An aluminum alloy plate surface Luoping line detection and evaluation method based on machine vision and feature parameter statistics

Abstract

The application discloses a kind of aluminum alloy plate surface Lu Ping line detection evaluation methods based on machine vision and characteristic parameter statistics, and establishes Lu Ping line detection based on morphological operation to the plate surface image obtained, and according to the size and distribution of each minimum outer rectangle in detection result, the statistics and output of Lu Ping line feature are carried out, and the quantitative evaluation of Lu Ping line is carried out according to the derived Lu Ping line feature.The application discloses a kind of aluminum alloy plate surface Lu Ping line detection evaluation methods based on machine vision and characteristic parameter statistics, provide a kind of convenient operation, standard clear Lu Ping line detection and evaluation method, make up the blank of relevant field and prior art, have very strong practical value, and solve the detection difficulty that Lu Ping line is broken, discontinuous and easily interfered by polishing scratch under current process, algorithm real-time is good, detection speed is fast, and detection cost is low.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of application of machine vision in surface quality detection of metal plates, and particularly relates to an aluminum alloy plate surface Roping line detection and evaluation method based on machine vision and feature parameter statistics. BACKGROUND

[0002] Vehicle lightweight design is one of the development directions of the automobile industry. Aluminum alloy has been widely used in high-end vehicles and new energy vehicles at home and abroad due to its small density, corrosion resistance, good formability, and high yield ratio.

[0003] Compared with traditional steel stamping parts, aluminum alloy sheet forming manufacturing has a series of significantly different characteristics. When aluminum alloy sheet is stamped into an automobile cover part, a series of closely spaced lines along the rolling direction on the surface of the part, called Roping line, are easily generated. The depth of the Roping line can reach dozens of microns, the corrugation width is about 1 millimeter, and the length can reach about 50 millimeters. The Roping line will cause significant surface quality defects of the automobile cover part, and the painting of the part cannot cover such defects and even the phenomenon of amplification occurs, thereby seriously affecting the appearance quality of the automobile outer plate part. Therefore, the detection of the Roping line of the aluminum alloy sheet is an important aspect of evaluating whether the sheet is qualified. At present, in the production application, the detection and evaluation of the Roping line of the aluminum alloy sheet still lacks an efficient and accurate process and method. The existing method mainly uses the naked eye of a technician to evaluate the number of Roping lines on the surface of the aluminum alloy sheet after the aluminum alloy sheet is stretched and slightly polished. This method is not only time-consuming and laborious, but also has a great subjective factor and requires a high level of experience.

[0004] Machine vision is a technology that uses machines to replace human vision, acquires and processes computer digital images, and makes further evaluation combined with actual needs. Machine vision can not only perform some dangerous or extreme work environments, but also liberate people from repetitive work in large-scale repetitive labor, improve automation and production efficiency. At the same time, machine vision can set fixed judgment standards to reduce human subjective factors in the judgment process. With the development of machine vision technology in recent years, machine vision plays an increasingly important role in product quality detection and evaluation.

[0005] At present, there are few research literatures on Roping line in China, and no patent on Roping line evaluation and detection has been proposed. And under the current process, the Roping line has the characteristics of fragmentation and discontinuity, and is easily disturbed by scratches caused by polishing. The traditional machine vision technology cannot solve the problem.

[0006] The application number is: CN 202011224823.4, the invention application discloses "a kind of metal plate surface defect detection method based on machine vision", specifically: red stripe light is projected to plate surface, so that pit, protrusion, scratch and other defects are highlighted, and CCD camera is used to collect stripe projection image;The color image is decomposed, the color information of light source is highlighted, and the color information of light source is highlighted;Extract the center of stripe, judge the distortion of line by algorithm, reflect the size of defect.It uses the lightening mode of stripe, eliminates the influence of mirror surface reflection effect, at the same time, can enhance the appearance of plate surface defect, and obtains defect image with high quality.

