A method for detecting hairy yarn defects in packaged filaments

Abstract

The present invention discloses a method for detecting fuzzy defects of packaged filaments. The method comprises: acquiring multiple collected images of the area to be detected of the packaged filaments; intercepting a target image corresponding to the inspected position in the area to be detected in the collected image; determining a continuous long curve in the target image and multiple short lines on the continuous long curve; dividing the target image into multiple sub-images, and determining the line geometry parameters of the short lines in each sub-image and the image geometry parameters of the sub-image; and determining whether the packaged filaments have fuzzy defects according to the line geometry parameters and the image geometry parameters. The method for detecting fuzzy defects of packaged filaments according to the embodiment of the present invention can reduce the error rate of manual visual inspection, improve detection efficiency, and save production costs.

Description

Technical Field

[0001] The invention relates to the technical field of surface detection of packaged filaments, and in particular to a method for detecting filament fuzz defects of packaged filaments. Background Art

[0002] Packaged filaments are filaments that are made into finished packages with a certain shape and capacity through a winding mechanism during the production process. The defects of packaged filaments are mainly divided into two parts: physical and chemical property defects of the filaments and appearance defects of the packaged filaments. Through practical production, it is found that the appearance defects of packaged filaments have a very important impact on the quality of the fabric and will lead to a decrease in the fabric yield. Therefore, during the production process, it is necessary to strengthen the detection of appearance defects of packaged filaments.

[0003] The fuzz defect of packaged filament is caused by unstable process and equipment. In the subsequent processing, it will affect the warping, weaving and dyeing performance. Therefore, the factory has very high requirements for the detection of fuzz. If fuzz appears on the packaged filament, fuzz will appear on its upper side, lower side and cylindrical surface at the same time.

[0004] Due to the deformability, multiple curved surfaces and large inspection surface of packaged filaments, it is difficult to extract unified standards for the appearance defect characteristics. The appearance inspection of packaged filaments has always affected the realization of long-process intelligent manufacturing, and hairy filament defects can only be detected through manual visual inspection.

[0005] Although manual visual inspection can detect some defects, it will increase the labor cost of production, and without unified and strict quality standards, it is difficult to ensure the accuracy of the inspection results. Summary of the invention

[0006] To solve the above problems, an embodiment of the present invention provides a method for detecting filament defects in packaged filaments, which can reduce the error rate of manual visual inspection, improve detection efficiency, and save production costs.

[0007] To achieve the above object, an embodiment of the present invention provides a method for detecting filament defects in packaged filaments, comprising:

[0008] Acquire multiple collected images of the area to be inspected of the packaged filament;

[0009] intercepting a target image corresponding to the inspected position in the area to be inspected from the acquired image;

[0010] Determine the continuous long curve and multiple short lines on the continuous long curve in the target image;

[0011] Divide the target image into multiple sub-images, and determine the line geometry parameters of the short lines in each sub-image and the image geometry parameters of the sub-image;

[0012] Whether there is a hair defect in the filament package is determined according to the line geometric parameters and the image geometric parameters.

[0013] Further, the inspected position includes a first conical surface of the packaged filament, and acquiring a plurality of collected images of the area to be inspected of the packaged filament includes:

[0014] A plurality of collected images including the first cone surface are respectively acquired along the circumferential direction of the packaged filament.

[0015] Furthermore, intercepting a target image corresponding to the inspected position in the area to be inspected in the acquired image includes:

[0016] obtaining a copy image of the acquired image;

[0017] De-noising the duplicated image, and using an image segmentation algorithm to obtain a plurality of edge lines in the de-noised duplicated image;

[0018] Determine multiple edge lines of the inspected position according to the multiple edge lines, and obtain coordinate values ​​of all pixel points of the multiple edge lines;

[0019] The target image is intercepted in the acquired image according to the coordinate values ​​of all pixel points of the plurality of edge lines.

[0020] Further, the plurality of edge lines of the inspected position are determined according to the plurality of side lines, including:

[0021] Determine a first sideline, a second sideline, and a third sideline of the first conical surface according to the plurality of sidelines, and determine a first intersection point of the first sideline with the second sideline, and a second intersection point of the first sideline with the third sideline;

[0022] Determine a reference line according to the first intersection point and the second intersection point, and rotate the copied image so that the reference line is set horizontally;

[0023] The coordinate values ​​of all pixel points of the rotated first intersection point, the rotated second intersection point and the rotated first edge line are determined, and a plurality of edge lines of the inspected position are determined.

