A data analysis-based intelligent detection method for textile fabric defects

Abstract

The application discloses a kind of based on data analysis's textile fabric flaw intelligent detection method, it is related to fabric flaw detection technical field.The based on data analysis's textile fabric flaw intelligent detection method, by screening suspected area, constructing distinguishing standard, extracting core feature and the logic of double quantitative determination, capture the essential difference of skipping and loose warp in warp continuity, interlacing tightness, floating height stability, avoid the interference of apparent similar of two, rely on the collaborative application of defect feature benchmark value, defect feature distribution interval and warp float warp continuation ratio, warp and weft interlacing cohesion, floating height warp fluctuation rate, and the double verification of distribution identification and fit degree calculation, solve the problem that skipping and loose warp in warp float class defects in woven fabric cannot be accurately distinguished due to apparent feature high similarity.

Description

Technical Field

[0001] This invention relates to the field of fabric defect detection technology, specifically to an intelligent method for detecting textile fabric defects based on data analysis. Background Technology

[0002] In the detection of warp yarn lifting defects in woven fabrics, both skipped yarns and loose warp yarns exhibit visual characteristics of deviating from the normal weaving trajectory and locally lifting. These defects are commonly affected by factors such as fabric texture density, defect severity, and environmental interference during photography. Existing detection technologies reveal the following deficiencies: First, existing technologies lack precise quantitative extraction of the implicit essential characteristics of the two types of defects, relying solely on superficial visual differences for rough judgment, failing to establish effective distinguishing criteria. Second, existing solutions lack core references for quantitative comparison, making it difficult to overcome the cognitive limitations of superficial similarity. Finally, existing technologies do not design multi-dimensional collaborative judgment logic, classifying defects solely through a single indicator or visual observation, failing to accurately identify the essential differences between the two types of defects in continuity, interweaving state, and floating stability, resulting in a high misjudgment rate.

[0003] Therefore, there is an urgent need for an intelligent detection method that possesses core feature precise quantification, benchmark reference system construction, and dual quantification judgment to accurately distinguish between skipped yarn and loose warp defects. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a data analysis-based intelligent detection method for textile fabric defects, which solves the problem of inaccurate differentiation between warp yarn float defects (skipped yarns and loose warp yarns) in woven fabrics due to their highly similar appearance.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a data analysis-based intelligent detection method for textile fabric defects, comprising the following steps: The fabric image is segmented into equal-width warp strip regions. The number of discontinuities per unit area and the proportion of interlacing texture in the warp strip regions are calculated to identify suspected floating defect areas.

[0006] Based on the yarn skipping characteristic parameters and the loose warp characteristic parameters, defect feature benchmark values ​​and defect feature distribution intervals corresponding to yarn skipping defects and loose warp defects relative to the benchmark features are constructed.

[0007] For suspected floating defect areas, warp tracking statistics of warp yarn pixel columns, pixel area identification of interlacing and interlocking areas, and height sampling statistics of warp equidistant points are used to obtain the warp yarn floating warp discontinuity ratio, warp and weft yarn interlacing cohesion degree, and floating height warp fluctuation rate.

[0008] Based on the defect feature benchmark value and defect feature distribution range, and combined with the warp breakage ratio of warp yarn float, warp and weft yarn interlacing cohesion, and warp fluctuation rate of float height, distribution identification and fit calculation are performed to determine the fabric defect type.

[0009] Compared with the prior art, the present invention has the following beneficial effects: This invention captures the essential differences between skipped warp and loose warp in terms of warp continuity, interlacing tightness, and float height stability by screening suspected areas, constructing differentiation criteria, extracting core features, and using dual quantitative judgment. This avoids interference from their similar appearances. Relying on the synergistic application of defect feature benchmark values, defect feature distribution ranges, warp float warp continuity ratio, warp and weft interlacing cohesion, and float height warp fluctuation rate, as well as dual verification of distribution identification and fit calculation, this invention solves the problem of inaccurate differentiation between skipped warp and loose warp in woven fabrics due to their highly similar appearances. Attached Figure Description

[0010] Figure 1 This is a flowchart of the intelligent detection method for textile fabric defects based on data analysis according to the present invention.