[0007] The application number is: CN 202110548337.6, the invention application discloses "a kind of surface defect detection model training method, surface defect detection method and system", wherein the surface defect detection model training method comprises: obtaining normal picture of product and external picture irrelevant to product;The normal picture and the external picture are input into the surface defect detection model based on deep neural network, the surface defect detection model based on deep neural network is trained, and the surface defect detection model is obtained. SUMMARY

[0008] To solve the above problems, the present application provides a kind of based on machine vision and feature parameter statistics's aluminum alloy plate surface Lu Ping line detection evaluation method, its technical scheme is as follows:

[0009] A kind of based on machine vision and feature parameter statistics's aluminum alloy plate surface Lu Ping line detection evaluation method, characterized in that:

[0010] The Lu Ping line detection based on morphological operation is established to the obtained plate surface image, and the statistics and output of Lu Ping line feature are carried out according to the size and distribution of each minimum outer rectangle in the detection result, and the Lu Ping line is quantitatively evaluated according to the obtained Lu Ping line feature.

[0011] According to the Lu Ping line detection evaluation method of the aluminum alloy plate surface based on machine vision and feature parameter statistics of the present application, it is characterized in that:

[0012] After obtaining the plate surface image, before establishing the Lu Ping line detection based on morphological operation, the obtained plate surface image is sequentially subjected to gray scale conversion, Gaussian blur and image cropping pretreatment.

[0013] According to the Lu Ping line detection evaluation method of the aluminum alloy plate surface based on machine vision and feature parameter statistics of the present application, it is characterized in that:

[0014] After Gaussian blur, before image cropping, the image is subjected to image enhancement processing based on brightness and contrast adaptive adjustment.

[0015] The application discloses an aluminum alloy plate surface Luoping line detection and evaluation method based on machine vision and feature parameter statistics.

[0016] The self-adaptive adjustment of the brightness and contrast is established by the linear transformation.

[0017] The linear transformation is expressed as follows:

[0018] O(r,c)=α×I(r,c)+β,0≤r<H,0≤c<W

[0019] The self-adaptive adjustment of the brightness and contrast is established by the linear transformation.

[0020]

[0021] β=I0-αI

[0022] Wherein,

[0023] Alpha: a parameter related to the contrast;

[0024] Beta: a parameter related to the brightness and contrast;

[0025] O: output image;

[0026] I: input image;

[0027] I0: target brightness;

[0028] C0: target contrast;

[0029] W: image column number;

[0030] H: image row number.

[0031] The application discloses an aluminum alloy plate surface Luoping line detection and evaluation method based on machine vision and feature parameter statistics.

[0032] The application discloses an aluminum alloy plate surface Luoping line detection and evaluation method based on machine vision and feature parameter statistics.

[0033] S1: binaryzation processing is performed on the image to obtain a binary image;

[0034] S2: morphological operation and threshold processing are performed on the binary image to obtain preliminary Luoping line detection based on minimum bounding rectangle representation;

[0035] S3: longitudinal and transverse optimization processing is respectively established on the preliminary Luoping line detection to form final Luoping line detection.

[0036] The application discloses an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics.

[0037] The roving line features include quantity, in-plane distribution density, average length and roving line density.

[0038] The application discloses an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics.

[0039] The binarization in step S1 is based on the maximum inter-class variance method.

[0040] The application discloses an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics.

[0041] The morphological operation in step S2 is specifically as follows.

[0042] First, a closing operation is performed, then an opening operation is performed, then a connected domain calculation is performed, and finally, a minimum bounding rectangle calculation is performed on each connected domain.

[0043] The application discloses an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics.

[0044] The threshold processing in step S2 is specifically as follows: a filtering processing is performed on the bounding rectangle which does not reach the longitudinal size setting threshold.

[0045] The application discloses an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics.

[0046] The longitudinal optimization is realized through the following steps.

[0047] SS1: the center coordinates [X c ,Y c ] of each bounding rectangle are calculated.

[0048] SS2: the roving line region after binarization is divided into sections, and the white pixel number N w (x,y) in each section is counted in the mode of x=X c .

[0049] SS3: according to the counting result and the longitudinal pixel number of the binary image, a discriminant of optimization or not is established, for the connected domain corresponding to the bounding rectangle which is crossed by the x=X c path and meets the discriminant, the minimum bounding rectangle is recalculated according to the result after merging; otherwise, the original processing is kept.

[0050] The application discloses an aluminum alloy plate surface Luoping line detection evaluation method based on machine vision and feature parameter statistics.

[0051] The transverse optimization is specifically:

[0052] The proportion of the minimum bounding rectangles in the transverse direction that coincide with each other is calculated, and the minimum bounding rectangles in the transverse direction are combined based on the calculation result.