[0024] Further, determining the continuous long curve and the plurality of short lines on the continuous long curve in the target image includes:

[0025] Perform denoising on the target image;

[0026] Filter the background pattern of the denoised target image to obtain an image to be analyzed;

[0027] The continuous long curves and multiple short lines in the image to be analyzed are determined using the dilation processing method.

[0028] Further, filtering the background pattern of the denoised target image to obtain an image to be analyzed includes:

[0029] Using an edge detection algorithm to determine pseudo edges of the inspected locations within the denoised target image;

[0030] Remove false edges in the target image;

[0031] The background of the target image with the pseudo-edges removed is filtered using a binarization method to obtain the image to be analyzed.

[0032] Further, using the dilation processing method to determine the continuous long curves and the multiple short lines in the image to be analyzed includes:

[0033] The region smaller than the predetermined area between each two adjacent first patterns in the image to be analyzed is filled by using the expansion processing method to obtain a continuous long curve and a plurality of short lines.

[0034] Furthermore, the line geometric parameters of the short line include the line length, the line width, and the angle between the short line and the vertical direction, and the image geometric parameters of the sub-image include the sub-image length.

[0035] Further, determining whether the packaged filament has a hair defect according to the line geometric parameters and the image geometric parameters includes:

[0036] Set a short line whose line length meets the first preset threshold, whose line width meets the second preset threshold, and whose angle meets the third preset threshold as a marking object;

[0037] Determining the sum of the number of marked objects in the sub-images of the same region of the plurality of acquired images;

[0038] Whether the packaged filaments have a filament defect is determined according to whether a first ratio of the sum of the quantities to the sub-image length meets a fourth preset threshold.

[0039] Further, after determining whether the target image has a hair defect, the method further includes:

[0040] When it is determined that the packaged filaments have hair defects, the hair defect level of the packaged filaments is determined according to the first ratio.

[0041] The method for detecting fuzzy defects in packaged filaments according to an embodiment of the present invention can be applied to a detection device on a production line, and image capture is performed on each packaged filament passing through a detection station to obtain a corresponding captured image. The image is then processed to extract defect features for determining fuzzy defects (i.e., geometric feature parameters of multiple short lines on a continuous long curve), and based on the properties of the defect features of different fuzzy defects, it is determined whether the packaged filaments have fuzzy defects.

[0042] The method for detecting fuzzy defects of packaged filaments of the embodiment of the present invention is suitable for detecting fuzzy defects of packaged filaments in the chemical fiber field. It can quickly analyze the appearance defects of packaged filaments, reduce errors caused by manual visual inspection, reduce false detection rate, improve detection efficiency, and reduce production costs through high-precision analysis and processing of images. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a method for detecting hairy yarn defects in packaged filaments according to an embodiment of the present invention;

[0044] Figure 2 This is a schematic structural diagram of a packaged filament according to an embodiment of the present invention;

[0045] Figure 3 is a schematic diagram of collecting images according to an embodiment of the present invention;

[0046] Figure 4 for Figure 1 Flow chart of the specific method of step S120;

[0047] Figure 5 for Figure 4 Flow chart of the specific method of step S123;

[0048] Figure 6 is a schematic diagram of a target image according to an embodiment of the present invention;

[0049] Figure 7 for Figure 1 Flow chart of the specific method of step S130;

[0050] Figure 8 for Figure 1 A flowchart of a specific method of an embodiment of step S150;

[0051] Fig. 9 for Figure 1 A flowchart of a specific method of another embodiment of step S150 in FIG. DETAILED DESCRIPTION

[0052] The structure and working principle of the embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0053] like Figure 1 As shown, a method for detecting filament hair defects in a packaged filament according to an embodiment of the present invention comprises:

[0054] S110, acquiring a plurality of collected images of the area to be detected of the packaged filament;

[0055] S120, intercepting a target image corresponding to the inspected position in the area to be inspected in the acquired image;

[0056] S130, determining a continuous long curve and a plurality of short lines on the continuous long curve in the target image;

[0057] S140, dividing the target image into a plurality of sub-images, and determining line geometry parameters of short lines in each sub-image and image geometry parameters of the sub-image;

[0058] S150, determining whether the packaged filaments have a hair defect according to the line geometry parameters and the image geometry parameters.

[0059] In the embodiment of the present invention, Figure 2 As shown, the inspected position includes the first conical surface 1 of the packaged filament, and the multiple acquired images of the area to be inspected of the packaged filament include:

[0060] A plurality of collected images including the first cone surface 1 are acquired respectively along the circumferential direction of the packaged filament.

