Method for identifying number of monofilaments based on chemical fiber section

Abstract

The invention belongs to the technical field of image recognition, and relates to a method for recognizing the number of monofilaments based on a chemical fiber section, which comprises the following steps of: after acquiring a target image and converting the target image into a grey-scale image, firstly, increasing a boundary distance between two adjacent detection objects in the grey-scale image through algorithm processing; the method comprises the following steps of: obtaining a grey-scale image, enhancing the boundary contrast between two adjacent detection objects in the grey-scale image, removing the background in the grey-scale image to obtain a grey-scale image without the background, carrying out binarization processing on the grey-scale image without the background to obtain a binarized image, and sequentially removing noise points and filling holes in the detection objects to obtain a binarized image. And finally, identifying the number of inscribed circles in the characteristic blocks of the filaments, namely obtaining the number of the monofilaments. The target image is a chemical fiber cross section picture shot by a microscope; the detection object is the area of each monofilament in the chemical fiber cross section picture. According to the method, manual counting one by one is not needed, the identification efficiency is high, and the accuracy is high.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition and relates to a method for identifying the number of single filaments based on the cross-section of chemical fibers. Background Art

[0002] In the production of chemical fibers, a single filament is first ejected from a spinneret plate into many fine filaments and then combined into one. So a single filament is composed of many, many fine filaments. In production, we need to regularly inspect the number of fine filaments. Currently, to inspect the number of fine filaments, the filament is first cut transversely, and a cross-sectional image is taken under a microscope. Then, the number of fine filaments is manually marked one by one by looking at the image. In some special filaments, there are often more than 700 fine filaments. Counting them one by one manually is time-consuming and laborious, and it is easy to miss some.

[0003] As Figure 1 shown, Patent CN201210280411.1 discloses a method for accurately identifying the geometric stiffness of a cable-bar cross-section. Among the three counting pictures, all are regular and closely arranged shapes, and the theoretical value of the number of roots is calculated through the diameter. However, for the actual cross-sectional pictures of chemical fibers, the arrangement shape of the filaments is completely irregular, and there are likely to be gaps (not closely arranged) in the middle. Using the identification method of this patent will result in large errors.

[0004] In addition, in Patent CN202211524389.0, only gray-scale and binarization operations are performed on the original picture, and then the counting operation starts without processing the picture. The contour matching method has a huge computer calculation amount and is only applicable to the identification and counting of a small number of targets, and is not applicable to the identification of the number of single filaments in the cross-sectional pictures of chemical fibers.

[0005] In actual production applications, due to different production environments, processes, etc., the collected cross-sectional pictures of chemical fibers are often not perfect, and there are various noise backgrounds in the pictures. Moreover, the brightness of the recognition targets is not uniform, and the brightness of the high-brightness background in some areas is greater than that of the low-brightness targets. In this case, direct binarization operation on the image cannot be performed. Currently, the existing related methods are mainly applicable in excellent environments such as laboratories and cannot be applied to the on-site production environment.

[0006] Therefore, it is of great significance to study a method for identifying the number of single filaments based on the cross-section of chemical fibers to solve the problems existing in the prior art. Summary of the Invention

[0007] The purpose of the present invention is to solve the problems existing in the prior art and provide a method for identifying the number of single filaments based on the cross-section of chemical fibers.

[0008] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0009] A method for identifying the number of single filaments based on the cross-section of chemical fiber. After obtaining the target image and converting it into a grayscale image, first, the boundary distance between adjacent two detection objects in the grayscale image is increased through algorithm processing, then the boundary contrast between adjacent two detection objects in the grayscale image is enhanced. After that, the background in the grayscale image is removed to obtain a grayscale image without background. Then, the grayscale image without background is binarized to obtain a binary image. Next, noise points are removed and holes in the detection objects are filled in turn. Finally, the number of inscribed circles in the characteristic block of the filament ( Figure 13 The red patches in

[0010] are the characteristic blocks of the filaments. Under normal circumstances, each red patch represents a single filament), that is, the number of single filaments is obtained;

[0011] The target image is a cross-sectional picture of chemical fiber taken by a microscope;

[0012] As a preferred technical solution:

