Textile cloth surface defect detection method based on photoelectric detector

Abstract

The invention discloses a textile cloth surface defect detection method based on a photoelectric detector, and relates to the technical field of online quality detection and control of textiles. Through the innovative design of a photoelectric signal processing chain, under the condition that production line hardware does not need to be transformed, the defect detection reliability of cloth with complex textures is remarkably improved; a dynamic reference template is constructed based on the inherent periodic characteristics of the cloth, so that the system automatically adapts to different patterns of fabrics such as lace and jacquard, and the generalization ability defect of a traditional fixed template is overcome; a pattern matching mechanism of normalized cross correlation coefficients is utilized, design features and real defects are separated from a physical signal level, and the fundamental contradiction of hole confusion is directly cracked; a dynamic threshold updating strategy further strengthens the anti-interference capability, and historical working condition data and real-time signal characteristics are fused to ensure that a judgment standard is adaptively adjusted along with cloth texture fluctuation, so that false alarms caused by environmental disturbance or process micro change are avoided.

Description

Technical Field

[0001] This invention relates to the field of online quality inspection and control technology for textiles, and in particular to a method for detecting defects in textile fabrics based on photoelectric detectors. Background Technology

[0002] In the high-end textile manufacturing sector, the proportion of specialty fabrics such as lace and jacquard continues to increase. These fabrics naturally contain regular holes or openwork textures due to design requirements. Existing fabric defect detection systems need to identify real defects such as holes and broken yarns in real time in a high-speed continuous production environment, while avoiding misjudging decorative structures as defects. Especially in the fields of medical and aerospace specialty textiles, micron-level defects may cause product malfunctions, placing stringent requirements on detection accuracy.

[0003] Current mainstream solutions mainly rely on two types of technologies: reflection signal analysis based on photodetectors and image recognition based on machine vision. The photodetector solution locates defects by detecting changes in the intensity of reflected light from the fabric surface. Its advantage lies in its millisecond-level response speed. However, in the case of lace fabric, the reflection signal waveforms generated by decorative holes and real holes are highly similar, making it difficult for the system to distinguish between design features and real damage. Although the machine vision solution can capture texture details, the complex light transmission and three-dimensional structure of lace cause significant fluctuations in the grayscale values ​​of the image, and the algorithm is easily affected by the texture of the substrate. Both technologies face the problem that defect signals can be submerged by periodic or random texture noise.

[0004] To address texture interference, some solutions enhance feature contrast through multispectral imaging or employ deep learning models to learn complex texture patterns. However, multispectral solutions require additional light sources in specific wavelengths and calibration costs, while deep learning models rely on massive amounts of labeled samples and lack the ability to generalize to the ever-changing patterns of lace. The key issue is that decorative holes and real defects lack a fundamental dimension for differentiation at the physical signal level, resulting in a persistent ambiguity in the algorithm's judgment logic. In addition, some solutions capture microstructural differences by increasing resolution, but their practical application effectiveness is limited by optical diffraction limits and production costs. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] This invention provides a method for detecting defects in textile fabrics based on photoelectric detectors, which solves the problem that it is difficult to distinguish between designed holes and actual holes in lace fabrics in reflected signals, and that existing solutions have a high false positive rate due to texture interference.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] This invention provides a method for detecting defects in textile fabrics based on photoelectric detectors, comprising:

[0009] Step S1: The reflected light signal of the moving fabric surface is collected by a linearly arranged array of photodetectors, the array covering the full width of the fabric.

[0010] Step S2: Extract the autocorrelation features of the reflected light signal to identify the repetition frequency of periodic holes on the fabric surface;

[0011] Step S3: Construct a reference signal template for periodic holes based on the repetition frequency;

[0012] Step S4: Calculate the normalized cross-correlation coefficient between the current reflected light signal and the reference signal template;

[0013] Step S5: When the normalized cross-correlation coefficient is lower than the dynamic threshold and is maintained for at least three consecutive cycles, it is determined that there is a hole defect.

