TAC film casting defect intelligent detection method based on machine vision

Abstract

The invention discloses a TAC film casting defect intelligent detection method based on machine vision, and relates to the technical field of film defect identification, and the method comprises the following steps: obtaining an initial abnormal contour in a real-time film grey-scale map based on a first grey-scale range value and a second grey-scale range value; constructing a real-time reference contour based on the initial abnormal region; constructing a real-time comparison value based on the initial abnormal contour and the real-time reference contour; acquiring a comparison threshold and a length threshold based on the real-time TAC film of the second number of non-plasticized point defects; dividing the initial abnormal contour into a first type defect and a second type defect based on a real-time comparison value and a comparison threshold value; dividing the second type of defects into a third type of defects and a fourth type of defects based on a length threshold value; the method is used for solving the problem that an existing film defect recognition technology cannot judge whether raw materials are processed disqualification or not based on defect characteristics, so that whether the raw materials are processed abnormally or not cannot be quickly recognized, and waste of the production raw materials is caused.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of film defect identification, in particular to a TAC film casting film defect intelligent detection method based on machine vision. BACKGROUND

[0002] The surface quality, such as cleanliness, flatness and uniformity, of a TAC film directly determines the yield and optical performance of downstream display products; in the flow casting production process of the TAC film, due to the influence of various complex factors such as raw material purity, production environment, process parameter fluctuation and equipment state, various micro defects are easily generated on the film surface, which are amplified in the subsequent polarizing plate bonding and display module assembly process, resulting in adverse phenomena such as bright and dark spots, halos and moire on the screen. If individual defects occur on the film, the production process is not greatly affected, but if the processing effect of the granular raw material is poor, defects occur, for example, the raw material is not completely plasticized, resulting in un-plasticized points on the TAC film, if the processing effect of the raw material is not improved, a large number of TAC films will be defective in the subsequent process, therefore, it is necessary to judge whether the defects are caused by unqualified processing of the raw material according to the defect characteristics; however, the existing defect identification cannot judge whether the defects are caused by unqualified processing of the raw material based on the defect characteristics, for example, in the patent application with the publication number CN120707568A, an optical film surface flaw defect intelligent detection method is disclosed, which only identifies the defects and cannot judge whether the defects are caused by unqualified processing of the raw material based on the defect characteristics, so that the causes of the defects cannot be quickly identified, resulting in waste of production raw materials, that is, the existing film defect identification technology cannot judge whether the defects are caused by unqualified processing of the raw material based on the defect characteristics, so that whether the defects are caused by abnormal processing of the raw material cannot be quickly identified, resulting in waste of production raw materials. SUMMARY

[0003] The application aims to at least solve one of the technical problems in the prior art, obtain an initial abnormal contour in a real-time film gray scale image based on a first gray scale range value and a second gray scale range value; construct a real-time reference contour based on the initial abnormal area; construct a real-time comparison value based on the initial abnormal contour and the real-time reference contour; obtain a comparison threshold value and a length threshold value based on a real-time TAC film of a second number of un-plasticized point defects; divide the initial abnormal contour into a first type of defect and a second type of defect based on the real-time comparison value and the comparison threshold value; and divide the second type of defect into a third type of defect and a fourth type of defect based on the length threshold value, so as to solve the problem that the existing film defect identification technology cannot judge whether the defects are caused by unqualified processing of the raw material based on the defect characteristics, so that whether the defects are caused by abnormal processing of the raw material cannot be quickly identified, resulting in waste of production raw materials.

