CN113012105B - Yarn hairiness detection and rating method based on image processing - Google Patents

Abstract

The invention relates to a yarn hairiness detection rating method based on image processing, which aims to accurately detect characteristic parameters of yarn hairiness, and the invention uses a digital video microscope RH2000 to perform image acquisition, performs wavelet illumination removal, binarization, noise removal and thinning connection processing on the acquired image to obtain hairiness images, divides different areas for each hairiness in the images, marks label values, respectively counts the pixel number and the hairiness number of each hairiness, finds the corresponding relation between the hairiness pixel number and the hairiness length, converts the hairiness pixel into the hairiness length, and further realizes hairiness rating by using the hairiness length and the hairiness number. The image processing result has consistency with the visual hairiness result, and the proposed method can be considered to improve the hairiness detection accuracy.

Description

Yarn hairiness detection and rating method based on image processing

Technical Field

The invention relates to the field of yarn hairiness detection methods, in particular to a yarn hairiness detection rating method based on image processing.

Background

The hairiness detection method comprises a photoelectric method and a manual visual inspection counting method, and the photoelectric hairiness method has high detection efficiency, is influenced by the hairiness form, detects the projection length of the hairiness in the vertical direction in a two-dimensional plane instead of the absolute length of the hairiness, cannot effectively count the bent hairiness, and leads to inaccurate results. The manual visual inspection counting method has the defects of less sampling and low efficiency,

Disclosure of Invention

The technical problem to be solved by the invention is to provide the yarn hairiness detection rating method with accurate detection result and high efficiency based on image processing.

The yarn hairiness detection and rating method based on image processing comprises the following steps:

step 1, acquiring an original yarn image;

step 2, carrying out gray processing on the collected original yarn image to obtain a yarn gray image;

Step 3, carrying out multi-layer wavelet decomposition reconstruction on the filtered yarn trunk image, approximately reconstructing an illumination information layer, eliminating the influence of uneven illumination through arithmetic operation, and obtaining a yarn image;

Step 4, performing an operation on the yarn image without the illumination influence in the step 3 to obtain a yarn trunk, and extracting a hairiness image without the yarn trunk by using mathematical operation; filtering and denoising the filter;

and 5, binarizing the yarn trunk image obtained in the step 4 by using an improved Ojin algorithm to obtain a clear hairiness image.

Step 6, refining hairiness, namely performing traversing treatment on a hairiness pixel matrix by using a3 multiplied by 3 template which comprises 9 pixel points, respectively performing two iterations by taking the pixel point with the value of 1 as the center of the template, wherein the first iteration is to change the value of the pixel point meeting a preset first condition into 0, and the second iteration is to change the median value of the rest pixel points into 0 after removing the pixel points meeting a second preset condition, and refining the hairiness into a fine line through two iterations to obtain a complete hairiness image;

And 7, dividing each hairiness in the binarized hairiness image matrix into different areas for statistics, and setting the communication area to be four, so that the cross hairiness can be prevented from being divided into one communication area. All hairiness pixels meeting the requirements of a preset communication area in the pixel matrix are marked to be the same label value, hairiness pixels in different communication areas are marked to be different label values, the total label number num is returned, and the marked hairiness image matrix L is obtained. The number of hairiness pixels in different areas of the pixel matrix is traversed through a loop. Calculating and storing the number of hairiness pixels of each connected region and the number of connected regions corresponding to the number of the pixels to obtain a matrix S;

Step 8, measuring the length Ls of yarns in the final hairiness image, wherein the hairiness diameters Lv and V represent the horizontal resolution of the final hairiness image, converting the hairiness length into a pixel number interval by finding the corresponding relation between the pixel number and the length and utilizing a formula sum=L×Lv×V 2/Ls 2, counting the matrix values of the pixel numbers of the matrix S in the corresponding interval by utilizing a cyclic algorithm to obtain the corresponding hairiness numbers of different hairiness lengths, for example, converting e mm into a pixel by the formula when calculating the number of the hairiness lengths in the [ e mm and f mm ] interval, converting f mm into b pixel, wherein the pixel number sum of the hairiness lengths in the [ e mm and f mm ] interval is [ a and b ], and counting the hairiness number m1 of the hairiness pixels in the [ a and b ] interval in the matrix S, wherein m1 represents the total number of the hairiness lengths in the [ e mm and f mm ] interval;

And 9, calculating the quantity of hairiness corresponding to different hairiness lengths in the yarn hairiness image according to the formula, respectively counting the quantity of hairiness with the yarn hairiness lengths L of 1,2,3,4, 5 and 6mm, and evaluating the yarn hairiness by calculating a cv value to finally obtain a hairiness evaluation result.