[0014] Optionally, defects can be cross-validated by combining infrared reflection signals.

[0015] As a preferred embodiment of the textile fabric surface defect detection method based on photoelectric detectors described in this invention, in step S4, the calculation of the normalized cross-correlation coefficient involves sequentially performing displacement alignment, unweighted cross-correlation, weight correction, and coefficient selection, specifically including:

[0016] The optimal alignment displacement of the signal segment is obtained by using a sliding window optimization method.

[0017] ,

[0018] in, This represents the optimal displacement, expressed in samples. To test the displacement, This represents the number of sampling points corresponding to the baseline template length, in units of samples. The displacement is The unweighted normalized cross-correlation coefficients calculated at that time are dimensionless.

[0019] Will Substituting the values ​​yields the aligned sequence. ,when Sometimes, ,

[0020] The unweighted cross-correlation coefficient is defined as:

[0021] ,

[0022] in, To align the current reflected light signal at the [number]th [time], The amplitude of the sampling point, in V. The arithmetic mean of the aligned signal segments, in V. The reference signal template is in the first The amplitude of the sampling point, in V. The arithmetic mean of the template sequence is expressed in V. The unweighted normalized cross-correlation coefficient is dimensionless.

[0023] when Time-triggered position weight correction, constructing Gaussian position weights:

[0024] ,

[0025] in, For the first The location weighting factor of the sampling points is dimensionless. For the sampling point index, 1- The unit is sample. is the standard deviation of the weighted distribution, in units of samples;

[0026] Weighted normalized cross-correlation coefficient:

[0027] ,

[0028] in, The cross-correlation coefficient is a weighted normalized coefficient, dimensionless.

[0029] Coefficient selection output:

[0030] ,

[0031] in, To output the correlation coefficient, dimensionless. The threshold is dimensionless.

[0032] As a preferred embodiment of the textile fabric surface defect detection method based on photoelectric detectors described in this invention, step S2 includes:

[0033] The reflected light signal is divided into equal-length sequence segments according to a fixed duration;

[0034] Calculate the peak value of the autocorrelation coefficient between adjacent sequence segments;

[0035] The repetition period of periodic holes is determined based on the peak interval distribution.

[0036] As a preferred embodiment of the textile fabric surface defect detection method based on photoelectric detectors described in this invention, step S3 includes:

[0037] Collect reflected light signals from multiple consecutive complete cycles;

[0038] Phase alignment processing is performed on each periodic signal;

[0039] The phase-aligned signals are superimposed and averaged to generate a reference signal template; phase alignment is a zero-crossing time displacement calibration.

[0040] As a preferred embodiment of the textile fabric surface defect detection method based on photodetectors described in this invention, the phase alignment process includes:

[0041] Identify the zero-crossing positions of signals in each cycle;

[0042] Time displacement alignment is performed based on the first zero-crossing point;

[0043] Extract signal sequences with the same phase interval.

[0044] In a preferred embodiment of the textile surface defect detection method based on a photoelectric detector described in this invention, step S5, determining the dynamic threshold, includes:

[0045] During the initial operation phase of the fabric, normal hole signals were collected for at least five complete cycles.

[0046] Calculate the historical mean of the normalized cross-correlation coefficient between the signal and the reference template;

[0047] Multiply the historical average by a preset coefficient to obtain the initial dynamic threshold.

[0048] As a preferred embodiment of the textile fabric surface defect detection method based on photoelectric detectors described in this invention, in step S5, the dynamic threshold setting method includes historical sample collection, initial threshold calculation, and exponential adaptive updating, specifically:

[0049] After the fabric enters a stable, uniform speed phase, continuous recording is performed. Correlation coefficient for each period:

[0050] ,

[0051] in, This indicates the number of cycles required for initialization, expressed in cycles. Indicates the first The correlation coefficient of the output over each period is dimensionless. For this The arithmetic mean of the correlation coefficients over a period of time, dimensionless;

[0052] An initial threshold is given based on the historical average level:

[0053] ,

[0054] in, The threshold scaling factor is dimensionless. The dynamic threshold obtained during initialization is dimensionless;

[0055] From the At the start of each cycle, the threshold is adjusted online using an exponentially weighted moving average, and the update formula is as follows:

[0056] , ,

[0057] in, To update the weights, use values ​​from 0 to 1, which are dimensionless. The dynamic threshold of the previous cycle is dimensionless. The correlation coefficient output for the current period is dimensionless. Indicates the current period number. The updated dynamic threshold is dimensionless.