[0004] To achieve the above-mentioned purpose, the application provides a TAC film casting film defect intelligent detection method based on machine vision, which comprises the following steps: An image of the TAC film to be detected is acquired, which is marked as a real-time film image; The real-time film image is subjected to grayscale processing to obtain a real-time film grayscale image; First and second grayscale range values are obtained based on the first number of normal TAC films; An initial abnormal contour in the real-time film grayscale image is obtained based on the first and second grayscale range values; A real-time reference contour is constructed based on the initial abnormal region; A real-time comparison value is constructed based on the initial abnormal contour and the real-time reference contour; A comparison threshold and a length threshold are obtained based on the second number of real-time TAC films with un-plasticized point defects; The initial abnormal contour is divided into first and second types of defects based on the real-time comparison value and the comparison threshold; The second type of defects is further divided into third and fourth types of defects based on the length threshold.

[0005] Further, obtaining the first and second grayscale range values based on the first number of normal TAC films includes the following sub-steps: Images of the first number of normal TAC films are acquired, which are marked as historical normal images; the historical normal images are subjected to grayscale processing to obtain historical normal grayscale images; Grayscale values in the historical normal grayscale images are obtained, which are marked as historical normal grayscale values; the historical normal grayscale values are regarded as target data, and a target threshold obtaining method is used to obtain first and second target range values, which are marked as the first and second grayscale range values, respectively.

[0006] Further, the target threshold obtaining method includes: The range of the target data is obtained, which is marked as a target range; The target range is evenly divided into a third number of ranges, which are marked as target division ranges; The frequency of the historical vertical coordinates in each target division range is obtained, which is marked as a target division frequency; The target division frequencies are sorted from left to right in ascending order according to the minimum values of the corresponding target division ranges; The sum of all the target division frequencies is obtained, which is marked as a target total frequency; The target total frequency is divided by the third number to obtain a target average frequency; an abnormal proportion value is set; the product of the target average frequency and the abnormal proportion value is obtained to obtain an abnormal division threshold; The target division frequencies less than the abnormal division threshold are marked as abnormal division frequencies; If the leftmost target division frequency is an abnormal division frequency, delete the leftmost abnormal division frequency until the leftmost is not an abnormal division frequency, and stop deleting; after stopping deleting, obtain the minimum value of the target division range corresponding to the leftmost target division frequency, and mark it as the first target range value; If the rightmost target division frequency is an abnormal division frequency, delete the rightmost abnormal division frequency until the rightmost is not an abnormal division frequency, and stop deleting; after stopping deleting, obtain the maximum value of the target division range corresponding to the rightmost target division frequency, and mark it as the second target range value.

[0007] Further, obtaining the initial abnormal contour in the real-time film gray scale image based on the first gray scale range value and the second gray scale range value includes the following sub-steps: Mark the area in the real-time film gray scale image whose gray scale value is not in the range between the first gray scale range value and the second gray scale range value as an abnormal area; Obtain the boundary contour of the abnormal area, and mark it as the initial abnormal contour.

[0008] Further, constructing a real-time reference contour based on the initial abnormal area includes the following sub-steps: Establish a plane rectangular coordinate system, and mark it as a target coordinate system; place the initial abnormal contour in the target coordinate system; Obtain a fourth number of coordinate points on the initial abnormal contour at random positions, and mark them as contour coordinate points; Fit the contour coordinate points with a circular standard function to obtain a function, and mark it as a real-time reference contour.

[0009] Further, constructing a real-time comparison value based on the initial abnormal contour and the real-time reference contour includes the following sub-steps: Obtain the center point of the real-time reference contour, and mark it as a reference center point; From the reference center point as the starting point, emit a fifth number of rays at different angles, and mark them as reference rays; Obtain the intersection points of each reference ray with the initial abnormal contour and the real-time reference contour, respectively, and mark them as first intersection points and second intersection points; Obtain the distance between the first intersection point and the second intersection point of each reference ray, and mark it as the offset distance; Obtain the mean value of all offset distances, and mark it as the real-time comparison value.

