Intelligent detection method of filament quality based on image processing

Through image processing technology, the grayscale image of the filament surface is solved, and the problem of manual inspection is time-consuming and laborious and poor accuracy in the prior art is solved, and the automation and high accuracy of filament quality detection is realized, ensuring the reliability of the detection results.

CN118644491BActive Publication Date: 2024-11-08SUZHOU FENGHUA XUEYUE INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411125369.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-08
Estimated Expiration
2044-08-16

AI Technical Summary

Technical Problem

In the prior art, the detection of filament quality depends on manual observation electron microscope, which is time-consuming and laborious, and has poor accuracy, and cannot effectively detect defects in the filament surface morphology.

Method used

Using an image-based processing method, by collecting grayscale images of the filament surface, analyzing the filament frame curve, spiral uniformity, reflective area and tungsten filament diameter, comprehensively calculating the appearance quality of the filament to achieve automated detection.

Benefits of technology

The accuracy and efficiency of filament quality detection are improved, and the winding uniformity, reflection conditions and diameter uniformity of the filament can be accurately identified, ensuring the reliability of the detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118644491B_ABST
    Figure CN118644491B_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of image recognition and processing, and in particular to an intelligent filament quality detection method based on image processing. The method comprises: collecting a grayscale image of a filament surface; obtaining a filament area in the grayscale image of the filament surface, and obtaining a filament skeleton curve according to the filament area; obtaining a filament spiral uniformity according to the filament skeleton curve; obtaining a suspected reflective area pixel point according to the filament area; obtaining a filament foot area and a filament spiral area in the filament area; obtaining a filament foot appearance qualification degree according to all suspected reflective area pixel points contained in the filament foot area; obtaining a tungsten filament diameter qualification degree according to all suspected reflective area pixel points contained in the filament spiral area; comprehensively obtaining the tungsten filament diameter qualification degree, the filament foot appearance qualification degree and the filament spiral uniformity to obtain a filament appearance quality qualification degree; obtaining a filament quality detection result according to the filament appearance quality qualification degree, and improving the accuracy of quality detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of image recognition and processing, and in particular to an intelligent detection method for filament quality based on image processing. Background Art

[0002] Filament refers to the light-emitting element in lighting equipment such as light bulbs. Since its birth, filament has become one of the necessities of human production and life. With the rapid development of modern artificial intelligence, Internet of Things and other technologies, humans are improving traditional filaments while also developing new filament materials. The most common materials are tungsten and molybdenum. The research and development of new materials not only increases the life of the filament, but is also more environmentally friendly.

[0003] At present, the quality inspection methods for filaments mainly inspect the resistance value, surface morphology, internal structure and material composition of the filaments. When inspecting the surface morphology of the filaments, the defects of the surface morphology of the filaments are often identified by observing them under an electron microscope. This method is not only time-consuming and labor-intensive, but also relies on human subjectivity, and the accuracy of quality inspection is poor. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an intelligent filament quality detection method based on image processing to solve the existing problems.

[0005] The intelligent filament quality detection method based on image processing of the present invention adopts the following technical solutions:

[0006] An embodiment of the present invention provides an intelligent filament quality detection method based on image processing, the method comprising the following steps:

[0007] Collect grayscale image of filament surface;

[0008] Obtain the filament area in the grayscale image of the filament surface, and obtain the filament skeleton curve according to the filament area; obtain the spiral area pixel points according to each pixel on the filament skeleton curve; obtain the central intersection point according to all the spiral area pixel points, and obtain the filament spiral uniformity by combining the Euclidean distance between the central intersection points; obtain the suspected reflective area pixel points according to the grayscale value of each pixel point in the filament area; obtain the filament surface grayscale template image, match the filament area with the filament surface grayscale template image, and obtain the filament foot area and the filament spiral area; obtain the suspected reflective area according to all the suspected reflective area pixel points contained in the filament foot area, and obtain the filament foot appearance qualification by combining the centroid of each suspected reflective area; obtain the suspected spiral reflective area according to all the suspected reflective area pixel points contained in the filament spiral area, and obtain the tungsten filament diameter qualification according to all the suspected spiral reflective areas; comprehensively consider the tungsten filament diameter qualification, the filament foot appearance qualification and the filament spiral uniformity to obtain the filament appearance quality qualification;

[0009] The test result of the filament quality is obtained according to the qualified degree of the filament appearance quality.

