Fabric fuzzing and pilling testing equipment based on images and testing method of fabric fuzzing and pilling testing equipment

Abstract

The invention discloses fabric fuzzing and pilling testing equipment based on an image and a testing method thereof, and belongs to the technical field of textile detection. The equipment comprises a rack, a rolling box with a cork lining, an environment regulation and control module and an image acquisition and processing system. The environment regulation and control module forms micro-positive pressure and directional airflow in the rolling box through an ion fan and an exhaust fan, so that friction static electricity is neutralized in real time, fiber chippings are removed, and the test environment is purified from the source. The image acquisition and processing system adopts D65, near ultraviolet and near infrared multispectral light sources for time-sharing illumination, and combines an unsupervised differential segmentation algorithm and a machine learning model to realize accurate segmentation, feature extraction and intelligent rating of a pilling area. According to the method, active environment regulation and control and objective quantitative analysis of results in the testing process are achieved, the accuracy, consistency and efficiency of rating are remarkably improved, and the defects that traditional manual visual evaluation is high in subjectivity and poor in repeatability are overcome.

Description

Technical Field

[0001] This invention relates to the field of textile performance testing technology, and in particular to an image-based fabric pilling and fuzzing testing device and method. Background Technology

[0002] The pilling performance of fabrics is an important indicator for evaluating textile quality. Currently, the widely used standard methods both domestically and internationally (such as GB / T 4802.3-2008 "Pilling Box Method") mainly rely on manual visual evaluation after the test. This method has significant drawbacks: First, human visual evaluation is easily affected by subjective factors such as fatigue and experience differences, leading to poor consistency and comparability of results; second, the evaluation results cannot be quantified, making precise comparison and analysis difficult.

[0003] More importantly, during the testing process, the fabric generates static electricity through continuous friction with the cork lining, the sample carrier tube, and itself. The accumulation of static electricity attracts fiber debris, leading to distorted pilling patterns: such as abnormal fiber adhesion or repulsion, which in turn interferes with subsequent image acquisition, increases background noise, and reduces the contrast between the pilling area and the substrate.

[0004] Therefore, it is necessary to provide an image-based fabric pilling and fuzzing testing device and method to solve the above problems. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides an image-based fabric pilling and fuzzing testing device and method.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: an image-based fabric pilling and fuzzing testing device, comprising: frame; A roller box, mounted on the frame, is motor-driven and cork-lined, for accommodating a sample carrier tube containing a fabric sample; and An environmental control module, comprising: An electrostatic neutralization unit includes an ion fan and an air supply chamber communicating with the roller box for injecting ionized air into the roller box; and The debris removal unit includes an exhaust fan and an extraction chamber communicating with the roller box for extracting air from inside the roller box; The air supply chamber and the air extraction chamber are connected to the inside of the roller box through multiple ion air outlets and air extraction ports, and the ion air outlets and air extraction ports are arranged in a staggered manner in space to form a directional airflow inside the roller box. The device also includes an image acquisition and processing system for acquiring multispectral images of the fabric sample after it has been processed by the rolling chamber, and for analyzing the images to output rating results.

[0007] In a preferred embodiment of the present invention, the air supply chamber and the air extraction chamber are annular sealed cavities arranged around the outer wall of the roller box; The axis of the ion air outlet points to the geometric center of the roller box and forms an angle of 10° to 20° with the normal direction of the inner wall of the roller box.

[0008] In a preferred embodiment of the present invention, the environmental control module further includes: a two-stage air supply duct connecting the ion fan and the air supply chamber, and a two-stage air extraction duct connecting the exhaust fan and the air extraction chamber. Both the dual-section air supply duct and the dual-section air extraction duct include: an inner section duct that rotates with the roller box, and an outer section duct; the inner section duct and the outer section duct are connected by a rotary joint.

[0009] In a preferred embodiment of the present invention, the rotary joint is a dual-channel air circuit rotary joint, comprising: a fixed cylinder that rotates synchronously with the roller box, and a spindle fixed to the frame; The fixed cylinder and the mandrel achieve relative rotation through bearings and a sealing structure, forming independent air inlet and outlet channels; The airflow in the dual-stage air supply duct is transmitted through the air inlet channel, and the airflow in the dual-stage air extraction duct is transmitted through the air outlet channel.

[0010] In a preferred embodiment of the present invention, the air supply volume of the ion fan is greater than the air extraction volume of the exhaust fan, so that the inside of the roller box is maintained at a slight positive pressure of 5-10 Pa.

[0011] In a preferred embodiment of the present invention, the image acquisition and processing system includes: Imaging box; An illumination module, located inside the imaging chamber, includes at least one set of visible light sources and one set of non-visible light sources, used for multi-band illumination of the sample at different times; An imaging module, located on top of the imaging box, is used to acquire multispectral images under each type of illumination. The image processing module is used to fuse the acquired multispectral images to segment the spherical region and extract features.