[0010] Further, obtaining a comparison threshold and a length threshold based on the real-time TAC film of the second number of un-plasticized point defects includes the following sub-steps: acquire the image of the real-time TAC film with the second number of un-plasticized point defects, marked as a historical defect image; take the historical defect image as the real-time film image to acquire a real-time reference contour and a real-time contrast value, marked as a historical reference contour and a historical contrast value respectively; take the historical contrast value as target data, and acquire a second target range value by using the target threshold acquisition method, marked as a contrast threshold.

[0011] Further, acquiring the contrast threshold and the length threshold based on the real-time TAC film with the second number of un-plasticized point defects further comprises the following sub-steps: acquire the radius of the historical reference contour, marked as a historical reference radius; take the historical reference radius as target data, and acquire a first target range value by using the target threshold acquisition method, marked as a length threshold.

[0012] Further, dividing the initial abnormal contour into the first kind of defects and the second kind of defects based on the real-time contrast value and the contrast threshold comprises the following sub-steps: if the real-time contrast value is greater than the contrast threshold, the initial abnormal contour is identified as the first kind of defects; if the real-time contrast value is less than or equal to the contrast threshold, the initial abnormal contour is identified as the second kind of defects.

[0013] Further, dividing the second kind of defects into the third kind of defects and the fourth kind of defects based on the length threshold comprises the following sub-steps: if it is the second kind of defects, acquire the radius of the real-time reference contour, marked as a real-time length; determine whether the real-time length is less than the length threshold, if yes, mark the second kind of defects as the third kind of defects; if not, mark the second kind of defects as the fourth kind of defects.

[0014] The present application has the following advantages: the present application acquires the initial abnormal contour in the real-time film gray scale image based on the first gray scale range value and the second gray scale range value; constructs the real-time reference contour based on the initial abnormal area; constructs the real-time contrast value based on the initial abnormal contour and the real-time reference contour; acquires the contrast threshold and the length threshold based on the real-time TAC film with the second number of un-plasticized point defects; divides the initial abnormal contour into the first kind of defects and the second kind of defects based on the real-time contrast value and the contrast threshold; divides the second kind of defects into the third kind of defects and the fourth kind of defects based on the length threshold, which has the advantage that whether the raw material is unqualified in processing can be determined based on the defect characteristics, whether the raw material is abnormal in processing can be quickly identified, and the waste of production raw materials is reduced. The present application acquires the contrast threshold and the length threshold based on the real-time TAC film with the second number of un-plasticized point defects, which has the advantage that whether the raw material is abnormal in processing can be determined based on the contrast threshold and the length threshold, and the waste of production raw materials is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 Flow chart of steps for the method of the present application; Figure 2 Schematic diagram of the initial abnormal profile of the present application; Figure 3 Schematic diagram of the real-time reference profile of the present application; Figure 4 Schematic diagram of the first intersection and the second intersection of the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0017] Embodiment 1, please refer to Figure 1 As shown in the figure, the present application provides a TAC film casting defect intelligent detection method based on machine vision, which comprises the following steps: Step S1, an image of a TAC film to be detected is acquired, which is marked as a real-time film image; the light environment should be kept consistent when acquiring the real-time film image to facilitate image comparison.

[0018] Step S2, the real-time film image is subjected to grayscale processing to acquire a real-time film grayscale image; the existing grayscale processing method can be used to acquire the real-time film grayscale image.