[0013] A method for identifying the number of single filaments based on the cross-section of chemical fiber as described above specifically includes the following steps:

[0014] Step 1: Obtain the target image and convert the target image into a grayscale image;

[0015] Step 2: Process the grayscale image obtained in Step 1 using the erosion algorithm to remove noise points while increasing the boundary distance between adjacent two detection objects;

[0016] Step 3: Perform an opening operation on the image processed in Step 2;

[0017] Step 4: For the image obtained in Step 3, use the image processing function convolution-Highlight Details to enhance the details and contrast of the image;

[0018] Step 5: Select multiple background points or background regions on the edge of the detection object;

[0019] Step 6: Based on the background points or background regions selected in Step 5, use the image processed in Step 4 as the reconstruction object for gray-scale morphological reconstruction to obtain the reconstructed image, that is, the background, so that the extracted background brightness can change according to the brightness change of the detection objects around the background pixel points;

[0020] Step 7: Remove the background extracted in Step 6 from the image obtained in Step 4 to obtain a grayscale image with the pixel value of the background below 20;

[0021] Step 8: Perform an opening operation on the grayscale image with the pixel value of the background below 20 in Step 7;

[0022] Step 9: Binarize the image obtained in Step 8 to obtain a binarized image;

[0023] Step 10: Process the binarized image obtained in Step 9 using an erosion algorithm to remove noise points;

[0024] Step 11: For the image processed in Step 10, use a dilation algorithm and a Fill holes algorithm respectively to fill the holes in the detected object;

[0025] Step 12: Use the Circle detection algorithm to detect the number of inscribed circles in the feature block of the identified wire, as well as the corresponding inscribed circle radius and the center coordinate positions;

[0026] Step 13: First, use the inscribed circle radius obtained in Step 12 and calculate the average radius R of all inscribed circles by the median averaging method; then, according to the center coordinate positions, calculate the center distance between each inscribed circle and all other inscribed circles. If there exists a center distance less than 2 times R, it indicates duplicate counting, and the counting of the inscribed circle with the smaller radius is excluded; finally, count the number of inscribed circles to obtain the number of single wires.

[0027] A method for identifying the number of single wires based on the cross-section of chemical fiber as described above, wherein the minimum width of the detected object in the target image of Step 1 is greater than 30 pixels.

[0028] A method for identifying the number of single wires based on the cross-section of chemical fiber as described above, wherein in Step 1, there is a situation where the background brightness of one or some detected objects in the target image is higher than that of other detected objects.

[0029] A method for identifying the number of single wires based on the cross-section of chemical fiber as described above, wherein in Step 2, after processing with the erosion algorithm, the boundary distance between two adjacent detected objects increases by 10 - 20 pixels.

[0030] A method for identifying the number of single wires based on the cross-section of chemical fiber as described above, wherein in Step 7, the Subtract function is used to remove the background to obtain a grayscale image with the pixel value of the background below 20.

[0031] A method for identifying the number of single wires based on the cross-section of chemical fiber as described above, wherein in Step 9, thresholdsetup is used for image binarization, and the pixels with a brightness exceeding 20 are converted to 1, and the others are 0.

[0032] A method for identifying the number of single wires based on the cross-section of chemical fiber as described in any one of the above, wherein the circularity of the cross-sectional shape of the single wire is greater than 0.9.

[0033] A method for identifying the number of single wires based on the cross-section of chemical fiber as described in any one of the above, which can identify more than 99% of the number of single wires.

[0034] Beneficial effects:

[0035] (1) The method for identifying the number of single filaments based on the cross-section of chemical fiber in the present invention does not require manual counting one by one, has high identification efficiency and high accuracy, and is applicable to on-site images in complex environments.