[0012] In a preferred embodiment of the present invention, the lighting module includes a D65 standard light source, a near-ultraviolet light source, and a near-infrared light source; The image processing module is configured to perform the following operations: The background image is obtained by converting a color image under a D65 standard light source to a grayscale image and then applying Gaussian blur. Calculate the difference map between the near-ultraviolet image and the background image; Gaussian blurring and normalization are applied to the near-infrared image to generate a thickness mask; Adaptive threshold segmentation and logical AND operation are performed on the difference map and the thickness mask to generate a binary segmentation map of the balling region.

[0013] In a preferred embodiment of the present invention, the image processing module is further configured to: calculate at least one morphological feature parameter among the number of balling regions, total area, average area, distribution density, and area variation coefficient based on the binary segmentation map, and input the feature parameter into a trained machine learning model to output an integer rating result of 1-5 levels.

[0014] An image-based method for testing fabric pilling and fuzzing includes the following steps: S1. The fabric sample is mounted on the sample carrier tube and placed into the roller box; S2. Start the motor to rotate the roller box, and simultaneously start the ion fan and the exhaust fan to simultaneously neutralize static electricity and remove fiber debris during the friction process; S3. After the test, remove the sample and place it in the imaging box to acquire multispectral images of the sample; S4. Process the multispectral image, segment the blistering area and extract features, and automatically output the rating result based on the extracted features.

[0015] In a preferred embodiment of the present invention, in step S2, the air supply volume of the ion fan is controlled to be greater than the air extraction volume of the exhaust fan, so that the inside of the roller box is maintained in a slightly positive pressure state; and / or In step S4, an unsupervised multispectral differential segmentation algorithm is used to generate a segmentation map of the balling region.

[0016] This invention addresses the shortcomings of the prior art and has the following beneficial effects: This invention provides an image-based fabric pilling and fuzzing testing device. It incorporates an environmental control module and an image acquisition and processing system. During testing, ion wind is injected in real-time to neutralize static electricity, and micro-positive pressure directional airflow is used to remove fiber debris, thus purifying the testing environment at its source. Subsequently, multispectral imaging technology is employed to acquire images of the sample in the visible, near-ultraviolet, and near-infrared bands. Furthermore, unsupervised differential segmentation algorithms and machine learning models are used to automatically extract and intelligently rate pilling features. This overall design ensures that the entire testing process is controllable and the results are objectively quantified. It completely solves the problems of poor rating consistency and inability to accurately quantify results caused by subjectivity, fatigue, and environmental interference in traditional manual visual assessment. Therefore, it provides reliable and repeatable data support for textile quality assessment and process optimization.

[0017] The environmental control module of this invention, through the synergistic action of an ion fan and an exhaust fan, constructs a stable micro-positive pressure and directional airflow field inside the roller box. The ion fan outlet and the exhaust port are arranged alternately and oppositely, forming an effective airflow path through the roller box. This structure can not only neutralize the static charge generated by friction in real time, preventing fibers from forming abnormal pilling due to electrostatic adsorption or repulsion, but also efficiently discharge suspended debris along a preset direction. Thus, it can continuously maintain the cleanliness of the sample surface during the test, significantly reduce background noise and interference during image acquisition, and improve the contrast between the pilling area and the substrate. Compared with the existing technology that relies solely on mechanical friction and ignores the influence of the environment, this invention fundamentally eliminates the interference of static electricity and debris on the test results, further ensuring the authenticity and reliability of subsequent image analysis data.

[0018] The image acquisition and processing system of this invention comprehensively utilizes D65 standard light source, near-ultraviolet light source, and near-infrared light source for multi-band time-division illumination, and uses an unsupervised differential segmentation algorithm to fuse the acquired multispectral images. By using the visible light image as a background reference and differentiating it with the near-ultraviolet image to enhance edge contrast, and then combining it with a thickness mask generated from the near-infrared image for logical AND operation, high-precision binary segmentation of the pilling area is achieved. This method can effectively distinguish overlapping fiber clumps and suppress noise, directly improving the accuracy and robustness of pilling recognition. Compared with traditional single visible light image processing methods, this invention is particularly suitable for low-contrast samples such as light-colored fabrics and complex textures. Furthermore, by extracting multi-dimensional morphological features such as the number, area, and density of pilling and inputting them into a machine learning model, fully automatic and objective output from image to rating is achieved, avoiding bias introduced by human factors.

