CN109377489B - Analysis method and analysis system for weave structure of shuttle fabric - Google Patents
Analysis method and analysis system for weave structure of shuttle fabric Download PDFInfo
- Publication number
- CN109377489B CN109377489B CN201811277402.0A CN201811277402A CN109377489B CN 109377489 B CN109377489 B CN 109377489B CN 201811277402 A CN201811277402 A CN 201811277402A CN 109377489 B CN109377489 B CN 109377489B
- Authority
- CN
- China
- Prior art keywords
- fabric
- image
- single tissue
- tissue
- warp
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/44—Analysis of texture based on statistical description of texture using image operators, e.g. filters, edge density metrics or local histograms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30124—Fabrics; Textile; Paper
Abstract
The invention relates to an analysis method and an analysis system for the weave structure of a woven fabric, and belongs to the technical field of image recognition. The analysis method comprises the following steps: receiving a fabric image; carrying out coarse adjustment on the longitude and latitude directions; identifying the distribution of different tissues on the fabric so as to obtain the distribution area of a single tissue in an image; cutting at least a single tissue segment slice from each single tissue distribution region; acquiring a circulation mode and a circulation number of a tissue structure in a single tissue pattern block slice; calculating the integral warp and weft density of the fabric to be analyzed according to the warp and weft densities of all the tissue structures on the fabric to be analyzed; and outputting the result. The fabric image is split into the tissue slices with single tissues, then the slices are subjected to artificial intelligent image recognition, the fabric tissue structure can be accurately recognized, the warp and weft density numerical value of the fabric is obtained, the subsequent sample copying and copying of the fabric product can be directly finished through a computer, and the method can be widely applied to the field of textile manufacturing.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a system for analyzing a weave structure of a woven fabric.
Background
The woven fabric is a fabric formed by vertically interweaving yarns in the warp and weft directions, the longitudinal yarns are called warp yarns, the transverse yarns are called weft yarns, and the warp yarns and the weft yarns are interwoven according to a certain organization rule. The weave structure of the woven fabric is the spatial position relationship of warp yarns and weft yarns, and refers to the law that the warp yarns and the weft yarns are staggered or float and sink in each other in the fabric. At the intersection of the warp and weft yarns, known as the stitch point (float); when the warp yarn floats on the weft yarn, the warp yarn is called a warp structure point (warp floating point); whenever a weft yarn is on a warp yarn, it is called a weft pattern point (weft float). When the warp and weft points float and sink regularly to reach the circulation, it is called a tissue circulation (or a complete tissue).
In the textile industry, the sample turning treatment is often required to be carried out on the fabric, the sample turning method of the existing fabric is usually completed through sample making, the sample directly prepared through sample making is easy to cause material waste, unnecessary cost is generated due to the occupied manpower, material resources and time required by sample making each time, the result is not necessarily satisfactory, and the sample turning cost of the fabric is greatly increased. In addition, the whole sample turning process is taken as a complicated working process, and operators are required to have certain fabric professional technical knowledge and experience to be competent.
Along with the development of computer technology and image processing technology, a plurality of software capable of simulating the appearance of the fabric are developed, and the appearance of the fabric is designed according to a reference sample model, so that the sample making is not needed when the sample making is carried out on the fabric, and the sample making cost is greatly reduced. However, the existing simulation software for the appearance of the fabric takes the yarns as basic units, analyzes each yarn in the reference sample, obtains the distribution of the tissue points of the reference sample, still has large workload, and also needs professional fabric professional technical knowledge.
In order to solve the above-mentioned technical problems, patent document CN104195713A discloses a method for generating a yarn distribution diagram of woven fabric based on a weave, in which a weave of a reference sample is directly used as a basic unit, a covering area of each type of yarn in each weave unit is analyzed, a corresponding area in a design sample is filled with the covering area of each type of yarn in each weave unit, and finally, a color is assigned to each type of yarn, that is, a yarn distribution diagram of the design sample is generated. The invention greatly improves the generation efficiency of the yarn distribution diagram of the design sample and can further improve the sample turning efficiency of the fabric. And the special technical knowledge of the fabric is basically not needed, the technical threshold of the fabric sample turning and production work is reduced, and the smoothness and the convenience of operation are improved.
However, in the practical application process of the method, the type of the fabric weave structure of the reference sample needs to be manually judged according to experience, then the fabric weave is cut from the fabric weave structure as a basic fabric pattern block according to a manual identification method, and then the pattern block of each fabric weave structure is identified and processed, which needs to be identified manually based on the preamble, and the efficiency is still low.
Disclosure of Invention
The invention aims to provide an analysis method of a shuttle fabric weave structure, which is used for reducing the manual workload in the weave structure reappearance in the textile sample-receiving and sample-turning process;
another object of the present invention is to provide a system for analyzing the weave structure of a woven fabric, which reduces the amount of manual work in the reconstruction of the weave structure during the sample application and the sample duplication of the textile.
