A fabric one-way moisture conducting detection method based on image processing

Abstract

The present application relates to a kind of fabric one-way wet detection methods based on image processing, the main steps of the present application are: first, water drop experiment is carried out to fabric, obtains the diffusion image of its in different time of fabric inner and outer layer, after pre-processing, the water drop contour map is segmented using image subtraction method, the radius and area of water drop at each time are calculated, obtain the diffusion rate of moisture in fabric, then the one-way wet ability of fabric is detected, and the correlation analysis of image method detection result and MMT detection result is carried out.Finally, the wet index of each item measured by image method is used as input function, the wet grade of fabric corresponding to MMT measured is used as output function, and it is substituted into the rating training of BP neural network, the rating of fabric wet performance is realized, the fabric wet theory and fluid equation are introduced into the mutual calculation of water absorption rate and diffusion rate, the diversity of image method performance index is increased, and the accuracy and scientificity of the detection result of this method are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fabric one-way moisture transfer level detection method, and particularly relates to a fabric one-way moisture transfer detection method based on image processing. BACKGROUND

[0002] The one-way moisture transfer fabric is obtained by making the water absorption performance of the inner and outer layers of the fabric different through certain physical or chemical methods to achieve the effect that water can only transfer in one direction. For example, the outer layer of the fabric is treated to be water and oil repellent; or during weaving, different raw materials are used, such as using polyester with poor water absorption in the outer layer and cotton, polypropylene and other fibers with strong water absorption in the inner layer; or using a special mesh structure. The above methods all make the inner and outer layers of the fabric different in water absorption performance to achieve the purpose of one-way moisture transfer. The fabric is divided into inner and outer layers. Since the sweat generated by human movement needs to be discharged from the inner layer to the outer layer of the fabric, the water absorption of the outer layer of the fabric is usually greater than that of the inner layer. In terms of the fiber content of the fabric of the clothes worn close to the body, the inner layer contains both hydrophilic and hydrophobic fibers, but the outer layer is entirely composed of hydrophilic fibers. When a person is sweating due to intense exercise, because the absorption effect of sweat on different layers of the fabric is different, that is, there is a difference in the chemical potential energy of sweat on different layers of the fabric, the sweat will be quickly absorbed from the hydrophilic area of the inner layer of the fabric and transferred to the outer layer of the fabric, and then quickly evaporated in the external environment. At the same time, during the absorption and dissipation of water, because a certain amount of hydrophobic fibers exist in the inner layer of the fabric, the sweat is also not in direct contact with the human body while being absorbed, avoiding the generation of adhesion, thereby increasing the rapid evaporation of sweat in the outer layer of the fabric to bring the wearer a comfortable cool feeling.

[0003] The image method for detecting the moisture transfer performance of the fabric is a method for obtaining the diffusion of water droplets on the inner and outer layers of the fabric at different time periods and calculating the moisture transfer performance. However, this method can only involve the diffusion rate of water droplets, and cannot detect the water absorption rate of the fabric, which is the core indicator of the moisture transfer performance of the fabric. SUMMARY

[0004] In view of the deficiencies of the existing mismatch detection methods, the present application provides a fabric one-way moisture transfer detection method based on image processing.

[0005] To solve the above technical problems, the present application adopts the following technical solutions:

[0006] A fabric one-way moisture transfer detection method based on image processing, comprising the following steps:

[0007] Step 1, adding liquid to the fabric, and obtaining the wetting images of the inner and outer layers of the fabric at each time point;

[0008] Step 2, the image enhancement, median filtering and improved binarization method are performed on the wetting images of the inner and outer layers of the fabric to obtain processed images;

[0009] Step 3, according to the processed images, the droplet wetting of the inner and outer layers of the fabric at each time is calculated to obtain the droplet wetting radius-time curve of the inner and outer layers of the fabric, and the method for calculating the droplet wetting radius is as follows: when the droplet penetration trace is a circle or an approximate circle with a radius difference of less than 20% in the warp and weft directions, the radius is taken as the wetting radius of the fabric corresponding to the radius of the circle with an equal wetting area; when the droplet penetration trace is an ellipse or a strip, the wetting radius is calculated according to the weight of the wetting radius in the warp and weft directions;

[0010] Step 4, the diffusion rate of water in the inner and outer layers of the fabric at each time is calculated using the droplet wetting radius-time curve of the inner and outer layers of the fabric, and then the diffusion rates obtained are summed to obtain the diffusion rate of water in the inner and outer layers of the fabric;

[0011] Step 5, the water absorption rate-time curve of the inner and outer layers of the fabric is obtained using the diffusion rate of water in the inner and outer layers of the fabric and the droplet wetting radius-time curve of the inner and outer layers of the fabric, and the average rate in the water absorption process of the inner and outer layers of the fabric is taken as the water absorption rate of the inner and outer layers of the fabric;

[0012] Step 6, the one-way wet transfer capacity RR of the fabric is calculated using the formula , wherein M i and N i respectively represent the water content of the outer layer and the inner layer of the fabric at the i-th time.

[0013] Step 7, the water absorption rate of the inner and outer layers of the fabric, the diffusion rate of water in the inner and outer layers of the fabric and the one-way wet transfer capacity of the fabric are input as input parameters of the training sample into the detection neural network, the detection neural network outputs the one-way wet level detection value of the training sample, the detection neural network is trained by minimizing the difference between the one-way wet level of the training sample measured by the MMT method and the one-way wet level detection value, and the best neural network model is obtained after training;

[0014] Step 8, the water absorption rate of the inner and outer layers of the fabric to be tested, the diffusion rate of water in the inner and outer layers of the fabric to be tested and the one-way wet transfer capacity of the fabric to be tested are calculated using the method of steps 1-6 and input into the final detection neural network to obtain the one-way wet level of the fabric to be tested.

