An intelligent management system based on modular clothing processing templates

Abstract

The invention relates to the technical field of garment processing management, and in particular to an intelligent management system based on modular garment processing templates, comprising: a garment acquisition module, used for acquiring garment images and processing parameters before and after garment processing in each process flow as a sample data set; the processing parameters include parameters corresponding to stitching quality, fabric defects, and color difference problems; a garment matching module, used for comparing the acquired garments with a preset image set according to the sample data set, determining defective areas of the garments in each process flow, and identifying defective labels corresponding to the defective areas; a garment recognition module, used for identifying defects in garment processing according to defective labels corresponding to the defective areas, and determining a defective parameter set; a quality assessment module, used for assessing garment quality according to the acquired defective parameter set, and determining the causes of the defects of the garments; and improving the accuracy and efficiency of defect recognition.

Description

Technical Field

[0001] The invention relates to the technical field of garment processing management, in particular to an intelligent management system based on modular garment processing templates. Background Art

[0002] The traditional garment processing industry relies on manual experience and intuitive judgment, and quality management is often based on subjective standards, lacking quantitative analysis and real-time feedback mechanisms. With the transformation of manufacturing to intelligence, the garment industry has also begun to explore the use of technical means such as big data and artificial intelligence to improve production management. However, due to the complexity and diversity of garment processing, how to effectively integrate these advanced technologies and build an intelligent management system that adapts to the characteristics of the industry has become a key issue that needs to be solved urgently.

[0003] For example, Chinese patent publication number CN115715609A discloses a ready-made garment production method, system, storage medium and intelligent terminal, which relate to the field of clothing, and include: if the placement image detection information is consistent with the image information of the garment piece, then controlling the preset lifting device to clamp and lift the garment piece; if the weight detection information falls into the weight information of the garment piece, then controlling the lifting device to move to the corresponding sewing station; determining the grabbing point from the image corresponding to the sewing image detection information, and controlling the preset lifting device to grab the grabbing point of the garment from the detection area to complete the lifting; if the image corresponding to the sewing image detection information is consistent with the image corresponding to the garment lifting posture image information, then controlling the lifting device to move to the next sewing station, until the lifting device moves to the preset finished product station to discharge the garment, and controlling the lifting device to move to the material delivery warehouse after discharging.

[0004] However, the prior art cannot determine the corresponding defects in the production of clothing, resulting in failure to timely discover the causes of the defects in the production of clothing according to the production process flow, resulting in reduced production quality. Summary of the invention

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] An intelligent management system based on modular clothing processing templates, comprising:

[0007] The clothing acquisition module is used to obtain clothing images and processing parameters before and after clothing processing in each process flow as a sample data set; the processing parameters include parameters corresponding to stitching quality, fabric defects, and color difference problems.

[0008] The clothing matching module is used to compare the acquired clothing with the preset image set according to the sample data set, determine the defective areas of the clothing in each process flow, and identify the defective labels corresponding to the defective areas.

[0009] The clothing recognition module is used to identify defects during clothing processing according to the defect labels corresponding to the defect areas and determine a set of defect parameters.

[0010] The quality assessment module is used to assess the quality of clothing and determine the causes of clothing defects based on the acquired defect parameter set.

[0011] The beneficial effects of the present invention are as follows: the present invention can accurately identify various defects in the clothing processing process, such as stitching quality, fabric defects, color difference problems, etc., through a clothing matching module and a clothing recognition module, using image recognition and classification technology, thereby greatly improving the accuracy and efficiency of defect recognition; it can evaluate the clothing quality according to the identified defect parameter set, and determine the cause of the defect, providing strong data support for subsequent process improvement and quality control; through monitoring and defect recognition of each process flow, problems in the production process can be discovered in time, and the production process can be optimized to reduce the defective rate and improve production efficiency and product quality; the system can intuitively understand the production situation and defect distribution of each process flow through modular design, providing a convenient tool for production management; the system adopts modular design, and can be customized and adjusted according to different production requirements and process flows. At the same time, the system also supports learning and identifying new defect types and processing parameters, thereby enhancing the adaptability and scalability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0013] Figure 1 It is a system schematic diagram based on the intelligent management system of modular clothing processing templates.

[0014] Figure 2 The invention is a flow chart of a clothing matching module based on a modular clothing processing template intelligent management system.

[0015] Figure 3 It is a flow chart of a classification model based on a modular clothing processing template intelligent management system. DETAILED DESCRIPTION

[0016] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used.