[0019] The environmental control module and the image acquisition and processing system of this invention have a synergistic enhancement effect. The environmental control module provides a physically stable and optically clean sample surface for image acquisition, ensuring the quality and consistency of multispectral images. Meanwhile, multispectral imaging and intelligent algorithms make full use of this optimized environment, and achieve refined extraction and quantitative rating of pilling features through multimodal information fusion. The two complement each other and together form a complete technical closed loop from physical environment purification to optical information acquisition and intelligent data analysis. This collaborative design not only overcomes the dual limitations of environmental interference and subjective rating in traditional methods, but also realizes the controllability of the testing process, the objectification of results, and the quantification of output, further promoting the technological leap of fabric pilling testing from experience-based judgment to data-driven approaches. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a structural diagram of the image-based fabric pilling and fuzzing testing device of the present invention. Figure 2 This is a three-dimensional structural diagram of the image-based fabric pilling and fuzzing testing device of the present invention. Figure 3 yes Figure 2 A magnified view of part A; Figure 4 This is a diagram showing the distribution of air holes inside the roller box of the present invention.

[0021] In the diagram: 1. Frame; 11. Sub-support; 2. Roller box; 21. Cork lining; 3. Ionizing fan; 4. Air supply chamber; 41. Ionizing air outlet; 5. Exhaust fan; 6. Extraction chamber; 61. Extraction port; 7. Two-stage air supply duct; 8. Two-stage extraction duct; 9. Rotary joint; 91. Fixed cylinder; 92. Mandrel; 93. Air inlet; 94. Air outlet; 95. Air inlet channel; 96. Air outlet channel. Detailed Implementation

[0022] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0024] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the scope of protection of this application.

[0025] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this application based on the specific circumstances.

[0026] This invention provides an image-based fabric pilling and fuzzing testing device, comprising: Bracket 1; Roller box 2, mounted on the support 1, is driven by a motor and lined with cork lining 21, for accommodating a sample carrier tube containing a fabric sample; and An environmental control module, comprising: An electrostatic neutralization unit includes an ion fan 3 and an air supply chamber 4 communicating with the roller box 2, for injecting ionized air into the roller box 2; and The debris removal unit includes an exhaust fan 5 and an extraction chamber 6 communicating with the roller box 2 for extracting air from inside the roller box 2. The air supply chamber 4 and the air extraction chamber 6 are connected to the inside of the roller box 2 through multiple ion air outlets 41 and air extraction ports 61, and the ion air outlets 41 and air extraction ports 61 are arranged in a staggered manner in space to form a directional airflow inside the roller box 2. The device also includes an image acquisition and processing system for acquiring multispectral images of the fabric sample after it has been processed by the roller 2, and analyzing the images to output rating results.

[0027] The core innovation of this invention lies in integrating an active environmental control module (static neutralization and debris removal) with a multispectral image intelligent analysis system into a traditional pilling test device. The basic principle is as follows: First, under the premise of fully complying with the mechanical friction conditions specified in national standards, ion wind is used to neutralize the static charge generated during friction in real time, and a micro-positive pressure directional airflow field is used to actively remove detached fiber debris. This creates a physically stable and optically clean testing environment at the source, completely eliminating the fundamental interference caused by electrostatic adsorption and debris contamination to subsequent image acquisition. Then, in this optimized environment, multispectral imaging technology is used to acquire image information of the fabric sample in the visible, near-ultraviolet, and near-infrared bands. An unsupervised differential segmentation algorithm is used to accurately extract the morphological features of pilling, and finally, an objective and quantifiable intelligent rating can be achieved through a machine learning model.

[0028] This technological approach, which involves proactive intervention in the physical environment, multimodal optical sensing, and intelligent algorithm analysis, generates a significant synergistic effect among features: environmental control is a prerequisite for accurate and reliable image analysis, while multispectral imaging and intelligent algorithms are the core means to achieve high-precision objective rating. The close integration of these three elements jointly solves long-standing technical problems in traditional methods, such as strong subjectivity, poor repeatability, and inability to quantify, achieving a paradigm shift from experience-based judgment to data-driven approaches.

[0029] The specific structure, connection relationship and system operation process of the above components are described in detail below with reference to the embodiments. Example

[0030] Figure 1 The diagram shows the structure of the image-based fabric pilling and fuzzing testing device of this embodiment. The device includes a support frame 1, a roller box 2, an environmental control module, and an image acquisition and processing system.

[0031] The support structure 1 consists of a base, a mounting platform on the base, and auxiliary supports 2 on both sides of the base. The auxiliary supports 2 are located on the side of the roller box 2 furthest from the main shaft, and a motor serving as the power source is mounted on the mounting platform of the support 1. This motor is preferably a servo motor and has a closed-loop speed control function with built-in encoder feedback, resulting in higher speed control accuracy to meet the stringent requirements of the standard for the speed stability of the roller box 2.

[0032] The motor's output shaft is connected to the main shaft via a transmission assembly. The transmission assembly is a synchronous belt pulley drive mechanism, including a driving synchronous pulley mounted on the motor's output shaft, a driven synchronous pulley mounted on the main shaft, and a synchronous toothed belt connecting the two. One end of the main shaft is fixed to a support plate on one side of the bracket 1 via a deep groove ball bearing seat, while the other end extends into the interior of the roller box 2 and is fixed to the roller box 2 via a flange, thereby reliably transmitting the motor's rotational motion to the roller box 2.