[0019] Step S3, first and second grayscale range values are acquired based on a first number of normal TAC films; step S3 comprises the following sub-steps: Step S301, images of the first number of normal TAC films are acquired, which are marked as historical normal images; the historical normal images are subjected to grayscale processing to acquire historical normal grayscale images; the conditions for acquiring the historical normal images are consistent with the conditions for acquiring the real-time film image; the first number of historical normal images are acquired in order to acquire the grayscale value distribution of the normal TAC films, so the greater the first number, the more accurate the data acquired, for example, the first number is 20; Step S302, obtaining the gray value in the historical normal gray image, marked as the historical normal gray value; taking the historical normal gray value as the target data, obtaining the first target range value and the second target range value by using the target threshold obtaining method, respectively marked as the first gray range value and the second gray range value; deleting the individual too small or too large historical normal gray value by using the target threshold obtaining method, and then obtaining the range of the accurate historical normal gray value; obtaining the first target range value and the second target range value by using the obtaining method; step S302 further includes the following sub-steps: Step S30201, obtaining the range of the target data, marked as the target range; Step S30202, uniformly dividing the target range into a third number of ranges, marked as the target division range; dividing the third number of target division ranges, in order to observe the distribution of the target data, therefore the third number of divisions is moderate, for example, the third number is 10; Step S30203, obtaining the frequency of the historical vertical coordinates in each target division range, marked as the target division frequency; Step S30204, sorting the target division frequency from left to right according to the minimum value of the corresponding target division range from small to large; Step S30205, obtaining the sum of all target division frequencies, marked as the target total frequency; Step S30206, dividing the target total frequency by the third number to obtain the target average frequency; setting an abnormal proportion value; obtaining the product of the target average frequency and the abnormal proportion value to obtain the abnormal segmentation threshold; the abnormal segmentation threshold is used to obtain smaller target division frequencies, therefore the abnormal proportion value should not be too large, for example, the abnormal proportion value is 0.01; the abnormal proportion value can be adjusted according to the size of the target average frequency, if the target average frequency is small, the abnormal proportion value can be increased, if the target average frequency is large, the abnormal proportion value can be decreased; in this way, individual too large or individual too small data can be better deleted; In actual application, for example, the target total frequency is 2000; dividing the target total frequency 2000 by the third number 10 to obtain the target average frequency 200; setting the abnormal proportion value to 0.01; obtaining the product of the target average frequency 200 and the abnormal proportion value 0.01 to obtain the abnormal segmentation threshold 2.

[0020] Step S30207, marking the target division frequency smaller than the abnormal segmentation threshold as the abnormal division frequency; Step S30208, if the leftmost target division frequency is an abnormal division frequency, delete the leftmost abnormal division frequency until the leftmost is not an abnormal division frequency; after stopping deleting, obtain the minimum value of the target division range corresponding to the leftmost target division frequency, and mark it as the first target range value; because the leftmost is the minimum distribution target division range, deleting the leftmost abnormal division frequency, i.e. deleting individual abnormal small target data, makes the obtained target data range more accurate; Step S30209, if the rightmost target division frequency is an abnormal division frequency, delete the rightmost abnormal division frequency until the rightmost is not an abnormal division frequency; after stopping deleting, obtain the maximum value of the target division range corresponding to the rightmost target division frequency, and mark it as the second target range value; because the rightmost is the maximum distribution target division range, deleting the rightmost abnormal division frequency, i.e. deleting individual abnormal large target data, makes the obtained target data range more accurate.

[0021] Step S4, obtaining an initial abnormal contour in the real-time film gray scale image based on the first gray scale range value and the second gray scale range value; step S4 includes the following sub-steps: Step S401, marking the area in the real-time film gray scale image whose gray scale value is not in the range between the first gray scale range value and the second gray scale range value as an abnormal area; the range between the first gray scale range value and the second gray scale range value is the gray scale value distribution of the normal TAC film, and when a defect occurs, the defect area will be displayed under light, i.e. the gray scale value of the defect area is different from the gray scale value distribution range of the normal TAC film, so the obtained abnormal area can be preliminarily determined as a defect area; Step S402, obtaining the boundary contour of the abnormal area, and marking it as an initial abnormal contour; obtaining the initial abnormal contour facilitates subsequent image processing.

[0022] Step S5, constructing a real-time reference contour based on the initial abnormal area; step S5 includes the following sub-steps: Step S501, establishing a plane rectangular coordinate system, and marking it as a target coordinate system; placing the initial abnormal contour in the target coordinate system; establishing the target coordinate system facilitates position comparison; Step S502, obtaining fourth quantity of coordinate points on the initial abnormal contour at random positions, and marking them as contour coordinate points; obtaining the fourth quantity of contour coordinate points is to facilitate contour fitting; therefore, the larger the fourth quantity is set, the more accurate the obtained data is, for example, the fourth quantity is 30; In actual application, please refer to Figure 2 the contour coordinate points obtained.