[0017] See also Figure 1, an intelligent management system based on modular clothing processing templates, including: a clothing acquisition module, a clothing matching module, a clothing recognition module, and a quality assessment module. The clothing acquisition module inputs clothing images and processing parameters into the clothing matching module, the clothing matching module inputs the identified defect areas and defect labels into the clothing recognition module, the clothing recognition module identifies the defect area and the defect parameter set corresponding to the defect label, and the quality assessment module determines the cause of the clothing defect from the defect parameter set.

[0018] The clothing acquisition module is used to obtain clothing images and processing parameters before and after clothing processing in each process flow as a sample data set; the processing parameters include parameters corresponding to stitching quality, fabric defects, and color difference problems.

[0019] The parameters of fabric defects obtained at this time include the size of the clothing and the properties of the materials used, such as the appearance, weight, color, etc. of the fabric. The fabric defects are mainly to determine whether there are defects in the clothing produced after the fabric is used, so that the fabric can be adjusted. At the same time, fabric defects also include defects caused by the production of each process; stitching quality is the image form of sewing stitches, sewing stitches, and sewing tension before and after completion to determine the required appearance style of the current clothing after sewing. Stitching quality is to determine whether the images before and after stitching correspond to the preset stitching pattern; the color difference problem is to determine whether the color of the current clothing changes after ironing or other dyeing processes, whether normal dyeing is completed, and whether there is no color migration after dyeing.

[0020] The selected process flow includes selecting the raw materials for garment processing, obtaining images of the garments before and after dyeing, peaking, and garment production, so as to determine the corresponding conditions of the garments under multiple processes before and after molding and finalization. The garment images obtained at this time can be garment images collected on a plane or platform, or garment images of the garments when the garments are clamped by a clamping device.

[0021] At the same time, for the clothing image to be recognized, it is necessary to judge the pixel intensity of the image. The pixel intensity of the image is used to represent the contrast of the image recognition during image processing, and is used to determine whether the current clothing needs image processing. If the current clothing form and style are single, it means that the difference in the middle of the image can be directly recognized. If the result of the clothing itself is more complicated, then it is necessary to perform specific recognition on the image to improve the accuracy of image recognition.

[0022] The clothing matching module is used to compare the acquired clothing with the preset image set according to the sample data set, determine the defective areas of the clothing in each process flow, and identify the defective labels corresponding to the defective areas.

[0023] The clothing recognition module is used to identify defects during clothing processing according to the defect labels corresponding to the defect areas and determine a set of defect parameters.

[0024] The quality assessment module is used to assess the quality of clothing and determine the causes of clothing defects based on the acquired defect parameter set.

[0025] The identified defects can be errors in the sewing, cutting, and related pattern combination of the garment, as well as problems with the appearance. At the same time, the corresponding process flow and each defect parameter set of the garment will be further determined based on the defects obtained. Based on this part of the content, quality assessment is then carried out to determine the cause of defects in garment processing, so as to improve the efficiency and quality of garment production.

[0026] The preset image set contains various clothing styles, sizes and craftsmanship to ensure that the currently recognized clothing image can accurately match the preset image set; the preset image set contains standard images before and after each process flow, which are used to compare the defects of the current clothing processing; at the same time, the preset image set also contains clothing images with obvious defects, and manually annotates the location and category of each defect; the annotation information should include the precise location of the defect area and the corresponding defect category label, and the identified location such as a bounding box, mask or key point to determine that after identifying that the current clothing has defects, the identified curve will be compared with the specific defect category to find out the specific style of the current defect.

[0027] like Figure 2 As shown, the methods for determining defective areas in each process flow in the clothing matching module include:

[0028] A1, obtain the production style and task sequence of clothing under each process flow, and extract clothing images according to the process flow from the sample image set according to the production style and task sequence of clothing.

[0029] A2, compare the clothing image with the preset image set in sequence according to the process flow to obtain the image features and position identification features corresponding to the clothing image; the position identification feature is used to indicate the position point where the abnormality appears in the image after the current comparison, as well as the specific size of the position point and the process flow after which the position appears in the image; the image feature is used to identify the relative features on the image when the defect currently occurs.

[0030] A3, input the acquired image features and location identification features into the classification model, and obtain the defect label corresponding to the defect area according to the output of the classification model.

[0031] For the position identification feature, the position identification feature is obtained according to the current process flow and the position of the corresponding defect; for example, the position identification feature in clothing production can be represented as the left sleeve sewing is completed at the 5th workstation of the production line, and an automatic sewing machine is responsible for sewing the connection between the sleeve and the shirt body; the defect occurs at the left cuff of the shirt, specifically at the inner edge of the cuff, about 2 cm away from the end of the cuff; at this time, the workstation, specific process, and position on the garment are the position identification features.