[0053] The application discloses an aluminum alloy plate surface Luoping line detection evaluation method based on machine vision and feature parameter statistics.

[0054] In step S3, after the optimization processing in the longitudinal and transverse directions is completed, a correction processing of the rectangle length of the optimization result is further established, and finally, the Luoping line detection is formed.

[0055] The application discloses an aluminum alloy plate surface Luoping line detection evaluation method based on machine vision and feature parameter statistics.

[0056] The discriminant is:

[0057]

[0058] In the formula,

[0059] N w : the number of white pixels in the corresponding section;

[0060] x: the horizontal coordinate of the white pixel;

[0061] X c : the horizontal coordinate of the center coordinate of the corresponding bounding rectangle;

[0062] H: the number of longitudinal pixels of the binary image.

[0063] The application discloses an aluminum alloy plate surface Luoping line detection evaluation method based on machine vision and feature parameter statistics.

[0064] The proportion of the minimum bounding rectangles in the transverse direction that coincide with each other is calculated according to the following formula:

[0065]

[0066] When R O is greater than a set value, the minimum bounding rectangles in the transverse direction are combined, otherwise the original processing is kept;

[0067] wherein,

[0068] w1: the transverse width of one of the bounding rectangles;

[0069] w2: transverse width of the other circumscribed rectangle;

[0070] w o : width of the overlap of the two circumscribed rectangles.

[0071] According to the aluminum alloy plate surface Luoping line detection evaluation method based on machine vision and feature parameter statistics of the application,

[0072] The correction processing of the rectangular length is completed by determining the correction coefficient a in the following formula,

[0073]

[0074] In the formula,

[0075] N w : number of white pixels in the corresponding section;

[0076] x: horizontal coordinate of the white pixel;

[0077] y: vertical coordinate of the white pixel;

[0078] X c : horizontal coordinate of the center coordinate of the corresponding circumscribed rectangle;

[0079] Y c : vertical coordinate of the center coordinate of the corresponding circumscribed rectangle;

[0080] w: transverse dimension of the circumscribed rectangle;

[0081] h: longitudinal dimension of the circumscribed rectangle;

[0082] a: correction coefficient.

[0083] According to the aluminum alloy plate surface Luoping line detection evaluation method based on machine vision and feature parameter statistics of the application,

[0084] The set value is 0.2.

[0085] The application discloses an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics.

[0086] In summary, the aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics provides a roving line detection and evaluation method which is convenient to operate and clear in standard, fills the blank of related fields and prior art, has strong practical value, solves the detection difficulties of broken, discontinuous and easy to be interfered by polishing scratches of the roving line under the current process, has good real-time algorithm, fast detection speed and low detection cost; through preparation of an aluminum alloy sample and image acquisition, the roving line on the surface of the aluminum alloy plate is analyzed and evaluated based on the machine vision algorithm and feature quantity statistics of the roving line, which has positive significance for objectively evaluating the surface quality of the aluminum alloy plate, improving the detection efficiency, reducing subjective factors and promoting the application of the aluminum alloy in the domestic automobile industry, and fills the blank in the field of roving line detection of the aluminum alloy plate surface. BRIEF DESCRIPTION OF DRAWINGS

[0087] Figure 1 It is an original image of a local part of the aluminum alloy plate sample with visible roving line on the polished surface in the embodiment of the application;

[0088] Figure 2 It is an algorithm flowchart of machine vision and morphological processing of the application;

[0089] Figure 3 It is a result based on adaptive adjustment of brightness and contrastivity in the embodiment of the application;

[0090] Figure 4 It is a result of the maximum inter-class variance method binarization in the embodiment of the application;

[0091] Figure 5 It is an open operation result in the embodiment of the application;

[0092] Figure 6 The minimum outer rectangle before merging in the embodiment of the present application;

[0093] Figure 7 The Luoping line detection result in the embodiment of the present application;

[0094] Figure 8 The parameter description diagram of the machine vision algorithm in the present application;

[0095] Figure 9 The machine vision algorithm program output in the embodiment of the present application;

[0096] Figure 10 The calculation diagram of the overlap ratio in the embodiment of the present application; w The longitudinal merging diagram of the rectangle;

[0097] Figure 11 The calculation diagram of the overlap ratio in the embodiment of the present application;

[0098] Figure 12 The calculation diagram of the length correction coefficient a in the embodiment of the present application. DETAILED DESCRIPTION

[0099] Hereinafter, a kind of Luoping line detection and evaluation method for aluminum alloy plate surface based on machine vision and feature parameter statistics according to the description drawings and specific embodiments of the present application is further specifically described.

[0100] A kind of Luoping line detection and evaluation method for aluminum alloy plate surface based on machine vision and feature parameter statistics,

[0101] Luoping line detection based on morphological operation is established to the obtained plate surface image, and the statistics and output of Luoping line feature are carried out according to the size and distribution of each minimum outer rectangle in the detection result, and the Luoping line is quantitatively evaluated according to the derived Luoping line feature.

[0102] Wherein,

[0103] After obtaining the plate surface image, before establishing Luoping line detection based on morphological operation, the obtained plate surface image is sequentially subjected to gray scale conversion, Gaussian blur and image cropping pretreatment.

[0104] Wherein,

[0105] After Gaussian blur, before image cropping, the image is subjected to image enhancement processing based on brightness and contrast adaptive adjustment.

[0106] Wherein,

[0107] The adaptive adjustment of brightness and contrast is established by setting a linear transformation;

[0108] The linear transformation is expressed as follows:

[0109] O(r,c)=α×I(r,c)+β,0≤r<H,0≤c<W

[0110] The adaptive adjustment of brightness and contrast is established by adjusting the following two parameters,

[0111]

[0112] β=I0-αI

[0113] Wherein,

[0114] α: a parameter related to contrast;

[0115] β: a parameter related to brightness and contrast;

[0116] O: output image;

[0117] I: input image;

[0118] I0: target brightness;

[0119] C0: target contrast;

[0120] W: number of image columns;

[0121] H: number of image rows.

[0122] Wherein,

[0123] The Luoping line detection based on morphological operation is established, and specifically includes the following steps:

[0124] S1: binaryzation processing is performed on the image to obtain a binary image;

[0125] S2: morphological operation and threshold processing are performed on the binary image to obtain preliminary Luoping line detection based on minimum bounding rectangle representation;

[0126] S3: longitudinal and transverse optimization processing is respectively established for the preliminary Luoping line detection to form the final Luoping line detection.

[0127] Wherein,

[0128] The Luoping line features include: number, in-plane distribution density, average length, and Luoping line density.

[0129] Wherein,

[0130] The binaryzation in step S1 is based on the maximum inter-class variance method.

[0131] Wherein,

[0132] The morphological operation in step S2 is specifically:

[0133] Firstly, a close operation is performed, then an open operation is performed, and then a connected domain calculation is performed, and finally a minimum bounding rectangle calculation is performed on each connected domain.

[0134] Wherein,

[0135] The threshold processing in step S2 is specifically: filtering processing is performed on the bounding rectangle that does not reach the longitudinal size setting threshold.

[0136] Wherein,

[0137] The longitudinal optimization is realized by the following steps:

[0138] SS1: Calculate the center coordinates [X c ,Y c ] of each bounding rectangle;

[0139] SS2: Divide the segmented section of the binarized Luoping line region, and count the number of white pixels N w (x,y) in each section according to x=X c ;

[0140] SS3: According to the statistical result and the number of longitudinal pixels of the binary image, a discriminant for optimization or not is established, and for the connected domain corresponding to the bounding rectangle passing through the path of x=X c after merging, the minimum bounding rectangle is recalculated; otherwise, the original processing is maintained. The discriminant is:

[0141]

[0142] In the formula,

[0143] N w : The number of white pixels in the corresponding section;

[0144] x: The horizontal coordinate of the white pixel;

[0145] X c : The horizontal coordinate of the center coordinate of the corresponding bounding rectangle;

[0146] H: The number of longitudinal pixels of the binary image.

[0147] The transverse optimization described above is specifically:

[0148] The proportion of the minimum bounding rectangle in the transverse direction is calculated, and the minimum bounding rectangle in the transverse direction is merged based on the calculation result. The proportion of the minimum bounding rectangle in the transverse direction is calculated according to the following formula:

[0149]

[0150] When R O If the value is greater than the set value, merge the smallest outer rectangle in the horizontal direction; otherwise, keep the original processing.

[0151] in,

[0152] w1: The horizontal width of one of the outer rectangles;

[0153] w2: The horizontal width of the other outer rectangle;

[0154] w o The width of the overlapping area of ​​the two outer rectangles.

[0155] The set value is 0.2.

[0156] In step S3 above, after completing the vertical and horizontal optimization processes, a correction process for the rectangle length is established based on the optimization results, thereby forming the final Luoping line detection. This correction process for the rectangle length is accomplished by determining the correction coefficient α in the following formula.

[0157]

[0158] In the formula,

[0159] N w The number of white pixels within the corresponding segment;

[0160] x: The horizontal coordinate of the white pixel;

[0161] y: The vertical coordinate of the white pixel;

[0162] X c : The x-coordinate of the center of the corresponding outer rectangle;

[0163] Y c : The ordinate of the center coordinate of the corresponding outer rectangle;

[0164] w: The horizontal dimension of the outer rectangle;

[0165] h: The vertical dimension of the outer rectangle;

[0166] α: Correction factor.

[0167] Working process, principle and implementation examples

[0168] For an understanding of the following statements, please refer to Figures 1 to 12 The specific process is as follows:

[0169] Step one: Preparation of aluminum alloy sample and image acquisition. The size of the aluminum alloy plate sample is 250 mm in length and 35 mm in width, where the length represents the vertical rolling direction and the width represents the rolling direction. The aluminum alloy plate sample is stretched to an engineering strain of 10% at a speed of 0.2 mm / s using a universal material testing machine, the surface to be tested is wiped with industrial alcohol, and then the surface is lightly polished several times along the length direction using an oilstone or sandpaper with a mesh size of about 800. The treated aluminum alloy plate is transferred to the image acquisition platform. The acquisition platform consists of a support, a light source, a camera, a calibration ruler, and a fixing module. The positions of the camera, light source, and aluminum alloy plate are adjusted at an incident angle of 45 degrees to obtain clear and distinct images of the aluminum alloy plate surface, which are transmitted to a digital image processing system.

[0170] Step two: Loping line detection method based on machine vision and morphological processing. The loping line in the image extends along the width direction and exhibits discontinuity and fragmentation due to the polishing process. At the same time, the line spacing is not clear and is easily disturbed by polishing scratches, making it difficult for existing machine vision methods to correctly detect the loping line. To effectively detect the loping line in the target image and calculate the characteristic parameters, the following specific processing scheme is established:

[0171] First, the image is read and converted to a grayscale image, and preprocessed to make the loping line in the image clear and distinct. The preprocessing steps include Gaussian blur and image enhancement based on brightness and contrast adaptive adjustment. The size of the resulting grayscale image is represented as W x H, where W is the number of pixels in the horizontal direction and H is the number of pixels in the vertical direction.

[0172] The image brightness is calculated by converting the RGB color space to the Y channel of the YUV color space:

[0173] Y = 0.299 x R + 0.587 x G + 0.114 x B

[0174] The contrast is calculated by summing the probabilities of the square of the gray level difference between adjacent pixels:

[0175]

[0176] where δ(i, j) = |i - j| represents the gray level difference between adjacent pixels, P δ (i, j) represents the probability distribution of the gray level difference δ between adjacent pixels.

[0177] The brightness and contrast are adjusted by linear transformation. Assuming that the input image is I, the number of columns is W, the number of rows is H, and the output image is O, the linear transformation of the image can be defined by the following formula:

[0178] O(r, c) = α x I(r, c) + β, 0 ≤ r < H, 0 ≤ c < W

[0179] According to the definition, the contrast is only related to a, and the brightness is affected by a and b. Given the target brightness I0 and the contrast C0, the specific parameters of a and b can be obtained, and the image is adaptively adjusted by the linear transformation formula:

[0180] a = sqrt(C0 / C)

[0181] b = I0 - aI

[0182] The calibration ruler in the image is identified, and the relationship between the image pixels and the actual size is established. The image is cropped to a specified size with the image center as the origin.

[0183] For the preprocessing result, the maximum inter-class variance method is used to binarize the image. Morphological operations are performed on the binary image, including closing operation and opening operation. The closing operation closes the smaller holes inside to ensure the continuity of the lines. The opening operation makes the distance between adjacent lines clear and reduces false merging. For the opening operation result, the connected domains are calculated, and the minimum bounding rectangle of each connected domain is calculated. The size of the bounding rectangle is w x h, w along the horizontal direction, and h along the vertical direction. Based on the threshold, filter the rectangles with too small h, and get the preliminary line detection result.

[0184] Due to the fragmentation and discontinuity, the same line is not surrounded by a single rectangular detection frame, and further merging of connected domains is needed according to the parallel rolling direction (vertical) of the line. The specific steps are to calculate the center coordinates [X c ,Y c ] of each rectangle, and define N w (x,y∈A) as the number of white pixels in a certain area A (the line area after binarization) of the binary image, calculate the number of white pixels N w (x=X c ) along the vertical direction of the rectangle center. If , then merge the connected domains corresponding to the rectangles along the x=X c path. For the merging result of the connected domains, recalculate the minimum bounding rectangle (see Figure 10 ).

[0185] In the horizontal direction, the line may also be identified by multiple bounding rectangles due to discontinuity. Calculate the overlap ratio R O between each pair of rectangular detection frames, and merge the horizontal bounding rectangles based on the overlap ratio (see Figure 11 ).

[0186]

[0187] where w1 and w2 are the horizontal widths of the two rectangular frames, and w O is the overlap width of the rectangular frames. If R O > 0.2, merge the connected domains of the horizontal bounding rectangles.

[0188] For the merger results to amend the length of the rectangle. Statistics of white pixels inside the rectangle, amend the length of the longitudinal length of the rectangle, so that it surrounds the distance from the center of the rectangle 95% of the white pixel points, in order to accurately locate the roping line (see Figure 12 ). That is to find the appropriate length correction coefficient alpha, meet:

[0189]

[0190] Step three: roping line grading based on feature statistics. After the above steps, the roping line is surrounded by the bounding rectangle, i.e. the detection frame. According to the size and distribution of the detection frame, the geometric features of the roping line can be statistically output, such as the number, in-plane distribution density, average length, etc. In the detection area, the number of roping lines is equal to the number of bounding rectangles, denoted as N R , let the longitudinal size of the detection area be L mm, then define the roping line density RN(Roping Density) as According to the value of RN, the roping line is quantitatively graded.

[0191] The present application is a kind of based on machine vision and feature parameter statistics's aluminum alloy plate surface roping line detection evaluation method, to solve the current process under the broken, not continuous and easily affected by polishing scratch interference of the detection difficulty of roping line. First, the acquired image is preprocessed, including denoising and enhancing image, identifying the calibration ruler in the image, establishing the relationship between image pixels and actual size, and cropping the image to a specified size. The binary image of the preprocessed image is obtained, and morphological operation is used, first close operation and then open operation. According to the number of critical pixel points along the rolling direction, the rolling direction recognition area is merged, and the recognition difficulty caused by the discontinuity of the roping line along the rolling direction is solved. According to the overlap ratio perpendicular to the rolling direction, the recognition area perpendicular to the rolling direction is merged, and the recognition difficulty caused by the change of the roping line perpendicular to the rolling direction is solved. The feature pixel distribution of the recognition area is statistically analyzed, and the area correction coefficient is determined according to the optimal ratio to accurately identify the length of the roping line. And based on the statistical geometric feature information of the roping line in the detection frame, such as average length and in-plane distribution density, the roping line density per unit length is calculated, and the evaluation of the surface quality of the aluminum alloy plate is completed.

[0192] The application provides an aluminum alloy plate surface roving line detection and evaluation method based on machine vision and feature parameter statistics, provides a roving line detection and evaluation method which is convenient to operate and clear in standard, fills the blank of related fields and prior art, has strong practical value, and solves the detection difficulties of roving line breaking, discontinuity and easy disturbance of polishing scratches under the current process, has good algorithm real-time performance, fast detection speed and low detection cost; through preparation and image acquisition of an aluminum alloy sample, roving line on the surface of the aluminum alloy plate is analyzed and evaluated based on a machine vision algorithm and feature quantity statistics of roving features, which has positive significance for objectively evaluating the surface quality of the aluminum alloy plate, improving detection efficiency, reducing subjective factors, improving the application of the aluminum alloy in the domestic automobile industry, and fills the blank in the field of aluminum alloy plate surface roving line detection.

Claims

1. A method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics, characterized in that: A morphological operation-based detection method for the surface image of the board material is established. The characteristics of the roping line are statistically analyzed and output based on the size and distribution of each minimum bounding rectangle in the detection results. The roping line is then quantitatively evaluated based on the obtained characteristics. The establishment of the morphological operation-based rhombus line detection method specifically includes the following steps: S1: Binarize the image to obtain a binary image; S2: Perform morphological operations and thresholding on the binary image to obtain preliminary Luo Ping line detection based on the minimum bounding rectangle representation; S3: Optimize the preliminary Luoping line detection in both the longitudinal and transverse directions to form the final Luoping line detection. In S3, the discriminant for vertical optimization is: ； In the formula, The number of white pixels within the corresponding segment; : The x-coordinate of the white pixel; : The x-coordinate of the center of the corresponding outer rectangle; The number of vertical pixels in a binary image; In S3, the formula for calculating the ratio of the pairwise overlap of the minimum outer rectangles in the horizontal direction in the lateral optimization process is as follows: ： when If the value is greater than the set value, merge the smallest outer rectangle in the horizontal direction; otherwise, keep the original processing. in, : The horizontal width of one of the outer rectangles; The horizontal width of the other outer rectangle; The width of the overlapping area of ​​the two outer rectangles.

2. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics as described in claim 1, characterized in that: After acquiring the surface image of the board material and before establishing the morphological operation-based Luo Ping line detection, the acquired surface image of the board material was preprocessed by grayscale conversion, Gaussian blurring and image cropping.

3. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 2, characterized in that: After Gaussian blurring and before image cropping, image enhancement processing based on adaptive adjustment of brightness and contrast was also performed on the image.

4. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics as described in claim 3, characterized in that: An adaptive adjustment of brightness and contrast is established through a set linear transformation; The linear transformation is expressed as follows: ； The adaptive adjustment of brightness and contrast is achieved by adjusting the following two parameters. ； ； in, α: A parameter related to contrast. β: A parameter related to brightness and contrast; Output image; Input image; Target brightness; Target contrast; Number of image columns; Number of rows in the image.

5. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: The characteristics of Luoping lines include: quantity, in-plane distribution density, average length, and Luoping line density.

6. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: The binarization in step S1 is performed based on the maximum inter-class variance method.

7. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics as described in claim 1, characterized in that: The morphological operations in step S2 are as follows: First, perform the closing operation, then the opening operation, then the connected component calculation, and finally calculate the minimum bounding rectangle for each connected component.

8. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: The threshold processing in step S2 specifically involves filtering out outer rectangles that do not meet the set threshold for the vertical dimension.

9. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: The vertical optimization is achieved through the following steps: SS1: Calculate the center coordinates of each outer rectangle. ; SS2: Divide the binarized Luoping line region into segments and count the number of white pixels within each segment. ,according to Statistical analysis was conducted in this manner; SS3: Based on the statistical results and the number of vertical pixels in the binary image, establish a discriminant to determine whether optimization is needed. For those that satisfy the discriminant, merge... The path passes through the connected components corresponding to the outer rectangle, and the minimum outer rectangle is recalculated based on the merged result; otherwise, the original processing is maintained.

10. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: The aforementioned lateral optimization specifically refers to: Calculate the ratio of the overlap between the two smallest outer rectangles in the horizontal direction, and merge the smallest outer rectangles in the horizontal direction based on the calculation results.

11. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: In step S3, after completing the vertical and horizontal optimization processes, a correction process for the rectangle length is established on the optimization results, thereby forming the final Luoping line detection.

12. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics as described in claim 11, characterized in that: The correction for the rectangle length is achieved by adjusting the correction coefficient in the following formula. The confirmation is complete. ； In the formula, The number of white pixels within the corresponding segment; : The x-coordinate of the white pixel; : The ordinate of the white pixel; : The x-coordinate of the center of the corresponding outer rectangle; : The ordinate of the center coordinate of the corresponding outer rectangle; : The horizontal dimension of the outer rectangle; : The vertical dimension of the outer rectangle; Correction factor.

13. The method for detecting and evaluating the surface flatness lines of aluminum alloy sheets based on machine vision and feature parameter statistics according to claim 1, characterized in that: The set value is 0.2.

Images