[0023] Step S503, fitting the contour coordinate points with a standard function of a circle to obtain a function, marked as a real-time reference contour; because the raw material is usually small particles, and when not fully plasticized, it is usually close to a circle; therefore, whether the raw material is not fully plasticized can be preliminarily judged by fitting into a circle through the similarity between the initial abnormal contour and the real-time reference contour; In practical applications, please refer to Figure 3 As shown in FIG. 6, the real-time reference contour is obtained.

[0024] Step S6, constructing a real-time comparison value based on the initial abnormal contour and the real-time reference contour; step S6 includes the following sub-steps: Step S601, obtaining a center point of the real-time reference contour, marked as a reference center point; Step S602, emitting a fifth number of rays at different angles from the reference center point, marked as reference rays; the fifth number of reference rays are obtained in order to obtain the similarity between the initial abnormal contour and the real-time reference contour, so the greater the fifth number is set, the more accurate the subsequent obtained data is, but the calculation amount will increase, for example, the fifth number is 20; Step S603, obtaining the intersection points of each reference ray with the initial abnormal contour and the real-time reference contour, respectively marked as a first intersection point and a second intersection point; Step S604, obtaining the distance between the first intersection point and the second intersection point of each reference ray, marked as an offset distance; that is, the offset of the initial abnormal contour and the real-time reference contour under each reference ray, and the smaller the offset distance is, the more similar the initial abnormal contour and the real-time reference contour are under this reference ray; Step S605, obtaining the mean value of all offset distances, marked as a real-time comparison value; the smaller the real-time comparison value is, the more similar the initial abnormal contour and the real-time reference contour are; In practical applications, please refer to Figure 4 As shown in FIG. 7, a schematic diagram of the first intersection point and the second intersection point of a reference ray is obtained; for example, the obtained real-time comparison value is 0.012 mm.

[0025] Step S7, obtaining a comparison threshold and a length threshold based on the real-time TAC film of the second number of un-plasticized point defects; step S7 includes the following sub-steps: Step S701, obtaining an image of the real-time TAC film of the second number of un-plasticized point defects, marked as a historical defect image; taking the historical defect image as the real-time film image to obtain a real-time reference contour and a real-time comparison value, respectively marked as a historical reference contour and a historical comparison value; the second number of historical defect images are used to obtain the distribution range of the real-time reference contour and the real-time comparison value, so the greater the second number is set, the more accurate the distribution range of the real-time reference contour and the real-time comparison value is obtained; for example, the second number is set to 100.

[0026] Step S702, taking the historical contrast value as target data, using the target threshold acquisition method to acquire a second target range value, marked as a contrast threshold; because the incompletely plasticized is closer to a circle, if greater than the contrast threshold, then exceeding the degree of the circle of the incompletely plasticized; Step S703, acquiring the radius of the historical reference contour, marked as a historical reference radius; thus by the dust particles also approximate to a circle, thus it is easy to confuse the two, by the incompletely plasticized particle is much larger than the dust particle, thus acquiring the historical reference radius; Step S704, taking the historical reference radius as target data, using the target threshold acquisition method to acquire a first target range value, marked as a length threshold.

[0027] In actual application, for example, the acquired contrast threshold is 0.032 mm, and the acquired length threshold is 0.2 mm.