In order to achieve the above object, the present invention provides a method for analyzing a weave structure of a woven fabric, comprising the steps of:
a receiving step, namely receiving an image of the fabric to be analyzed;
a coarse adjustment step, namely adjusting the image to the longitude and latitude direction of the fabric to be analyzed to be arranged along the transverse direction and the longitudinal direction of the screen by taking the longitude and latitude direction appointed by the pointing operation as the current longitude and latitude direction of the fabric to be analyzed according to the received pointing operation aiming at the longitude and latitude direction on the image;
a dividing step, namely extracting gray scale features and texture features from the image after coarse adjustment, dividing different tissues by using an unsupervised clustering algorithm, judging the number of the tissues by using a contour coefficient, and identifying the distribution of the different tissues on the fabric to obtain a distribution area of a single tissue in the image;
a cutting step, cutting at least one single tissue pattern block slice from each single tissue distribution area according to the distribution condition of the single tissue;
fine adjustment, namely performing primary correction on the longitude and latitude directions of the single tissue image block slices based on image pattern recognition and regression algorithm so as to arrange the tissue image blocks along the transverse direction and the longitudinal direction of a screen;
the identification step, based on color features and texture features, using a random forest identification model generated by pre-training to perform image mode identification, frequency domain analysis and global optimization on the finely adjusted single tissue pattern block, and acquiring a circulation mode and a circulation number of an organization structure in a single tissue pattern block slice;
calculating, namely calculating the thread count and the longitude and latitude density of the yarn of the current texture structure based on the acquired cycle number, the physical size of a single cycle and the cycle mode, and calculating the integral thread count and longitude density of the fabric to be analyzed according to the thread count and longitude densities of all texture structures on the fabric to be analyzed;
and an output step, outputting the yarn distribution diagram of the fabric to be analyzed and the calculated warp and weft density.
According to the invention, the fabric image is divided into tissue slices with single tissues, and then artificial intelligent image recognition is carried out on the slices, so that the fabric tissue structure can be accurately recognized, and the warp and weft density numerical value of the fabric is obtained, thus the fabric tissue structure can be directly expressed in a digital form, the subsequent sample copying and copying of the fabric product can be directly finished by a computer, and the manual workload is reduced.
The specific scheme is that the image is a fabric overall scanning image with the scanning precision of 1200 dpi.
In the dividing step, the gray scale feature and the color feature are extracted from the image after coarse adjustment, different tissues are divided by using an unsupervised clustering algorithm, the number of the tissues is judged by using a contour coefficient, and the distribution of the different tissues on the fabric is identified so as to obtain the distribution area of a single tissue in the image.
Another preferred embodiment is a method for cutting at least a single tissue segment slice from each single tissue distribution region, comprising: and cutting out the maximum inscribed rectangle slice in the distribution area of the single tissue as the slice of the single tissue block.
More preferably, the step of cutting out the maximum inscribed rectangle slice in the distribution region of the single tissue comprises: (1) in a gray scale image corresponding to an image of the fabric to be analyzed, representing the regional distribution of different fabric structures on the whole image by using different gray scale values; (2) and identifying the boundaries of different gray value areas based on the acquired gray level images, cutting a single tissue slice with the largest area from the image of the fabric to be analyzed, and cutting a maximum inscribed rectangle slice from the single tissue slice.
Another preferred embodiment is that the fine-tuning step comprises: and performing spectral analysis on the single tissue image block slice after the primary correction through performing the gray level projection in different directions, and performing secondary correction on the longitude and latitude direction on the single tissue image block slice based on the result of the spectral analysis.
Another preferred embodiment is a method for obtaining a cycle count of tissue structures in a single tissue section slice comprising: and carrying out tissue classification on the single tissue pattern block slices according to the period, the phase difference and the warp and weft yarn width in the acquired tissue circulation mode, and then using the closest value of the tissue circulation number calculated by the period and the phase difference as the circulation number of the tissue structure in the single tissue pattern block slices.
Another preferred embodiment is a method for acquiring a cyclic pattern of tissue structure in a single slice of a tissue segment, comprising: (1) dividing or column division is carried out according to the gray projection distribution of the slice of the single tissue image block; (2) distinguishing the warp and weft yarns by using a pre-trained random forest model based on the texture characteristics according to the color and the texture characteristics of the slice of the single tissue picture block by taking a line or a column as a unit; (3) carrying out local smoothing treatment on the distinguished warp and weft yarn distribution diagram by using a hidden Markov random field method; (4) performing global optimization processing on the warp and weft yarn distribution map subjected to local smoothing processing by utilizing a particle swarm algorithm in combination with frequency domain analysis; (5) and calculating the width and the period of the warps and the wefts and the translation offset between two adjacent rows based on the integral warp and weft yarn distribution map on the optimized fabric to be analyzed, so as to obtain the circulation mode of the weave structure.
In the calculation step, the warp and weft density of the single tissue is equal to the product of the number of warp and weft yarns in the unit circulation and the circulation number in the unit length, wherein the step of acquiring the number of the warp and weft yarns in the unit circulation comprises the steps of calculating the number of the warp and weft yarns after intercepting the single circulation from the warp and weft yarn distribution diagram; when the overall warp and weft density of the fabric to be analyzed is calculated, the highest warp density of each single tissue is taken as the overall warp density, and the weft density is the product of the density of each single weft yarn in the single tissue and the number of weft yarn groups.