[0032] like Figure 3 As shown, the processing methods for the classification model include:

[0033] B1, the image features and position identification features identified from the clothing image are used as the target feature set; image features include but are not limited to color, texture, shape, edge, etc.; position identification features are used to determine the exact location of defects in the image, which can be pixel coordinates, ROI bounding box or other forms of spatial positioning information; this set describes the appearance and location of defects in the image.

[0034] B2, sequentially acquiring first sample images, wherein the first sample images are sample images for identifying the locations of defective areas; selecting first sample images known to contain defects, and these images should have been marked or the defective areas should have been clearly identified by other means.

[0035] B3, compare the defect features of the first sample image with the target feature set, determine the first similarity between the target feature set and the defect features of the first sample image, and identify the defect area corresponding to the first similarity; at this time, compare the target feature set with the defect features in the first sample image, calculate the first similarity; based on the first similarity, identify the most matching defect area in the target feature set; high similarity indicates that the target feature has a strong match with the defect feature in the sample image.

[0036] B4, extracting a second sample image from the defect area corresponding to the first similarity, the second sample image is used to represent the relevant image features that may appear in the defect area in various situations; that is, obtaining an image related to the corresponding defect in the target feature set as the second sample image, the second sample image is used to extract the defect area and all related defect images of the defect from the defect area where the first similarity exists, to identify the main source of the current defect, the purpose is to compare the defect area again, from being able to identify more subtle defects in the defect area, compare the defect features of the target feature set with the second sample image, and obtain the second similarity; at this time, extract the image features related to the defects in the target feature set from the defect area corresponding to the first similarity, these features can describe the appearance of the defects in more detail; based on the extracted features, generate or select the second sample image, these images emphasize the details of the defect area, can be defect magnification or a view of a specific angle; compare the defect features in the target feature set with the second sample image, and obtain the second similarity; to take into account the differences in defect details.

[0037] B5, based on the comparison result of the first similarity and the second similarity, determine the output defect label.

[0038] The implementation method of the comparison result of the first similarity and the second similarity includes:

[0039] If the second similarity of the target feature set is greater than the first similarity, the sample label of the second sample image is compared with the target feature set, and the sample label with the maximum value of the second similarity after comparison is used as the output defect label; if the second similarity of the target feature set is less than the first similarity, the sample label of the first sample image is compared with the target feature set; the sample label with the maximum value of the first similarity after comparison is used as the output defect label.

[0040] According to the acquired defect label, a pixel coordinate set on the corresponding position identification feature is acquired, and the area of ​​the defect region is determined according to the acquired pixel coordinate set.

[0041] When obtaining the first similarity and the second similarity, the image features that best match the current defect situation are screened out by comparing the obtained target feature set with the first sample image and the second sample image in sequence.

[0042] The first similarity is determined by determining the cosine similarity measure between the target feature set and the first sample image, with the aim of obtaining the image that is most similar to the current feature. The second similarity is calculated in the same way as the first similarity. The second similarity is determined by comparing the sample image with the types and locations of defects that may occur in the known process flow to determine which process the current defect may come from.

[0043] At this time, when the clothing matching module compares the sample data set with the preset image set, it first determines the style of clothing that needs to be completed in each process, as well as the corresponding task order in each process, so as to facilitate the comparison of the corresponding sample data set with the preset image set one by one to determine the specific form of expression of the image.

[0044] Therefore, the method of comparing the clothing images in the sample data set with the preset image set in sequence according to the process flow includes:

[0045] Based on the production style of clothing in each process flow, image features are extracted. The Sobel operator can be used to detect changes in clothing images, and the difference points of the edges of the clothing images in the horizontal and vertical directions are determined respectively, and the pixel values ​​of the difference points are determined. According to the difference points between the clothing image and the preset image set, the defective areas at the corresponding positions are marked according to the positions of the difference points.

[0046] The Sobel operator consists of two 3x3 matrices, one for detecting the horizontal gradient G x , and the other one is used to detect the vertical gradient G y .

[0047]

[0048] By applying these two operators to each pixel of the image, two result images can be obtained, representing the gradient changes in the horizontal and vertical directions. Then, the gradient magnitude G of each pixel is calculated:

[0049]

[0050] After determining the gradient amplitude, a threshold T can be set to identify potential defect areas; if the gradient amplitude of a pixel exceeds this threshold, the pixel is considered to belong to the defect area.