[0061] In the embodiment of the present invention, since the hair defect will appear simultaneously on the first conical surface 1 at the upper end of the packaged filament, the second conical surface 2 at the lower end of the packaged filament, and the cylindrical surface 3 of the packaged filament, any one of the surfaces can be selected for detection. Since the hair defect can be more easily determined based on the first conical surface 1, in the embodiment of the present invention, the first conical surface 1 can be selected as the inspection position, and the area of ​​the packaged filament including the first conical surface 1 is used as the area to be detected.

[0062] Therefore, a surface array camera can be set at a position corresponding to the area to be detected. The surface array camera can be set perpendicular to the first cone surface 1 and can obtain Figure 3 The captured image shown. The use of an area array camera can reduce the number of shots and save time. In addition, since the position of the area array camera is fixed and the shooting angle is limited, the image capture of the entire circumference of the filament package can be completed by rotating the filament package and capturing the captured image of the area to be detected while it rotates one circle. In one embodiment, the shooting time can be 2.2s, and 30 captured images of the area to be detected are obtained at equal time intervals during the shooting time.

[0063] Since low-angle lighting can highlight defect features higher than hair defects on the surface of the inspection object, in an embodiment of the present invention, a low-angle lighting source is used to illuminate the area to be inspected, and the light of the light source is collected in the same direction as the area array camera.

[0064] In the embodiment of the present invention, Figure 4 As shown, the target image corresponding to the inspected position in the area to be inspected in the intercepted acquired image includes:

[0065] S121, obtaining a copy image of the collected image;

[0066] S122, performing denoising processing on the copied image, and obtaining a plurality of edge lines in the denoised copied image using an image segmentation algorithm;

[0067] S123, determining multiple edge lines of the inspected position according to the multiple edge lines, and obtaining coordinate values ​​of all pixel points of the multiple edge lines;

[0068] S124. Capturing a target image in the acquired image according to the coordinate values ​​of all pixel points of the plurality of edge lines.

[0069] Among them, Figure 5 As shown, the multiple edge lines of the inspected position determined according to the multiple edge lines in step S123 may include:

[0070] S210, determining a first sideline 13, a second sideline 11, and a third sideline 12 of the first conical surface 1 according to the plurality of sidelines, and determining a first intersection point 14 of the first sideline 13 and the second sideline 11, and a second intersection point 15 of the first sideline 13 and the third sideline 12;

[0071] S220, determining a reference line 16 according to the first intersection point 14 and the second intersection point 15, and rotating the copied image so that the reference line 16 is set horizontally;

[0072] S230, determining coordinate values ​​of all pixel points of the rotated first intersection point 14, the rotated second intersection point 15 and the rotated first edge line 13, and determining a plurality of edge lines of the inspected position.

[0073] In the embodiment of the present invention, by intercepting the target image in the acquired image, interference from non-processing objects in the image can be eliminated. By adopting the above steps, the situation where the inspected position is deviated or the inspected position range is different can be handled, thereby improving the accuracy of image interception.

[0074] In an embodiment of the present invention, the captured image may be copied first to obtain a copied image, and the copied image may be processed to obtain an edge line, thereby preventing the subsequent processing of the original captured image from being affected by the processing of the original captured image.

[0075] Then, the copied image can be denoised by filtering. In this embodiment, the filtering method can preferably be Gaussian filtering. The denoised copied image can obtain multiple edge lines by an image segmentation algorithm. The image segmentation algorithm can be an edge detection algorithm or a binarization method, and the edge line can be a sudden change position of the grayscale brightness change of the edge of the first conical surface 1 of the packaged filament.

[0076] Secondly, the edge line in the copied image closest to the top of the image can be selected as the first edge line 13, the edge line in the copied image closest to the left of the image can be selected as the second edge line 11, and the edge line in the copied image closest to the right of the image can be selected as the third edge line 12, and the first intersection point 14 of the first edge line 13 and the second edge line 11, and the second intersection point 15 of the first edge line 13 and the third edge line 12 can be determined, and at the same time, the coordinates (x1, y1) of the first intersection point 14 and the coordinates (x2, y2) of the second intersection point 15 can be obtained.

[0077] Again, the first intersection 14 and the second intersection 15 are connected to obtain a reference line 16. The angle φ between the reference line 16 and the horizontal reference line 16 is measured, and the copied image is rotated with the first intersection 14 as the rotation axis according to the angle φ of the reference line 16, so that the reference line 16 is set horizontally. At this time, the coordinates of the second intersection 15 become (x3, y1).