[0010] Further, the step of acquiring the filament region in the grayscale image of the filament surface and obtaining the filament skeleton curve according to the filament region includes:

[0011] The grayscale image of the filament surface is segmented using the region growing method to obtain a binary image; the foreground area in the binary image is used as the filament area;

[0012] The k3M skeleton extraction algorithm is used on the filament area to obtain the filament skeleton curve.

[0013] Further, the step of obtaining the pixel points of the spiral region according to each pixel on the filament skeleton curve includes:

[0014] Calculate the second-order derivative at each pixel point on the filament skeleton curve; count the number of pixels where the second-order derivative is zero, and record it as the number of turns;

[0015] Calculate the curvature of the filament skeleton curve at each pixel point; arrange the curvatures of all pixel points on the filament skeleton curve in descending order to obtain a curvature change vector; take the pixel points corresponding to the first q elements in the curvature change vector as turning points; where q is the number of turns;

[0016] The Euclidean distance between any two turning points is calculated, and the pixel points on the filament skeleton curve between the two turning points with the largest Euclidean distance are recorded as the spiral region pixel points.

[0017] Furthermore, obtaining the central intersection points according to all the pixels in the spiral area and obtaining the filament spiral uniformity in combination with the Euclidean distance between the central intersection points includes:

[0018] The least square method is used to fit all the pixels in the spiral area with the straight line to obtain the center line of the spiral area; the pixel points in the spiral area located on the center line of the spiral area are recorded as the center intersection points;

[0019] For the hth center intersection point, calculate the Euclidean distance between the hth center intersection point and the h+1th center intersection point; calculate the mean of the Euclidean distances of all center intersection points; obtain the difference between the Euclidean distance between the hth center intersection point and the h+1th center intersection point and the mean; calculate the square of the difference; obtain the mean of the squares of all center intersection points, recorded as the first mean; calculate the sum of the first mean and the number 1; and use the ratio of the number 1 to the sum as the filament spiral uniformity.

[0020] Furthermore, obtaining the pixel point of the suspected reflection area according to the gray value of each pixel point in the filament area includes:

[0021] The filament area is segmented by using a dual-threshold segmentation algorithm to obtain a first segmentation threshold and a second segmentation threshold; wherein the first segmentation threshold is less than the second segmentation threshold; and pixels in the filament area whose grayscale values ​​are greater than or equal to the second segmentation threshold are recorded as pixels in the suspected reflective area.

[0022] Further, the step of acquiring the grayscale template image of the filament surface, matching the filament area with the grayscale template image of the filament surface, and obtaining the filament foot area and the filament spiral area includes:

[0023] A grayscale template image of the filament surface is obtained, and the filament area is matched with the grayscale template image of the filament surface using a BBS image matching algorithm to obtain a filament foot area within the filament area; all areas within the filament area except the filament foot area are taken as the filament spiral area.

[0024] Further, obtaining the suspected reflective area according to all the suspected reflective area pixels contained in the filament foot area, and obtaining the appearance qualification of the filament foot in combination with the centroid of each suspected reflective area, includes:

[0025] Performing connected domain analysis on all the pixels of the suspected reflective area contained in the filament foot area to obtain each connected domain, and taking each connected domain as each suspected reflective area; counting the number of suspected reflective areas; and obtaining the centroid of each suspected reflective area;

[0026] Record any suspected reflective area as the area to be analyzed, calculate the absolute value of the difference between the centroid of the area to be analyzed and the centroid of each suspected reflective area on the y-axis; obtain the minimum value of all the absolute values ​​of the difference of the area to be analyzed; obtain the mean of the minimum values ​​of all suspected reflective areas, recorded as the second mean; calculate the difference between the minimum value and the second mean; obtain the sum of the squares of the differences of all suspected reflective areas, recorded as the first sum of squares;

[0027] Obtain the mean of the x-axis coordinates of the centroids of all suspected reflective areas, recorded as the third mean; calculate the difference between the x-axis coordinate of the centroid of the area to be analyzed and the third mean, recorded as the first difference; obtain the sum of the squares of the first differences of all suspected reflective areas, recorded as the second sum of squares;

[0028] Obtain the sum of the first square sum and the second square sum; calculate the ratio of the sum to the number of suspected reflective areas; obtain the sum of the ratio and the number 1, recorded as the first sum; and use the ratio of the number 1 to the first sum as the appearance qualification of the filament foot.