[0058] Will The continuous period determination logic in step S5 is fed into the correlation coefficient. The comparison is performed, and when the latter is continuously below the threshold for more than three cycles, the hole defect monitoring stage is entered.

[0059] As a preferred embodiment of the textile fabric surface defect detection method based on photoelectric detectors described in this invention, it further includes:

[0060] When a hole defect is determined to exist, the infrared band reflection signal is checked simultaneously.

[0061] Defect determination is confirmed only when the infrared band energy integral decrease condition is met; the energy integral decrease is the signal area attenuation within the window.

[0062] As a preferred embodiment of the textile fabric surface defect detection method based on photoelectric detectors described in this invention, the infrared band energy integration decrease condition includes:

[0063] Calculate the integral value of infrared signal energy within a local time window;

[0064] A judgment is triggered when the score drops by more than a preset percentage from the baseline.

[0065] In a preferred embodiment of the textile fabric defect detection method based on photodetectors described in this invention, the photodetector array comprises miniature detection units.

[0066] The size of each unit's photosensitive area is no greater than 0.1mm × 0.1mm.

[0067] The spacing between array units is less than the minimum detectable defect size on the fabric surface; the minimum detectable defect size refers to the minimum hole diameter or yarn breakage width allowed in the production line quality specifications.

[0068] The beneficial effects of this invention are as follows: Through the innovative design of the photoelectric signal processing chain, this invention significantly improves the reliability of defect detection for complex textured fabrics without requiring modifications to the production line hardware; based on the inherent periodic characteristics of the fabric, a dynamic reference template is constructed, enabling the system to automatically adapt to different patterns of lace, jacquard, and other fabrics, overcoming the generalization limitations of traditional fixed templates; and by utilizing the pattern matching mechanism of normalized cross-correlation coefficients, design features and actual defects are separated at the physical signal level, directly resolving the fundamental contradiction of hole confusion.

[0069] The dynamic threshold update strategy of this invention further enhances the anti-interference capability. By integrating historical working condition data with real-time signal characteristics, it ensures that the judgment criteria are adaptively adjusted according to the fluctuation of fabric texture, avoiding false alarms caused by environmental disturbances or slight changes in the process. For high reflective special fabrics, it introduces infrared band energy integration as an auxiliary verification method, and uses the differences in physical response of different spectral dimensions to double-lock the real defects and filter optical noise.

[0070] At the level of micro-defect detection, the synergistic effect of miniature photodetector arrays and signal enhancement algorithms breaks through the spatial resolution limitations of traditional photoelectric solutions; by using weighted cross-correlation calculations to focus on the core segment of the signal, the interference of edge noise on the judgment results is suppressed.

[0071] Upgrading photoelectric detection technology from a single intensity threshold judgment to multi-dimensional feature collaborative decision-making, while retaining the advantage of millisecond-level response speed, endows the system with intelligent decoupling capability for complex textures, providing a reliable technical foundation for full-speed online quality inspection of high-end textiles. Attached Figure Description

[0072] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a flowchart illustrating the textile fabric defect detection method based on a photodetector in Example 1. Detailed Implementation

[0074] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0075] 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, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0076] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0077] Example 1, referring to Figure 1 This embodiment provides a method for detecting defects in textile fabrics based on photodetectors, including:

[0078] Step S1: The reflected light signal of the moving fabric surface is collected by a linearly arranged photodetector array, which covers the full width of the fabric.

