distribution of the first image feature and the second image feature in the production image, and obtains the third image feature based on the identification of the outer fiber area; the third image feature includes the distribution data of the outer fiber area and the image pixel data of the distribution area.
[0061] The present invention constructs a layer-by-layer image feature recognition model based on the structural characteristics of air-coated yarns. First, network nodes with obvious characteristics are identified. Based on the distribution of network nodes, the distribution of spandex inner cores is identified. Based on the distribution of network nodes and spandex inner cores, outer fibers with various shapes and complex distributions are further identified. The complex image analysis tasks are split, the weight of a single recognition task is reduced, and the accuracy and robustness of recognition are improved.
[0062] The quality recognition unit recognizes the production image in combination with the equipment parameters to obtain the image stretch coefficient; and obtains the quality parameters of the air-covered yarn based on the image stretch coefficient and the production image feature recognition.
[0063] The process of obtaining the image stretching coefficient is as follows: Identify the production image and obtain the production image features; Identifying the second image feature in the production image feature, and obtaining the stretching feature data of the spandex inner core based on the distribution path, distribution type, and distribution structure of the spandex inner core; The image stretch coefficient of the air covered yarn in the production image is identified and determined based on the stretch feature data in combination with the first image feature, the third image feature and the equipment parameters.
[0064] This application proposes an image stretching coefficient, which is used to identify the stretching state of the air-covered yarn in the image, and to adjust the recognition of production image features, so that the benchmark for standard judgment is more in line with the actual state of the current sequence, reducing misjudgments caused by tension changes and improving recognition accuracy.
[0065] The quality parameter identification process of the air covered yarn includes: Constructing a covered yarn quality recognition model; identifying the image stretch coefficient and production image features through the covered yarn quality recognition model to determine quality parameters; The training process of the covered yarn quality recognition model includes: collecting an air-covered yarn training set, wherein the air-covered yarn training set includes an air-covered yarn production image and an air-covered yarn quality label; identifying the air-covered yarn production image to obtain production image features and image stretch coefficients of the air-covered yarn production image; The quality recognition model of air-covered yarn is obtained through training based on the production image features, image stretching coefficient and quality labels of air-covered yarn production images.
[0066] The present invention constructs a covered yarn quality recognition model; the image stretching coefficient and production image characteristics are identified by the covered yarn quality recognition model, and the recognition process of production feature data is adjusted by the image stretching coefficient to accurately determine the quality parameters of the air-covered yarn in the image.
[0067] The equipment early warning unit identifies and obtains equipment fluctuation data based on the difference between equipment parameters and standard parameters; and identifies and obtains equipment abnormality coefficients based on production image features, image stretching coefficients and equipment fluctuation data obtained from continuous production.
[0068] The process of obtaining the equipment abnormality coefficient includes: Collect production images and equipment parameters at corresponding moments according to time periods, and identify equipment fluctuation data based on the differences between the equipment parameters and standard parameters; the standard parameters are preset normal equipment parameters; Identify the production image to obtain the production image features and image stretching coefficient; According to the spatiotemporal variation characteristics of production image features, combined with image stretching coefficient and equipment fluctuation data, identification is performed to determine the equipment anomaly coefficient.
[0069] The spatiotemporal variation characteristics of the production image features include spatial variations among the first image features, the second image features, and the third image features, and temporal variations among the image features of consecutive production images.
[0070] The present invention continuously monitors the air-covered yarn production line, collects production images and equipment parameters at corresponding moments according to a time period; obtains production image features and image stretch coefficients; identifies equipment fluctuation data based on the differences between equipment parameters and standard parameters; identifies the equipment abnormality coefficient based on the spatiotemporal variation characteristics of the production image features in combination with the image stretch coefficient and equipment fluctuation data, thereby accurately identifying equipment abnormalities through production defects of the air-covered yarn.
[0071] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.