[0011] Figure 2 This is a flowchart illustrating the determination of the defect feature benchmark value and defect feature distribution range in the intelligent detection method for textile fabric defects based on data analysis of the present invention.

[0012] Figure 3 This is a flowchart illustrating the process of determining the type of fabric defect in the intelligent detection method for textile fabric defects based on data analysis, as described in this invention. Detailed Implementation

[0013] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. Please refer to the accompanying drawings. Figure 1 This invention provides a technical solution: an intelligent detection method for textile fabric defects based on data analysis, comprising the following steps: S1. Divide the fabric image into equal-width warp strip regions, calculate the number of discontinuities per unit area and the proportion of interlacing texture in the warp strip regions, and identify suspected floating defect areas.

[0014] Considering that warp yarn lifting disrupts its continuous warp extension, the number of discontinuities per unit area directly reflects the frequency of warp yarn breakage or discontinuity. Furthermore, the interlacing state of warp and weft yarns changes due to warp yarn lifting, resulting in more drastic changes in pixel grayscale at the interlacing points in both warp and weft directions. Additionally, lifted warp yarns are further from the fabric plane and reflect light more strongly, with the average grayscale value reflecting brightness differences. Therefore, the process for identifying suspected lifting defect areas is as follows: S101. Perform a row-by-row pixel scan of the warp strip area along the warp direction of the fabric. That is, from the top to the bottom of the warp strip area, traverse each pixel point horizontally, count the number of discontinuities, and take the ratio of the total number of discontinuities to the area of ​​the warp strip area as the number of discontinuities per unit area. Among them, discontinuities are pixels that do not satisfy the 4-neighborhood connectivity relationship.

[0015] S102. For each pixel in the current warp strip area, calculate the relative gray value in the radial and weft directions respectively. The relative gray value in the warp direction is the absolute difference between the gray value of the pixel and the adjacent warp (vertical) pixel. The relative gray value in the weft direction is the absolute difference between the gray value of the pixel and the adjacent weft (left-right) pixel.

[0016] S103. If the relative grayscale difference of a pixel in the warp strip area is greater than a set threshold in both the radial and weft directions, it is marked as a pixel with interwoven texture. The ratio of the number of pixels with interwoven texture to the total number of pixels in the warp strip area is taken as the proportion of interwoven texture. The set threshold is determined based on the statistical analysis of the relative grayscale difference in the warp and weft directions of flawless fabric: the warp strip area of ​​the flawless fabric image is scanned pixel by pixel, the relative grayscale difference in the warp direction and the relative grayscale difference in the weft direction of all pixels are calculated, and the 95th percentile of the two types of differences is taken as the set threshold.

[0017] S104. Calculate the grayscale value of all pixels in each warp strip area and calculate the average grayscale value of each warp strip area.

[0018] S105. If the number of discontinuities per unit area and the proportion of interlacing texture in a certain warp strip area are both greater than those in the adjacent warp strip areas (i.e., 1-2 warp strip areas that are directly adjacent after being divided), and the average gray value is greater than that in the adjacent warp strip areas, then it is determined to be a suspected floating defect area.

[0019] It should be noted that the number of discontinuities per unit area characterizes the continuity of the warp yarn in the warp direction. It directly reflects the frequency of continuous interruption caused by the lifting of the warp yarn in each unit area. The larger the number of discontinuities per unit area, the more obvious the disruption of the continuous warp extension state.

[0020] The proportion of interlaced texture characterizes the tightness of the interlacing of warp and weft yarns and the abnormality of the interlacing state. When the warp and weft yarns interlac, the pixel grayscale will have a large difference in both the warp and weft directions; while when the warp yarns float, the pixel grayscale parameters at the interlacing point will change significantly and the number of pixels will change.