[0079] Step S2: Extract the autocorrelation features of the reflected light signal to identify the repetition frequency of periodic holes on the fabric surface;

[0080] Step S2 includes:

[0081] The reflected light signal is divided into equal-length sequence segments according to a fixed duration;

[0082] Calculate the peak value of the autocorrelation coefficient between adjacent sequence segments;

[0083] The repetition period of periodic holes is determined based on the peak interval distribution;

[0084] Step S3: Construct a reference signal template for periodic holes based on the repetition frequency;

[0085] Step S3 includes:

[0086] Collect reflected light signals from multiple consecutive complete cycles;

[0087] Phase alignment processing is performed on each periodic signal;

[0088] The phase-aligned signals are superimposed and averaged to generate a reference signal template;

[0089] Phase alignment processing includes:

[0090] Identify the zero-crossing positions of signals in each cycle;

[0091] Time displacement alignment is performed based on the first zero-crossing point;

[0092] Extract signal sequences with the same phase interval;

[0093] Step S4: Calculate the normalized cross-correlation coefficient between the current reflected light signal and the reference signal template;

[0094] In step S4, during the calculation of the normalized cross-correlation coefficient, the following steps are performed sequentially: shift alignment, unweighted cross-correlation, weight correction, and coefficient selection. Specifically, this includes:

[0095] The optimal alignment displacement of the signal segment is obtained by using a sliding window optimization method.

[0096] ,

[0097] in, This represents the optimal displacement, expressed in samples. To test the displacement, This represents the number of sampling points corresponding to the baseline template length, in units of samples. The displacement is The unweighted normalized cross-correlation coefficients calculated at that time are dimensionless.

[0098] Will Substituting the values ​​yields the aligned sequence. ,when Sometimes, ,

[0099] The unweighted cross-correlation coefficient is defined as:

[0100] ,

[0101] in, To align the current reflected light signal at the [number]th [time] The amplitude of the sampling point, in V. The arithmetic mean of the aligned signal segments, in V. The reference signal template is in the first The amplitude of the sampling point, in V. The arithmetic mean of the template sequence is expressed in V. The unweighted normalized cross-correlation coefficient is dimensionless.

[0102] when Time-triggered position weight correction, constructing Gaussian position weights:

[0103] ,

[0104] in, For the first The location weighting factor of the sampling points is dimensionless. For the sampling point index, 1- The unit is sample. The standard deviation of the weighted distribution is given by the following values: The unit is sample;

[0105] Weighted normalized cross-correlation coefficient:

[0106] ,

[0107] in, The cross-correlation coefficient is a weighted normalized coefficient, dimensionless.

[0108] Coefficient selection output:

[0109] ,

[0110] in, To output the correlation coefficient, dimensionless. The threshold value is dimensionless.

[0111] Specifically, a sliding correlation maximization strategy is adopted to automatically eliminate phase deviations caused by process disturbances, ensuring that the signal sequence and the template remain consistent on the time axis. The subsequently calculated unweighted cross-correlation coefficient can directly reflect the degree of agreement between the repeating features of the holes and the template when the fabric surface is stable. When yarn vibration or light fluctuations cause increased sampling noise at both ends and reduce the coefficient, Gaussian weights are introduced, giving higher weight to the periodic information in the center of the sequence, and the influence of edge noise on the results is quickly attenuated. The dual-channel output logic balances computational speed and adaptability to abnormal distributions, providing stable and reliable correlation indicators for subsequent dynamic threshold determination.

[0112] Step S5: When the normalized cross-correlation coefficient is lower than the dynamic threshold and is maintained for at least three consecutive cycles, it is determined that there is a hole defect.

[0113] Also includes:

[0114] When a hole defect is determined to exist, the infrared band reflection signal is checked simultaneously.

[0115] Defect determination is confirmed only when the infrared band energy integration decrease condition is met.