[0036] (2) The present invention can achieve automatic counting of more than 99% (individually up to 100%) of the pictures. Manual labor only needs to mark the remaining unrecognized filaments, which can save more than 90% of the manual operation time. Description of the drawings

[0037] Figure 1 Schematic diagrams of the cross-sections of steel wire cable rods with different shapes in Patent CN201210280411.1;

[0038] Figure 2 Flow chart of the method for identifying the number of single filaments based on the cross-section of chemical fiber in the present invention;

[0039] Figure 3 Identification result of Application Scenario Example 1 of the present invention; among them, the green dots are the number recognized by the software, and the red dots are the number added manually;

[0040] Figure 4 Identification result of Application Scenario Example 2 of the present invention; among them, the green dots are the number recognized by the software, and the red dots are the number added manually;

[0041] Figure 5 Is a grayscale image;

[0042] Figure 6 Is a grayscale image (left figure) and the image after being processed in Step 2 (right figure);

[0043] Figure 7 Is the image before being processed in Step 3 (left figure) and the image after being processed in Step 3 (right figure);

[0044] Figure 8 Is the image before being processed in Step 4 (left figure) and the image after being processed in Step 4 (right figure);

[0045] Figure 9 Is after being processed in Step 6;

[0046] Figure 10 Is after being processed in Step 7;

[0047] Figure 11 Is the image before being processed in Step 8 (left figure) and the image after being processed in Step 8 (right figure);

[0048] Figure 12 Is the image after binarization in Step 9;

[0049] Figure 13 Before the processing in Step 10 (left figure) and the image after the processing in Step 10 (right figure);

[0050] Figure 14 Before the processing using the dilation algorithm in Step 11 (left figure) and the image after the processing (right figure);

[0051] Figure 15 Before the processing using the Fill holes algorithm in Step 11 (left figure) and the image after the processing (right figure);

[0052] Figure 16 Results of the inscribed circle recognition algorithm in Step 12; among them, the green is the schematic diagram of the recognized inscribed circle, and the red is the binary feature block of each filament;

[0053] Figure 17 Schematic diagram of repeated recognition in the results of the algorithm in Step 12; among them, the green is the schematic diagram of the recognized inscribed circle, and the red is the feature block of each filament;

[0054] Figure 18 Recognition results of Application Scenario Example 3 of the present invention; among them, the green dots are the number recognized by the software, and the red dots are the number added manually;

[0055] Figure 19 Recognition results of Application Scenario Example 4 of the present invention; among them, the green dots are the number recognized by the software, and the red dots are the number added manually;

[0056] Figure 20 Recognition results of Application Scenario Example 5 of the present invention; among them, the green dots are the number recognized by the software, and the red dots are the number added manually;

[0057] Figure 21 Schematic diagram of the pixel origin of the picture, that is, the red dot at the upper left corner vertex of the picture; unless otherwise specified, the image origin (0, 0) in all examples of the present invention is at the upper left corner vertex;

[0058] Among them, Figures 3 - 4 the upper left corner numbers in 17 - 21 represent the recognition results of the number of single filaments by the software and manually. Specific Embodiments

[0059] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

[0060] A method for recognizing the number of single filaments based on the cross-section of chemical fibers, the specific steps are as follows:

[0061] Step 1: Obtain the target image and convert the target image into a grayscale image; the target image is a cross-sectional picture of chemical fiber taken by a microscope; the minimum width of the detection object in the target image is greater than 30 pixels.

[0062] Step 2: Process the grayscale image obtained in Step 1 using the erosion algorithm in the "gray-scale morphology algorithm" to remove noise points and increase the boundary distance between two adjacent detection objects at the same time;

[0063] After processing with the erosion algorithm, the boundary distance between two adjacent detection objects increases by 10 - 20 pixels; the detection object is the area where each single fiber is located in the cross-sectional picture of the chemical fiber; the circularity of the cross-sectional shape of the single fiber is greater than 0.9;

[0064] Step 3: Perform an opening operation on the image processed in Step 2;

[0065] Step 4: For the image obtained in Step 3, use the image processing function convolution - Highlight Details to enhance the details and contrast of the image;

[0066] Step 5: Select multiple background points or background regions on the edge of the detection object;

[0067] Step 6: Based on the background points or background regions selected in Step 5, perform gray-scale morphological reconstruction with the image processed in Step 4 as the reconstruction object to obtain the reconstructed image, that is, the background;

[0068] Step 7: Use the Subtract function to remove the background extracted in Step 6 from the image obtained in Step 4 to obtain a grayscale image with the pixel value of the background below 20;

[0069] Step 8: Perform an opening operation on the grayscale image with the pixel value of the background below 20 in Step 7;