[0021] The mean grayscale value represents the overall reflective brightness level of the warp strip area. When the warp yarns are lifted and illuminated by a light source, the reflection is stronger, and the corresponding pixel grayscale value in that area of ​​the image is generally higher. Therefore, this parameter directly quantifies the brightness difference of the area and can distinguish between the lifted defect area and the normal area.

[0022] This embodiment uses equal-width warp strip area segmentation, combined with the determination of three parameters: the number of discontinuities per unit area, the proportion of interlacing texture, and the average gray value, to screen out warp floating defect areas and effectively eliminate interference from normal areas and non-floating defects.

[0023] S2. Based on the skipped yarn characteristic parameters and loose warp characteristic parameters, construct the defect characteristic benchmark value and defect characteristic distribution range corresponding to the skipped yarn defect and loose warp defect relative to the benchmark feature, respectively.

[0024] Considering that although skipped warp and loose warp both belong to the warp float category, their characteristic parameters exhibit implicit typical patterns. Clustering based on numerical similarity can be used to select the most representative parameter groups. Furthermore, considering that there must be differences between the baseline characteristics and defects of flawless fabrics, therefore... Figure 2 As shown, the specific process for determining the defect feature baseline value and the defect feature distribution range is as follows: S201, the warp breakage ratio of the reference warp yarn float, the interlacing cohesion of the reference warp and weft yarns, and the warp fluctuation rate of the reference float height of the flawless fabric are used as reference characteristics.

[0025] S202. Cluster the multiple sets of yarn skipping characteristic parameters and loose warp characteristic parameters respectively to determine the optimal yarn skipping characteristic parameter group and the optimal loose warp characteristic parameter group.

[0026] S203. Calculate the mean of the relative deviation values ​​of the parameters in the optimal yarn skipping feature parameter group and the optimal loose warp feature parameter group with the corresponding parameters in the benchmark feature to obtain the corresponding defect feature benchmark value.

[0027] S204. Based on the relative deviation values ​​of each parameter in the yarn skipping characteristic parameters and loose warp characteristic parameters, the fluctuation range of each parameter is determined by the interquartile range method. After removing the extreme values ​​outside the interquartile range, the minimum value is used as the lower boundary and the maximum value is used as the upper boundary to obtain the corresponding defect characteristic distribution interval.

[0028] The optimal set of feature parameters for skipped yarn and the optimal set of feature parameters for loose warp, selected after clustering, can reflect the general characteristics of the two types of skipped yarn and loose warp to the greatest extent, ensuring the typicality of the defect feature benchmark values. Furthermore, the calculation based on the mean of the relative deviation values ​​of the benchmark features makes the feature differences between skipped yarn and loose warp more comparable. In addition, the defect feature distribution range determined by the interquartile range method after removing extreme values ​​can effectively cover the reasonable fluctuation range of skipped yarn and loose warp.

[0029] The process of calculating the mean of the relative deviation values ​​for the optimal yarn skipping feature parameter group is as follows: For the warp breakage ratio of the warp yarn float corresponding to each sample in the optimal yarn skipping feature parameter group, calculate the relative deviation value between it and the reference warp breakage ratio of the warp yarn float. Then, calculate the mean of the relative deviation values ​​corresponding to all warp breakage ratios of the warp yarn float. The same calculation method is used for the warp and weft yarn interlacing cohesion and the warp fluctuation rate of the float height in the same yarn skipping feature parameters.

[0030] Of course, the process of calculating the mean of the relative deviation values ​​of the optimal pine needle characteristic parameter set is the same as the process described above.

[0031] Taking the characteristic relative deviation value corresponding to the warp breakage ratio in the warp yarn float characteristic parameters as an example, the process of determining the corresponding defect characteristic distribution range using the interquartile range method is as follows: First, sort the warp breakage ratios in ascending order, calculate the first and third quartiles, and obtain the interquartile range by the difference between the third and first quartiles.

[0032] Then, the range for determining extreme values ​​is: ,in, It is the first quartile. Interquartile range, It is the third quartile. This is the lower limit of the range for determining extreme values. This is the upper limit of the range for determining extreme values.