[0033] Furthermore, the roller box 2 is the working cavity, a rectangular box structure made of rolled and welded metal plates. The inner wall of the roller box 2 is lined with cork lining 21 at intervals, and is used to accommodate the sample carrier tube on which the fabric sample is mounted. The sample carrier tube is a rigid PVC pipe, and the fabric sample to be tested (which can be cut to 125mm x 125mm) is tightly sewn or wrapped around the surface of the sample carrier tube to ensure that it does not slip or loosen during the high-speed rotation of the roller box 2.

[0034] The testing process of the roller test chamber 2 involves simulating the pilling and fuzzing process during actual wear through multi-directional friction between the fabric and cork, the sample carrier tube, and other samples at a set rotation speed and time. However, the static electricity and fiber debris generated during the friction process significantly interfere with image analysis. To address this issue, this invention integrates an environmental control module into the traditional roller test chamber 2 structure.

[0035] The environmental control module includes an electrostatic neutralization unit and a debris removal unit. The electrostatic neutralization unit includes an ion fan 3 and an air supply chamber 4 connected to the roller box 2, used to inject ionized air into the roller box 2; the debris removal unit includes an exhaust fan 5 and an extraction chamber 6 connected to the roller box 2, used to extract air from the roller box 2. The air supply chamber 4 and the extraction chamber 6 are connected to the inside of the roller box 2 through multiple ionized air outlets 41 and extraction ports 61, respectively, and the ionized air outlets 41 and extraction ports 61 are arranged in a staggered, opposite manner in space to create a directional airflow inside the roller box 2.

[0036] Furthermore, the initial coordination process between the environmental control module and the roller chamber 2 is as follows: When the motor drives the roller chamber 2 and the sample carrier tube inside to rotate, and friction occurs between the sample and the cork lining 21 and other samples, the ion fan 3 starts. The ion wind generated by the ion fan 3, rich in positive and negative ions, is transported to the air supply chamber 4 through the air supply pipe and injected into the roller chamber 2 through the ion wind outlet 41. This ion wind can effectively neutralize the static charge generated on the sample surface due to friction. At the same time, the exhaust fan 5 starts, and draws air from inside the roller chamber 2 through the exhaust pipe and the exhaust chamber 6. By precisely controlling the air supply and exhaust volume, a micro-positive pressure environment of 5-10 Pa is maintained inside the roller chamber 2. This micro-positive pressure environment can prevent external dust from entering, and on the other hand, it can drive the fiber debris that has been neutralized and "lifted" by the ion wind to migrate to the exhaust port 61 along the preset airflow path determined by the relative positions of the ion wind outlet 41 and the exhaust port 61, and finally be discharged from the roller chamber 2. This initial coordination enabled the active purification and electrostatic control of the test environment to be completed simultaneously with the standard friction test, laying a solid foundation for subsequent image acquisition.

[0037] like Figure 2 and Figure 4 As shown, both the air supply chamber 4 and the air extraction chamber 6 are sealed cavities located on the outer wall of the roller box 2, each consisting of an annular cavity and multiple strip-shaped cavities connected to the annular cavity. The annular cavity of the air supply chamber 4 and the annular cavity of the air extraction chamber 6 are symmetrically arranged, and the corresponding strip-shaped cavities are spaced apart along the circumference of the roller box 2.

[0038] The axis of the ion air outlet 41 points to the geometric center of the roller box 2, and forms an angle of 10° to 20° with the normal direction of the inner wall of the roller box 2. The exhaust port 61 and the ion air outlet 41 are arranged in an axially staggered and radially opposed relationship. That is, if a certain ion air outlet 41 is located at a circumferential angle θ and an axial position z of the roller box 2, then the corresponding exhaust port 61 is located at a circumferential angle θ + 180° and an axial position z + 50mm. This "staggered blowing" layout ensures that the airflow injected from the ion air outlet 41 can effectively carry the suspended fiber debris through the central area of ​​the roller box 2 and be captured by the exhaust port 61 located slightly below on the opposite side, forming a complete and efficient airflow migration path, minimizing the formation of airflow short circuits or local dead zones.

[0039] Furthermore, the gas supply and exhaust of the environmental control module are achieved through a two-stage air supply duct 7 and a two-stage exhaust duct 8. The two-stage air supply duct 7 includes an inner section duct, an outer section duct, and a rotary joint 9. The inner section duct rotates synchronously with the roller box 2, with one end connected to the air supply chamber 4 and the other end connected to the rotary joint 9. The outer section duct is fixed, with one end connected to the ion fan 3 and the other end connected to the rotary joint 9. The two-stage exhaust duct 8 has a symmetrical structure, with the inner section duct connected to the exhaust chamber 6 and the outer section duct connected to the exhaust fan 5.