[0028] Step S8, based on the real-time contrast value and the contrast threshold, dividing the initial abnormal contour into a first kind of defect and a second kind of defect; step S8 includes the following sub-steps: Step S801, if the real-time contrast value is greater than the contrast threshold, identifying the initial abnormal contour as the first kind of defect; if the real-time contrast value is less than or equal to the contrast threshold, identifying the initial abnormal contour as the second kind of defect; first, judging whether it is approximate to a circle through the real-time contrast value and the contrast threshold; the first kind of defect is a defect that is not approximate to a circle, for example, a scratch; the second kind of defect is a defect that is approximate to a circle, for example, an incompletely plasticized point, dust interference, etc.; In actual application, the real-time contrast value is 0.012 mm, which is less than the contrast threshold 0.32 mm, so the initial abnormal contour is identified as the second kind of defect.

[0029] Step S9, based on the length threshold, further dividing the second kind of defect into a third kind of defect and a fourth kind of defect; step S9 includes the following sub-steps: Step S901, if it is the second kind of defect, acquiring the radius of the real-time reference contour, marked as a real-time length; Step S902, judging whether the real-time length is less than the length threshold, if yes, marking the second kind of defect as the third kind of defect; if not, marking the second kind of defect as the fourth kind of defect; the third kind of defect is a dust interference defect, and the fourth kind of defect is very likely to be an incompletely plasticized point of raw material; In actual application, the acquired real-time length is 1.2 mm, so the real-time length 1.2 mm is greater than the length threshold 0.2 mm; thus the second kind of defect is marked as the fourth kind of defect; the appearance of the fourth kind of defect is very likely to be an incompletely plasticized point of raw material, and the plasticizing effect of the raw material particles needs to be improved.

[0030] In some embodiments, the electronic device can include a processor, a communication interface, a memory, and a communication bus. The processor, the communication interface, and the memory can communicate with each other through the communication bus. The memory can store computer-readable instructions. The processor can invoke the instructions in the memory. When the computer-readable instructions are executed by the processor, the steps in the method for intelligently detecting defects in a TAC film casting film based on machine vision can be executed to achieve the following functions: obtaining an image of a TAC film to be detected, which is labeled as a real-time film image; performing grayscale processing on the real-time film image to obtain a real-time film grayscale image; obtaining a first grayscale range value and a second grayscale range value based on a first number of normal TAC films; obtaining an initial abnormal contour in the real-time film grayscale image based on the first grayscale range value and the second grayscale range value; constructing a real-time reference contour based on the initial abnormal region; constructing a real-time comparison value based on the initial abnormal contour and the real-time reference contour; obtaining a comparison threshold value and a length threshold value based on a second number of real-time TAC films with un-plasticized point defects; dividing the initial abnormal contour into a first type of defect and a second type of defect based on the real-time comparison value and the comparison threshold value; and dividing the second type of defect into a third type of defect and a fourth type of defect based on the length threshold value.

[0031] In addition, the logic instructions in the memory described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0032] In embodiment 3, the application further provides a computer program product, which comprises a computer program stored on a computer readable storage medium, and the computer program comprises program instructions, when the program instructions are executed by a computer, the computer can execute the machine vision based TAC film casting defect intelligent detection method provided by the above method, and the method comprises the following steps: acquiring an image of a TAC film to be detected, and marking the image as a real-time film image; performing gray processing on the real-time film image to acquire a real-time film gray image; acquiring a first gray range value and a second gray range value based on a first number of normal TAC films; acquiring an initial abnormal contour in the real-time film gray image based on the first gray range value and the second gray range value; constructing a real-time reference contour based on the initial abnormal region; constructing a real-time contrast value based on the initial abnormal contour and the real-time reference contour; acquiring a contrast threshold and a length threshold based on a second number of real-time TAC films with un-plasticized point defects; dividing the initial abnormal contour into a first type of defect and a second type of defect based on the real-time contrast value and the contrast threshold; and dividing the second type of defect into a third type of defect and a fourth type of defect based on the length threshold.