[0051] The defect area D(x,y) is defined as:

[0052] Where G(x,y) is the gradient amplitude at the position (x,y), T is the gradient amplitude threshold; all pixels with a gradient amplitude value of 1 are regarded as the identified defect area; the feature vector corresponding to the current target feature set is obtained from the identified defect area.

[0053] The first similarity can be expressed as obtaining the feature vector of the target feature set and the feature vector of the first sample image to determine the first similarity:

[0054]

[0055] Among them, S1 is the first similarity, D and R represent the vectorized representation of the target feature set and the first sample image respectively, and D i The i-th element in the feature vector representing the target feature set; this can be the grayscale value, color value, gradient value, texture feature, etc. of a pixel in the image; R i Represents the i-th element in the feature vector of the first sample image; used for comparison with the target feature set, N is the number of feature vectors, the feature vector of the target feature set and the feature vector of the first sample image correspond to each other, i=1,2,,,N.

[0056] The analysis content of the second similarity is as follows: the second similarity is obtained based on the feature vector of the second sample image, the feature vector of the target feature set, and the process feature vector.

[0057]

[0058] in, is the feature vector of the target feature set, is the feature vector of the second sample image, is the process feature vector; is the similarity between the feature vector of the target feature set and the feature vector of the second sample image, It is the similarity between the feature vector of the target feature set and the process feature vector.

[0059] The similarities between the feature vector of the target feature set and the feature vector of the second sample image and the similarities between the feature vector of the target feature set and the process feature vector are traversed, and the maximum value of the two is taken as the second similarity.

[0060] After the defective area is identified, a weight is set for the currently identified defective area to obtain the product of the defective part and the weight in each defective area, and the defect label corresponding to the defective area is set according to the product of the defective part and the weight in each defective area.

[0061] According to the location and type of the identified defects, the probability of defect occurrence and the impact of the defect on the appearance of clothing are determined. The impact on the appearance of clothing is expressed as whether the currently identified defect can have a direct impact on the clothing, and whether the impact can be directly observed. At this time, weights are set for the defects according to the parts of the current clothing. The weight is expressed as the weight of the defect for each part of the clothing, and the weight is normalized and weighted averaged to obtain the impact on the appearance of clothing.

[0062] First, determine the range of the mean value of pixels at the defect position and the difference between pixels at the defect position. Based on the range of the mean value of pixels and the difference between pixels in the defect area, take the target area for current identification, compare the target area with the preset image set, and determine the first similarity of the target area. The first similarity is used to indicate the similarity between the defect on the target area and the defect in the preset image set, as well as the corresponding label of the target area under the similarity corresponding to the first similarity. Based on the identification of the label of the target area, determine the defect label corresponding to the defect area under the similarity of the first similarity.

[0063] During the identification, it also includes determining the second similarity between the target area and the process flow. The second similarity is used to determine whether the target area where the defect occurs is related to the current production process flow, and whether it is related to the completed production process flow, so as to find out the factors that can mainly affect the occurrence of the target area, and thus obtain the defect label related to the defect area and the second similarity.

[0064] The methods for identifying defects during garment processing and determining a set of defect parameters include:

[0065] The method of obtaining the defect parameter set includes determining the image information corresponding to each process flow during garment processing;

[0066] According to the corresponding output results in the clothing matching module, the process flow where the defects occur is determined, and the corresponding parameters of each process flow are determined. The determined parameters include but are not limited to style confirmation, size specifications, stitching quality, fabric defects, color difference problems, and accessory integrity in the garment inspection stage.

[0067] Style confirmation: consistency with confirmed sample.

[0068] Size specifications: in accordance with the requirements of the workmanship sheet and sample garments.

[0069] Suture quality: whether the sewing is neat and smooth, and whether there are any skipped stitches or broken threads.

[0070] Fabric defects: Check whether there are defects, oil stains, etc. on the fabric.

[0071] Color difference problem: whether there is color difference in the same garment.

[0072] Integrity of accessories: whether the trademark, care label, size label, etc. are complete and in the correct position.

[0073] When this module is performing clothing recognition, it mainly determines the stitching quality, fabric defects and color difference problems that occur during the processing of the current clothing based on the recognized feature information to determine the inconsistencies in the clothing during image recognition. At this time, the style confirmation and size specifications are the labels when the clothing is matched, and the integrity of the accessories is to determine whether the current clothing is a complete production clothing.