Further, the preset communication areas are four communication areas.

Further, in the step 3, 7 layers of multi-layer wavelet decomposition is performed.

Further, the preset first condition includes the following three conditions, and the condition that three conditions are satisfied is determined as satisfying the preset first condition: the condition 1, the sum of the pixel values around the center point and w can meet that w is more than or equal to 2 and less than or equal to 6, and w is equal to the sum of the rest pixel values except the center of the template in the template;

When traversing the pixel values around the central point anticlockwise under the condition 2, the times of occurrence of the continuous two pixel values of 0 and 1 are 1 respectively;

The product of the pixel points of the upper, lower right and left positions of the center point and the lower, left and right positions is 0 under the condition 3.

Further, the preset second condition includes the following three conditions, and the condition that the three conditions are satisfied is determined as the condition that the preset second condition is satisfied:

the condition 1, the sum of the pixel values around the center point and w can meet that w is more than or equal to 2 and less than or equal to 6,w, and the sum of the rest pixel values except the center of the template in the template;

When traversing the pixel values around the central point anticlockwise under the condition 2, the times of occurrence of the continuous two pixel values of 0 and 1 are 1 respectively;

The product of the pixel values of the condition 3, the upper and lower left and the lower and upper left and right positions on the center point is 0.

The beneficial effects of the invention are as follows: the image processing method provided by the invention removes the influence of uneven illumination, extracts complete hairiness information, and calculates the hairiness length by dividing each hairiness into different areas and counting the pixel quantity of each hairiness. Compared with the conventional method, the method can eliminate the influence of uneven illumination when the yarn image is acquired, avoid the morphological influence of hairiness and improve the accuracy of hairiness detection.

Drawings

FIG. 1 is an original yarn image;

FIG. 2 is a schematic diagram of the process of obtaining a final hairiness image;

FIG. 3 is a yarn gray scale image;

FIG. 4 is a view of the yarn image after the wavelet has been decomposed for different layers;

FIG. 5 is a yarn image with the illumination effect removed;

FIG. 6 is a hairiness image without yarn backbone;

fig. 7 is a hairiness image after the filtering process;

FIG. 8 is a binarized hairiness image;

FIG. 9 is a final hairiness image;

FIG. 10 is a matrix generated image;

Fig. 11 is a partial enlarged view of fig. 10.

Detailed Description

The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are a partial embodiment of the present invention, not a full embodiment. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

1 Yarn image acquisition

And acquiring yarn images by using an RH2000 video microscope, wherein the acquired yarn images are clear, complete and have referential property. The proper size is also ensured in the case of ensuring definition, otherwise the processing speed is affected. Therefore, the resolution of the image is 1920×1200 with an accuracy of 1.0 um. While ensuring that the yarn is in the middle of the camera field of view, the axis remains horizontal. The hairiness image is clearest by adjusting the light source and the focal length. An original yarn image is shown in fig. 1.

2 Yarn image processing

2.1 Yarn image processing flow chart

The video microscope is used for acquiring the images of the yarns, and the acquired yarn hairiness images are uneven in brightness distribution due to the intersection of the yarn hairiness during the image acquisition. In the acquisition process, a small amount of noise points can appear due to the influence of hairiness background, the subsequent yarn hairiness image processing can be influenced, and in order to eliminate the influence, the acquired yarn hairiness image needs to be preprocessed. The flow chart is shown in figure 2.

2.2 Hairiness texture extraction

And (3) carrying out low-frequency decomposition and reconstruction on the collected hairiness image by utilizing wavelet transformation to acquire illumination layer information, decomposing the image information into each layer by wavelet decomposition, and selecting seven layers of low-frequency decomposition as the more the layers of decomposition are, the less the information represented by the low-frequency image is. As shown in fig. 4, the sixth layer can be approximately regarded as illumination information by contrast analysis. The sixth layer of signal representing illumination information is reconstructed from the multi-layer two-dimensional wavelet decomposition detail coefficients, and the influence of uneven illumination on subsequent image processing is eliminated by means of arithmetic operation, so that Mao Yute signs are more prominent, and hairiness is conveniently extracted. The image after removal of the illumination is shown in fig. 5. And (3) carrying out morphological open operation on the fifth graph to obtain a trunk, subtracting the trunk from the illumination yarn image by using the graph 5, and extracting a hairiness image without the trunk, as shown in the graph 6.

2.3 De-noising yarn hairiness

Because a small amount of noise points exist in the image acquisition process, direct binarization can lead to the segmentation elimination of the part Mao Yute, so that the image of fig. 6 is subjected to binarization segmentation more accurately, filtering is used for filtering the noise points, the hairiness characteristic is more obvious, and the result is shown in fig. 7.