[0033] In embodiment 4, the application further provides a computer readable storage medium, and the application provides a storage medium, which stores a computer program, and when the computer program is executed by a processor, the steps of the above machine vision based TAC film casting defect intelligent detection method are run to realize the following functions: acquiring an image of a TAC film to be detected, and marking the image as a real-time film image; performing gray processing on the real-time film image to acquire a real-time film gray image; acquiring a first gray range value and a second gray range value based on a first number of normal TAC films; acquiring an initial abnormal contour in the real-time film gray image based on the first gray range value and the second gray range value; constructing a real-time reference contour based on the initial abnormal region; constructing a real-time contrast value based on the initial abnormal contour and the real-time reference contour; acquiring a contrast threshold and a length threshold based on a second number of real-time TAC films with un-plasticized point defects; dividing the initial abnormal contour into a first type of defect and a second type of defect based on the real-time contrast value and the contrast threshold; and dividing the second type of defect into a third type of defect and a fourth type of defect based on the length threshold.

[0034] Through the above description of the embodiments, the embodiments of the application can be provided as a method, a system or a computer program product. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in each embodiment or some parts of the embodiment.

[0035] In the embodiments of the present application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are only illustrative, for example, the division of modules or units is only a logical function division, and other division manners can be used in actual implementation, for example, a plurality of modules or units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some communication interface, the indirect coupling or communication connection between the system, the module and the unit can be electrical, mechanical or other forms.

[0036] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A machine vision-based intelligent detection method for TAC film casting film defects, characterized in that, The method comprises the following steps: An image of the TAC film to be detected is acquired, which is marked as a real-time film image; The real-time film image is subjected to grayscale processing to obtain a real-time film grayscale image; First and second grayscale range values are obtained based on the first number of normal TAC films; An initial abnormal contour in the real-time film grayscale image is obtained based on the first and second grayscale range values; A real-time reference contour is constructed based on the initial abnormal region; A real-time comparison value is constructed based on the initial abnormal contour and the real-time reference contour; A comparison threshold value and a length threshold value are obtained based on the second number of real-time TAC films with un-plasticized point defects; The initial abnormal contour is divided into a first type of defect and a second type of defect based on the real-time comparison value and the comparison threshold value; The second type of defect is further divided into a third type of defect and a fourth type of defect based on the length threshold value. 2.The machine vision-based TAC film casting film defect intelligent detection method according to claim 1, characterized in that, The first and second grayscale range values are obtained based on the first number of normal TAC films, which comprises the following sub-steps: Images of the first number of normal TAC films are acquired, which are marked as historical normal images; the historical normal images are subjected to grayscale processing to obtain historical normal grayscale images; Grayscale values in the historical normal grayscale images are obtained, which are marked as historical normal grayscale values; the historical normal grayscale values are regarded as target data, and a first target range value and a second target range value are obtained by using a target threshold value acquisition method, which are marked as the first and second grayscale range values, respectively. 3.The machine vision based TAC film casting film defect intelligent detection method according to claim 2, characterized in that, The target threshold value acquisition method comprises: A range of the target data is obtained, which is marked as a target range; The target range is uniformly divided into a third number of ranges, which are marked as target division ranges; The frequency of historical vertical coordinates in each target division range is obtained, which is marked as a target division frequency; The target division frequencies are sorted from left to right in ascending order according to the minimum values of the corresponding target division ranges; The sum of all target division frequencies is obtained, which is marked as a target total frequency; The target total frequency is divided by the third number to obtain a target average frequency; an abnormal proportion value is set; and the product of the target average frequency and the abnormal proportion value is obtained to obtain an abnormal segmentation threshold value; Target division frequencies less than the abnormal segmentation threshold value are marked as abnormal division frequencies; If the leftmost target division frequency is an abnormal division frequency, the leftmost abnormal division frequency is deleted until the leftmost target division frequency is not an abnormal division frequency, and then the deletion is stopped; after the deletion is stopped, the minimum value of the target division range corresponding to the leftmost target division frequency is obtained, which is marked as the first target range value; If the rightmost target division frequency is an abnormal division frequency, the rightmost abnormal division frequency is deleted until the rightmost target division frequency is not an abnormal division frequency, and then the deletion is stopped; after the deletion is stopped, the maximum value of the target division range corresponding to the rightmost target division frequency is obtained, which is marked as the second target range value. 4.The machine vision-based TAC film casting film defect intelligent detection method according to claim 3, characterized in that, The initial abnormal contour in the real-time film grayscale image is obtained based on the first and second grayscale range values, which comprises the following sub-steps: Regions in the real-time film grayscale image with grayscale values not in the range between the first and second grayscale range values are marked as abnormal regions; The boundary contour of the abnormal region is obtained, which is marked as the initial abnormal contour. 5.The machine vision based TAC film casting film defect intelligent detection method according to claim 4, characterized in that, The real-time reference contour is constructed based on the initial abnormal region, which comprises the following sub-steps: A plane rectangular coordinate system is established and marked as a target coordinate system; the initial abnormal profile is placed in the target coordinate system; A fourth number of coordinate points are obtained at random positions on the initial abnormal profile and marked as profile coordinate points; The profile coordinate points are fitted with a standard function of a circle to obtain a function, which is marked as a real-time reference profile. 6.The machine vision based TAC film casting film defect intelligent detection method according to claim 5, characterized in that, Based on the initial abnormal profile and the real-time reference profile, a real-time comparison value is constructed, including the following sub-steps: A center point of the real-time reference profile is obtained and marked as a reference center point; From the reference center point, a fifth number of rays are emitted at different angles and marked as reference rays; The intersection points of each reference ray with the initial abnormal profile and the real-time reference profile are obtained and marked as first intersection points and second intersection points, respectively; The distance between the first intersection point and the second intersection point of each reference ray is obtained and marked as an offset distance; The mean value of all offset distances is obtained and marked as a real-time comparison value. 7.The machine vision based TAC film casting film defect intelligent detection method according to claim 6, characterized in that, Based on the real-time TAC film of the second number of un-plasticized point defects, a comparison threshold value and a length threshold value are obtained, including the following sub-steps: An image of the real-time TAC film of the second number of un-plasticized point defects is obtained and marked as a historical defect image; the historical defect image is regarded as a real-time film image to obtain a real-time reference profile and a real-time comparison value, which are marked as a historical reference profile and a historical comparison value, respectively; The historical comparison value is regarded as target data, and a second target range value is obtained using a target threshold value acquisition method and marked as a comparison threshold value. 8.The machine vision based TAC film casting film defect intelligent detection method according to claim 7, characterized in that, Based on the real-time TAC film of the second number of un-plasticized point defects, a comparison threshold value and a length threshold value are obtained, including the following sub-steps: The radius of the historical reference profile is obtained and marked as a historical reference radius; The historical reference radius is regarded as target data, and a first target range value is obtained using a target threshold value acquisition method and marked as a length threshold value. 9.The machine vision based TAC film casting film defect intelligent detection method according to claim 8, characterized in that, Based on the real-time comparison value and the comparison threshold value, the initial abnormal profile is divided into a first type of defect and a second type of defect, including the following sub-steps: If the real-time comparison value is greater than the comparison threshold value, the initial abnormal profile is identified as a first type of defect; if the real-time comparison value is less than or equal to the comparison threshold value, the initial abnormal profile is identified as a second type of defect. 10.The machine vision based TAC film casting film defect intelligent detection method according to claim 9, characterized in that, Based on the length threshold value, the second type of defect is further divided into a third type of defect and a fourth type of defect, including the following sub-steps: If it is a second type of defect, the radius of the real-time reference profile is obtained and marked as a real-time length; It is determined whether the real-time length is less than the length threshold value; if yes, the second type of defect is marked as a third type of defect; if no, the second type of defect is marked as a fourth type of defect.

Images