In order to achieve the above another object, the present invention provides an analyzing system for weave structure of woven fabric, including a processor and a memory, wherein the memory stores a computer program, and the computer program is capable of implementing the following steps when executed by the processor:
a receiving step, namely receiving an image of the fabric to be analyzed;
a coarse adjustment step, namely adjusting the image to the longitude and latitude direction of the fabric to be analyzed to be arranged along the transverse direction and the longitudinal direction of the screen by taking the longitude and latitude direction appointed by the pointing operation as the current longitude and latitude direction of the fabric to be analyzed according to the received pointing operation aiming at the longitude and latitude direction on the image;
a dividing step, namely extracting gray scale features and texture features from the image after coarse adjustment, dividing different tissues by using an unsupervised clustering algorithm, judging the number of the tissues by using a contour coefficient, and identifying the distribution of the different tissues on the fabric to obtain a distribution area of a single tissue in the image;
a cutting step, namely cutting at least a single tissue pattern block from each single tissue distribution area according to the distribution condition of the single tissue;
fine adjustment, namely performing primary correction on the longitude and latitude directions of the single tissue image blocks based on image pattern recognition and regression algorithm so as to arrange the image blocks along the transverse direction and the longitudinal direction of the screen;
the identification step, based on color features and texture features, using a random forest identification model generated by pre-training to perform image mode identification, frequency domain analysis and global optimization on the finely adjusted single tissue pattern block, and acquiring a circulation mode and a circulation number of an organization structure in a single tissue pattern block slice;
calculating, namely calculating the thread count and the longitude and latitude density of the yarn of the current texture structure based on the acquired cycle number, the physical size of a single cycle and the cycle mode, and calculating the integral thread count and longitude density of the fabric to be analyzed according to the thread count and longitude densities of all texture structures on the fabric to be analyzed;
and an output step, outputting the yarn distribution diagram of the fabric to be analyzed and the calculated warp and weft density.
According to the invention, the fabric image is divided into tissue slices with single tissues, and then artificial intelligent image recognition is carried out on the slices, so that the fabric tissue structure can be accurately recognized, and the warp and weft density numerical value of the fabric is obtained, thus the fabric tissue structure can be directly expressed in a digital form, the subsequent sample copying and copying of the fabric product can be directly finished by a computer, and the manual workload is reduced.
Drawings
FIG. 1 is a flow chart of the operation of an embodiment of the analysis method of the present invention;
FIG. 2 is a schematic view of a fabric image received in the receiving step and a schematic view of the pointing operation in the coarse adjustment step in an embodiment of the analysis method of the present invention;
FIG. 3 is a schematic diagram of the same tissue structure represented by different gray-scale values in the segmentation step according to an embodiment of the analysis method of the present invention;
FIG. 4 is a schematic illustration of the distribution of the cut pieces on different tissue structures during the cutting step in an embodiment of the analysis method of the present invention;
FIG. 5 is a schematic diagram of the cutting step of the analysis method of the present invention to cut a slice of a single tissue segment;
FIG. 6 shows a single slice of a tissue block cut by the cutting step as a subject of a subsequent analysis in an embodiment of the analysis method of the present invention;
FIG. 7 is a warp and weft yarn distribution map obtained in the obtaining step in an embodiment of the analysis method of the present invention;
FIG. 8 is a diagram illustrating a minimum cycle chart obtained in the obtaining step in the analysis method according to the embodiment of the present invention.
Detailed Description
The invention is further illustrated by the following examples and figures.
Example of analytical method
As shown in fig. 1, the analyzing method of the weave texture of the present invention includes a receiving step S1, a coarse adjusting step S2, a dividing step S3, a trimming step S4, a fine adjusting step S5, a recognition step S6, a calculating step S7, and an output step S8.
A receiving step S1, receiving an image of the fabric to be analyzed.
In the present embodiment, as shown in fig. 2, a flat bed scanner is used to scan the overall structure of the woven fabric, so as to obtain the overall fabric image with a resolution of 1200dpi or more at one time as the object of the subsequent analysis. Of course, the specific mode for acquiring the image is not limited to the method of the embodiment, and for example, the resolution of the fabric image may be 1200dpi or more, and may also be a fabric image captured by a camera, a mobile phone, or acquired from a network.
And a coarse adjustment step S2, wherein according to the received directional operation aiming at the longitude and latitude directions on the image, the longitude and latitude directions appointed by the directional operation are used as the current longitude and latitude directions of the fabric to be analyzed, and the longitude and latitude directions of the image to the other fabric to be analyzed are adjusted to be arranged along the transverse direction and the longitudinal direction of the screen.