[0070] Step 9: Use threshold setup to binarize the image obtained in Step 8, convert the pixels with pixel brightness exceeding 20 to 1, and the others to 0 to obtain a binary image;

[0071] Step 10: Process the binary image obtained in Step 9 using the erosion algorithm to remove noise points;

[0072] Step 11: For the image processed in Step 10, use the dilation algorithm and Fill holes algorithm respectively to fill the holes in the detection object to avoid affecting the calculation of the inscribed circle algorithm in the subsequent steps;

[0073] Step 12: Use the Circle detection algorithm to detect the number of inscribed circles in the characteristic block of the identified filaments, as well as the corresponding inscribed circle radii and the coordinates of the circle centers;

[0074] Step 13: For two independent circular objects, the distance between the objects is not less than the sum of the radii of the two circles. According to this principle, first use the inscribed circle radii obtained in Step 12 to calculate the average radius R of all inscribed circles by the median averaging method; then, based on the coordinates of the circle centers, calculate the distance between the centers of each inscribed circle and all other inscribed circles. If there is a center distance less than 2 times R, it means there is double counting, and the inscribed circle with the smaller radius is removed from the count; finally, count the number of inscribed circles to obtain the number of single filaments.

[0075] The present invention can identify more than 99% of the number of single filaments.

[0076] The following uses specific embodiments to illustrate a method for identifying the number of single filaments based on the cross-section of chemical fibers of the present invention, as follows:

[0077] Embodiment 1

[0078] A method for identifying the number of single filaments based on the cross-section of chemical fibers, as Figure 2 shown, the specific steps are as follows:

[0079] Step 1: Obtain the target image and convert the target image into a grayscale image, as Figure 5 shown; the target image is a cross-sectional picture of polyester chemical fiber taken by a microscope; the minimum width of the detection object in the target image is in the range of 35 - 40 pixels.

[0080] Step 2: Use the erosion algorithm to process the grayscale image obtained in Step 1 to remove noise points and increase the boundary distance between adjacent two detection objects; wherein, the algorithm parameters are: size = 5, structuring element = circle, Iteration = 2;

[0081] As Figure 6 shown, after using the erosion algorithm, the boundary distance between adjacent two detection objects increases by 10 pixels; the detection object is the area where each single filament is located in the cross-sectional picture of the chemical fiber; the circularity of the cross-sectional shape of the single filament is 0.96 - 1;

[0082] Step 3: As Figure 7 shown, perform an opening operation on the image processed in Step 2 to make the boundary processed in Step 2 clearer and the pixels in the middle of the detection object more uniform; wherein, the algorithm parameters are: size = 15, structuring element = circle;

[0083] Step 4: AsFigure 8 As shown, for the image obtained in Step 3, use the image processing function convolution - Highlight Details to enhance the details and contrast of the image; among them, the algorithm parameters: kernelsize (kernel size) = 9;

[0084] Step 5: Manually select the background points (984, 763) and (867, 719) on the picture;

[0085] Step 6: Based on the background points selected in Step 5, use the gray - scale morphological reconstruction with the image processed in Step 4 as the reconstruction object to obtain the reconstructed image as shown, that is, the background. As can be seen from, the pixel brightness in different regions is different; among them, the algorithm parameters: size X = 3, size Y = 3, reconstruction = bright regions; Figure 9 shown, that is, the background. From Figure 9 it can be seen that the pixel brightness in different regions is different; among them, the algorithm parameters: size X = 3, size Y = 3, reconstruction = bright regions;

[0086] Step 7: Use the Subtract function to remove the background extracted in Step 6 from the image obtained in Step 4 to obtain a grayscale image with the pixel value of the background below 20; as shown, it can be seen that the original hazy background is basically eliminated and the boundary becomes clear; Figure 10 shown, it can be seen that the original hazy background is basically eliminated and the boundary becomes clear;

[0087] Step 8: As shown, perform an opening operation on the grayscale image without background in Step 7 to make the edges of the detection object clear and reduce edge noise; among them, the algorithm parameters: size = 3, matrix structure = circular; Figure 11 shown, perform an opening operation on the grayscale image without background in Step 7 to make the edges of the detection object clear and reduce edge noise; among them, the algorithm parameters: size = 3, matrix structure = circular;