2.4 Binarization processing

The hairiness image in fig. 7 is divided by improving the method of the Sedrin method to obtain a proper threshold value, the gray probability p (i) is added on the basis of the original Sedrin algorithm, the influence of the gray probability on the average gray is reduced, the proper threshold value is obtained, the gray value lower than the threshold value is changed to 0, and the hairiness pixels after processing are more prominent. And further obtaining a yarn hairiness binary image shown in the figure. The formula is as follows

Where p (i) represents the gray value probability, f (x, y) represents the pixel value of a certain point, and m×n represents the image size. Mu ₀ represents the average gray scale of the foreground portion, mu ₁ represents the average gray scale of the background portion, and mu represents the overall gray scale average.

2.5 Morphological treatments

The obtained binarized hairiness image comprises 9 pixel points by using a 3X3 template, traversing is carried out on a hairiness pixel matrix, two iterations are respectively carried out by taking the pixel point with the value of 1 as the center of the template, and when the first iteration can meet the following condition, the pixel point value is changed to 0.

A1	A2	A3
			A8	P	A4
A7	A6	A5

Condition 1:2< = a1+a2+a3+a4+a5+a6+a7+a8< = 6;

condition 2: when traversing the pixel values around the central point anticlockwise, the times of occurrence of the continuous two pixel values of 0 and 1 are 1 respectively;

condition 3: a2×a4×a6=0 and

A4*A6*A8＝0；

The second iterative removal satisfies the following condition to change the pixel point with the value of 1 to 0.

Condition 1:2< = a1+a2+a3+a4+a5+a6+a7+a8< = 6,

Condition 2: when traversing the pixel values around the central point anticlockwise, the times of occurrence of the continuous two pixel values of 0 and 1 are 1 respectively;

condition 3: a2×a4×a8=0 and

A2*A6*A8＝0；

Hairiness is refined into a thin line through two iterations. Obtaining a complete hairiness image; as shown in fig. 9.

3 Yarn hairiness characteristic parameter extraction and analysis

3.1 Hairiness pixel statistics

The pixel number with the value of 1 in the binarized hairiness image matrix of fig. 9 is divided into different areas according to the connection condition, the pixel number with the value of 1 in the upper, lower, left and right directions of the connection condition is set to be divided into one connection area, different label numbers of marks meeting the requirements of the connection area are met, and finally the matrix L and the total label number num are returned, wherein the label values are 1,2 and num. By calculating the maximum regional pixel number maxArea, constructing a zero matrix S with one row and maxArea columns, by traversing to calculate the different label regional pixel numbers sum (num) in the L matrix, calculating the regional pixel numbers sum (num) as column values for constructing the zero matrix S, and adding 1 at the position of the matrix as the hairiness number. And counting the binarized yarn hairiness matrix to finally obtain a matrix S with the columns as the hairiness pixel number sum (num) and the value at the corresponding position representing the hairiness number. The matrix S image is shown in fig. 10 and 11.

3.2 Yarn hairiness statistics

Yarn hairiness length in an image is measured by a two-dimensional tool provided by a video microscope and is recorded as Ls (mm), hairiness diameter is recorded as Lv (mm), and V represents the number of pixels forming the yarn length by a single pixel, namely a picture level pixel value, and in the embodiment, V is 1920. The hairiness length L is taken (1 mm,2mm,3mm,4mm,5mm,6 mm) and the hairiness pixel number sum is calculated by multiplying the formula pixel number M by the hairiness diameter pixel number. The hairiness length is divided by the number sum of pixels corresponding to different hairiness lengths, so that the hairiness rating is realized. The calculation formula for calculating the hairiness number is as follows:

sum＝L*Lv*V^2/Ls^2

The matrix values of the S matrix in each interval are counted by using a circulation algorithm to obtain the hairiness length and the corresponding hairiness quantity because the S matrix array value corresponds to the hairiness pixel number by dividing the pixel number intervals corresponding to different hairiness lengths. Ranking hairiness by coefficient of variation cv

The experiment is carried out by taking raw yarn with yarn density of 18.7tex, selecting 6 yarn segments, collecting 50 frames of images from each yarn segment, wherein the pixels corresponding to each frame of images after image processing are 1920 multiplied by 1200, the actual length of each corresponding frame of yarn is about 10.25mm, and the actual length of each yarn segment is about 5m. According to the yarn hairiness detection method, the image detection result is shown in table 1 under the condition that hairiness cross and special hairiness curl of the yarn are not considered.