The specific example process is as follows: (1) the computer with the specific analysis software installed displays the whole image of the fabric to be analyzed, which is received in the receiving step S1 and is shown in fig. 2, on the display of the computer; (2) specifically, when the display screen is a touch screen, the operator can click two end points of a longitude line where a schematic vertical black and thick line shown in fig. 2 is located on the touch screen, the connecting line of the two end points to form the trend of the longitude line, click two end points of a latitude line where a schematic horizontal black and thick line shown in fig. 2 is located on the touch screen, and the connecting line of the two end points to form the trend of the latitude line; or when the display screen is a non-touch screen, a mouse cursor is used for appointing a first point and a second point on the display area of the display screen, the connecting line of the two points to form a warp direction, then a third point and a fourth point are appointed on the display area of the display screen, and the connecting line of the two points to form a weft direction; (3) according to the longitude and latitude direction appointed by an operator, the image of the fabric to be analyzed is subjected to angle adjustment processing, so that the longitude and latitude direction on the image of the fabric to be analyzed is approximately arranged along the transverse longitudinal direction of the display screen, the precision of subsequent analysis is improved, and the workload of subsequent image processing is reduced.
And a dividing step S3, extracting gray scale and texture features from the image after coarse adjustment, dividing different tissues by using an unsupervised clustering algorithm, judging the number of the tissues by using a contour coefficient, and identifying the distribution of the different tissues on the fabric to obtain the distribution area of a single tissue in the image. The specific method comprises the following steps:
(1) reducing the picture resolution of the fabric image after coarse adjustment to about 300dpi, specifically 300 dpi;
(2) extracting a color histogram and an lbp (uniform pattern) in a region of size N × N pixel points, with a pixel as a unit: wherein the color histogram uses LAB colors, each channel is 8 bins, and the number of pixels in each Bin is made approximately equal throughout the map; n is a configurable variable, and the larger N is, the larger the processable single circulation area of the tissue is; in addition, the method is realized by using an integral graph to accelerate the calculation of the histogram;
(3) performing MiniBatchKMean clustering on all pixel points in the picture according to the features extracted in the step (2), wherein the clustering number is from 2 to 10, so as to obtain 9 clustering results, and in the embodiment, the default fabric comprises 10 maximum organization numbers;
(4) and (4) scoring the clustering result in the step (3) by using a contour Coefficient (Silhouette coeffient). Because the operation amount of the contour coefficient is too large, in the embodiment, 1000 points are specifically sampled for each clustering result to be calculated;
(5) based on the principle of avoiding missing tissues as much as possible in the initial tissue division stage, the clustering Score result Score in the aforementioned step (4)iWhen close to, wherein, ScoreiFor a Score representing the number of clusters as i, the larger the number of clusters, the higher the priority, the higher the Score by 2 Score thresholds (Score _ High, Score _ Low) are used to assist in selecting the best cluster number: when the Score of the cluster number exceeds Score _ High, taking the cluster number with the larger cluster number as the optimal cluster number; when no cluster number Score is higher than Score _ High, the cluster number with the larger cluster number when the following conditions are met is taken as the optimal cluster number:
Scorei>Score_Low and Scorei>Scorei-1*0.6
based on the unsupervised clustering algorithm, the image of the fabric to be analyzed is converted into a gray-scale map, and in the gray-scale map, the area distribution of different fabric structures on the whole image is represented by different gray-scale values, and in this embodiment, the representation result is shown in fig. 3.
And a cutting step S4, wherein at least a single tissue section block slice is cut from each single tissue distribution area according to the distribution condition of the single tissue.
Because different weaves are manufactured on the same overall fabric according to a preset scheme, such as satin weave, twill weave, plain weave and the like, (1) the position areas of the different weaves are represented by different gray values by utilizing the weave structure identified in the identification step S3; (2) based on the acquired gray-scale image, identifying the boundaries of different gray-scale value regions, and cutting out the single tissue slice with the largest mask, for example, the different black rectangular frame regions shown in fig. 4 are all single tissue slices; specifically, the largest inscribed rectangle slice is cut out from the single tissue slice, and the single tissue block slice in this embodiment is configured to be a processing object for subsequent analysis. By cutting the largest local graph as a subsequent analysis object, the accuracy of the subsequent analysis can be improved, and the local graph can be better used for representing the tissue structure of the tissue structure in the fabric.
For example, the black frame area in the local area shown in fig. 5 is cut according to the area coordinates obtained in the previous step, and a single tissue block slice shown in fig. 6 is obtained as the object of the subsequent analysis step, so as to obtain the tissue structure parameters in the tissue block slice.
A fine adjustment step S5, once corrects the longitude and latitude of a single tissue patch slice to be arranged longitudinally in the lateral direction of the screen. Taking the weft direction as an example, the specific method is as follows:
(1) converting the tissue pattern block slices cut in the step S4 into LAB color channels, and performing Gaussian blur on L channel images in the LAB color channels, wherein the window size is 11 and the standard deviation is 1.0;
(2) in the range of [ -15 degrees, 15 degrees ], taking 1 circle as a step length, and rotating the result picture obtained in the step (1) for multiple times;
(3) projecting the brightness of the result picture obtained in the step (2) in the horizontal direction, and removing the image black edge generated in the rotation process during calculation;
(4) fourier transform is carried out on the projection result obtained in the step (3), and the highest intensity IMax of the low-frequency signal (2 Hz-30 Hz) is recordediSubscript i is the corresponding rotation angle;
(5) take max (IMax)i)，i∈[-15°,15°]The slice of the tissue segment is rotated by the corresponding i, i.e., the rotation angle.