[0033] Finally, all data outside the extreme value judgment range are removed to obtain the effective feature relative deviation value. The smallest effective feature relative deviation value is taken as the lower boundary of the defect feature distribution interval, and the largest effective feature relative deviation value is taken as the upper boundary of the defect feature distribution interval.

[0034] Of course, the interlacing cohesion of warp and weft yarns and the floating height of warp yarns, as well as the relative deviation values ​​of each parameter in the loose warp characteristic parameters, are also determined using the above-mentioned interquartile range method to determine their corresponding defect characteristic distribution range.

[0035] Furthermore, the process of determining the optimal yarn skipping feature parameter set and the optimal warp loosening feature parameter set based on clustering is as follows: First, the yarn skipping feature parameters and the loose warp feature parameters are standardized separately, and the standardized yarn skipping feature parameters and loose warp feature parameters are treated as independent datasets.

[0036] Then, a clustering algorithm based on parameter numerical similarity, such as K-means clustering, is used to cluster the two independent datasets separately. It should be noted that K-means clustering is an existing technique; the value of K is determined by the elbow rule and is not changed during application. Taking the standardized skipping feature parameters as an example, the clustering process is as follows: Calculate the numerical similarity of warp yarn rise ratio, warp and weft yarn interlacing cohesion, and rise height warp fluctuation rate in each group of skipped yarn characteristic parameters. Group multiple groups of skipped yarn characteristic parameters with similar numerical characteristics into the same skipped yarn characteristic parameter cluster group, and finally form multiple skipped yarn characteristic parameter cluster groups.

[0037] The standardized pine diameter feature parameters are clustered using the same method.

[0038] Finally, counts were performed on all the yarn skipping feature parameter clusters and all the warp loosening feature parameter clusters. Among all the yarn skipping feature parameter clusters, the yarn skipping feature parameter cluster with the largest number of samples was selected as the optimal yarn skipping feature parameter cluster. Similarly, the warp loosening feature parameter cluster with the largest number of samples was selected as the optimal warp loosening feature parameter cluster.

[0039] S3. For suspected floating defect areas, warp tracking statistics of warp yarn pixel columns, pixel area identification of interlacing and interlocking areas, and sampling statistics of warp equidistant points are used to obtain the warp yarn floating warp discontinuity ratio, warp and weft yarn interlacing cohesion degree, and floating height warp fluctuation rate.

[0040] The specific process is as follows: S301. Perform continuous connected domain tracing along the warp direction for the warp yarn pixel column and along the weft direction for the weft yarn pixel column to determine the uninterrupted continuous segments of the warp yarn and the weft yarn.

[0041] S302. The ratio of the sum of the warp pixel lengths of the uninterrupted continuous warp segments to the total warp pixel length is used as the warp float discontinuity ratio. It should be noted that the total warp pixel length is the sum of the pixel lengths of all warp yarns within the warp strip area, and the pixel length of a single warp yarn is the sum of the warp pixel lengths of the uninterrupted continuous warp segments and the warp pixel lengths of the discontinuous regions, where the discontinuous regions are the areas formed by connecting discontinuities.

[0042] S303. Combine the uninterrupted continuous segments of warp yarns and weft yarns to determine the pixel area in which the warp and weft yarns interweave and engage. Use the ratio of the pixel area to the theoretical total pixel area as the warp and weft yarn interweaving cohesion degree.

[0043] S304. Sampling is performed at equal intervals along the warp direction. The range and arithmetic mean of the pixel height values ​​at the sampling points are calculated. The ratio of the range to the arithmetic mean is used as the warp fluctuation rate of the float height. The number of sampling points is determined based on the warp length of the warp strip area, and in principle, one sampling point is set for every 10 pixels.

[0044] Among them, the warp lift warp discontinuity ratio characterizes the degree of continuity and integrity of the warp yarn in the warp direction. The closer the ratio is to 1, the better the continuous extension of the warp yarn in the warp direction and the fewer the discontinuities. The lower the warp lift warp discontinuity ratio, the more serious the discontinuity caused by the lifting of the warp yarn.