[0040] like Figure 3 As shown, the rotary joint 9 is a double-channel gas-tight rotary joint 9, which is a key component for dynamic sealing. The rotary joint 9 includes a fixed cylinder 91, with a spindle 92 centrally located on both sides of the fixed cylinder 91. Ball bearings, a static seal ring, and a dynamic seal ring are installed between the fixed cylinder 91 and the spindle 92, allowing relative rotation between them. One end of the fixed cylinder 91 is fixedly connected to the outer wall of the roller box 2 and is coaxial with the rotation axis of the roller box 2. A flange is installed at the end of the outer ring of the spindle 92 away from the roller box 2. The spindle 92 is fixedly connected to the auxiliary support 2 via the flange, thus keeping the spindle 92 stationary relative to the support 1.

[0041] Specifically, an air inlet 93 and an air outlet 94 are respectively provided at the top and bottom of the fixed cylinder 91. An air inlet channel 95 and an air outlet channel 96 are respectively provided on one side of the spindle 92 near its top and bottom ends. Two O-rings are provided on the outer ring of the spindle 92 inside the fixed cylinder 91 for installing sealing rings. The positions of the two O-rings correspond to the positions of the air inlet 93 and the air outlet 94, respectively, so that the air inlet 93 is connected to the air inlet channel 95 through one sealing ring, and the air outlet 94 is connected to the air outlet channel 96 through the other sealing ring, and the air inlet channel 95 and the air outlet channel 96 are independent of each other.

[0042] The inner section of the dual-stage air supply duct 7 is connected to the air inlet 93 on the fixed cylinder 91, and the outer section is connected to the air inlet channel 95 on the spindle 92. The inner section of the dual-stage air extraction duct 8 is connected to the air outlet 94 on the fixed cylinder 91, and the outer section is connected to the air outlet channel 96 on the spindle 92.

[0043] During operation, the ionizing air generated by the ionizing fan 3 is delivered to the air supply chamber 4 through the outer section pipe, the air inlet channel 95 and air inlet 93 of the rotary joint 9, and the inner section pipe, and finally injected into the roller box 2 through the ionizing air outlet 41 to neutralize static electricity. The exhaust fan 5 draws air through the air intake port 61, the air extraction chamber 6, the air outlet 94 of the rotary joint 9, and the air outlet channel 96, forming a slight positive pressure of 5-10 Pa, which drives the debris to be discharged along the directional airflow. The rotary joint 9 ensures continuous sealing of the air passage during the rotation of the roller box 2.

[0044] Specifically, during equipment operation, the air supply volume is set to 250-300 m³ / h via the PLC control system. 3 / h, exhaust volume set at 220-250m³ / h 3 Within a range of / h, ensuring that the supply air volume is always greater than the extraction air volume, this controllable flow difference allows the inside of the roller chamber 2 to stably establish and maintain a micro-positive pressure of 5-10Pa (which can be monitored in real time by a differential pressure sensor on the chamber wall and fed back to the control system, forming a closed-loop regulation). Under the guidance of the uniform driving force of this micro-positive pressure and the spiral airflow path formed by the staggered opposing layout of the ion air outlet 41 and the exhaust port 61, the fiber debris suspended inside the chamber and neutralized by the ion air is forced to migrate efficiently from the supply air side to the exhaust air side along a preset directional path. The debris finally enters the exhaust chamber 6 through the exhaust port 61, passes through the outlet air passage of the dual-stage exhaust pipe 8 and the rotary joint 9, is extracted by the exhaust fan 5, and is filtered by the HEPA high-efficiency air filter connected in series at the end for clean discharge, thus providing a continuously clean physical environment for the sample during the test.

[0045] The image acquisition and processing system of this embodiment includes: an imaging box, an illumination module disposed inside the imaging box, an imaging module disposed on the top of the imaging box, and an image processing module. The illumination module includes at least one set of visible light sources and one set of non-visible light sources for time-division multi-band illumination of the sample; the imaging module is used to acquire multispectral images under each light source illumination; and the image processing module is used to perform fusion processing on the acquired multispectral images to segment the blistering region and extract features.

[0046] Specifically, the imaging box has a cubic structure with the inner walls completely coated with a matte black light-absorbing coating, which has excellent light-sealing performance and effectively isolates stray light interference.

[0047] The lighting module includes a D65 standard light source, a near-ultraviolet light source, and a near-infrared light source.

[0048] The image processing module is configured to perform the following operations: convert the color image under the D65 standard light source into a grayscale image and perform Gaussian blur processing to obtain a background image; calculate the difference map between the near-ultraviolet image and the background image; perform Gaussian blur and normalization on the near-infrared image to generate a thickness mask; and perform adaptive threshold segmentation and logical AND operation on the difference map and the thickness mask to generate a binary segmentation map of the spherical region.