[0029] Further, obtaining the suspected spiral reflective area according to all the suspected reflective area pixel points contained in the filament spiral area, and obtaining the tungsten filament diameter qualification according to all the suspected spiral reflective areas, includes:

[0030] Performing connected domain analysis on all pixel points of the suspected reflective area contained in the filament spiral area to obtain each connected domain, and using each connected domain as each suspected spiral reflective area;

[0031] For each suspected spiral reflective area, obtain the minimum circumscribed rectangle of the suspected spiral reflective area, and use the width of the minimum circumscribed rectangle as the tungsten wire diameter corresponding to the suspected spiral reflective area; obtain the mean of the tungsten wire diameters corresponding to all suspected spiral reflective areas; calculate the difference between the tungsten wire diameter corresponding to each suspected spiral reflective area and the mean; obtain the square of the difference; calculate the mean of the squares of all suspected spiral reflective areas, recorded as the fourth mean; obtain the sum of the fourth mean and the number 1; and use the ratio of the number 1 to the sum as the tungsten wire diameter qualification degree.

[0032] Furthermore, the comprehensive tungsten filament diameter qualification, filament foot appearance qualification and filament spiral uniformity are used to obtain the filament appearance quality qualification, including:

[0033] The sum of the tungsten filament diameter qualification, the filament foot appearance qualification and the filament spiral uniformity is calculated, and the ratio of the sum to the number 3 is taken as the filament appearance quality qualification.

[0034] Further, the test result of the filament quality is obtained according to the qualified degree of the filament appearance quality, including:

[0035] If the filament appearance quality qualification is greater than the preset quality threshold, the filament quality is qualified; otherwise, the filament quality is unqualified.

[0036] The present invention has at least the following beneficial effects:

[0037] The present invention processes the collected grayscale image of the filament surface by analyzing the morphological changes when there are quality problems in the filament, and analyzes the quality of the filament; firstly, the qualified filaments are twisted into a spiral shape after being wound together, the filament skeleton curve in the grayscale image of the filament surface is identified, and the uniformity of the filament spiral is calculated according to the filament skeleton curve, reflecting the uniformity of the filament winding; further, the reflection of the filament in different position areas is analyzed, and the filament foot area and the filament spiral area are identified; by analyzing the vertical shape of the filament foot, the appearance qualification of the filament foot is obtained according to the difference between the reflection positions of the filament foot area, reflecting the overall qualification of the filament foot shape; according to the reflection area of ​​the filament spiral area, the tungsten filament diameter corresponding to each restaurant position is obtained, and the tungsten filament diameter qualification is calculated according to the difference in tungsten filament diameter, reflecting the uniformity of the thickness of the filament; finally, the tungsten filament diameter qualification, the filament foot appearance qualification and the filament spiral uniformity are comprehensively considered to obtain the filament appearance quality qualification, the quality of the filament appearance is detected, and the detection accuracy is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flowchart of the steps of the intelligent detection method of filament quality based on image processing provided by the present invention;

[0040] Figure 2 Schematic diagram of the center line of the spiral area. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of the intelligent detection method of filament quality based on image processing proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the intelligent filament quality detection method based on image processing provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] An embodiment of the present invention provides an intelligent filament quality detection method based on image processing. Specifically, the following intelligent filament quality detection method based on image processing is provided. Figure 1 , the method comprises the following steps:

[0045] Step S001, collecting the RGB image of the surface of the filament by a CCD camera, and performing preprocessing to obtain a grayscale image of the surface of the filament.

[0046] A CCD camera is used to obtain an image of the filament surface to obtain an RGB image of the filament surface. In the process of obtaining the image of the filament surface, a ring-shaped LED light source is set above the filament surface to enhance the characteristics of the filament surface in subsequent images. The RGB image of the filament surface is processed to reduce the influence of external noise and light. In this embodiment, the RGB image of the filament surface is denoised by Gaussian filtering. Gaussian filtering is a well-known technology and will not be repeated here. The implementer may use other denoising methods. The RGB of the filament surface after denoising is grayed to obtain a grayscale image F of the filament surface.