[0045] The warp and weft interlacing cohesion degree characterizes the tightness of the interlacing between warp and weft yarns. The higher the warp and weft interlacing cohesion degree, the tighter the interlacing, the wider the effective interlacing range, and the closer the interlacing state is to normal. The lower the warp yarn float ratio, the more disordered the interlacing relationship is.

[0046] The warp variability of the lift height characterizes the stability of the warp lift height in the warp extension direction. The higher the warp variability of the lift height, the more drastic the fluctuation of the lift height during the warp extension process, and the more unstable the lift state; the lower the warp variability of the lift height, the more stable the lift height.

[0047] The reasons for choosing warp yarn float ratio, warp and weft yarn interlacing cohesion, and warp fluctuation rate of float height as the distinction between loose warp and skipped warp are as follows: The warp yarn lift ratio reflects the warp continuity. Loose warp yarns, due to only slack without significant breaks or jumps, have longer continuous warp sections and a relatively higher warp yarn lift ratio. Skipped warp yarns, due to warp yarns skipping multiple weft yarns, cause significant discontinuities and have a lower warp yarn lift ratio. The warp and weft yarn interlacing cohesion reflects the tightness and effective range of interlacing. Loose warp yarns only result in looser warp yarns and shallower interlacing, with a slight reduction in the effective interlacing area, resulting in a lower but not extreme warp and weft yarn interlacing cohesion. Skipped warp yarns, due to warp yarns directly skipping weft yarns, cause localized interlacing gaps, resulting in an even lower and more pronounced warp and weft yarn interlacing cohesion. The warp fluctuation rate of the lift height reflects the stability of the lift height. Loose warp yarns typically exhibit uniform overall slack and lift with gradual height changes, resulting in a lower warp fluctuation rate. Skipped warp yarns, due to sudden jumps in some areas of the warp yarns, cause rapid increases and decreases in lift height, resulting in a significantly higher warp fluctuation rate.

[0048] The warp yarn float ratio, warp and weft yarn interlacing cohesion, and warp fluctuation of float height work together to form characteristic dimensions that can distinguish between loose warp and skipped yarn defects.

[0049] The process of determining the uninterrupted continuous segments of warp and weft yarns is as follows: Perform yarn pixel grayscale statistics along the warp and weft directions respectively to obtain the main pixel columns corresponding to the warp and weft textures.

[0050] Starting from the two main pixel columns, if a warp pixel or a weft pixel does not satisfy the 4-neighbor connectivity relationship, it is determined to be a discontinuity.

[0051] Using the discontinuity point, the start point and the end point of the main pixel column as coordinate boundaries, the warp-connected pixel intervals and weft-connected pixel intervals between two adjacent coordinate boundaries are respectively marked as warp-uninterrupted continuous segments and weft-uninterrupted continuous segments.

[0052] Furthermore, the process of determining the pixel area where the warp and weft yarns interlock by combining the uninterrupted continuous segments of the warp and weft yarns is as follows: Geometric intersection operations are performed on the sets of two-dimensional pixel coordinates corresponding to the uninterrupted continuous warp and weft segments to obtain the overlapping pixel region.

[0053] The number of pixels in overlapping pixel regions is counted, and the pixel area is obtained based on the conversion relationship between a single pixel and its physical area. This conversion relationship is a pre-stored fixed value. For example, the acquisition process is as follows: a flawless textile fabric with known physical dimensions is selected as a standard sample. Images are captured under an environment with a light intensity of 500-1000 lux and a shooting distance of 50-100cm. The images are divided into equal-width warp strip regions, and the total number of pixels in each region is counted. The actual physical area of ​​the standard sample's warp strip region is divided by the total number of pixels to obtain the physical area corresponding to a single pixel. This value is pre-stored in the detection system as the conversion relationship.

[0054] In this embodiment, geometric intersection operation, as a mature coordinate filtering technology, can directly lock the overlapping part of the two-dimensional pixel coordinate sets corresponding to the uninterrupted continuous warp and weft segments; by counting the number of overlapping pixel areas, the visually interwoven areas can be transformed into quantifiable values, providing a basis for area calculation.