[0074] The implementation of defect parameter set in clothing matching module is as follows:

[0075] The parameter set corresponding to the defect label in the current defect area is input into the loss function to determine the loss rate of the defect label; according to the loss rate of the defect label, the parameter set with a loss rate less than the preset loss rate is output as the defect parameter set.

[0076] The corresponding loss function is expressed as follows: Get the value of the parameter set corresponding to the defect label and the predicted probability of the defect label to determine the loss rate of the defect label:

[0077]

[0078] Where L is the loss rate of defective labels, γ j is the value of the parameter set corresponding to the j-th defect label, p j is the predicted probability of the jth defect label, and m is the number of defect labels; j = 1, 2,,,m.

[0079] The parameter set corresponding to the defect label with the smallest output value of the loss function is used as the output defect parameter set.

[0080] The quality assessment module is used to assess the quality of the clothing according to the acquired parameter set and determine the causes of the defects of the clothing.

[0081] Methods for determining the causes of clothing defects include:

[0082] The probability of occurrence of the defect parameter set is obtained, and the probability distribution of the defect parameter set in each process flow is determined. According to the probability distribution of the defect parameter set in each process flow, the cause of the defect of the clothing is determined.

[0083] The probability distribution of the defect parameter set in each process flow is determined by sorting the occurrence probability of the defect parameter set according to each process flow, determining the overlap rate between the current defect parameter set and the defect parameter set when the defect occurred last time each time a defect occurs, and taking the weighted average of the overlap rate according to the weights of the corresponding parameters in the defect parameter set. The weighted average overlap rate is used as the value of the probability distribution of the defect parameter set in each process flow, and the parameter corresponding to the maximum overlap rate after weighted averaging is used as the defect cause of the clothing.

[0084] First, count the number of times each defect parameter (such as shrinkage, color difference, thread ends, size error, etc.) appears in each process, and calculate the probability of each defect parameter in each process; next, sort the probability of occurrence of the defect parameter set according to the process flow to obtain the probability distribution of the defect parameter set in each process; when a defect occurs again, such as shrinkage, calculate the overlap rate between the current defect parameter set and the parameter set when the shrinkage defect occurred before. The overlap rate refers to the proportion of the same parameters in two sets; according to the weight of each parameter in the defect parameter set, it may be based on historical data, such as the importance of the parameter or the degree of influence on the quality of the final product, to obtain a weighted average overlap rate; find the process with the largest weighted average overlap rate, and the parameter with the largest weighted average overlap rate in this process is regarded as the main cause of the defect.

[0085] It is assumed that shrinkage problems have a higher probability of occurring in both the dyeing and ironing processes, but the weighted average overlap rate is the highest in the dyeing process.

[0086] Parameters: In the dyeing process, water temperature and dye dosage are the parameters with the highest overlap rate with shrinkage defects.

[0087] Conclusion: Therefore, it is preliminarily judged that the cause of the shrinkage problem may be improper control of water temperature and dye dosage in the dyeing process.

[0088] First defect: When the shrinkage problem first occurs, the defect parameter set includes water temperature too high and too much dye.

[0089] Second defect: Shrinkage problem occurs again, and the defect parameter set is water temperature too high and drying temperature too high.

[0090] Calculation of overlap rate: The overlap rate of two defect parameter sets is 50% (only considering excessively high water temperature).

[0091] Weighted average: Assuming the weight of water temperature is 0.7, and the weight of dye dosage and drying temperature is 0.3, the weighted average overlap rate is 0.7*50%=35%.

[0092] Through the above steps, it can be determined that the water temperature control in the dyeing process is the main cause of shrinkage defects, so that targeted improvement measures can be taken, such as adjusting the water temperature control standard, strengthening employee training or improving the dyeing process.

[0093] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention and they are still covered by the protection scope of the present invention.

Claims

1. An intelligent management system based on modular clothing processing templates, characterized in that: include: The clothing acquisition module is used to obtain clothing images and processing parameters before and after clothing processing in each process flow as sample data sets; the processing parameters include parameters corresponding to stitching quality, fabric defects, and color difference problems; The clothing matching module is used to compare the acquired clothing with the preset image set according to the sample data set, determine the defective areas of the clothing in each process flow, and identify the defect labels corresponding to the defective areas; The clothing recognition module is used to identify defects during clothing processing according to the defect labels corresponding to the defect areas and determine the defect parameter set; A quality assessment module is used to assess the quality of clothing and determine the causes of clothing defects based on the acquired defect parameter set; Methods for determining defective areas of garments in various process flows include: Obtain the production style and task sequence of clothing under each process flow, and extract clothing images according to the process flow from the sample image set according to the production style and task sequence of clothing; The clothing image is compared with the preset image set in sequence according to the process flow to obtain the image features and position identification features corresponding to the clothing image; The acquired image features and location identification features are input into the classification model, and the defect label corresponding to the defect area is obtained according to the output of the classification model; The processing methods for classification models include: The image features and position identification features recognized from the clothing image are used as the target feature set; Sequentially acquiring first sample images, wherein the first sample images are sample images for identifying positions of defective areas; Comparing the defect features of the first sample image with the target feature set, determining a first similarity between the target feature set and the defect features of the first sample image, and identifying a defect area corresponding to the first similarity; Extracting a second sample image from the defect area corresponding to the first similarity, comparing the target feature set with the defect feature of the second sample image, and obtaining a second similarity; If the second similarity of the target feature set is greater than the first similarity, the sample label of the second sample image is compared with the target feature set, and the sample label with the maximum value of the second similarity after comparison is used as the output defect label; if the second similarity of the target feature set is less than the first similarity, the sample label of the first sample image is compared with the target feature set; the sample label with the maximum value of the first similarity after comparison is used as the output defect label.

2. According to claim 1, the intelligent management system based on modular clothing processing templates is characterized in that: The garment images are compared with the preset image set in sequence according to the process flow, including: Based on the production style of clothing in each process flow, image features are extracted. The Sobel operator is used to detect changes in clothing images, and the difference points of the edges of the clothing images in the horizontal and vertical directions are determined respectively, and the pixel values ​​of the difference points are determined. According to the difference points between the clothing image and the preset image set, the defective areas at the corresponding positions are marked according to the positions of the difference points.

3. According to claim 1, the intelligent management system based on modular clothing processing templates is characterized in that: The first similarity is expressed as obtaining the feature vector of the target feature set and the feature vector of the first sample image to determine the first similarity: ； in, is the first similarity, The i-th element in the feature vector representing the target feature set; The i-th element in the feature vector representing the first sample image; is the number of eigenvectors, .

4. According to claim 1, the intelligent management system based on modular clothing processing templates is characterized in that: The analysis content of the second similarity is as follows: based on obtaining the feature vector of the second sample image, the feature vector of the target feature set and the process feature vector, the second similarity is obtained; ； ； in, is the feature vector of the target feature set, is the feature vector of the second sample image, is the process feature vector; is the similarity between the feature vector of the target feature set and the feature vector of the second sample image, is the similarity between the feature vector of the target feature set and the feature vector of the process; The similarities between the feature vector of the target feature set and the feature vector of the second sample image and the similarities between the feature vector of the target feature set and the process feature vector are traversed, and the maximum value of the two is taken as the second similarity.

5. According to claim 1, the intelligent management system based on modular clothing processing templates is characterized in that: The defect parameter set in the clothing recognition module is implemented as follows: the parameter set corresponding to the defect label in the current defect area is input into the loss function to determine the loss rate of the defect label; according to the loss rate of the defect label, the parameter set with a loss rate less than the preset loss rate is output as the defect parameter set; The parameter set corresponding to the defect label with the smallest output value of the loss function is used as the output defect parameter set.

6. The intelligent management system based on modularized garment processing templates according to claim 5 is characterized in that: The corresponding loss function is expressed as: obtaining the value of the parameter set corresponding to the defect label and the predicted probability of the defect label, and determining the loss rate of the defect label: ； in, is the loss rate of defective labels, is the value of the parameter set corresponding to the j-th defect label, is the predicted probability of the jth defect label, is the number of defect labels; .

7. The intelligent management system based on modularized garment processing templates according to claim 1 is characterized in that: Methods for determining the causes of clothing defects include: Obtain the occurrence probability of the defect parameter set, and determine the probability distribution of the defect parameter set in each process flow, and determine the cause of the defect of the clothing according to the probability distribution of the defect parameter set in each process flow; The probability distribution of the defect parameter set in each process flow is determined by sorting the occurrence probability of the defect parameter set according to each process flow, determining the overlap rate between the current defect parameter set and the defect parameter set when the defect occurred last time each time a defect occurs, and taking the weighted average of the overlap rate according to the weights of the corresponding parameters in the defect parameter set. The weighted average overlap rate is used as the value of the probability distribution of the defect parameter set in each process flow, and the parameter corresponding to the maximum overlap rate after weighted averaging is used as the defect cause of the clothing.

Images