And an obtaining step S6, obtaining a circulation mode and a circulation number of the tissue structure in the slice of the single tissue image block based on image mode identification, frequency domain analysis and a global optimization algorithm.
In this step, a circulation mode of the tissue structure is obtained first, and then a circulation number of the tissue structure is obtained according to the circulation mode, specifically including the steps of:
the step of acquiring a cyclic pattern of the tissue structure comprises:
(1) the single tissue section slice obtained in the previous cutting step S4 is divided or column-divided according to the gray projection distribution thereof, for example, as shown in fig. 6.
In this step, the law of low brightness of yarn edge is utilized, and proper waveform optimization is performed according to the brightness projection of the tissue slice in the vertical or horizontal direction, so as to find out a proper trough position, thereby determining the dividing position of a row or a column, namely the dividing position of weft yarn-weft yarn or warp yarn-warp yarn.
(2) And (2) preliminarily distinguishing the warps and the wefts by using a fabric clustering analysis model obtained by training according to the color and texture characteristics of the single tissue image block slices by using the rows or the columns divided in the step (1) as units.
In the embodiment, the specific modes include 2 types of texture division and color division, which can be selected according to actual needs; for texture division, a pre-trained random forest model based on texture features is used for judging the distribution of warp and weft yarns in each row or column to obtain a result of a probability value; for color division, firstly, using an unsupervised clustering algorithm (FCM) based on color features to divide yarns in each row or column into 2 types, obtaining a result of a probability value, then using a random forest model in texture division to judge the probability that the 2 types of yarns are respectively warps and wefts, and taking the high probability as a division result.
The extraction method of the texture features comprises the following steps:
(a) extracting the horizontal and vertical gradients of the tissue block slices;
(b) for each column of pixels S in each row in the sliceijExtracting texture features, namely for a weft yarn, wherein i represents a row number, and j represents a column including maximum value, minimum value, mean value, standard deviation, variation trend from top to bottom, upper end, lower end and columnThe ratio of the brightness of the middle to the overall average brightness of the slice of the tissue segment;
(c) for a random forest model, training is carried out by using the texture characteristics and the label values of responses as input data, the label values are binary data of warps and wefts, 100 trees are used, the sample number of the minimum leaf nodes is set to be 10, and S is output during predictionijCorresponding warp yarn judgment result probability value Pij∈[0,1]。
And the way of extracting color features is as follows:
(a) tissue section slices were converted to LAB color channels.
(b) For each column of pixels S in each row in the sliceijExtracting color features, i.e. a weft yarn, wherein i represents a row number and j represents a column, comprising: maximum, minimum, mean, standard deviation of LAB three channel values.
(3) Carrying out local smoothing treatment on the distinguished warp and weft yarn distribution diagram by using a hidden Markov random field method;
(4) based on the characteristic that the distribution of the warp and weft yarns is in a single cycle in a single organization, the distribution diagram of the warp and weft yarns is subjected to global optimization processing by utilizing a particle swarm algorithm and combining frequency domain analysis. The concrete mode is as follows:
(a) using Fourier transform to obtain the result obtained in the step (2), namely the distribution probability value of warp and weft yarns in each row, obtaining the intensity value of a low-frequency signal of 2 Hz-15 Hz, and counting the result obtained in each row to obtain the frequency value with the highest intensity;
(b) if the frequency value is less than 3, directly using the local smoothing result of the step (3) as a final output result;
(c) if the frequency value is more than 3, the wavelength (Interval, the unit is the number of pixels) corresponding to the frequency is taken as the sum of the warp yarn span and the weft yarn span;
(d) using a particle swarm optimization algorithm, the optimization objects are 4 variables, which are respectively: the Warp yarn span Weft _ Width and the Weft yarn span Warp _ Width in each row, the starting x coordinate Start _ x of the Warp yarns in the first row and the relative position of the distribution of the Warp yarns in the two rows are Offset by Offset _ x;
(e) in the particle swarm optimization algorithm, the number of the used particles is 500, the initial value of each particle is uniformly distributed in the value interval of each variable, and the limiting conditions of the value interval of each variable are as follows:
Weft_Width+Warp_Width＝Interval；
Start_x<Interval；
Offset_x<Interval；
(f) the loss function used by the particle swarm optimization algorithm is: reconstructing the warp and weft yarn distribution map (binary distribution of 0, 1) of the single tissue by using the corresponding optimization object variable, and calculating the difference value of the result (probability value distribution of 0-1) acquired in the step (2), wherein the formula is as follows:
Loss＝Mean(Max(0，|Rij–Pij|-0.5))
wherein R isijE {0, 1}, which is a reconstructed warp and weft yarn distribution diagram; pij∈[0,1]，
The probability of the warp and weft yarns is judged according to the texture and the color characteristics; i is the row number and j is the column;
(g) reconstructing the warp and weft yarn distribution diagrams of the weave as the final output result by the optimal variables obtained by the particle swarm optimization algorithm, for example, obtaining the warp and weft yarn distribution diagrams of the weave cut block shown in fig. 6 as shown in fig. 7, wherein the black boxes represent warp yarns, the white boxes represent weft yarns, and the minimum cycle diagram is shown in fig. 8;
(5) and calculating the width and the period of the warps and the wefts and the phase difference between two adjacent rows based on the integral warp and weft yarn distribution map on the optimized fabric to be analyzed, so as to obtain the circulation mode of the weave structure.