[0049] After the test, based on the processed images of the balling area, professionals will conduct a visual rating according to standard sample photos, or use a large model for automatic rating.

[0050] Through the above structural design in this embodiment, when the fabric sample is subjected to standard friction in the rolling box 2, the electrostatic potential on its surface can be effectively controlled, and the generated fiber debris is removed in real time, thus providing hardware guarantee for obtaining a real and clean sample surface. Example

[0051] This embodiment further refines and optimizes the image acquisition and processing system based on Embodiment 1.

[0052] The image acquisition and processing system includes an imaging box, an illumination module, an imaging module, and an image processing module.

[0053] The illumination module in this embodiment is located inside the imaging chamber and includes at least three independently controllable light sources: a D65 standard light source, a near-ultraviolet light source, and a near-infrared light source. These light sources are configured to illuminate the fabric sample placed on the stage inside the imaging chamber at different times using different wavelengths.

[0054] The D65 standard light source is installed on the upper front side of the inner wall of the imaging chamber. The preferred model is VeriVide CAC50 with a color temperature of 6500K and a color rendering index Ra > 95. The vertical height of the D65 standard light source from the stage is approximately 300mm, which is used to simulate sunlight illumination conditions and obtain the true color and texture information of the sample.

[0055] The near-ultraviolet (NIUV) light source is mounted on the left side of the imaging chamber and consists of an array of multiple NIUV LEDs (such as Nichia NCSU276A), with a center wavelength of 395 nm and a full width at half maximum (FWHM) of 10 nm. The NIUV light source is approximately 250 mm from the sample plane. Ultraviolet light is sensitive to the microstructure and chemical properties of the fiber surface and can enhance the contrast between the pilling area and the substrate.

[0056] The near-infrared light source is mounted on the right side of the imaging chamber and consists of an array of multiple near-infrared LEDs (such as OSRAM SFH 4775S), with a center wavelength of 850 nm and a full width at half maximum (FWHM) of 20 nm. The near-infrared light source is approximately 250 mm from the sample plane. Near-infrared light has a certain degree of penetrability; however, the thicker fiber buildup in the pilling area leads to reduced transmittance, resulting in a dark area in the image. This information can be used to infer the thickness or density of the pilling.

[0057] Each light source is powered by a constant current drive power supply, and the light intensity is adjustable (D65: 0-1000 lux, NUV: 0-500 μW / cm²). 2 NIR: 0-800μW / cm 2 And it is controlled in a time-sharing manner through an image acquisition controller.

[0058] The imaging module in this embodiment is at least one high-resolution color CMOS camera mounted on the central observation window at the top of the imaging box, with its optical axis perpendicular to the stage below.

[0059] After the pilling and fuzzing test is completed inside the rolling chamber 2, the sample removed from the rolling chamber 2 is placed on the stage of the imaging chamber. The image acquisition process of the image acquisition and processing system is as follows: 1. Turn off the near-ultraviolet and near-infrared light sources, and turn on the D65 standard light source. Wait 2 seconds for the light source to stabilize, then trigger the camera to acquire a color image, which is recorded as follows. .

[0060] 2. Turn off the D65 standard light source and near-infrared light source, and turn on the near-ultraviolet light source. After stabilizing for 2 seconds, trigger the camera to acquire an image, which is recorded as [image name missing]. .

[0061] 3. Turn off the D65 standard light source and near-ultraviolet light source, and turn on the near-infrared light source. After stabilizing for 2 seconds, trigger the camera to acquire an image, which is recorded as follows: .

[0062] All images are stored in 12-bit TIFF format, preserving the full dynamic range.

[0063] Furthermore, based on the acquired images, an unsupervised multispectral difference segmentation method is performed, including: 1. Transfer the color image Convert to grayscale image The conversion formula is: ; For grayscale images Gaussian blurring (kernel size 15x15, σ=3) is applied to obtain a smooth background image. , used to estimate the background signal of the fabric substrate; Calculate the difference plot: The proportionality coefficient α is an empirical constant, taken as 0.85, used to compensate for the intensity difference between the near-ultraviolet light source and the visible light channel.

[0064] This step is used to enhance the response of the blistering area relative to the background in the ultraviolet band.

[0065] 2. Regarding Performing Gaussian blurring with the same parameters yields the following results. ; Calculate the thickness of the mask: .

[0066] Because the fibers are thick at the pilling area, the near-ultraviolet light transmission is poor and the value is low. After this transformation, the pilling area is... The value is high.

[0067] 3. Set adaptive threshold ; Generate a preliminary binary segmentation map. If and only if and hour, (Pilling), otherwise 0 (background), ensuring that only areas that simultaneously meet the ultraviolet enhancement and near-infrared thickness characteristics are identified as pilling, effectively suppressing noise; right A morphological closing operation (5x5 circular structuring element) is performed to fill the small holes inside the spheres, and then connected component analysis is performed to remove false noise points with an area of ​​less than 50 pixels, resulting in the final accurate binary mask of the sphere region.