[0015] Further, in the improved binarization method in step 2, the method for obtaining the optimal segmentation threshold is as follows:

[0016] Step 2.1, the probability of pixel point i in the image is:

[0017]

[0018] MxN represents the total number of pixels of the image, f i (x, y) represents the gray value corresponding to pixel i, and {0, L-1} is denoted as GL;

[0019] The image is divided into foreground (0, t) and background (t, L-1) by setting threshold t, the proportion of the foreground region is ω0(t), the proportion of the background region is ω1(t), the average gray value of the foreground part is μ0(t), the average gray value of the background part is μ1(t), and the overall average of the image is:

[0020] μ = ω0(t) x μ0(t) + ω1(t) x μ1(t)

[0021] Step 2.2, the improved variance formula is: During operation, t is sequentially assigned in the GL range, when t is a certain value A The value of the variance is the maximum, which indicates that the value of t at this time is the optimal segmentation threshold.

[0022] Further, in step 3, when the droplet penetration trace has a radius difference of less than 20% in the warp and weft directions, i.e., the shape is a circle or approximately a circle, the calculation formula of the droplet wetting radius is: S is the wetting area of the droplet;

[0023] When the droplet penetration trace is an ellipse or a strip, the calculation formula of the droplet wetting radius is:

[0024]

[0025] Where, r j and r w are the maximum wetting radii of the droplet in the warp and weft directions, respectively.

[0026] Further, in step 5, the calculation formula of the water diffusion rate v is:

[0027]

[0028] In the formula, t i At the i-th moment, ΔR represents the difference between the wetting radius of the droplet at the i-th moment and the i-1-th moment.

[0029] Further, in step 5, the method for calculating the water absorption rate of the inner / outer layer of the fabric at any time is to calculate the water injection rate of the inner / outer layer of the fabric at that time, and to regard the water injection rate as the water absorption rate. The calculation formula of the water injection rate at any time is: q θ = ρdr(1-ξ)v; In the formula, q θFor the water injection rate, p is the liquid density, d is the fabric thickness, r is the wetting radius, ξ is the fabric tightness, and v is the fabric diffusion rate at that time.

[0030] Compared with the prior art, the present application has the following advantages:

[0031] The present application uses the improved Otsu algorithm to remove the influence of the factor pixel ratio ω, directly divides the optimal threshold by maximizing the average gray variance, realizes the accurate segmentation of the image, obtains the wetting profile of the fabric at each time, and finally measures the actual wetting radius and area of the fabric through the quantitative calculation of the pixels; The present application introduces the fabric moisture transfer theory and fluid equation into the mutual calculation of the water absorption rate and the diffusion rate, thereby increasing the diversity of the performance indicators of the image method, improving the accuracy and scientificity of the detection results of this method, and at the same time, the test process is efficient, this method is applied to the fabric moisture transfer detection in the textile industry, can improve the work efficiency, and the test data is more accurate, which provides an efficient and practical scheme for the moisture transfer performance test of the one-way moisture transfer fabric.

[0032] The present application will be described in detail below in combination with the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The present application is a method overall flowchart;

[0034] Figure 2 The present application is a schematic diagram of the experimental table used;

[0035] Figure 3 The present application is a binary effect diagram of the improved Otsu algorithm;

[0036] Figure 4 The present application is a fitting diagram of the inner and outer layer moisture diffusion rate;

[0037] Figure 5 The present application is a fabric water absorption rate-time curve diagram;

[0038] Figure 6 The present application is a fitting diagram of the inner and outer layer moisture absorption rate;

[0039] Figure 7 The present application is a one-way moisture transfer ability data fitting diagram;

[0040] Figure 8 The present application is a detection neural network training flowchart. DETAILED DESCRIPTION

[0041] The principles and features of the present application will be described below in combination with the drawings, and the examples are only used to explain the present application, and are not used to limit the scope of the present application.

[0042] The method of the present application is verified by specific experiments, and the embodiment can be divided into two parts. The first part is image method detection: it is to obtain the diffusion images of the water droplets in the inner and outer layers of the fabric at different times, and after processing, the water droplet contour map is segmented by appropriate method (the final image subtraction method is adopted), the radius and area of the water droplet at each time are calculated, and the diffusion rate of water in the fabric is obtained. Then, by simplifying the theoretical model of fabric moisture transfer and limiting the experimental conditions, the fluid mechanics equation is introduced into the calculation of fabric diffusion rate and absorption rate. Based on this, the one-way moisture transfer capacity of the fabric is detected, and the correlation analysis between the image method detection results and the MMT detection results is made. Finally, the moisture transfer indexes measured by the image method are taken as the input function, and the moisture transfer grade of the fabric measured by the MMT is taken as the output function, which is brought into the rating training of the BP neural network, so as to realize the rating of the moisture transfer performance of the fabric, such as Figure 1 The overall design flowchart of the method of the embodiment is shown in the following figure.