[0047] Step S002, obtain the filament area in the grayscale image of the filament surface, obtain the filament skeleton curve according to the filament area; obtain the filament spiral uniformity according to the filament skeleton curve; obtain the filament foot area and the filament spiral area in the filament area; obtain the filament foot appearance qualification according to the filament foot area; obtain the tungsten filament diameter qualification according to the filament spiral area; comprehensively consider the tungsten filament diameter qualification, the filament foot appearance qualification and the filament spiral uniformity to obtain the filament appearance quality qualification.

[0048] During the filament production process, if the diameter of the filament is uneven, the light emission effect will be uneven during use, which will affect the brightness and life of the filament. Uneven diameter can also cause problems such as filament breakage during use. If the filament is more uneven, it will also break during the later processing of the filament. For example, after the filament is drawn, it will break during the cutting and welding process. When cleaning the filament later, the dirt and oxides on the surface of the filament may not be cleaned due to improper operation and other problems, which will affect the quality of the filament.

[0049] The filament is usually composed of two tungsten wires, which are twisted into a spiral after being intertwined with each other. The surface of a qualified filament is smooth, and a reflective area will be produced on the surface of the filament under the light.

[0050] For images To facilitate subsequent analysis, the image is segmented by the region growing method. Ten pixels are randomly selected from the image F as the starting points. The growth criterion is that the absolute value of the gray value difference between the next growing point and the current growing point is less than the threshold =6, and the stop condition is to traverse all the pixels in the image F. According to the result of region growing, the foreground region and the background region in the binary image BF are obtained, and the foreground region is used as the filament region. The specific implementation process of the region growing method is a well-known technology and will not be repeated here.

[0051] For the filament area in the binary image BF, The skeleton extraction algorithm processes the filament area in the binary image BF and identifies the filament skeleton curve. , which is the skeleton of the filament; The skeleton extraction algorithm is a well-known technology and will not be described in detail. , due to the filament skeleton curve It is obtained by performing skeleton recognition on the filament area. In the filament area, the middle part of the filament is screw-shaped, while the two legs of the filament are close to straight lines.

[0052] Therefore, in order to further identify the filament, the filament skeleton curve Segmentation and statistics of filament skeleton curves The number of pixels on , for the filament skeleton curve Pixel points on the filament skeleton curve The second-order derivative at each pixel. If the second-order derivative is zero, the pixel is the filament skeleton curve. The inflection point is the number of turns in the filament spiral; the number of pixels where the second-order derivative is zero is counted and recorded as the number of turns . Calculate pixel points Curvature , and so on, traverse the filament skeleton curve Arrange the curvatures of all the pixels on the filament skeleton curve from large to small to obtain the curvature change vector , . For the curvature change vector Since the curvature of the turning point of the filament spiral is large, the first element of the curvature change vector is taken. The pixel points corresponding to the elements are taken as turning points. The Euclidean distance between any two turning points is calculated, such as Figure 2 As shown, the filament skeleton curve The pixels between the two turning points with the largest Euclidean distance are recorded as the spiral region pixels. The least squares method is used to fit all the spiral region pixels with the straight line to obtain the spiral region center line. , will be located at the center line of the spiral area The pixels in the spiral area on the graph are recorded as the central intersection points, and the number of central intersection points is counted. , get the Euclidean distance between any adjacent central intersection points. When the spiral shape of the filament is a uniform spiral, the straight line The distance between adjacent intersections of the spiral lines is also uniform, and the uniformity of the filament spiral is calculated. , the formula is:

[0053]

[0054] in, is the filament spiral uniformity; is the Euclidean distance between the hth center intersection point and the h+1th center intersection point; is the mean of the Euclidean distances of all adjacent central intersections; is the number of central intersections. The more uniform the spiral shape of the filament is, the closer the uniformity of the filament spiral is to 1; conversely, the more uneven the spiral shape of the filament is, the smaller the uniformity of the filament spiral is.

[0055] For the pixels in the filament area in image F, the pixels in the filament area are divided into three categories according to the grayscale value: the first category is the pixels in the reflective area on the filament, the second category is the pixels in the shadow area on the filament, and the third category is the pixels in the color area of ​​the filament itself. The specific division method is as follows:

[0056] The filament area is segmented by a double threshold segmentation algorithm to obtain two segmentation thresholds. The first segmentation threshold and the second segmentation threshold ,in Less than , for the pixels in the filament area, the grayscale value is less than or equal to The pixels are identified as suspected shadow area pixels, and the gray value is and The pixels between are identified as suspected filament color area pixels, and the grayscale values ​​are greater than or equal to The pixel point is identified as a pixel point in the suspected reflective area; the dual threshold segmentation algorithm is a well-known technology and will not be described in detail.

[0057] Analyze the pixels in the suspected reflective area. When the filament is good, that is, the thickness is uniform, it is not broken, and there is no oxidation or pollution on the filament, the bright area in the filament is evenly distributed and the direction of the bright area is consistent. Segment the filament area in image F through BBS image matching. The specific segmentation method is as follows:

[0058] In order to further identify the filament foot area, a straight filament is selected, and the cut length is kept consistent with the filament foot length. Different bulbs have different filament foot length standards, and the implementer can select the standard length according to the actual situation. The cut filament is placed vertically and its image is obtained in the same way. The image is grayed to obtain the grayscale template image of the filament surface.

[0059] Image F is matched with the grayscale template image of the filament surface through BBS image matching. The specific matching process is a well-known technology and will not be described in detail. The filament foot area in the filament area in image F is obtained through BBS image matching, and the unsuccessful matching area in the filament area is set as the filament spiral area. The centroid coordinates of each filament foot area are obtained; the connected domain analysis is performed on all the pixels of the suspected reflective area contained in the filament foot area to obtain each connected domain, and each connected domain is used as each suspected reflective area, and the number of suspected reflective areas is counted. . Get the centroid coordinates of each suspected reflective area.

[0060] Since the filament foot area as a whole is usually in the shape of a vertical straight line, the suspected reflective areas in the filament foot area are distributed on the same vertical line, that is, the centroid coordinates of the suspected reflective areas are The axis direction remains consistent, and The distance between the axis coordinates is also relatively uniform, and the distance between the suspected reflective areas is calculated. The difference of the axis centroid coordinates, randomly select a suspected reflective area, and find its centroid Axis coordinates and other suspected reflective area centroids The absolute value of the axis coordinate difference is set to the centroid of the other suspected reflective areas. The minimum absolute value of the axis coordinate difference is , and so on, traverse all the suspected reflective areas in the filament foot area, and find the corresponding If the filament foot area is in good condition, that is, there is no oxidation, pollution, breakage, etc., then the corresponding reflective area is distributed on a vertical line, that is, the centroid The coordinates of the axes remain consistent, and the adjacent reflective areas The difference between the axis coordinates is equal, and the shape of the filament foot is a straight line as a whole. , the formula is:

[0061]

[0062] in, The appearance qualification of the filament foot; is the x-axis coordinate of the centroid of the vth suspected reflection area in the filament foot area; is the mean x-axis coordinate of the centroid of all suspected reflective areas in the filament foot area; The minimum absolute value of the difference between the y-axis coordinate of the centroid of the vth suspected reflective area in the filament foot area and the y-axis coordinates of the centroids of other suspected reflective areas; The average of the minimum absolute values ​​of the differences between the y-axis coordinates of the centroids of all suspected reflective areas in the filament foot area and the y-axis coordinates of the centroids of other suspected reflective areas; The number of suspected reflective areas. If the filament pin is in good condition, without breakage, oxidation or contamination, the pass rate of the filament pin bend is close to 1.

[0063] For the pixel points in the filament spiral area, the suspected reflective area of ​​the filament spiral area is obtained in the same way. The filament is made of two tungsten wires. In the spiral area, a part of the filament exposed on the surface will produce all reflections, while the exposed parts of the filament in other positions will not produce all reflections. Connected domain analysis is performed on all the pixel points in the suspected reflective area contained in the filament spiral area to obtain each connected domain, and each connected domain is used as each suspected spiral reflective area, and the number of suspected spiral reflective areas is counted. Obtain the minimum circumscribed rectangle of each suspected spiral reflection area, and use the width of the minimum circumscribed rectangle as the tungsten wire diameter corresponding to each suspected spiral reflection area. , the calculation formula is:

[0064]

[0065] in, The qualified degree of tungsten wire diameter; is the diameter of the tungsten wire corresponding to the jth suspected spiral reflection area; is the average value of the tungsten wire diameter corresponding to all suspected spiral reflection areas; The more uniform the diameter distribution in the filament, that is, the diameter of the tungsten wire remains consistent throughout the filament, the closer the qualified degree of the tungsten wire diameter is to 1.

[0066] Constructing the quality of filament appearance , the formula is:

[0067]

[0068] in, The quality of the filament appearance is acceptable; The qualified degree of tungsten wire diameter; The appearance qualification of the filament foot; is the uniformity of the filament spiral.

[0069] Filament appearance quality qualification It reflects the quality of the filament. The higher the quality of the filament, the higher the appearance quality of the filament. The higher the filament diameter meets the requirements, the uniform diameter of the filament, the surface free of scratches, fractures, oxidation and pollution, and the uniform spiral shape. The closer to 1.

[0070] At this point, the filament appearance quality is obtained .

[0071] Step S003, completing the inspection of the filament quality according to the quality qualification of the filament appearance.

[0072] In order to comprehensively identify the quality of the filament appearance, the quality threshold σ is set to 0.8. If the filament appearance quality qualification of the filament is greater than the quality threshold, the filament quality meets the requirements; otherwise, the filament quality does not meet the requirements and is treated as a defective product.

[0073] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0074] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0075] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifications to the technical solutions recorded in the aforementioned embodiments, or equivalent replacement of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An intelligent filament quality detection method based on image processing, characterized in that: The method comprises the following steps: Collect grayscale image of filament surface; Obtain the filament area in the grayscale image of the filament surface, and obtain the filament skeleton curve according to the filament area; obtain the spiral area pixel points according to each pixel on the filament skeleton curve; obtain the central intersection point according to all the spiral area pixel points, and obtain the filament spiral uniformity by combining the Euclidean distance between the central intersection points; obtain the suspected reflective area pixel points according to the grayscale value of each pixel point in the filament area; obtain the filament surface grayscale template image, match the filament area with the filament surface grayscale template image, and obtain the filament foot area and the filament spiral area; obtain the suspected reflective area according to all the suspected reflective area pixel points contained in the filament foot area, and obtain the filament foot appearance qualification by combining the centroid of each suspected reflective area; obtain the suspected spiral reflective area according to all the suspected reflective area pixel points contained in the filament spiral area, and obtain the tungsten filament diameter qualification according to all the suspected spiral reflective areas; comprehensively consider the tungsten filament diameter qualification, the filament foot appearance qualification and the filament spiral uniformity to obtain the filament appearance quality qualification; Obtain the test result of the filament quality according to the qualified degree of the filament appearance quality; The step of obtaining the pixel points of the spiral region according to the pixels on the filament skeleton curve includes: Calculate the second-order derivative at each pixel point on the filament skeleton curve; count the number of pixels where the second-order derivative is zero, and record it as the number of turns; Calculate the curvature of the filament skeleton curve at each pixel point; arrange the curvatures of all pixel points on the filament skeleton curve in descending order to obtain the curvature change vector; The pixel points corresponding to the elements are taken as turning points; among them, is the number of turns; Calculate the Euclidean distance between any two turning points, and record the pixel points on the filament skeleton curve between the two turning points with the largest Euclidean distance as the spiral area pixel points; The method of obtaining the central intersection points according to all the pixels in the spiral area and obtaining the uniformity of the filament spiral by combining the Euclidean distance between the central intersection points includes: The least square method is used to fit all the pixels in the spiral area with the straight line to obtain the center line of the spiral area; the pixel points in the spiral area located on the center line of the spiral area are recorded as the center intersection points; For the hth center intersection point, calculate the Euclidean distance between the hth center intersection point and the h+1th center intersection point; calculate the mean of the Euclidean distances of all center intersection points; obtain the difference between the Euclidean distance between the hth center intersection point and the h+1th center intersection point and the mean; calculate the square of the difference; obtain the mean of the squares of all center intersection points, recorded as the first mean; calculate the sum of the first mean and the number 1; and use the ratio of the number 1 to the sum as the filament spiral uniformity; The method of obtaining the suspected reflective area according to all the suspected reflective area pixels contained in the filament foot area, and obtaining the appearance qualification of the filament foot in combination with the centroid of each suspected reflective area, includes: Performing connected domain analysis on all the pixels of the suspected reflective area contained in the filament foot area to obtain each connected domain, and taking each connected domain as each suspected reflective area; counting the number of suspected reflective areas; and obtaining the centroid of each suspected reflective area; Any suspected reflective area is recorded as the area to be analyzed, and the centroid of the area to be analyzed and the centroid of each suspected reflective area are calculated. The absolute value of the difference of the axis coordinates; obtaining the minimum value of all the absolute values ​​of the difference of the area to be analyzed; obtaining the mean of the minimum values ​​of all suspected reflective areas, recorded as the second mean; calculating the difference between the minimum value and the second mean; obtaining the sum of the squares of the differences of all suspected reflective areas, recorded as the first sum of squares; Get the centroid of all suspected reflective areas The mean of the axis coordinates is recorded as the third mean; calculate the centroid of the area to be analyzed The difference between the axis coordinate and the third mean is recorded as the first difference; the sum of the squares of the first differences of all suspected reflective areas is obtained, which is recorded as the second sum of squares; Obtain the sum of the first square sum and the second square sum; calculate the ratio of the sum to the number of suspected reflective areas; obtain the sum of the ratio and the number 1, and record it as the first sum; and use the ratio of the number 1 to the first sum as the appearance qualification of the filament foot; The method of obtaining the suspected spiral reflective area according to all the suspected reflective area pixel points contained in the filament spiral area, and obtaining the tungsten filament diameter qualification according to all the suspected spiral reflective areas, includes: Performing connected domain analysis on all pixel points of the suspected reflective area contained in the filament spiral area to obtain each connected domain, and using each connected domain as each suspected spiral reflective area; For each suspected spiral reflective area, obtain the minimum circumscribed rectangle of the suspected spiral reflective area, and use the width of the minimum circumscribed rectangle as the tungsten wire diameter corresponding to the suspected spiral reflective area; obtain the mean of the tungsten wire diameters corresponding to all suspected spiral reflective areas; calculate the difference between the tungsten wire diameter corresponding to each suspected spiral reflective area and the mean; obtain the square of the difference; calculate the mean of the squares of all suspected spiral reflective areas, recorded as the fourth mean; obtain the sum of the fourth mean and the number 1; and use the ratio of the number 1 to the sum as the tungsten wire diameter qualification degree.