[0068] 4. Based on the final binary mask, calculate a set of quantifiable morphological features: Number of balls N: The total number of connected components; Total balling area The sum of all pixels marked as 1; Average sphere area ; Pilling density ,in The effective imaging area of ​​the sample (e.g., 120mm x 120mm = 14400mm) 2 ); area variation coefficient ,in It is the standard deviation of the area of ​​a single ball, reflecting the uniformity of the ball size; This constitutes a five-dimensional feature vector. .

[0069] 5. Input the feature vector F into the pre-trained machine learning model and output integer rating results from level 1 to 5. .

[0070] Specifically, a support vector machine classification model was adopted, using a radial basis function kernel with a penalty parameter C=10 and a kernel coefficient γ=0.1. The model was trained on a sample set containing 500 manually labeled samples covering various fabric types.

[0071] In another embodiment, a lightweight convolutional neural network model is employed, comprising: three convolutional layers (with 32, 64, and 128 filters and 3x3 kernels), each followed by ReLU activation; two max-pooling layers (stride 2); one global average pooling layer; and a fully connected output layer containing 5 neurons (corresponding to 5 levels), using the Softmax activation function. This model is used with the registered three-channel image. As direct input, the end-to-end output rating probability distribution is used to obtain the final rating through the argmax operation.

[0072] In summary, the image acquisition and processing system constructed in Example 2 achieves accurate quantitative characterization of fabric pilling morphology through deep collaboration between multi-band optical imaging and unsupervised intelligent algorithms. The system utilizes time-division illumination from a D65 standard light source, a near-ultraviolet light source, and a near-infrared light source to capture the apparent texture of the sample in the visible light band, the surface chemical response in the ultraviolet band, and the fiber packing thickness information in the near-infrared band, respectively. An unsupervised differential segmentation algorithm is used, with the visible light image serving as the background reference, and differential operations are performed with the near-ultraviolet image to enhance the edge and detail contrast of the pilling area. Simultaneously, a thickness mask generated from the near-infrared image is used to distinguish overlapping fiber clusters. This multimodal information fusion and cross-validation based on different physical principles (optical reflection, ultraviolet fluorescence characteristics, and near-infrared absorption) significantly improves the accuracy and robustness of pilling area segmentation. Especially for light-colored fabrics, complex textures, or low-contrast samples, its recognition accuracy is far superior to traditional image processing methods that rely on a single visible light image.

[0073] Furthermore, this system eliminates the reliance on large amounts of manually labeled data, automatically extracting multi-dimensional morphological features such as the number of pills, total area, average area, distribution density, and area variation coefficient through unsupervised methods. These features not only directly support objective and repeatable intelligent ratings (outputting levels 1-5) based on support vector machines or convolutional neural network models, but more importantly, they provide a refined and quantifiable data foundation for textile quality analysis, process improvement, and product development, realizing a paradigm shift from subjective, discrete human experience judgment to objective, continuous data-driven decision-making. Example

[0074] This embodiment provides an image-based method for testing the pilling and fuzzing of fabrics. The method uses the equipment described in Embodiment 1 and can preferably be combined with the image analysis method described in Embodiment 2.

[0075] This testing method includes the following steps: Step S1: Cut the fabric sample according to the standard requirements and install it on the sample carrier tube, and then put the sample carrier tube into the rolling box 2.

[0076] Step S2: Start the motor to rotate the roller box 2 according to the set parameters (e.g., 60 rpm, 1800 s) to perform a standard friction test. At the same time, start the ion fan 3 and the exhaust fan 5. The ionized air generated by the ion fan 3 is sent into the roller box 2 through the dual-stage air supply pipe 7 and the rotary joint 9 to neutralize the static electricity from the friction.

[0077] By controlling the air supply volume to be slightly greater than the exhaust volume, a slight positive pressure is formed in the roller box 2, driving the debris to be discharged from the exhaust port 61 along the directional airflow, thus achieving real-time cleaning during the test process.

[0078] Step S3: After the test, remove the sample and place it on the stage of the imaging chamber. Turn on the D65 standard light source, near-ultraviolet light source, and near-infrared light source in sequence, and simultaneously trigger the CMOS camera to acquire the corresponding visible light images. Near-ultraviolet images and near-infrared images .

[0079] Step S4: Process the acquired multispectral image. Using the unsupervised multispectral differential segmentation algorithm described in Example 2, generate an accurate segmentation mask for the balling region. Based on this mask, extract morphological feature parameters and input these parameters into the trained machine learning model to automatically output rating results of 1-5 levels.

[0080] The testing method provided in this embodiment seamlessly integrates active environmental control with intelligent image analysis, forming a complete automated testing process that ensures objectivity, repeatability, and high efficiency throughout the entire process from sample preparation to result output.