(II) acquiring the number of cycles of the tissue structure:
(1) according to the period, the phase difference and the width of the warp and weft yarns in the obtained organization circulation mode, carrying out organization classification on the single organization pattern block slices;
(2) and then using the nearest value of the tissue cycle number calculated by the period and the phase difference as the cycle number of the tissue structure in the slice of the single tissue block.
And a calculating step S7, calculating the thread count and the longitude density of the yarn of the current organizational structure based on the acquired cycle number, the physical size of a single cycle and the cycle mode, calculating the overall thread count and the longitude density of the fabric to be analyzed according to the thread count and the longitude density of all organizational structures on the fabric to be analyzed, wherein the calculation result of the cutting block shown in the figure 6 is that the cycle number is 5, the warp density is 61.8918 pieces/cm, the density of single group of weft threads is 22.9836 pieces/cm, and the reliability of the distribution diagram is 92.3015%.
In the step, the warp and weft density of the single tissue is equal to the product of the number of warp and weft yarns in the unit circulation and the number of circulation in the unit length, wherein the step of acquiring the number of the warp and weft yarns in the unit circulation comprises the steps of intercepting the single circulation from a warp and weft yarn distribution diagram and then calculating to acquire the number of the warp and weft yarns in the unit circulation; and when the overall warp and weft density of the fabric to be analyzed is calculated, the highest warp density of each single tissue is taken as the overall warp density, and the weft density is the product of the density of each single weft yarn in the single tissue and the number of weft yarn groups.
And an output step S8, outputting the yarn distribution diagram of the fabric to be analyzed and the calculated warp and weft density.
And displaying the acquired organization structure on a display screen in a graph and data mode, displaying the organization structure type in a character and picture mode, and displaying the longitude and latitude density in a data mode.
Based on the related data obtained in the previous steps, a warp and weft yarn distribution map which can reflect the overall characteristics of the tissue slice can be synthesized, and the subsequent sample copying and copying of the fabric product can be directly finished through a computer, so that the manual workload is reduced.
Analysis System embodiment
The analysis system for the weave structure of the woven fabric comprises a processor and a memory, wherein the specific carrier is a terminal device such as a tablet computer, a desktop computer or a notebook computer, and the memory stores a computer program, and when the computer program is executed by the processor, the receiving step S1, the rough adjusting step S2, the dividing step S3, the cutting step S4, the fine adjusting step S5, the identifying step S6, the calculating step S7 and the outputting step S8 shown in fig. 1 can be implemented, and specific contents of the foregoing 8 steps are specifically described in the foregoing analysis method embodiment, and are not described again here.
Claims (10)
1. A method for analyzing the weave structure of a woven fabric, wherein the weave structure comprises the weave structure type and the warp and weft density, and the method comprises the following steps:
a receiving step, namely receiving an image of the fabric to be analyzed;
a coarse adjustment step, namely adjusting the longitudinal and latitudinal directions of the fabric to be analyzed to be longitudinally and transversely arranged along a screen by taking the longitudinal and latitudinal directions specified by the pointing operation as the current longitudinal and latitudinal directions of the fabric to be analyzed according to the received pointing operation aiming at the longitudinal and latitudinal directions on the image;
a dividing step, namely extracting gray scale features and texture features from the image after coarse adjustment, and identifying the distribution of different tissues on the fabric so as to obtain the distribution area of a single tissue in the image;
a cutting step, cutting at least one single tissue pattern block slice from each single tissue distribution area according to the distribution condition of the single tissue;
fine adjustment, namely performing primary correction on the longitude and latitude directions of the single tissue image block slice based on an image pattern recognition and regression algorithm so as to be arranged along the longitudinal direction and the transverse direction of the screen;
the method comprises the steps of identification, based on color features and texture features, using a random forest identification model generated by pre-training to perform image mode identification, frequency domain analysis and global optimization on a single tissue pattern block after fine adjustment, and acquiring a circulation mode and a circulation number of an organization structure in a single tissue pattern block slice;
calculating the thread count and the longitude count of the yarn of the current organizational structure based on the acquired number of cycles, the physical size of a single cycle and a cycle mode, and calculating the overall thread count and the longitude count of the fabric to be analyzed according to the thread count and the longitude count of all organizational structures on the fabric to be analyzed;
an output step, namely outputting the yarn distribution map of the fabric to be analyzed and the calculated warp and weft density;
in the dividing step, gray scale features and color features are extracted from the image after coarse adjustment, different tissues are divided by using an unsupervised clustering algorithm, the number of the tissues is judged by using a contour coefficient, and the distribution of the different tissues on the fabric is identified so as to obtain the distribution area of a single tissue in the image.