2. The intelligent filament quality detection method based on image processing according to claim 1, characterized in that: The step of acquiring a filament region in a grayscale image of the filament surface and obtaining a filament skeleton curve according to the filament region includes: The grayscale image of the filament surface is segmented using the region growing method to obtain a binary image; the foreground area in the binary image is used as the filament area; Use for the filament area Skeleton extraction algorithm to obtain the filament skeleton curve.

3. The intelligent filament quality detection method based on image processing according to claim 1, characterized in that: The step of obtaining the pixel points in the suspected reflection area according to the grayscale value of each pixel point in the filament area includes: The filament area is segmented by using a dual-threshold segmentation algorithm to obtain a first segmentation threshold and a second segmentation threshold; wherein the first segmentation threshold is less than the second segmentation threshold; and pixels in the filament area whose grayscale values ​​are greater than or equal to the second segmentation threshold are recorded as pixels in the suspected reflective area.

4. The intelligent filament quality detection method based on image processing according to claim 1, characterized in that: The step of acquiring the grayscale template image of the filament surface, matching the filament area with the grayscale template image of the filament surface, and obtaining the filament foot area and the filament spiral area includes: A grayscale template image of the filament surface is obtained, and the filament area is matched with the grayscale template image of the filament surface using a BBS image matching algorithm to obtain a filament foot area within the filament area; all areas within the filament area except the filament foot area are taken as the filament spiral area.

5. The intelligent filament quality detection method based on image processing according to claim 1, characterized in that: The above-mentioned comprehensive tungsten filament diameter qualification, filament foot appearance qualification and filament spiral uniformity are used to obtain the filament appearance quality qualification, including: The sum of the tungsten filament diameter qualification, the filament foot appearance qualification and the filament spiral uniformity is calculated, and the ratio of the sum to the number 3 is taken as the filament appearance quality qualification.

6. The intelligent filament quality detection method based on image processing according to claim 1, characterized in that: The method of obtaining the test result of the filament quality according to the filament appearance quality qualification degree comprises: If the filament appearance quality qualification is greater than the preset quality threshold, the filament quality is qualified; otherwise, the filament quality is unqualified.