[0055] S4. Based on the defect feature benchmark value and defect feature distribution range, combined with the warp breakage ratio of warp yarn float, warp and weft yarn interlacing cohesion and float height warp fluctuation rate, the distribution identification and fit calculation are performed to determine the fabric defect type.

[0056] like Figure 3 As shown, the specific process is as follows: S401. Compare the warp breakage ratio of warp yarn float, warp and weft yarn interlacing cohesion, and warp fluctuation rate of float height with the defect feature distribution intervals corresponding to skipped yarn defects and loose warp defects, respectively, to obtain the membership degree of the skipped yarn interval and the membership degree of the loose warp interval.

[0057] S402. Based on the warp break ratio of warp yarn float, warp and weft yarn interlacing cohesion, and warp fluctuation rate of float height, Euclidean distance is calculated with the corresponding defect feature benchmark values ​​of skipped yarn defect and loose warp defect, respectively, to obtain the matching degree of skipped yarn defect and the matching degree of loose warp defect.

[0058] S403. When the membership degree of the skipped yarn interval is not less than 2 and the fit degree of the skipped yarn defect is less than the fit degree of the loose warp defect, the fabric defect type is skipped yarn defect.

[0059] S404. When the membership degree of the loose warp interval is not less than 2 and the fit degree of loose warp defect is less than the fit degree of skipped yarn defect, the fabric defect type is loose warp defect.

[0060] If the above criteria are not met, the defect is directly identified as a defect to be identified.

[0061] Both the membership degree of the skipped yarn interval and the membership degree of the loose warp interval characterize the warp yarn floating warp direction discontinuity ratio, warp and weft yarn interlacing cohesion degree, and floating height warp direction fluctuation rate of the suspected floating defect area, as well as the degree of matching and fit with the defect characteristic distribution intervals of skipped yarn defects and loose warp defects.

[0062] The membership values ​​for both the skipped warp interval and the loose warp interval range from 0 to 3. A value of 0 indicates that the warp breakage ratio, warp and weft interlacing cohesion, and warp fluctuation rate of the rise height do not fall within the defect characteristic distribution range of the corresponding defect, suggesting that the parameter fluctuation range of the suspected rise defect area does not match the defect. A value of 3 indicates that the warp breakage ratio, warp and weft interlacing cohesion, and warp fluctuation rate of the rise height all fall within the defect characteristic distribution range of the corresponding defect, suggesting that the parameter fluctuation range of the suspected rise defect area matches the defect height.

[0063] The fit degree of the skipped yarn defect and the fit degree of the loose warp defect are both calculated using Euclidean distance. They characterize the relative deviations of three features in the suspected floating defect area and the degree of similarity between these features and the baseline values ​​of the defect features of the skipped yarn defect and the loose warp defect. The closer the values ​​of the fit degree of the skipped yarn defect and the loose warp defect are to 0, the more similar the relative deviations of the features in the suspected floating defect area are to the baseline values ​​of the defect features.

[0064] Membership degree can screen out suspected areas whose parameter fluctuation ranges conform to the defect rules by judging whether the warp yarn float ratio, warp and weft yarn interlacing cohesion, and float height warp fluctuation rate fall into the characteristic distribution range of the corresponding defect, thus ensuring the basic adaptability of the judgment; while the fit degree uses cosine similarity to quantify the similarity between the suspected area features and the typical features of the two types of defects, capturing the subtle differences between the two in core features.

[0065] The process of obtaining the membership degree of the yarn skipping interval and the membership degree of the loose warp interval is as follows: If the warp yarn float ratio, warp and weft yarn interlacing cohesion, and float height warp fluctuation rate fall within the defect characteristic distribution range corresponding to the skipped yarn defect and loose warp defect, then the membership degree of the corresponding skipped yarn range or loose warp range is increased by 1; otherwise, it is not increased by 1.