2. The analytical method of claim 1, wherein:
the image is a fabric overall scanning image with the scanning precision of 1200 dpi.
3. The analysis method of claim 2, wherein the step of excising at least one single tissue segment slice from each single tissue distribution region comprises:
and cutting out a maximum inscribed rectangle slice in the distribution area of the single tissue as the slice of the single tissue block.
4. The method of claim 3, wherein the step of cutting out a largest inscribed rectangle slice within the distributed region of the single tissue comprises:
in a gray scale image corresponding to the image of the fabric to be analyzed, representing the regional distribution of different fabric structures on the whole image by using different gray scale values;
and identifying the boundaries of different gray value areas based on the acquired gray level images, cutting a single tissue slice with the largest area from the image of the fabric to be analyzed, and cutting the largest inscribed rectangle slice from the single tissue slice.
5. The analysis method of claim 1, wherein the step of excising at least one single tissue segment slice from each single tissue distribution region comprises:
and cutting out a maximum inscribed rectangle slice in the distribution area of the single tissue as the slice of the single tissue block.
6. The assay of any one of claims 1 to 5, wherein the fine-tuning step comprises:
and performing spectral analysis on the single tissue image block slice after the primary correction through performing the gray level projection in different directions, and performing secondary correction on the longitude and latitude direction on the single tissue image block slice based on the result of the spectral analysis.
7. The method of any one of claims 1 to 5, wherein the step of obtaining a cycle number of tissue structures in a single tissue segment slice comprises:
and carrying out tissue classification on the single tissue pattern block slices according to the period, the phase difference and the width of the longitude and latitude yarns in the acquired tissue circulation mode, and then using the closest value of the tissue circulation number calculated by the period and the phase difference as the circulation number of the tissue structure in the single tissue pattern block slices.
8. The analysis method according to any one of claims 1 to 5, wherein the step of acquiring a cyclic pattern of tissue structures in the single tissue segment slice comprises:
dividing rows or columns according to the gray projection distribution of the single tissue image block slice;
distinguishing the warp and weft yarns by using a pre-trained random forest model based on the texture characteristics according to the color and the texture characteristics of the slice of the single tissue picture block by taking a line or a column as a unit;
carrying out local smoothing treatment on the distinguished warp and weft yarn distribution diagram by using a hidden Markov random field method;
performing global optimization processing on the warp and weft yarn distribution map subjected to local smoothing processing by utilizing a particle swarm algorithm in combination with frequency domain analysis; and obtaining the width and the period of the warp and weft yarns in a single weave pattern block and the translational offset between two adjacent rows after optimization, thereby obtaining the circulation mode of the weave structure.
9. The assay of any one of claims 1 to 5, wherein:
in the calculating step, the warp and weft density of the single tissue is equal to the product of the number of warp and weft yarns in unit circulation and the circulation number in unit length, wherein the step of acquiring the number of the warp and weft yarns in the unit circulation comprises the steps of acquiring the single circulation by calculation after intercepting the single circulation from a warp and weft yarn distribution diagram;
and when the overall warp and weft density of the fabric to be analyzed is calculated, taking the highest warp density of each single tissue as the overall warp density, wherein the weft density is the product of the density of each single weft yarn in the single tissue and the number of weft yarn groups.
10. An analysis system for the weave architecture of a woven fabric, the weave architecture including a weave architecture type and a thread count, the analysis system comprising a processor and a memory, the memory storing a computer program, wherein the computer program when executed by the processor is operable to perform the steps of:
a receiving step, namely receiving an image of the fabric to be analyzed;
a coarse adjustment step, namely adjusting the image to the longitudinal and latitudinal directions of the fabric to be analyzed along the transverse longitudinal arrangement of the screen by taking the longitudinal and latitudinal directions specified by the pointing operation as the current longitudinal and latitudinal directions of the fabric to be analyzed according to the received pointing operation aiming at the longitudinal and latitudinal directions on the image;
a dividing step, namely extracting gray scale features and color features from the image after coarse adjustment, and identifying the distribution of different tissues on the fabric to obtain the distribution area of a single tissue in the image;
a cutting step, cutting at least one single tissue pattern block from each single tissue distribution area according to the distribution condition of the single tissue;
fine adjustment, namely, performing primary correction on the longitude and latitude directions of the single tissue image block based on an image pattern recognition and regression algorithm so as to be arranged along the transverse direction and the longitudinal direction of the screen;
the method comprises the steps of identification, based on color features and texture features, using a random forest identification model generated by pre-training to perform image mode identification, frequency domain analysis and global optimization on a single tissue pattern block after fine adjustment, and acquiring a circulation mode and a circulation number of an organization structure in a single tissue pattern block slice;