[0043] I. Measurement of liquid wetting area

[0044] 1.1 Design of experimental platform and image acquisition

[0045] 1.1.1 Design of experimental platform

[0046] In the traditional part of the liquid experiment, the fabric moisture transfer performance is often measured by drawing concentric circles or squares on the surface of the fabric, placing it on the experimental platform, then dropping a suitable amount of liquid on the fabric with a rubber dropper, observing and recording the time of water wetting each circle or square. This method obtains a rough wetting radius and has a certain subjectivity. The most important thing is that the fabric and the experimental platform are attached, which makes the platform guide the water from the lower layer of the fabric to overflow, affecting the absorption and diffusion of water on the surface of the fabric. At the same time, the existence of copper ring in the liquid water management tester (MMT) also has a similar effect, which makes a certain amount of water remain on the copper ring after the detection, affecting the accuracy of the experimental results. To solve this defect, the experimental platform is redesigned as follows Figure 2As shown, the design, the object table adopts the design of openwork, using two 8*8 cm openwork area of iron sheet composition. Openwork can ensure that the lower layer of fabric after wetting liquid will not because of the role of flow and spill. At the same time, the use of two iron sheet to ensure that the fabric can be laid on the object table, and give the fabric a certain tension, also can avoid some fabric wet part because of the water droplet quality and appear droop and affect the fabric diffusion. Finally, and in the object table directly above and directly below each installation a high-definition camera to obtain the diffusion of moisture in the fabric at each time on the upper and lower layers; in addition, in order to control the moisture can be fully absorbed by the fabric, strictly control the liquid drop speed and weight. Ensure that every three seconds drop liquid, each drop in the fabric surface water droplet total weight in 1 g ± 0.05 g, and let all the liquid in 20 seconds evenly drop finally let the liquid in the fabric continue to wet 40 seconds to ensure the full absorption of moisture.

[0047] 1.1.2 image acquisition

[0048] In the acquisition of fabric image, in order to more truly and objectively reflect the one-way wetting performance of fabric, during the experiment, a high-definition camera is placed on both sides of the fabric to obtain the diffusion of water in the fabric. The obtained fabric image is transmitted to the computer in the form of digital signal for subsequent processing. Since the invention involves timed photographing of fabric and batch acquisition and processing of a large number of images, in general, the higher the resolution of the camera, the clearer the fabric image obtained by recording, and poor quality fabric images will seriously interfere with the subsequent recognition effect. At the same time, when the resolution of the camera exceeds a certain range, the fabric image pixels will be too large, increasing the computer's calculation load and increasing the performance requirements of the experimental equipment. For the invention, the resolution of the acquired fabric image should be above 600 dpi, which meets the requirements of most experiments. In addition, when using a high-definition camera to acquire fabric images, unnecessary interference factors should be reduced. For example, the experimental cloth should be as clean as possible without hair and defects, and the fabric should be placed in the center of the test table, and the fabric should be given appropriate tension by using a holding device to make the surface of the fabric smooth and flat during the experiment.

[0049] 1.2 fabric image preprocessing Due to the relationship between the equipment, different noises are often generated during the acquisition and extraction of fabric images, which reduces the image quality. To this end, the preprocessing technology of image is used to remove these irrelevant noises and information, retain or restore useful information, thereby enhancing the detectability of target information and simplifying complex data as much as possible, and increasing the reliability of feature signal extraction, image threshold segmentation, etc. For example, for images with low definition, the definition of the image can be improved by changing the gray scale distribution of the image; for the influence of light source on the image, the uneven illumination of the image can be improved by homomorphic filtering.

[0050] 1.2.1 Extraction of water droplet contour

[0051] Due to the existence of yarn, the surface of fabric presents unevenness, and the gray value of fabric itself at each position of the image has a large difference. At the same time, with the continuous diffusion of moisture in the fabric, the water content in the edge area of the wet area is significantly lower than that in the center area, and this part of moisture has a weak effect on the gray value of the image. Therefore, it is difficult to accurately obtain the water droplet contour using traditional segmentation algorithms, which can easily make the binary image overall "thin" to the original image.

[0052] However, after the droplet, the fabric will increase the transmittance of the wet part due to the effect of water, and the gray value will decrease, and the closer to the center area, the more obvious the decrease in gray value. The subtraction operation of the image can well highlight this feature. After the original fabric image and the image at each time after wetting are grayed, the original fabric image and the wet image group are subtracted, which can effectively highlight the change of gray value before and after the fabric absorbs water, eliminate the influence of the difference in gray value of the fabric itself, and realize the extraction of the water droplet contour. Since the gray value of the fabric after wetting decreases after absorbing water, the gray value of the subtracted image is positive, and the image is bright. The gray value of the unwetted part is unchanged, and the gray value of the subtracted image is 0, and the image is black. Some colored fabric after wetting is similar to the color of the fabric itself, and the difference after subtraction is small, and the image will still be dark, or for some images with unclear water droplet contour, similar situations will also occur. For this, the subtracted image can be multiplied by a constant greater than 1. In this way, the wet part of the image can be overall brightened, increasing the difference between the fabric before and after wetting, which is conducive to the binary processing of the image in the later stage.

[0053] 1.2.2 Enhancement of image

[0054] After the subtraction operation, the gray value of the fabric image will be reduced, so the image will be obviously darkened, and the gray difference between pixels will also be reduced, and the gray value is between 90 and 120. When the gray value is greater than 120 or less than 90, the pixel distribution is obvious. In this case, the gray value of the image is too narrow in space, the brightness is low, and the pixel difference is small, so some subtle losses may occur during image segmentation. Therefore, the intensity of the image and the difference between the pixels must be increased, and the gray value of the image must be adjusted. The gray histogram can accurately show the gray distribution of the image, and can well show the proportion of each gray point in the image and the number of each gray level pixel in the image. The contrast enhancement of the fabric image is usually achieved by using the distribution of the gray value in the whole image. Histogram equalization is usually used. Histogram equalization means that the pixels with relatively more gray levels in the image are expanded as much as possible, and the pixels with relatively less gray levels are reduced, so that the entire gray region is averaged, the gray change range of the fabric is improved, and the contrast is improved. In this embodiment, on the basis of using equalization filtering, a target histogram is established by using specification, and the initial histogram is related, that is, the initial image is changed into the required histogram by using the gray mapping function. Let r and z be the gray levels of the original fabric image and the fabric image after specification respectively, and the specific steps of specification are as follows:

[0055] (1) The original fabric image is processed by using histogram equalization method to obtain the conversion function s=T(r), and s is the gray level of the equalized initial image.

[0056] (2) The step 1 is continuously equalized to obtain a new conversion function v=G(r); v is the new equalized gray level.

[0057] (3) Because the equalization operation is applied to the same fabric image, the results should be the same, so s=v, and the composite function z=G-1(s)=G-1(T(r)) is obtained.

[0058] (4) The mutual relationship between the obtained composite functions can be used to specify the original fabric image.

[0059] After the modification of the histogram, the gray distribution of the image changes from the original relative concentration to the uniform distribution, so that the brightness of the image is obviously improved, and the difference between the image gray values is also obviously increased, which can effectively avoid the loss of important details in the subsequent image segmentation process.

[0060] 1.2.3 Image median filter processing

[0061] During the diffusion process of water droplets on fabric, the water absorption at the edge gradually decreases, and the effect on the fabric gray value also decreases. The edge of the water droplet is relatively blurred. At the same time, the water droplet is easily affected by pulse noise during the transmission of the fabric image, resulting in a large number of white small noise points, which affects the extraction of the contour. Median filtering can better improve the above situation. Median filtering is a typical nonlinear filtering, and its core idea is to arrange the gray values of the adjacent neighborhood in a single channel according to size, and select the median value of all gray values in the neighborhood to replace the original pixel brightness value. It is the best way to eliminate salt and pepper noise today.

[0062] The principle of the median filtering algorithm is to select an m*n filtering window, then move the two-dimensional filtering window in the gray image according to a specific order, then replace the original pixel gray value with the median value of all pixel gray values in the moving window, and finally output the filtering result. The definition is as follows:

[0063] Let A represent a two-dimensional filtering sliding window, X(i,j) represent the gray value of a certain point in the filtering window, where (i,j) belongs to A, Y(i,j) represents the output value after two-dimensional median filtering, then:

[0064]

[0065] 1.2.4 Binaryzation of image

[0066] Binaryzation of image is to convert the gray value of all points on the gray image to 0 or 255, that is, to make the image only show black and white effects in color. That is, the binary image obtained by selecting an appropriate threshold from the original image with 256 gray levels can still reflect the overall and local features of the image. In image processing technology, binary processing plays an extremely important role. Especially for processing operations that need to obtain specific parts of the image, the processing system based on binaryzation technology is widely used. To analyze a binary image, first, the original image gray value is binaryzation, and the binary image is obtained. Such a change only needs to analyze the image pixel position coordinates and the positions corresponding to the gray values 0 and 255, which simplifies the original processing process.

[0067] Otsu algorithm is a common binarization method, which is proposed by Kittler and Otsu, and the maximum inter-class variance method (Otsu) is further derived based on two-dimensional inter-class variance method and least square method. The core idea is that: the image is divided into two parts, i.e. the background part and the target part, by setting a threshold value, and the optimal selection method of the threshold value is to calculate the variance value of the gray value of the two parts, i.e. the threshold value corresponding to the maximum variance is the best threshold value, and the principle formula is as follows:

[0068] Suppose f(x, y) is the gray value of a certain point of the image MxN, and the gray level is L, then f(x, y) ∈ {0, L-1,} is recorded as GL.

[0069] The probability of gray value i in the image is:

[0070]

[0071] MxN represents the total number of pixel points of the image, f i (x, y) represents the gray value corresponding to pixel point i, the threshold value t is set to divide the image into foreground (0, t) and background (t, L-1) two parts, then the foreground area ratio is ω0(t), the background area ratio is ω1(t), the average gray value of the foreground part is μ0(t), and the average gray value of the background part is μ1(t), and the overall average is:

[0072] μ=ω0(t)×μ0(t)+ω1(t)×μ1(t) Equation (3)

[0073] The variance between them is the best threshold value, which is expressed as:

[0074]

[0075] During the running process, t is assigned in GL range in turn, when t is a value, the value of is the maximum, which indicates that the t value at this time is the best segmentation threshold value. According to the analysis of the above Otsu algorithm, the average gray value of the target and the background is required to be farthest from the average gray value, and for the defects on the fabric, the target has obvious deviation relative to the background, which leads to that the above Otsu algorithm is difficult to divide the target and the background, and the ideal threshold value is obtained.

[0076] Therefore, firstly, the influence of the factor pixel ratio ω is removed, as shown in formula 5, and the best threshold value is directly divided by maximizing the average gray variance. Compared with the original algorithm, the threshold value of this algorithm is smaller than that of the Otsu algorithm, so that the processed target pixel is more prominent; in addition, based on formula 5, as shown in formula 6, a new factor gray probability P is introduced to reduce the influence of the gray probability on the average gray value, so that the best threshold value is more deviated.Figure 3 The effect pictures after fabric binaryzation segmentation by three methods are shown, and from left to right in the picture are the original fabric picture, the effect pictures of Otsu algorithm, improved method 1 and improved method 2. As can be seen from the picture, after improvement, the segmentation effect of the image has been obviously improved, and various impurity noise points in the image have been better removed, but the comparison between (b) and (c) shows that the effect of improved method 1 is not as good as that of improved method 2.

[0077]

[0078]

[0079] According to the analysis of the above Otsu algorithm, the algorithm requires the average gray values of the target and the background to be farthest from the average gray value, and for the defects on the fabric, the target has obvious deviation relative to the background, which leads to the fact that it is difficult to divide the target and the background by the above Otsu algorithm to obtain an ideal threshold value.

[0080] 1.2.5 Pixel ratio method for wetting area

[0081] After the image subtraction operation, filtering and binaryzation operation in the early stage, the water droplet contour can be approximately obtained. In the final processed picture, only the white part representing the wetting area and the black part representing the non-wetting area exist. The number of white pixels and the total number of image pixels are counted respectively, and the ratio of the number of white pixels to the total number of image pixels multiplied by the actual area of the fabric can obtain the actual area of the contour.

[0082] After the morphological operation of the image, the improved Otsu algorithm for segmented image and other processing methods, the complete image of the fabric in the wetting process is obtained, and then the pixel size of the fabric is converted into the actual area by using the quantitative calculation of pixels, and the area measured by the image method is compared with the area measured by the weighing method. The results show that the improved Otsu algorithm improves the segmentation effect of the original algorithm on the image. At the same time, the wetting area obtained by the image method is close to the actual result, but the measurement time is much less than that of the weighing method.

[0083] II. Calculation of the water diffusion rate of the fabric

[0084] 2.1.1 Determination of water droplet radius

[0085] The traditional image method is limited to making area-time curve after binarization and conversion of fabric wetting area. The curve can directly reflect the dynamic diffusion process of water in the fabric. The actual wetting of water is not a uniform diffusion process to all directions, and sometimes there is obvious directional difference. In the index evaluation of the liquid water management tester, the index is the accumulation of the diffusion rate before the wetting radius reaches the maximum. In order to make the index obtained by the present application scientific, the same method is used to calculate the performance index, which is convenient for the correlation study of the two methods in the later stage.

[0086] The diffusion of water in the fabric can be divided into two stages. The first stage is the wicking stage from one side of the fabric to the other side. For some pure cotton and viscose fabrics with good water absorption performance, the time is usually short, and it is difficult to capture by manual or image processing method, so this process is usually not studied. The second stage is the capillary transport stage. After wicking, water has diffused into the fabric, and there is no water on the surface. The fabric is a geometric body composed of many fibers, and there are capillaries between the pores and gaps of the fibers. Because there are hydrophilic groups in the molecules of the fibers, the water molecules in the pores and gaps are subjected to force, so that the water molecules permeate and diffuse along the capillaries. According to Poiseuille's law, it is found that the flow rate of the fluid conforms to formula 7,

[0087]

[0088] When the control water injection speed is the same, the 2.5th power of the fluid diffusion radius is positively correlated with the porosity of the fabric, and the greater the porosity of the fabric, the greater the gap between the fabrics, and the smaller the density of the fabric. Therefore, due to the difference in density of the fabric in the warp and weft / horizontal and vertical directions, and the greater the difference in density of the fabric in the warp and weft / horizontal and vertical directions, the greater the difference in water flow in each direction. Through the drop experiment of different density fabrics, it is found that the penetration trace of water droplet can be roughly divided into four types, which are circular, strip, ellipse and special flower pattern.

[0089] In view of the great difference between the diffusion radius of the liquid droplet in the warp and weft directions in the above four cases, different methods are adopted to calculate the radius under different conditions. When the radius difference of the liquid droplet penetration trace in the warp and weft directions is less than 20%, that is, the shape is circular or approximately circular, the radius is taken as the radius corresponding to the wetting area of the fabric, that is, as shown in formula 8; when the liquid droplet penetration trace is elliptical or strip-shaped, the wetting radius is calculated according to the weight of the wetting radius in the warp and weft directions, that is, as shown in formula 9: in the formula, r j , r w are the maximum wetting radii of the liquid droplet in the warp and weft directions, respectively.

[0090]

[0091]

[0092] 2.1.2 Calculation of the water diffusion rate of fabric

[0093] After obtaining the fabric radius-time wetting curve as shown in the above figure, the design refers to the calculation method of the diffusion rate in MMT, and the diffusion rate of water at each time is calculated first, then the obtained rate is accumulated and summed, and the rate is taken as the final reference evaluation index of the diffusion of water in the fabric, and the calculation formula is shown as formula 10. In order to verify the accuracy of the results of this method, the design will analyze the error between the experimental results and the results obtained by MMT while detecting by image method. The comparison of part of the results measured by the two methods is shown in the table. As can be seen from the table, the results obtained by the two methods are close, and most of the errors are within 10%.

[0094]

[0095] In the formula, t i , S i respectively represent the moisture wetting radius corresponding to the i-th time, i=1,2,3...

[0096] 2.1.3 Correlation analysis of diffusion rate

[0097] In the measurement of moisture diffusion rate by MMT, when the liquid slowly drops on the fabric, the fabric starts to permeate to the surrounding due to the moisture absorption effect. In this process, each time a group of copper rings is contacted, it will cause a resistance change, and the system will record the time of the resistance change at the same time. By recording the change time for many times, the diffusion rate at each time can be measured. In the image method, the diffusion radius at different times is measured. Therefore, the former speed is the time spent for measuring a certain length, and the latter belongs to the distance passed in a certain time. The measurement methods of the two are different, which is the biggest reason for the error of the measurement results. In order to find the quantitative relationship between the two, as Figure 4 shown, data fitting analysis is performed on the results obtained by the two methods. As can be seen from the fitting graph, there is a strong linear relationship between the two, and the correlation coefficient is above 0.98.

[0098] 2.1.4 Calculation of the water absorption rate of fabric

[0099] The water absorption of a fabric is the rate of change of water content in the fabric over a certain period of time. It is the main measure of the fabric's ability to absorb water. The penetration of a liquid into a fabric is essentially the diffusion of a liquid phase material, i.e. the dispersion of the material. A fabric is a porous medium composed of yarns and the spaces between them, and the transmission of a liquid in this porous medium is very complex, involving the transmission of molecules and heat, and the influence of field forces and volume forces. In practice, these factors often exist simultaneously. The weight of the influence of these factors on the transmission of water varies, and if all the factors are taken into account, it will be very difficult to study the theory. Through analysis of the penetration characteristics and dynamic characteristics of the fibers, it is found that the fibers are anisotropic, and their moisture transmission characteristics are different in different directions. In addition, due to the type and concentration of the fluid, the properties of the fibers, the structure of the yarns, the structure of the fabric, the external air conditions during the test, etc., it is not practical to include all these factors in a formula. Therefore, the above factors should be analyzed simply, and only the fluid flow velocity, the filling rate and the thickness of the fabric on the surface of the fabric are analyzed, and the performance of the fabric and the air conditions are limited:

[0100] First, the selected fabric should be a uniform medium; second, the liquid should be fully absorbed by the fabric and not spill over. Third, the relative humidity of the atmosphere is 100%, and there is no liquid vaporization.

[0101] Under the above assumptions, if the water evaporation is not calculated, the liquid dropped from the center of the fabric will spread from the center of the fibers to the periphery. Assuming that the rate of drop is very low, the liquid drop on the fabric will flow along the cross section of the fibers without leaking from the other side of the fabric. On this basis, the fluid mechanics equation can be solved.

[0102] When water is injected, the liquid is continuously and uniformly injected into the fabric along point o, moves in the capillary channel of the fiber, and tests the fluid inside the angle AOB of the fabric. The diffusion displacement OA of the liquid is L, and the thickness of the fabric is d. In the case where the propagation radius of the liquid in all directions along the fiber is the same, the liquid flow in the arc surface AOB;

[0103] According to the principle of fluid mechanics, the inflow of liquid on the pipe section is equal to the outflow of liquid. If AB arc section is the liquid diffusion front position, the thickness of the fabric is h, and the filling coefficient of the fabric section is ξ, then the inflow of liquid on the section is equal to the outflow of liquid, i.e.

[0104] q θ = p - d - r (1 - ξ) v (equation 11)

[0105] In the formula, q θwhere V is the water injection rate, p is the liquid density, d is the fabric thickness, r is the wetting radius, ξ is the fabric tightness, and v is the fabric diffusion rate.

[0106] The theory explores the theoretical equation of water injection rate and liquid diffusion rate when the droplet is injected by simplifying the model. When the water injection rate is small enough and the fabric water absorption performance is high enough, the water is quickly absorbed by the fabric, and the water injection rate is equal to the fabric water absorption rate. The equation can be considered as the theoretical equation of fabric water absorption rate and diffusion rate, that is, the water absorption speed of the droplet can be calculated by the diffusion speed. With the increase of water injection rate, the liquid on the surface of the fabric will accumulate and wait to be absorbed due to the limitation of the fabric moisture absorption performance. However, with the increase of time, the remaining liquid will be completely absorbed. At this time, the water absorption rate will be less than the water injection rate. For this case, in order to measure the fabric water absorption rate, the process can be differentiated, and the entire moisture absorption process can be divided into several small processes. For each small process, part of the liquid will be quickly and completely absorbed by the fabric, that is, the liquid drop speed (water injection rate) and the fabric water absorption rate are equal in each small process. In the previous chapter, the image method can accurately extract the contour of the liquid droplet at each time during the wetting process and its diffusion rate. Therefore, the theoretical equation is used in each process of wetting, that is, the water absorption rate is calculated by the diffusion rate of the fabric moisture at each time. Thus, the disadvantage of the traditional image method that can only measure the diffusion rate can be overcome.

[0107] The drawing of the water absorption rate curve is as follows. In the previous steps, the image preprocessing and segmentation can obtain the wetting contour of the fabric at each time during the entire wetting process. The wetting radius at each time is measured by pixel quantization, and the corresponding diffusion rate is calculated. Then, the fabric thickness is measured by density division and fabric thickness instrument, and the fabric number is measured by the splitting method. Thus, all the parameters required for fabric water absorption rate in the theoretical equation are obtained, and the water absorption rate at each time can be calculated by the water injection rate calculation formula. The water absorption rate curve of each fabric at each time can be drawn as shown in FIG. 1. Figure 5

[0108] The water absorption rate curve of the fabric will be very different due to the influence of the fabric raw material composition, density and other factors. For example, the water absorption rate curves of four fabrics are shown in FIG. 1. Figure 5 ​There is a certain regularity in (a) and (b). When the droplet is dripped, the fabric starts to absorb water, and as the water continues to be dripped, the water absorption rate of the fabric gradually increases and gradually reaches a maximum value; but as time increases, the water content in the fabric gradually increases, and the water capture ability of the fibers in the fabric starts to decrease, and the water absorption rate starts to decrease. The rapid increase and decrease of the rate is completed in about the first 20s. After that, the outside does not drip liquid into the fabric, and the fabric starts to absorb the surface accumulated water, and the rate starts to slowly decrease until it is completely absorbed, and the rate will be 0; and for Figure 5 (c) and (d), there is no regularity as Figure 5 (a) and (b), the water absorption rate shows a large fluctuation during water injection.

[0109] 2.1.5 Final evaluation of fabric water absorption rate index

[0110] To detect the water absorption rate of the fabric, first of all, the water content of the fabric must be measured. There are two methods for detecting the water content of the fabric: direct method and indirect method. The direct method is to directly weigh the mass of the fabric in dry and wet states and calculate it into the formula, but this method is to detect the final wet state of the fabric rather than the real-time state; the indirect method is to indirectly measure the water content by using some external radiation such as resistance, dielectric coefficient, etc. For example, in the microwave attenuation method, the attenuation of microwave energy is mainly caused by water in the fabric, and there is a quantitative relationship, so by detecting the change of microwave energy in the experiment, the water content of the fabric can be realized. The resistance method detects the water content of the fabric according to the exponential function relationship between the water content of the fabric and the conductivity, which is the most commonly used online humidity detection method at home and abroad. For all electrical properties, the relationship between the conductivity of the fabric and its water content is the closest, and a few percentage difference in water content can cause more than ten times change in the conductivity of the fabric.

[0111] Although there are many methods for measuring the water content of the fabric, most of them are indirect methods, and when defining the water absorption rate of the fabric, it refers to the change rate before and after the fabric absorbs water, which is usually expressed in percentage. When the slope k = tan 15, the corresponding time is 24s, and the area around the curve at each time before 24s is calculated by integral method, that is, the water content-time curve of the fabric can be drawn. According to the definition, the rate curve under the definition can be obtained. Finally, the average rate in the water absorption process is the same as MMT, which is used as the final evaluation index of the water absorption rate.

[0112] After obtaining the water absorption rate of each fabric by the above method, in order to explore the quantitative relationship between the detection results of the image method and the MMT method, the present application uses multiple functions, trigonometric functions and logarithms to fit and analyze the results of the two methods, and finally finds that when the image method result is taken logarithm and the MMT method result is taken 0.5 power, then the fitting effect is best after using trigonometric function analysis, which has a function relationship as shown in formula 12 and formula 13. Figure 6 As shown in formula 12 and formula 13, the fitting effect after conversion by formula is shown in the figure; it can be seen from the fitting graph that after fitting, the data results obtained by the image method are close to the results measured by the MMT method, and the general error is within 5%, so it can be considered that the two methods have the function relationship shown in the formula:

[0113]

[0114]

[0115] III. Calculation of fabric one-way wet transfer capacity

[0116] The fabric one-way wet transfer capacity measures the difference between the cumulative water content of the inner and outer layers of the fabric, but in the actual MMT detection results, it is found that the index is in %. Finally, after consulting the instrument instruction manual and consulting the instrument manager, it is known that it is the change of copper ring resistance caused by water, and through the relationship between resistance and water content, the water content-time curve of the inner and outer layers of the fabric is drawn, the water content change percentage of the inner and outer layers of the fabric at each time is calculated and accumulated, and finally the difference between the two accumulated values is taken as the evaluation standard of the index. In the calculation of the water absorption rate above, the fabric water content curve has been drawn as an intermediate process, and the present application calculates the fabric one-way wet transfer capacity by referring to the calculation method of MMT (i.e. as shown in formula 17). At the same time, in order to explore the correlation between the measurement results of the two methods, the fitting analysis of the two results is also carried out. The final fitting relationship is shown in formula 15. In the formula, g(t) is the modified function of the relationship between the two, which is mainly because the total measurement time of MMT is 120s, which is longer than the measurement time of the present application, and during the extra time, due to the evaporation of water, there will still be a certain water content change rate of the fabric, which will cause the difference between the fitting results of the two. As shown in formula 15, Figure 7 After taking the logarithm and 0.5 power transformation of the data of this index measured by the MMT method and the image method respectively, substituting into formula 15, the fitting graph of the two is shown in the figure, and it can be seen from the figure that after the transformation, the two have a good quadratic function relationship.

[0117]

[0118] In the formula, M and N respectively represent the water content of the outer and inner layers of the fabric at the i-th moment, and RR is the fabric one-way wet transfer capacity

[0119] log(y)=1.03x2 -2.42x + 4.28 + g(t)

[0120] g(t) = (120 - t) tan(3) Equation (15)

[0121] In the formula, x, y are the one-way moisture transfer capacity of the image method and the MMT method, respectively, and g(t) is the correction function.

[0122] The results show that the moisture diffusion rate measured by the image method has a strong linear relationship with the value measured by the MMT; the logarithm of the water absorption rate and the one-way moisture transfer capacity data measured by the image method has a good function relationship with the 0.5 power of the corresponding results measured by the MMT, the former is approximately a trigonometric function, and the latter has a strong quadratic relationship, but there is an additional function related to time t to correct the linear relationship.

[0123] Four, BP neural network and fabric moisture transfer capacity rating training

[0124] 4.1 Basic principles of neural network

[0125] As shown in Figure 8 , the BP neural network is composed of an input layer, an intermediate layer and an output layer. It continuously trains the entire network model through pre-prepared training samples and training functions, continuously adjusts the weight and threshold matrix of the network to obtain the optimal performance neural network model parameters, and finally approximates the corresponding relationship between the input and output of the training samples. The input samples use the pre-set initial weight and threshold to gradually calculate the corresponding weight, threshold and output matrix of each layer of neurons, and the training function calculates the difference between the actual output and the theoretical output of each layer and adjusts the weight and threshold of each layer of neurons according to the size of the difference, finally making the error tend to be minimal. The structure and principle of BP neural network are shown in the following figure: it is a typical three-layer BP network structure diagram. The structure of this network model is simple, easy to implement, and has good scalability. The training process of the standard BP neural network mainly includes two stages of forward propagation of input information and backward propagation of error signal. When the learning sample is input into the neural network, it begins to propagate in the direction of input layer, hidden layer and output layer, and finally obtains the actual output according to the weight and bias vector of each layer.

[0126] 4.2 BP neural network training results

[0127] After the parameter setting of the BP neural network is completed, the training phase of the neural network can be performed. In the present application, by introducing the moisture transfer theory of the fabric and referring to the calculation method of the performance index of the liquid flow state management tester (MMT), the moisture diffusion rate of the inner and outer layers of the fabric, the moisture absorption rate of the inner and outer layers of the fabric and the moisture transfer index of the fabric are measured. The five performance indexes are also the main factors affecting the comprehensive moisture transfer index of the fabric. The five characteristic value parameters are input into the training model of the BP neural network as the input function to start training the network model. Through the nonlinear change of the hidden layer, the actual value is output through the output layer, and then the weights between the layers are continuously corrected through the reverse correction, and the output result of the output layer is the fabric moisture transfer performance grade. The fabric moisture transfer performance grade measured by the MMT is put into the BP network as the input function to train and evaluate the grade, and the results show that the training results of the network for the moisture transfer grades of 2-4 are good, and all can be stable at an accuracy rate of more than 90%, and the highest accuracy rate reaches 95%.

[0128] The above is an example of the best embodiment of the present application, wherein the parts not described in detail are the common knowledge of ordinary skilled in the art. The protection scope of the present application is subject to the content of the claims, and any equivalent transformation based on the technical inspiration of the present application is also within the protection scope of the present application.

Claims

1. A fabric unidirectional moisture conducting detection method based on image processing, characterized in that, The method comprises the following steps: Step 1, liquid is added dropwise to the fabric, and the wetting images of the inner and outer layers of the fabric at different times are obtained; Step 2, the wetting images of the inner and outer layers of the fabric are processed by image enhancement, median filtering and improved binarization method to obtain processed images; Step 3, according to the processed images, the wetting radii of the inner and outer layers of the fabric at different times are calculated to obtain the wetting radius-time curve of the inner and outer layers of the fabric, and the calculation method of the wetting radius of the droplet is as follows: when the droplet penetration trace is a circle or an approximate circle with a radius difference of less than 20% in the warp and weft directions, the radius is taken as the wetting radius of the fabric corresponding to the radius of the circle with the same wetting area; when the droplet penetration trace is an ellipse or a strip, the wetting radius is calculated according to the weight of the wetting radius in the warp and weft directions; Step 4, the diffusion rate of the water in the inner and outer layers of the fabric at different times is calculated by using the wetting radius-time curve of the inner and outer layers of the fabric, and then the diffusion rates of the water in the inner and outer layers of the fabric are summed to obtain the diffusion rate of the water in the inner and outer layers of the fabric; Step 5, the water absorption rate-time curve of the inner and outer layers of the fabric is obtained by using the diffusion rate of the water in the inner and outer layers of the fabric and the wetting radius-time curve of the inner and outer layers of the fabric, and the average rate in the water absorption process of the inner and outer layers of the fabric is taken as the water absorption rate of the inner and outer layers of the fabric; Step 6, using the formula to calculate the one-way wicking transfer capability RR of the fabric, where, and Wi and W2i represent the water content of the outer and inner layers of the fabric at the i-th moment, respectively. Step 7, the water absorption rate of the inner and outer layers of the fabric, the diffusion rate of the water in the inner and outer layers of the fabric and the one-way moisture transfer capacity of the fabric are taken as the input parameters of the training sample to input the detection neural network, the one-way moisture level detection value of the training sample is output by the detection neural network, the detection neural network is trained by minimizing the difference between the one-way moisture level of the training sample measured by the MMT method and the one-way moisture level detection value, and the best neural network model is obtained after training; Step 8, the water absorption rate of the inner and outer layers of the fabric, the diffusion rate of the water in the inner and outer layers of the fabric and the one-way moisture transfer capacity of the fabric are calculated by the method of steps 1-6, and the one-way moisture level of the fabric is obtained by inputting the best neural network model.

2. The image processing based fabric unidirectional wetness detection method according to claim 1, wherein, In the improved binarization method of step 2, the method for obtaining the optimal segmentation threshold is as follows: Step 2.1, the probability of pixel point i in the image is: ; M x N represents the total number of pixels of the image, represents the gray value corresponding to the pixel i, and {0, L-1} is denoted as GL; Setting threshold t divides the image into foreground (0, t) and background (t, L-1) two parts, the foreground area ratio is , the background area ratio is , the average gray of foreground part is , the average gray of background part is , and the overall average of the image is: ; Step 2.

2. Improved Variance Formula is: ; During the running process, t is assigned values in GL range in turn. When t is a value that makes the variance of the value maximum, it indicates that the t value at this time is the optimal segmentation threshold value.

3. The image processing based fabric unidirectional wetness detection method according to claim 1, wherein, In step 3, when the droplet penetration trace is a circle or an approximate circle with a radius difference of less than 20% in the warp and weft directions, the calculation formula of the droplet wetting radius is as follows: , S is the droplet wetting area. When the droplet penetration trace is an ellipse or a strip, the calculation formula of the wetting radius of the droplet is: ; where r j and r w are the maximum wetting radii of the droplet in the warp and weft directions, respectively.

4. The image processing based fabric unidirectional wetness detection method according to claim 1, wherein, In step 5, the calculation formula of the diffusion rate of the water v is: ; In the formula, t i At the i-th moment, represents the difference between the wet radii of the droplets at the i-th moment and the i-1-th moment.

5. The image processing based fabric unidirectional wetness detection method according to claim 1, wherein, In step 5, the method for calculating the water absorption rate of the inner / outer layer of the fabric at any time is to calculate the water injection rate of the inner / outer layer of the fabric at the time, and the water injection rate is regarded as the water absorption rate. The calculation formula of the water injection rate at any time is: ; wherein, is the water injection rate, p is the liquid density, d is the fabric thickness, r is the wetting radius, ξ is the fabric tightness, and v is the moisture diffusion rate at the time.

Images