[0078] Next, the average y coordinate value of all pixel points of the rotated first edge line 13 is obtained, and the edge line coordinate y3 of the top of the first edge line 13 is obtained by adding the average y coordinate value and the preset coordinate difference value, and the edge line coordinate y4 of the bottom of the first edge line 13 is obtained by subtracting the average y coordinate value from the preset coordinate difference value. Therefore, the coordinate values ​​of all pixel points in the left edge line x=x1, the right edge line x=x3, the upper edge line y=y3, and the lower edge line y=y4 in the copied image can be obtained.

[0079] Finally, the coordinate values ​​of all the pixel points of the edge line in the copy image can be brought into the acquired image, and the coordinate values ​​of the same pixel points in the acquired image can be extracted. The extracted coordinate values ​​can be used to obtain multiple segmentation edge lines of the acquired image, including the left edge line, the right edge line, the upper edge line, and the lower edge line. Then, the image within the segmentation edge line is intercepted to obtain the following: Figure 6 Target image shown.

[0080] In the embodiment of the present invention, Figure 7 As shown, determining the continuous long curve in the target image and the multiple short lines on the continuous long curve includes:

[0081] S131, performing denoising processing on the target image;

[0082] S132, filtering the background pattern of the target image after denoising to obtain an image to be analyzed;

[0083] S133. Determine the continuous long curves and multiple short lines in the image to be analyzed by using a dilation processing method.

[0084] In an embodiment of the present invention, the noise points in the target image may be analyzed first, and then the target image may be filtered according to the existing noise points. Since the noise points of the target image are concentrated in the high-frequency region, the filtering process may use a Butterworth low-pass filter. The Butterworth low-pass filter has a stable amplitude characteristic and can obtain an image with isolated noise points eliminated, thereby improving the deterioration of the image quality caused by various interference factors during the process of the area array camera collecting the original signal to the signal conversion and signal transmission.

[0085] In an embodiment of the present invention, filtering the background pattern of the denoised target image to obtain the image to be analyzed includes:

[0086] Using an edge detection algorithm to determine pseudo edges of the inspected locations within the denoised target image;

[0087] Remove false edges in the target image;

[0088] The background of the target image with the pseudo-edges removed is filtered using a binarization method to obtain the image to be analyzed.

[0089] Specifically, edge detection can be performed using the Prewitt operator + logarithmic form or the Roberts operator + logarithmic form to obtain several edge lines. Then, according to the grayscale difference between the upper and lower and left and right neighboring points of the pixel point, the edge is detected at the extreme value at the edge, and some pseudo edges are removed, which can make the image enhancement effect obvious without losing the original details of the target image, so as to achieve the best image enhancement effect.

[0090] In the embodiment of the present invention, the target image after edge detection can be binarized according to a preset optimal threshold value, thereby obtaining an image with only red and black colors to remove background interference. The optimal threshold value can be set to 90-95.

[0091] In an embodiment of the present invention, determining the continuous long curve and the plurality of short lines in the image to be analyzed by using the dilation processing method includes:

[0092] The region smaller than the predetermined area between each two adjacent first patterns in the image to be analyzed is filled by using the expansion processing method to obtain a continuous long curve and a plurality of short lines.

[0093] In the embodiment of the present invention, the first patterns are connected by the expansion processing method to obtain a continuous long curve and a plurality of short lines, which can avoid unnecessary interference to the image to be analyzed after filtering the background pattern. Specifically, the image to be analyzed can be divided into multiple blocks, and then the red color image can be connected according to the red image of the predetermined area to obtain a continuous long curve and a plurality of short lines. Since the red image of the predetermined area can effectively fill the area between each two adjacent first patterns that is smaller than the predetermined area, the boundary of the line can be smoothed without obviously changing the line, which is convenient for the subsequent extraction of the geometric parameters of the line.

[0094] In the embodiment of the present invention, the line geometric parameters of the short line include the line length, the line width, and the angle between the short line and the vertical direction, and the image geometric parameters of the sub-image include the sub-image length.

[0095] In one embodiment, each target image may be divided into three sub-images, namely a first sub-image, a second sub-image and a third sub-image.

[0096] In the embodiment of the present invention, Figure 8 As shown, determining whether there is a hair defect in the packaged filament according to the line geometry parameters and the image geometry parameters includes:

[0097] S151, setting a short line whose line length meets a first preset threshold, whose line width meets a second preset threshold, and whose angle meets a third preset threshold as a marking object;

[0098] S152, determining the total number of marked objects in the sub-images of the same area of ​​the multiple acquired images;

[0099] S153: Determine whether the filament package has a filament defect according to whether a first ratio of the sum of the quantities to the sub-image length meets a fourth preset threshold.