[0081] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. An image-based fabric pilling and fuzzing testing device, characterized in that, include: frame; A roller box, mounted on the frame, is driven by a motor and lined with cork, used to hold a sample carrier tube on which a fabric sample is mounted; as well as An environmental control module, comprising: An electrostatic neutralization unit includes an ion fan and an air supply chamber communicating with the roller box for injecting ionized air into the roller box; and The debris removal unit includes an exhaust fan and an extraction chamber communicating with the roller box for extracting air from inside the roller box; The air supply chamber and the air extraction chamber are connected to the inside of the roller box through multiple ion air outlets and air extraction ports, and the ion air outlets and air extraction ports are arranged in a staggered manner in space to form a directional airflow inside the roller box. The device also includes an image acquisition and processing system for acquiring multispectral images of the fabric sample after it has been processed by the rolling chamber, and for analyzing the images to output rating results.

2. The image-based fabric pilling and fuzzing testing device according to claim 1, characterized in that, The air supply chamber and the air extraction chamber are annular sealed cavities arranged around the outer wall of the roller box; The axis of the ion air outlet points to the geometric center of the roller box and forms an angle of 10° to 20° with the normal direction of the inner wall of the roller box.

3. The image-based fabric pilling and fuzzing testing device according to claim 1, characterized in that, The environmental control module further includes: a two-stage air supply duct connecting the ion fan and the air supply chamber, and a two-stage air extraction duct connecting the exhaust fan and the air extraction chamber. Both the dual-section air supply duct and the dual-section air extraction duct include: an inner section duct that rotates with the roller box, and an outer section duct; the inner section duct and the outer section duct are connected by a rotary joint.

4. The image-based fabric pilling and fuzzing testing device according to claim 3, characterized in that, The rotary joint is a dual-channel air-path rotary joint, comprising: a fixed cylinder that rotates synchronously with the roller box, and a spindle fixed to the frame; The fixed cylinder and the mandrel achieve relative rotation through bearings and a sealing structure, forming independent air inlet and outlet channels; The airflow in the dual-stage air supply duct is transmitted through the air inlet channel, and the airflow in the dual-stage air extraction duct is transmitted through the air outlet channel.

5. The image-based fabric pilling and fuzzing testing device according to claim 1, characterized in that, The air supply volume of the ion fan is greater than the air extraction volume of the exhaust fan, so that the inside of the roller box is maintained at a slight positive pressure of 5-10 Pa.

6. The image-based fabric pilling and fuzzing testing device according to claim 1, characterized in that, The image acquisition and processing system includes: Imaging box; An illumination module, located inside the imaging chamber, includes at least one set of visible light sources and one set of non-visible light sources, used for multi-band illumination of the sample at different times; An imaging module, located on top of the imaging box, is used to acquire multispectral images under each type of illumination. The image processing module is used to fuse the acquired multispectral images to segment the spherical region and extract features.

7. The image-based fabric pilling and fuzzing testing device according to claim 6, characterized in that, The lighting module includes a D65 standard light source, a near-ultraviolet light source, and a near-infrared light source; The image processing module is configured to perform the following operations: The background image is obtained by converting a color image under a D65 standard light source to a grayscale image and then applying Gaussian blur. Calculate the difference map between the near-ultraviolet image and the background image; Gaussian blurring and normalization are applied to the near-infrared image to generate a thickness mask; Adaptive threshold segmentation and logical AND operation are performed on the difference map and the thickness mask to generate a binary segmentation map of the balling region.

8. The image-based fabric pilling and fuzzing testing device according to claim 7, characterized in that, The image processing module is further configured to: calculate at least one morphological feature parameter among the number of balling regions, total area, average area, distribution density, and area variation coefficient based on the binary segmentation map, and input the feature parameter into a trained machine learning model to output an integer rating result of 1-5 levels.

9. An image-based method for testing fabric pilling, employing the image-based fabric pilling testing equipment as described in any one of claims 1-8, characterized in that, Includes the following steps: S1. The fabric sample is mounted on the sample carrier tube and placed into the roller box; S2. Start the motor to rotate the roller box, and simultaneously start the ion fan and the exhaust fan to simultaneously neutralize static electricity and remove fiber debris during the friction process; S3. After the test, remove the sample and place it in the imaging box to acquire multispectral images of the sample; S4. Process the multispectral image, segment the blistering area and extract features, and automatically output the rating result based on the extracted features.

10. The image-based fabric pilling and fuzzing test method according to claim 9, characterized in that, In step S2, the airflow of the ion fan is controlled to be greater than the airflow of the exhaust fan, so that the inside of the roller box is maintained in a slightly positive pressure state; and / or In step S4, an unsupervised multispectral differential segmentation algorithm is used to generate a segmentation map of the balling region.

Images