[0088] Step 9: Use threshold setup to binarize the image obtained in Step 8, convert the pixels with pixel brightness exceeding 20 to 1, and the others to 0, to obtain the binarized image as shown; Figure 12 shown;

[0089] Step 10: As shown, use the erosion algorithm to process the binarized image obtained in Step 9 to remove noise points; among them, the algorithm parameters: size = 3, matrix structure = circular, number of iterations = 1; Figure 13 shown, use the erosion algorithm to process the binarized image obtained in Step 9 to remove noise points; among them, the algorithm parameters: size = 3, matrix structure = circular, number of iterations = 1;

[0090] Step 11: For the image processed in Step 10, as shown in and, use the dilation algorithm and Fillholes algorithm respectively to fill the holes in the detection object; among them, the algorithm parameters: Dilate size (dilation size) = 3, number of iterations = 2, matrix structure = circular; Figure 14 and 15 shown, use the dilation algorithm and Fillholes algorithm respectively to fill the holes in the detection object; among them, the algorithm parameters: Dilate size (dilation size) = 3, number of iterations = 2, matrix structure = circular;

[0091] Step 12: As shown, Figure 16As shown in the figure, use the Circle detection algorithm to detect the number of inscribed circles in the characteristic block of the identified wire. The detection range of the inscribed circle radius is greater than or equal to 6 and less than or equal to 30, as well as the corresponding inscribed circle radius and the center coordinate position.

[0092] Step 13: As Figure 17 shown in the figure, for two independent circular objects, the distance between the objects will not be less than the sum of the radii of the two circles. According to this principle, first use the average radius R of all inscribed circles calculated by the median averaging method using the inscribed circle radius obtained in Step 12; then, according to the center coordinate position, calculate the center distance between each inscribed circle and the centers of all other inscribed circles. If there is a center distance less than 2 times R, it means there is double counting, and the count of the inscribed circle with the smaller radius is excluded; finally, count the number of inscribed circles to obtain the number of single wires.

[0093] As Figure 3 、 Figure 21 shown in the application scenario Example 1, there are a total of 768 wires, the algorithm identifies 767, and 1 is manually added for identification. The software recognition rate is 99.9%.

[0094] Example 2

[0095] A method for identifying the number of single wires based on the chemical fiber cross-section is basically the same as Example 1, except that: in Step 1, the minimum width of the detection object in the target image is in the range of 35 - 45 pixels, and the points manually selected in Step 5 are (862, 1074), (706, 303), and (383, 724).

[0096] As Figure 4 shown in the application scenario Example 2, there are a total of 768 wires, the algorithm identifies 765, and 3 are manually added for identification. The software recognition rate is 99.6%.

[0097] Example 3

[0098] A method for identifying the number of single wires based on the chemical fiber cross-section is basically the same as Example 1, except that: in Step 1, the minimum width of the detection object in the target image is in the range of 40 - 50 pixels, and the points manually selected in Step 5 are (764, 695) and (638, 1177).

[0099] As Figure 18 shown in the application scenario Example 3, there are a total of 384 wires, the algorithm identifies 384, and 0 are manually added for identification. The software recognition rate is 100%.

[0100] Example 4

[0101] A method for identifying the number of single filaments based on the cross-section of chemical fiber is basically the same as that in Embodiment 1, except that: in step 1, the minimum width of the detection object in the target image is in the range of 40 to 50 pixels, and in step 5, the manually selected points are (293, 638), (675, 653), and (388, 825).

[0102] As Figure 19 shown in Application Scenario Embodiment 3, there are a total of 382 filaments, the algorithm identifies 382, the manual additional identification is 0, and the software identification rate is 100%.

[0103] Embodiment 5

[0104] A method for identifying the number of single filaments based on the cross-section of chemical fiber is basically the same as that in Embodiment 1, except that: in step 1, the minimum width of the detection object in the target image is in the range of 40 to 50 pixels, and in step 5, the manually selected points are (764, 695) and (638, 1177).

[0105] As Figure 20 shown in Application Scenario Embodiment 3, there are a total of 382 filaments, the algorithm identifies 382, the manual additional identification is 0, and the software identification rate is 100%.

[0106] In summary, the present invention can identify more than 99% of the number of single filaments in the picture.

Claims

1. A method for identifying the number of monofilaments based on the cross section of a chemical fiber, characterized in that: After acquiring the target image and converting it into a grayscale image, the boundary distance between two adjacent detection objects in the grayscale image is first increased through algorithm processing, and then the boundary contrast between two adjacent detection objects in the grayscale image is enhanced. After that, the background in the grayscale image is removed to obtain a grayscale image without background, and then the grayscale image without background is binarized to obtain a binary image, and then the noise points are removed and the holes in the detection object are filled in turn. Finally, the number of inscribed circles in the characteristic block of the silk is identified, that is, the number of single silk roots is obtained; The target image is a cross-section image of a chemical fiber taken with a microscope; The detection object is the area where each single fiber is located in the chemical fiber cross-section image.

2. A method for identifying the number of single fibers based on the cross section of a chemical fiber according to claim 1, characterized in that: The specific steps are as follows: Step 1: Obtain the target image and convert it into a grayscale image; Step 2: Use the corrosion algorithm to process the grayscale image obtained in step 1, remove noise points and increase the boundary distance between two adjacent detection objects; Step 3: Perform an opening operation on the image processed in step 2; Step 4: For the image obtained in step 3, use the image processing function convolution-Highlight Details to enhance the details and contrast of the image; Step 5: Select multiple background points or background areas around the detection object; Step 6: Based on the background points or background areas selected in step 5, grayscale morphological reconstruction is performed with the image processed in step 4 as the reconstruction object to obtain a reconstructed image, i.e., the background; Step 7: Remove the background extracted in step 6 from the image obtained in step 4 to obtain a grayscale image with a background pixel value below 20; Step 8: Perform an opening operation on the grayscale image of the background whose pixel value is below 20 in step 7; Step 9: Binarize the image obtained in step 8 to obtain a binary image; Step 10: Use the corrosion algorithm to process the binary image obtained in step 9 to remove noise points; Step 11: For the image processed in step 10, use the dilation algorithm and the Fill holes algorithm to fill the holes in the detected object respectively; Step 12: Use the Circle detection algorithm to detect the number of inscribed circles in the identified silk feature block, as well as the corresponding inscribed circle radius and circle center coordinate position; Step 13: First, use the radius of the inscribed circle obtained in step 12 to calculate the average radius R of all inscribed circles using the median average method; then, based on the coordinate position of the center of the circle, calculate the center distance between each inscribed circle and all other inscribed circles. If the distance between the centers of two circles is less than 2 times R, it means that there are repeated counts, and the count of the inscribed circle with the smaller radius is eliminated; finally, count the number of inscribed circles to obtain the number of single fibers.

3. A method for identifying the number of single filaments based on the cross section of a chemical fiber according to claim 2, characterized in that: The minimum width of the detected object in the target image of step 1 is greater than 30 pixels.

4. A method for identifying the number of single fibers based on the cross section of a chemical fiber according to claim 2, characterized in that: In the target image of step 1, there is a situation where the background brightness of the detection object is higher than the brightness of other detection objects.

5. The method for identifying the number of single fibers based on the cross section of chemical fiber according to claim 2, characterized in that: In step 2, the boundary distance between two adjacent detection objects increases by 10 to 20 pixels after being processed using the corrosion algorithm.

6. A method for identifying the number of single filaments based on the cross section of a chemical fiber according to claim 2, characterized in that: In step 7, the Subtract function is used to remove the background and obtain a grayscale image with a background pixel value below 20.

7. A method for identifying the number of single filaments based on the cross section of a chemical fiber according to claim 2, characterized in that: In step 9, the threshold setup is used to binarize the image, so that pixels with brightness greater than 20 are converted to 1 and the others are 0.

8. A method for identifying the number of single fibers based on a chemical fiber cross section according to any one of claims 1 to 7, characterized in that: The cross-sectional shape of the monofilament has a circularity greater than 0.

9.

9. A method for identifying the number of single filaments based on the cross section of a chemical fiber according to claim 8, characterized in that: Able to identify more than 99% of the number of single filaments.

Images