[0100] In the embodiment of the present invention, the first preset threshold value ≥ 2 mm, the second preset threshold value ≥ 0.04 mm, and the third preset threshold value can be set to be -20° to 20°. If all the line geometric parameters of the short line do not meet the above thresholds, that is, the line length is less than 2 mm or the line width is less than 0.04 mm or the angle is between 0° and 20°, it is considered not to be a marked object.

[0101] Taking the case where the number of acquired images is 30 and the target image corresponding to each acquired image is divided into 3 sub-images, after the marked objects in each sub-image are determined, the number of marked objects in each sub-image is obtained, and then the total number of marked objects in the 30 first sub-images, 30 second sub-images and 30 third sub-images is obtained respectively.

[0102] Then, taking the sum of the number of marked objects in 30 first sub-images as an example, the first ratio of the sum of the number and the length of the first sub-image can be compared with a preset fourth preset threshold to determine whether there is a hair defect in the packaged filament.

[0103] Taking AAA grade packaged filament as an example, the fourth preset threshold value can be set to ≤1.8. When the first ratio meets the fourth preset threshold value, it is determined that there is no hair defect. When the first ratio does not meet the fourth preset threshold value, it is determined that there is a hair defect.

[0104] In another embodiment of the present invention, Fig. 9 As shown, after determining whether the target image has a hair defect, the following steps are also included:

[0105] S154, when it is determined that the packaged filaments have fuzz defects, determining the fuzz defect level of the packaged filaments according to the first ratio.

[0106] In one embodiment of the present invention, 30 images are collected, and the target image corresponding to each collected image is divided into 3 sub-images. The detection of hairy defects in AAA-grade packaged filament is taken as an example. When the first ratio is ≤1.8, there is no hairy defect; when the first ratio ranges from 1.8 to 2.4, it is a hairy defect that is downgraded from grade A; when the first ratio is ≥2.4, it is a hairy defect that is downgraded from grade A.

[0107] If one or more of the first ratios of the 30 first sub-images, the 30 second sub-images and the 30 third sub-images meet the B-level downgrade requirement, the first conical surface 1 of the packaged filament is judged to be B-level downgrade; if 0 meet the B-level downgrade and at the same time, 1 or more meet the A-level downgrade requirement, the first conical surface 1 of the packaged filament is judged to be A-level downgrade; if 0 meet the B-level downgrade and 0 meet the A-level downgrade, the first conical surface 1 of the packaged filament is judged to be AAA-level, i.e., there is no hair defect.

[0108] In one embodiment of the present invention, 30 images are collected, and the target image corresponding to each collected image is divided into 3 sub-images. The detection of hairy defects in AAA-grade packaged filament is taken as an example. When the first ratio is ≤2.1, there is no hairy defect; when the first ratio is in the range of 2.1-2.7, it is a hairy defect that is downgraded from Grade A; when the first ratio is ≥2.7, it is a hairy defect that is downgraded from Grade A.

[0109] If one or more of the first ratios of the 30 first sub-images, the 30 second sub-images, and the 30 third sub-images meet the B-grade downgrade, the first conical surface 1 of the packaged filament is judged to be B-grade downgrade; if 0 meet the B-grade downgrade and at the same time, 1 or more meet the A-grade downgrade, the first conical surface 1 of the packaged filament is judged to be A-grade downgrade; if 0 meet the B-grade downgrade and 0 meet the A-grade downgrade, the first conical surface 1 of the packaged filament is judged to be AA-grade, that is, there is no fuzzy defect. According to the fuzzy defect detection method of the packaged filament of the embodiment of the present invention, the correct detection rate of fuzzy defects can reach 98.2%, and the efficiency is 4 seconds / spindle, which is better than the manual visual inspection method, realizes the online automatic detection of the appearance quality of the packaged filament, and reduces the manual workload by 95%.

[0110] In summary, the method for detecting fuzzy defects of packaged filaments of the present invention can be widely applied to the online detection of fuzzy defects of packaged filaments in the field of chemical fibers, and is easy to analyze appearance defects. Moreover, by extracting and judging the characteristics of fuzzy defects, fuzzy defects can be detected, and the defect level can be automatically identified, thereby improving the detection accuracy and reducing the error caused by manual visual inspection. At the same time, in the image processing process, interference can be eliminated, fuzzy defects can be quickly identified, and the false detection rate can be reduced.

[0111] The above is only an illustrative description of the present invention. Those skilled in the art should know that various improvements can be made to the present invention without departing from the working principle of the present invention, which all fall within the protection scope of the present invention.

Claims

1. A method for detecting filament defects in packaged filaments, characterized in that: include: A low-angle illumination light source is used to illuminate the area to be inspected of the packaged filament, and a fixed-position area array camera is used to collect images while the packaged filament rotates one circle, and obtain multiple collected images of the area to be inspected of the packaged filament; wherein the light of the low-angle illumination light source is in the same collection direction as that of the area array camera; intercepting a target image corresponding to the inspected position in the area to be inspected from the acquired image; Determine a continuous long curve in the target image and a plurality of short lines on the continuous long curve; Dividing the target image into a plurality of sub-images, and determining line geometry parameters of the short lines in each of the sub-images and image geometry parameters of the sub-images; Determining whether the packaged filament has a hairy filament defect according to the line geometric parameters and the image geometric parameters; The inspected position includes a first conical surface of the packaged filaments, and acquiring a plurality of acquired images of the to-be-inspected area of ​​the packaged filaments includes: acquiring a plurality of acquired images including the first conical surface respectively along the circumference of the packaged filaments; Intercepting the target image corresponding to the inspected position in the area to be inspected in the acquired image includes: obtaining a copy image of the acquired image; performing denoising on the copy image, and obtaining a plurality of edge lines in the denoised copy image using an image segmentation algorithm; determining a plurality of edge lines of the inspected position according to the plurality of edge lines, and obtaining coordinate values ​​of all pixel points of the plurality of edge lines; intercepting the target image in the acquired image according to the coordinate values ​​of all pixel points of the plurality of edge lines; Determining the multiple edge lines of the inspected position according to the multiple edge lines includes: determining the first edge line, the second edge line and the third edge line of the first conical surface according to the multiple edge lines, and determining the first intersection point of the first edge line with the second edge line, and the second intersection point of the first edge line with the third edge line; determining a reference line according to the first intersection point and the second intersection point, rotating the copied image so that the reference line is set horizontally; determining the coordinate values ​​of all pixel points of the rotated first intersection point, the rotated second intersection point and the rotated first edge line, and determining the multiple edge lines of the inspected position; Determining a continuous long curve in the target image and a plurality of short lines on the continuous long curve comprises: Performing denoising processing on the target image; Filtering the background pattern of the target image after denoising to obtain an image to be analyzed; The continuous long curve and the plurality of short lines in the image to be analyzed are determined by using a dilation processing method.

2. The method for detecting filament yarn defects in packaged filaments according to claim 1, characterized in that: Filtering the background pattern of the target image after denoising to obtain an image to be analyzed includes: Determining a pseudo edge of the detected position in the target image after denoising using an edge detection algorithm; Deleting the pseudo edge in the target image; The background of the target image with the pseudo edge deleted is filtered using a binarization method to obtain the image to be analyzed.

3. The method for detecting filament yarn defects in packaged filaments according to claim 2, characterized in that: Determining the continuous long curve and the plurality of short lines in the image to be analyzed by using a dilation processing method includes: The region smaller than a predetermined area between every two adjacent first patterns in the image to be analyzed is filled by using a dilation processing method to obtain the continuous long curve and the plurality of short lines.

4. The method for detecting filament yarn defects in packaged filaments according to claim 1, characterized in that: The line geometric parameters of the short line include the line length, the line width, and the angle between the short line and the vertical direction, and the image geometric parameters of the sub-image include the sub-image length.

5. The method for detecting filament yarn defects in packaged filaments according to claim 4, characterized in that: Determining whether the packaged filament has a hair defect according to the line geometric parameter and the image geometric parameter includes: Set a short line whose line length meets the first preset threshold, whose line width meets the second preset threshold, and whose angle meets the third preset threshold as a marking object; Determining the sum of the numbers of the marked objects in the sub-images of the same area of ​​the plurality of acquired images; Whether the packaged filament has a filament defect is determined according to whether a first ratio of the sum of the quantities to the length of the sub-image meets a fourth preset threshold.

6. The method for detecting filament yarn defects in packaged filaments according to claim 5, characterized in that: After determining whether the target image has a hair defect, the method further includes: When it is determined that the packaged filaments have fuzz defects, the fuzz defect level of the packaged filaments is determined according to the first ratio.

Images