[0116] The conditions for energy integral decrease in the infrared band include:

[0117] Calculate the integral value of infrared signal energy within a local time window;

[0118] A judgment is triggered when the score value drops by more than a preset percentage from the baseline.

[0119] In step S5, determining the dynamic threshold includes:

[0120] During the initial operation phase of the fabric, normal hole signals were collected for at least five complete cycles.

[0121] Calculate the historical mean of the normalized cross-correlation coefficient between the signal and the reference template;

[0122] Multiply the historical average by a preset coefficient to obtain the initial dynamic threshold;

[0123] In step S5, the dynamic threshold setting method includes historical sample collection, initial threshold calculation, and exponential adaptive updating, specifically as follows:

[0124] After the fabric enters a stable, uniform speed phase, continuous recording is performed. Correlation coefficient for each period:

[0125] ,

[0126] in, This indicates the number of cycles required for initialization, expressed in cycles. Indicates the first The correlation coefficient of the output over each period is dimensionless. For this The arithmetic mean of the correlation coefficients over a period of time, dimensionless;

[0127] An initial threshold is given based on the historical average level:

[0128] ,

[0129] in, The threshold scaling factor is dimensionless. The dynamic threshold obtained during initialization is dimensionless;

[0130] From the At the start of each cycle, the threshold is adjusted online using an exponentially weighted moving average, and the update formula is as follows:

[0131] , ,

[0132] in, To update the weights, use values ​​from 0 to 1, which are dimensionless. The dynamic threshold of the previous cycle is dimensionless. The correlation coefficient output for the current period is dimensionless. Indicates the current period number. The updated dynamic threshold is dimensionless.

[0133] Will The continuous period determination logic in step S5 is fed into the correlation coefficient. The comparison is performed, and when the latter is continuously below the threshold for more than three cycles, the hole defect monitoring stage is entered.

[0134] Specifically, during the initialization phase, five complete periods of samples are introduced to avoid threshold bias caused by accidental fluctuations in a single period, and the scaling factor... Setting the threshold below the average correlation level allows for the capture of early, subtle anomalies; the exponentially weighted update mechanism automatically integrates recent information in each new cycle, ensuring that the threshold follows a smooth trajectory as the process environment gradually changes without jumps or fluctuations, while also minimizing... It retains the memory of historical working conditions and suppresses excessive threshold reduction caused by instantaneous interference; the three-cycle continuous judgment strategy further filters random noise and realizes timely and robust triggering conditions for hole defects.

[0135] A photodetector array contains miniature detection units.

[0136] The size of each unit's photosensitive area is no greater than 0.1mm × 0.1mm.

[0137] The spacing between array units is less than the minimum detectable defect size on the fabric surface; the minimum detectable defect size refers to the minimum hole diameter or yarn breakage width allowed in the production line quality specifications.

[0138] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting defects in textile fabrics based on photoelectric detectors, characterized in that, Includes the following steps: Step S1: The reflected light signal of the moving fabric surface is collected by a linearly arranged array of photodetectors, the array covering the full width of the fabric. Step S2: Extract the autocorrelation features of the reflected light signal to identify the repetition frequency of periodic holes on the fabric surface; Step S3: Construct a reference signal template for periodic holes based on the repetition frequency; Step S4: Calculate the normalized cross-correlation coefficient between the current reflected light signal and the reference signal template; Step S5: When the normalized cross-correlation coefficient is lower than the dynamic threshold and is maintained for at least three consecutive cycles, it is determined that there is a hole defect.

2. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 1, characterized in that, In step S4, the calculation of the normalized cross-correlation coefficient involves sequentially performing shift alignment, unweighted cross-correlation, weight correction, and coefficient selection, specifically including: The optimal alignment displacement of the signal segment is obtained by using a sliding window optimization method. ， in, This represents the optimal displacement, expressed in samples. To test the displacement, This represents the number of sampling points corresponding to the baseline template length, in units of samples. The displacement is The unweighted normalized cross-correlation coefficients calculated at that time are dimensionless. Will Substituting the values ​​yields the aligned sequence. ,when Sometimes, , The unweighted cross-correlation coefficient is defined as: ， in, To align the current reflected light signal at the [number]th [time] The amplitude of the sampling point, in V. The arithmetic mean of the aligned signal segments, in V. The reference signal template is in the first The amplitude of the sampling point, in V. The arithmetic mean of the template sequence is expressed in V. The unweighted normalized cross-correlation coefficient is dimensionless. when Time-triggered position weight correction, constructing Gaussian position weights: ， in, For the first The location weighting factor of the sampling points is dimensionless. For the sampling point index, 1- The unit is sample. is the standard deviation of the weighted distribution, in units of samples; Weighted normalized cross-correlation coefficient: ， in, The cross-correlation coefficient is a weighted normalized coefficient, dimensionless. Coefficient selection output: ， in, To output the correlation coefficient, dimensionless. The threshold is dimensionless.

3. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 1, characterized in that, Step S2 includes: The reflected light signal is divided into equal-length sequence segments according to a fixed duration; Calculate the peak value of the autocorrelation coefficient between adjacent sequence segments; The repetition period of periodic holes is determined based on the peak interval distribution.

4. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 1, characterized in that, Step S3 includes: Collect reflected light signals from multiple consecutive complete cycles; Phase alignment processing is performed on each periodic signal; The phase-aligned signals are superimposed and averaged to generate a reference signal template.

5. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 4, characterized in that, The phase alignment process includes: Identify the zero-crossing positions of signals in each cycle; Time displacement alignment is performed based on the first zero-crossing point; Extract signal sequences with the same phase interval.

6. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 1, characterized in that, In step S5, determining the dynamic threshold includes: During the initial operation phase of the fabric, normal hole signals were collected for at least five complete cycles. Calculate the historical mean of the normalized cross-correlation coefficient between the signal and the reference template; Multiply the historical average by a preset coefficient to obtain the initial dynamic threshold.

7. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 6, characterized in that, In step S5, the dynamic threshold setting method includes historical sample collection, initial threshold calculation, and exponential adaptive updating, specifically as follows: After the fabric enters a stable and uniform speed phase, continuous recording is performed. Correlation coefficient for each period: ， in, This indicates the number of cycles required for initialization, expressed in cycles. Indicates the first The correlation coefficient of the output over each period is dimensionless. For this The arithmetic mean of the correlation coefficients over a period of time, dimensionless; An initial threshold is given based on the historical average level: ， in, The threshold scaling factor is dimensionless. The dynamic threshold obtained during initialization is dimensionless; From the At the start of each cycle, the threshold is adjusted online using an exponentially weighted moving average, and the update formula is as follows: ， ， in, To update the weights, use values ​​from 0 to 1, which are dimensionless. The dynamic threshold of the previous cycle is dimensionless. The correlation coefficient output for the current period is dimensionless. Indicates the current period number. The updated dynamic threshold is dimensionless. Will The continuous period determination logic in step S5 is fed into the correlation coefficient. The comparison is performed, and when the latter is continuously below the threshold for more than three cycles, the hole defect monitoring stage is entered.

8. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 1, characterized in that, Also includes: When a hole defect is determined to exist, the infrared band reflection signal is checked simultaneously. Defect determination is confirmed only when the condition of energy integral decrease in the infrared band is met.

9. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 8, characterized in that, The infrared band energy integral decrease condition includes: Calculate the integral value of infrared signal energy within a local time window; A judgment is triggered when the score drops by more than a preset percentage from the baseline.

10. The method for detecting defects in textile fabrics based on photoelectric detectors as described in claim 1, characterized in that, The photodetector array contains miniature detection units. The size of each unit's photosensitive area is no greater than 0.1mm × 0.1mm. The spacing between array units is less than the minimum detectable defect size on the fabric surface; the minimum detectable defect size refers to the minimum hole diameter or yarn breakage width allowed in the production line quality specifications.

Images