[0066] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0067] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0068] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0069] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0070] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data analysis-based intelligent detection method for textile fabric defects, characterized in that, Includes the following steps: The fabric image is segmented into equal-width warp strip regions. The number of discontinuities per unit area and the proportion of interlacing texture in the warp strip regions are calculated to identify suspected floating defect areas. Based on the yarn skipping characteristic parameters and the loose warp characteristic parameters, the defect characteristic benchmark values ​​and defect characteristic distribution intervals corresponding to yarn skipping defects and loose warp defects relative to the benchmark characteristics are constructed respectively. For suspected floating defect areas, warp yarn pixel column warp direction tracking statistics, interlacing and interlocking area pixel area identification, and warp direction equidistant point height sampling statistics are used to obtain warp yarn floating warp direction discontinuity ratio, warp and weft yarn interlacing cohesion degree, and floating height warp direction fluctuation rate. Based on the defect feature benchmark value and defect feature distribution range, and combined with the warp breakage ratio of warp yarn float, warp and weft yarn interlacing cohesion, and warp fluctuation rate of float height, distribution identification and fit calculation are performed to determine the fabric defect type.

2. The intelligent detection method for textile fabric defects based on data analysis according to claim 1, characterized in that, The process of calculating the number of discontinuities per unit area and the proportion of interlacing texture in the warp strip region is as follows: The warp strip area is scanned pixel by pixel along the warp direction of the fabric, and the number of discontinuities is counted. The ratio of the total number of discontinuities to the area of ​​the warp strip area is taken as the number of discontinuities per unit area. If the relative difference in grayscale values ​​of a pixel in the warp strip area is greater than a set threshold in both the radial and weft directions, it is marked as a pixel with interwoven texture. The ratio of the number of pixels with interwoven texture to the total number of pixels in the warp strip area is used as the proportion of interwoven texture.

3. The intelligent detection method for textile fabric defects based on data analysis according to claim 1, characterized in that, The process for determining the suspected floating defect area is as follows: Calculate the average gray value of each warp strip region; If the number of discontinuities per unit area and the proportion of interlacing texture in a certain warp strip area are both greater than those in the adjacent warp strip areas, and the average gray value is greater than that in the adjacent warp strip areas, then it is identified as a suspected floating defect area.

4. The intelligent detection method for textile fabric defects based on data analysis according to claim 1, characterized in that, The process of constructing the defect feature reference value and defect feature distribution interval corresponding to the skipped yarn defect and loose warp defect relative to the reference feature based on the skipped yarn feature parameter and loose warp feature parameter is as follows: The baseline warp yarn lift warp discontinuity ratio, baseline warp and weft yarn interlacing cohesion, and baseline lift height warp fluctuation rate of flawless fabric are used as baseline characteristics. Multiple sets of yarn skipping feature parameters and warp loosening feature parameters are clustered to determine the optimal yarn skipping feature parameter group and the optimal warp loosening feature parameter group. The average value of the relative deviation between the parameters in the optimal yarn skipping feature parameter group and the optimal loose warp feature parameter group and the corresponding parameters in the benchmark feature is calculated to obtain the corresponding defect feature benchmark value. Based on the relative deviation values ​​of each parameter in the yarn skipping characteristic parameters and loose warp characteristic parameters, the fluctuation range of each parameter is determined by the interquartile range method. After removing the extreme values ​​outside the interquartile range, the minimum value is used as the lower boundary and the maximum value is used as the upper boundary to obtain the corresponding defect characteristic distribution range.

5. The intelligent detection method for textile fabric defects based on data analysis according to claim 4, characterized in that, The process of determining the optimal yarn skipping characteristic parameter set and the optimal warp loosening characteristic parameter set is as follows: The number of samples in the cluster groups for the yarn skipping feature parameters and the cluster groups for the yarn warp feature parameters were counted separately. The cluster group of yarn skipping feature parameters with the largest number of samples and the cluster group of loose warp feature parameters are respectively taken as the optimal yarn skipping feature parameter group and the optimal loose warp feature parameter group.

6. The intelligent detection method for textile fabric defects based on data analysis according to claim 1, characterized in that, The process for obtaining the warp yarn lift warp discontinuity ratio, warp and weft yarn interlacing cohesion, and lift height warp fluctuation rate is as follows: Continuous connected component tracing is performed along the warp direction for the warp yarn pixel column and along the weft direction for the weft yarn pixel column to determine uninterrupted continuous segments of the warp yarn and weft yarn; The ratio of the sum of the warp pixel lengths of uninterrupted continuous warp segments to the total warp pixel length is taken as the warp breakage ratio. The pixel area where the warp and weft yarns interlock is determined by combining the uninterrupted continuous segments of the warp and weft yarns, and the ratio of the pixel area to the theoretical total pixel area is taken as the warp and weft yarn interlocking degree. Equidistant pixel sampling is performed along the warp extension direction. The range and arithmetic mean of the pixel height values ​​at the sampling points are calculated. The ratio of the range to the arithmetic mean is used as the warp fluctuation rate of the floating height.

7. The intelligent detection method for textile fabric defects based on data analysis according to claim 6, characterized in that, The process of determining the uninterrupted continuous segments of warp yarns and weft yarns is as follows: Perform yarn pixel grayscale statistics along the warp and weft directions respectively to obtain the main pixel columns corresponding to the warp texture and weft texture respectively; Starting from the two main pixel columns respectively, if a warp pixel or a weft pixel does not satisfy the 4-neighbor connectivity relationship, it is determined to be a discontinuity. Using the discontinuity point, the start point and the end point of the main pixel column as coordinate boundaries, the warp-connected pixel intervals and weft-connected pixel intervals between two adjacent coordinate boundaries are respectively marked as warp-uninterrupted continuous segments and weft-uninterrupted continuous segments.

8. The intelligent detection method for textile fabric defects based on data analysis according to claim 6, characterized in that, The process of determining the pixel area where the warp and weft yarns interlock by combining uninterrupted continuous warp and weft yarn segments is as follows: Geometric intersection operation is performed on the two-dimensional pixel coordinate sets corresponding to the uninterrupted continuous warp and weft segments to obtain the overlapping pixel region; The number of pixels in the overlapping pixel region is counted, and the pixel area is obtained based on the conversion relationship between a single pixel and the physical area.

9. The intelligent detection method for textile fabric defects based on data analysis according to claim 1, characterized in that, The process of performing distribution identification and fit calculation to determine the type of fabric defect is as follows: The warp yarn float ratio, warp and weft yarn interlacing cohesion, and float height warp fluctuation rate are compared with the defect feature distribution intervals corresponding to skipped yarn defects and loose warp defects, respectively, to obtain the membership degree of the skipped yarn interval and the membership degree of the loose warp interval. Based on the warp break ratio of warp yarn float, warp and weft yarn interlacing cohesion, and warp fluctuation rate of float height, Euclidean distance is calculated by combining the benchmark features with the benchmark values ​​of the defect features corresponding to skipped yarn defects and loose warp defects, respectively, to obtain the fit degree of skipped yarn defects and the fit degree of loose warp defects. If the membership degree of the skipped yarn interval is not less than 2 and the fit degree of the skipped yarn defect is less than the fit degree of the loose warp defect, then the fabric defect type is skipped yarn defect. If the membership degree of the loose warp interval is not less than 2 and the fit degree of the loose warp defect is less than the fit degree of the skipped yarn defect, then the fabric defect type is loose warp defect.

10. The intelligent detection method for textile fabric defects based on data analysis according to claim 9, characterized in that, The process of obtaining the membership degree of the skipped yarn interval and the membership degree of the loose warp interval is as follows: If the warp yarn float ratio, warp and weft yarn interlacing cohesion, and float height warp fluctuation rate fall within the defect characteristic distribution range corresponding to the skipped yarn defect and loose warp defect, then the membership degree of the corresponding skipped yarn range or loose warp range is increased by 1; otherwise, it is not increased by 1.

Images