Claims
1. A real-time detection system for spandex covered yarn core defects based on industrial vision, characterized in that: include: The data acquisition unit captures the production image of the air-covered yarn on the production line; based on the acquisition time of the production image, the equipment parameters are collected in the production equipment control system; The image processing unit identifies the network nodes of the air-covered yarn in the production image to obtain a first image feature; identifies the spandex inner core in the production image based on the first image feature to obtain a second image feature; identifies the outer covering fiber in the production image based on the first image feature and the second image feature to obtain a third image feature; and obtains a production image feature based on the first image feature, the second image feature, and the third image feature; The quality recognition unit recognizes the production image in combination with the equipment parameters to obtain the image stretch coefficient; the quality parameters of the air-covered yarn are obtained based on the image stretch coefficient and the production image feature recognition; The equipment early warning unit identifies and obtains equipment fluctuation data based on the difference between equipment parameters and standard parameters; and identifies and obtains equipment abnormality coefficients based on production image features, image stretching coefficients and equipment fluctuation data obtained from continuous production.
2. The real-time detection system for spandex covered yarn core defects based on industrial vision according to claim 1 is characterized by: The equipment parameters include raw material feeding parameters, spraying and pressing processing parameters, traction molding parameters and equipment environment parameters; The raw material feeding parameters include the spandex yarn feeding roller speed, the spandex yarn drafting multiple setting value, the spandex yarn feeding tension, the outer fiber feeding roller speed, the outer fiber overfeed rate and the outer fiber feeding tension; The spraying processing parameters include the network nozzle air supply pressure, compressed air flow, compressed air temperature and compressed air humidity; The traction forming parameters include yarn output speed, godet speed, winding tension and winding speed; The device environmental parameters include the temperature and humidity of the environment in which the device is located.
3. The real-time detection system for spandex covered yarn core defects based on industrial vision according to claim 1 is characterized by: Constructing an image feature recognition model to recognize production images, including a first feature recognition layer, a second feature recognition layer, and a third feature recognition layer; The first feature recognition layer recognizes the network node area in the production image, extracts the distribution position of the network node in the image and the image features of the distribution area, and obtains the first image feature; the image feature of the distribution area of the network node is the image pixel data of the network node distribution area; The second feature recognition layer recognizes the distribution of the first image feature in the production image to determine the distribution path of the spandex inner core; recognizes the distribution path of the spandex inner core to obtain exposed inner core data and covered inner core data; and obtains the second image feature based on the exposed inner core data and covered inner core data; The third feature recognition layer identifies the outer fiber area of the air-covered yarn based on the distribution of the first image feature and the second image feature in the production image, and obtains the third image feature based on the identification of the outer fiber area; the third image feature includes the distribution data of the outer fiber area and the image pixel data of the distribution area.
4. The real-time detection system for spandex covered yarn core defects based on industrial vision according to claim 1 is characterized in that: The process of obtaining the image stretching coefficient is as follows: Identify the production image and obtain the production image features; Identifying the second image feature in the production image feature, and obtaining the stretching feature data of the spandex inner core based on the distribution path, distribution type, and distribution structure of the spandex inner core; The image stretch coefficient of the air covered yarn in the production image is identified and determined based on the stretch feature data in combination with the first image feature, the third image feature and the equipment parameters.
5. The real-time detection system for spandex covered yarn exposed core defects based on industrial vision according to claim 1 is characterized by: Constructing a covered yarn quality recognition model, identifying the image stretch coefficient and production image features through the covered yarn quality recognition model to determine quality parameters; The training process of the covered yarn quality recognition model includes: collecting an air-covered yarn training set, wherein the air-covered yarn training set includes an air-covered yarn production image and an air-covered yarn quality label; identifying the air-covered yarn production image to obtain production image features and image stretch coefficients of the air-covered yarn production image; The training is performed based on the production image features, image stretching coefficient and quality labels of the air covered yarn production images to obtain a trained covered yarn quality recognition model.
6. The real-time detection system for spandex covered yarn core defects based on industrial vision according to claim 1 is characterized in that: The process of obtaining the equipment abnormality coefficient includes: Collect production images and equipment parameters at corresponding moments according to time periods, and identify equipment fluctuation data based on the differences between the equipment parameters and standard parameters; the standard parameters are preset normal equipment parameters; Identify the production image to obtain the production image features and image stretching coefficient; According to the spatiotemporal variation characteristics of production image features, combined with image stretching coefficient and equipment fluctuation data, identification is performed to determine the equipment anomaly coefficient.