Table 1: image detection method results

To verify the accuracy of the hairiness detection result of the method, hairiness of the 6 yarn hairiness fragments in table 1 was counted by visual inspection according to a length of 1m, and the result is shown in table 2. The comparison shows that the error of the test results is within 6% compared with the visual inspection method.

Table 2: image detection and visual inspection

Conclusion 4

The video microscope imaging provided by the article uses the yarn hairiness detection system of image processing, so that the reliability of yarn hairiness detection can be effectively improved. Compared with the common hairiness detection method, the method can eliminate the influence of uneven illumination during image acquisition, is not influenced by hairiness morphology, and improves the accuracy of hairiness detection. The test result shows that the method is practical and feasible, and can provide reference for future design of yarn hairiness detection systems.

The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims (5)

1. The yarn hairiness detection and rating method based on image processing is characterized by comprising the following steps of:

step 1, acquiring an original yarn image;

step 2, carrying out gray processing on the collected original yarn image to obtain a yarn gray image;

step3, carrying out multi-layer wavelet decomposition reconstruction on the filtered yarn trunk image, approximately reconstructing an illumination information layer, eliminating the influence of uneven illumination through arithmetic operation, and obtaining a yarn image;

step 4, performing an operation on the yarn image without the illumination influence in the step 3 to obtain a yarn trunk, and extracting a hairiness image without the yarn trunk by using mathematical operation; filtering and denoising the filter;

step 5, binarizing the yarn trunk image obtained in the step 4 by using an improved Ojin algorithm to obtain a clear hairiness image;

Step 6, refining hairiness, namely performing traversing treatment on a hairiness pixel matrix by using a 3 multiplied by 3 template which comprises 9 pixel points, respectively performing two iterations by taking the pixel point with the value of 1 as the center of the template, wherein the first iteration is to change the value of the pixel point meeting a preset first condition into 0, and the second iteration is to change the median value of the rest pixel points into 0 after removing the pixel points meeting a second preset condition, and refining the hairiness into a fine line through the two iterations to obtain a complete hairiness image;

Step 7, dividing each hairiness in the binarized hairiness image matrix into different areas for statistics, marking all hairiness pixels meeting the requirements of preset connected areas in the pixel matrix as the same label value, marking the hairiness pixels of the different connected areas as different label values, returning the total label number num, circularly traversing the marked hairiness image matrix L to obtain the quantity of the hairiness pixels of the different areas in the pixel matrix, and calculating and storing the quantity of the hairiness pixels of each connected area and the quantity of the connected areas corresponding to the quantity of the hairiness pixels to obtain a matrix S;

Step 8, measuring the length Ls of yarns in the final hairiness image, wherein the hairiness diameters Lv and V represent the horizontal resolution of the final hairiness image, converting the hairiness length into a pixel number interval by using a formula sum=L×Lv×V 2/Ls 2 by finding the corresponding relation between the pixel number and the length, and counting the matrix values of the pixel numbers of the corresponding interval of a matrix S by using a cyclic algorithm to obtain the corresponding hairiness numbers of different hairiness lengths;

And 9, calculating the quantity of hairiness corresponding to different hairiness lengths in the yarn hairiness image according to the formula, and evaluating the yarn hairiness by calculating a cv value to finally obtain a hairiness evaluation result.

2. The method for detecting and grading yarn hairiness based on image processing according to claim 1, wherein the preset connected areas are four connected areas.

3. The method for detecting and grading yarn hairiness based on image processing according to claim 1, wherein the step3 is 7 layers of wavelet decomposition.

4. The image processing-based yarn hairiness detection and rating method according to claim 1, wherein the preset first condition comprises three conditions, and the satisfaction of the three conditions is determined as satisfaction of the preset first condition: the condition 1, the sum of the pixel values around the center point and w can meet that w is more than or equal to 2 and less than or equal to 6,w, and the sum of the rest pixel values except the center of the template in the template;

when traversing the pixel values around the central point anticlockwise under the condition 2, the times of occurrence of the continuous two pixel values of 0 and 1 are 1 respectively;

The product of the pixel points of the upper, lower right and left positions of the center point and the lower, left and right positions is 0 under the condition 3.

5. The image processing-based yarn hairiness detection and rating method according to claim 1, wherein the preset second condition comprises three conditions, and the satisfaction of the three conditions is determined as satisfaction of the preset second condition:

The condition 1, the sum of the pixel values around the center point and w can meet that w is more than or equal to 2 and less than or equal to 6,w, and the sum of the other pixel values except the center of the template in the template;

when traversing the pixel values around the central point anticlockwise under the condition 2, the times of occurrence of the continuous two pixel values of 0 and 1 are 1 respectively;

The product of the pixel values of the condition 3, the upper and lower left and the lower and upper left and right positions on the center point is 0.