calculating the thread count and the longitude count of the yarn of the current organizational structure based on the acquired number of cycles, the physical size of a single cycle and a cycle mode, and calculating the overall thread count and the longitude count of the fabric to be analyzed according to the thread count and the longitude count of all organizational structures on the fabric to be analyzed;
an output step, namely outputting the yarn distribution map of the fabric to be analyzed and the calculated warp and weft density;
in the dividing step, gray scale features and color features are extracted from the image after coarse adjustment, different tissues are divided by using an unsupervised clustering algorithm, the number of the tissues is judged by using a contour coefficient, and the distribution of the different tissues on the fabric is identified so as to obtain the distribution area of a single tissue in the image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811277402.0A CN109377489B (en) | 2018-10-30 | 2018-10-30 | Analysis method and analysis system for weave structure of shuttle fabric |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811277402.0A CN109377489B (en) | 2018-10-30 | 2018-10-30 | Analysis method and analysis system for weave structure of shuttle fabric |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109377489A CN109377489A (en) | 2019-02-22 |
| CN109377489B true CN109377489B (en) | 2020-12-11 |
Family
ID=65390458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811277402.0A Active CN109377489B (en) | 2018-10-30 | 2018-10-30 | Analysis method and analysis system for weave structure of shuttle fabric |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109377489B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110390675A (en) * | 2019-07-26 | 2019-10-29 | 常州弘仁智能科技有限公司 | A kind of fabric weft inclination detection method based on image analysing computer |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550660A (en) * | 2015-12-26 | 2016-05-04 | 河北工业大学 | Woven fabric weave structure type identification method |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101096819B (en) * | 2006-06-27 | 2014-07-09 | 中国纺织科学研究院 | Organization discrimination method of fabrics |
| CN103106645B (en) * | 2013-03-15 | 2015-08-12 | 天津工业大学 | A kind of recognition method for woven fabric structure |
| CN104195713B (en) * | 2014-08-27 | 2015-08-26 | 杭州易上弘网络技术有限公司 | Based on the yarn distribution map generalization method of the shuttle-woven fabric of fabric tissue and the processing of shuttle-woven fabric and outward appearance display methods |
| CN104715477B (en) * | 2015-03-05 | 2017-08-01 | 浙江工业大学之江学院 | The image analysis method of woven fabric count based on multiple dimensioned and multizone |
| US10198804B2 (en) * | 2015-04-15 | 2019-02-05 | Halliburton Energy Services, Inc. | Method for determining fabric and upscaled properties of geological sample |
| CN106485288B (en) * | 2016-12-21 | 2023-11-28 | 上海工程技术大学 | Automatic identification method for colored fabric tissue |
| CN108647687B (en) * | 2018-04-23 | 2021-09-24 | 浙江大学 | Fabric tissue identification method based on translation subtraction method |
-
2018
- 2018-10-30 CN CN201811277402.0A patent/CN109377489B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550660A (en) * | 2015-12-26 | 2016-05-04 | 河北工业大学 | Woven fabric weave structure type identification method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109377489A (en) | 2019-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109961049B (en) | Cigarette brand identification method under complex scene | |
| US10445557B2 (en) | Learning pixel visual context from object characteristics to generate rich semantic images | |
| CN107705306B (en) | Fabric defect detection method based on multi-feature matrix low-rank decomposition | |
| CN104834912A (en) | Weather identification method and apparatus based on image information detection | |
| CN104376556B (en) | A kind of rock CT images Target Segmentation method | |
| CN104463882B (en) | SAR image segmentation method based on shape completion administrative division map and feature coding | |
| CN104036529A (en) | Image analysis method for embroidery fabric design colors | |
| CN1844550A (en) | Textile and yarn analysis system based on two-side scanning technology | |
| US7747079B2 (en) | Method and system for learning spatio-spectral features in an image | |
| CN110309781A (en) | Damage remote sensing recognition method in house based on the fusion of multi-scale spectrum texture self-adaption | |
| US8249342B1 (en) | Color analytics for a digital image | |
| CN114004834B (en) | Method, equipment and device for analyzing foggy weather condition in image processing | |
| CN103530644A (en) | Automatic identification method for weave structures of striped fabric | |
| CN113255434B (en) | Apple identification method integrating fruit characteristics and deep convolutional neural network | |
| CN109829507B (en) | Aerial high-voltage transmission line environment detection method | |
| CN112348831B (en) | Shale SEM image segmentation method based on machine learning | |
| CN108986109B (en) | Automatic segmentation method for serialized visible human body slice images | |
| CN112102224A (en) | Cloth defect identification method based on deep convolutional neural network | |
| KR20120000732A (en) | An automatic segmentation method for object-based analysis using high resolution satellite imagery | |
| CN108830857A (en) | A kind of adaptive Chinese character rubbings image binaryzation partitioning algorithm | |
| CN109377489B (en) | Analysis method and analysis system for weave structure of shuttle fabric | |
| CN113689431B (en) | Industrial product appearance defect detection method and device | |
| CN114842262A (en) | Laser point cloud ground object automatic identification method fusing line channel orthographic images | |
| CN113139557B (en) | Feature extraction method based on two-dimensional multi-element empirical mode decomposition | |
| JP2020160543A (en) | Information processing system and information processing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |