An intelligent inkjet positioning method for marble tiles based on image recognition

Abstract

The present invention discloses an intelligent inkjet positioning method for marble tiles based on image recognition, which belongs to the technical field of image processing. The present invention constructs a first and a second texture image by shooting a tile image and performing adjacent row and column operations, and then extracts clear texture features through binarization processing. Then, the binarized image is skeletonized to obtain a more simplified texture skeleton image. By analyzing the similarity of the curved segments and straight segments of the texture line, the positioning line is screened out, and finally the positioning point of the tile is determined and its deviation position is identified. The present invention finds the corresponding positioning line from both texture skeleton images, combines the two positioning lines, obtains the positioning point, realizes the re-positioning of the tile, accurately identifies the deviation position of the tile, and reduces the texture incoherence problem caused by position deviation.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to an intelligent inkjet positioning method for marble tiles based on image recognition. Background Art

[0002] In the field of modern ceramic production, marble-textured tiles are highly favored by the market for their realistic natural stone appearance and diversity. At present, the inkjet process of marble-textured tiles has achieved mechanized production, but the existing inkjet positioning technology still has significant deficiencies. Traditional inkjet equipment moves tiles to designated positions through conveyor belts. However, due to objective factors such as mechanical vibrations during the production process and slight displacement of the conveyor belt, the actual position of the tiles often deviates from the expected position. In this case, the tiles to be inkjetted need to be repositioned to reduce defects such as discontinuous and misaligned tile textures. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides an intelligent inkjet positioning method for marble tiles based on image recognition, which solves the problem that the prior art cannot accurately identify the deviated position of tiles.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: an intelligent inkjet positioning method for marble tiles based on image recognition, comprising the following steps:

[0005] S1, capturing an image of the placed marble tiles, performing adjacent row operations and adjacent column operations on the image, and constructing a first texture image and a second texture image;

[0006] S2, binarizing the first texture image and the second texture image respectively to obtain a first texture binarized image and a second texture binarized image;

[0007] S3, skeletonizing the textures in the first texture binary image and the second texture binary image to obtain a first texture skeleton image and a second texture skeleton image;

[0008] S4, selecting a positioning line from each texture line according to the similarity of curved segments and straight segments of texture lines in the first texture skeleton image and the second texture skeleton image;

[0009] S5. Determine a positioning point according to the positioning lines on the first texture skeleton image and the second texture skeleton image, and obtain the deviation position of the marble tile based on the positioning point.

[0010] Furthermore, S1 comprises the following sub-steps:

[0011] S11, taking an image of the placed marble tiles, taking the pixel value of each row in the image as a row vector, and taking the pixel value of each column as a column vector;

[0012] S12, subtract the nth row vector from the n+1th row vector bit by bit, and take the absolute value of each subtraction result to obtain the nth row texture vector, and form a first texture image with the texture vectors of each row, where n is a positive integer;

[0013] S13. Subtract the m column vector from the m+1 column vector bit by bit, and take the absolute value of each subtraction result to obtain the mth column texture vector, and use the column texture vectors to form a second texture image, where m is a positive integer.

[0014] Further, S2 includes the following sub-steps:

[0015] S21, extracting the pixel value of each pixel point on the first texture image, calculating the average value of each pixel value to obtain a first average value, setting the pixel values ​​greater than the first average value to 1, and setting the pixel values ​​of other pixels to 0, to obtain a first texture binary image;

[0016] S22, extracting the pixel value of each pixel point on the second texture image, calculating the average value of each pixel value to obtain a second average value, setting the pixel values ​​greater than the second average value to 1, and setting the pixel values ​​of other pixels to 0, to obtain a second texture binary image.

[0017] Furthermore, S3 includes the following sub-steps:

[0018] S31, taking each connected area formed by the pixel points with a pixel value of 1 on the first texture binary image as a texture contour, retaining the pixel values ​​of the edge pixels in each texture contour, and setting the pixel values ​​of other pixels to 0, to obtain a first texture skeleton image;

[0019] S32, taking each connected area formed by the pixel points with a pixel value of 1 on the second texture binary image as a texture contour, retaining the pixel values ​​of the edge pixels in each texture contour, and setting the pixel values ​​of other pixels to 0, to obtain a second texture skeleton image.

[0020] Furthermore, the processing of the first texture skeleton image and the second texture skeleton image in S4 includes the following sub-steps:

[0021] S41, taking pixel points with a pixel value of 1 on the texture skeleton image as texture points, and connecting adjacent texture points to obtain a plurality of texture lines;

[0022] S42, calculating the shape factor of the pixel point on each texture line;

[0023] S43, dividing the texture line into curved segments and straight segments according to the shape factor of the pixel point;

[0024] S44, calculating the similarity of the curved segments according to the number and curvature of the curved segments on each texture line;

[0025] S45, calculating the straight line segment similarity according to the number and linearity of the straight line segments on each texture line;

[0026] S46, adding the similarity of the curved segment and the similarity of the straight segment on the same texture line to obtain a similarity value;

[0027] S47, selecting the texture line with the largest similarity value as the positioning line of the tile.

[0028] Furthermore, the formula for calculating the shape factor of the pixel in S42 is: , where α i is the shape factor of the i-th pixel, x i is the horizontal coordinate of the i-th pixel on the texture line, y i is the ordinate of the i-th pixel on the texture line, i and j are positive integers, x i+j is the horizontal coordinate of the i+jth pixel on the texture line, y i+j is the ordinate of the i+jth pixel on the texture line, x i+j-1 is the horizontal coordinate of the i+j-1th pixel on the texture line, y i+j-1 is the ordinate of the i+j-1th pixel on the texture line, x i+N is the horizontal coordinate of the i+Nth pixel on the texture line, y i+N is the ordinate of the i+Nth pixel on the texture line, the pixel (x i ,y i ) is the starting point, (x i+N , y i+N ) is the end point, and N is the number of pixels from the start point to the end point.

[0029] Further, S43 includes the following sub-steps:

[0030] S431, when the shape factor of the i-th pixel is greater than a threshold, mark the area between the i-th pixel and the (i-1)-th pixel as a segmentation point, where i is a positive integer;

[0031] S432, segmenting the texture line according to each segmentation point, when a segment contains multiple pixels, the segment is marked as a straight line segment, and when a segment contains one pixel, the pixel is marked as a suspected corner point;

[0032] S433: Discard isolated suspected corner points, and connect adjacent suspected corner points to obtain a curved segment.

[0033] Furthermore, the curvature in S44 is: the average curvature of each pixel point on the curved segment;

[0034] The formula for calculating the similarity of curved segments in S44 is: , where S c is the similarity of curved segments, K is the number of curved segments on the texture line, | | is the absolute value operation, c k is the curvature of the kth curved segment on the texture line, c tar is the target curvature, k is a positive integer, and f is a normalized function.

[0035] Furthermore, the linearity in S45 is: the average value of the shape factor of each pixel point on the straight line segment;

[0036] The formula for calculating the similarity of straight line segments in S45 is: , where S r is the straight line segment similarity, f is the normalization function, L is the number of straight line segments on the texture line, r k is the linearity of the kth straight line segment on the texture line, r tar is the target linearity, || is the absolute value operation, and k is a positive integer.

[0037] Further, S5 includes the following sub-steps:

[0038] S51, extracting a geometric center point from the positioning line on the first texture skeleton image to obtain a first center point;

[0039] S52, extracting a geometric center point from the positioning line on the second texture skeleton image to obtain a second center point;

[0040] S53, taking the average of the first center point and the second center point to obtain a positioning point: , , where x p is the horizontal coordinate of the positioning point, y p is the ordinate of the positioning point, x o,1 is the horizontal coordinate of the first center point, y o,1 is the ordinate of the first center point, x o,2 is the horizontal coordinate of the second center point, y o,2 is the ordinate of the second center point;

[0041] S54, subtracting the positioning point from the stored positioning point to obtain the deviation position of the marble tile.

[0042] The beneficial effects of the present invention are as follows: the present invention constructs the first and second texture images by shooting the tile image and performing adjacent row and column operations, and then extracts clear texture features through binarization processing. Next, the binarized image is skeletonized to obtain a more simplified texture skeleton image. By analyzing the similarity between the curved segments and the straight segments of the texture line, the positioning line is screened out, and finally the positioning point of the tile is determined and its deviation position is identified. The present invention finds the corresponding positioning line from both texture skeleton images, combines the two positioning lines, obtains the positioning point, realizes the re-positioning of the tile, accurately identifies the deviation position of the tile, and reduces the texture incoherence problem caused by position deviation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of an intelligent inkjet positioning method for marble tiles based on image recognition;

[0044] Figure 2 Schematic diagram of the formation process of the connected region;

[0045] Figure 3 A schematic diagram of the tile outline. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, a marble tile intelligent inkjet positioning method based on image recognition includes the following steps:

[0048] S1, capturing an image of the placed marble tiles, performing adjacent row operations and adjacent column operations on the image, and constructing a first texture image and a second texture image;

[0049] S2, binarizing the first texture image and the second texture image respectively to obtain a first texture binarized image and a second texture binarized image;

[0050] S3, skeletonizing the textures in the first texture binary image and the second texture binary image to obtain a first texture skeleton image and a second texture skeleton image;

[0051] S4, selecting a positioning line from each texture line according to the similarity of curved segments and straight segments of texture lines in the first texture skeleton image and the second texture skeleton image;

[0052] S5. Determine a positioning point according to the positioning lines on the first texture skeleton image and the second texture skeleton image, and obtain the deviation position of the marble tile based on the positioning point.

[0053] In this embodiment, S1 includes the following sub-steps:

[0054] S11, taking an image of the placed marble tiles, taking the pixel value of each row in the image as a row vector, and taking the pixel value of each column as a column vector;

[0055] S12, subtract the nth row vector from the n+1th row vector bit by bit, and take the absolute value of each subtraction result to obtain the nth row texture vector, and form a first texture image with the texture vectors of each row, where n is a positive integer;

[0056] S13. Subtract the m column vector from the m+1 column vector bit by bit, and take the absolute value of each subtraction result to obtain the mth column texture vector, and use the column texture vectors to form a second texture image, where m is a positive integer.

[0057] In step S12, after the nth row vector is bitwise subtracted from the n+1th row vector, the n+1th row vector is bitwise subtracted from the n+2th row vector, and so on, until the original image is traversed, and the nth row texture vector in the first texture image is used as the pixel value of the nth row.

[0058] In step S23, after the m column vector is bitwise subtracted from the m+1 column vector, the m+1 column vector is bitwise subtracted from the m+2 column vector, and so on, until the original image is traversed, and the mth column texture vector in the second texture image is used as the pixel value of the mth column. The present invention subtracts adjacent rows and adjacent columns to enhance the pixel texture in both the horizontal and vertical directions.

[0059] In this embodiment, S2 includes the following sub-steps:

[0060] S21, extracting the pixel value of each pixel point on the first texture image, calculating the average value of each pixel value to obtain a first average value, setting the pixel values ​​greater than the first average value to 1, and setting the pixel values ​​of other pixels to 0, to obtain a first texture binary image;

[0061] S22, extracting the pixel value of each pixel point on the second texture image, calculating the average value of each pixel value to obtain a second average value, setting the pixel values ​​greater than the second average value to 1, and setting the pixel values ​​of other pixels to 0, to obtain a second texture binary image.

[0062] Since the pixel values ​​between adjacent rows and adjacent columns are subtracted in step S1, the larger the pixel value of the pixel point in the first texture image and the second texture image, the greater the possibility that the pixel point is on the texture. The average value of the pixel values ​​on the first texture image and the second texture image is taken, the pixel value greater than the average value is selected, and the pixel value is set to 1, so that the texture area is distinguished from the smooth area.

[0063] In this embodiment, S3 includes the following sub-steps:

[0064] S31, taking each connected area formed by the pixel points with a pixel value of 1 on the first texture binary image as a texture contour, retaining the pixel values ​​of the edge pixels in each texture contour, and setting the pixel values ​​of other pixels to 0, to obtain a first texture skeleton image;

[0065] S32, taking each connected area formed by the pixel points with a pixel value of 1 on the second texture binary image as a texture contour, retaining the pixel values ​​of the edge pixels in each texture contour, and setting the pixel values ​​of other pixels to 0, to obtain a second texture skeleton image.

[0066] In the present invention, the process of obtaining the connected area is as follows: randomly select a pixel point with a pixel value of 1 from the texture binary image as the starting point, search from the neighborhood of the starting point whether there are other pixel points with a pixel value of 1, if so, classify the starting point and other pixel points into one area, then search for a pixel point that has not been a starting point in the area as a new starting point, then search from the neighborhood of the new starting point whether there are other unpartitioned pixel points with a pixel value of 1, if so, classify the other unpartitioned pixel points with a pixel value of 1 into the area, until all the pixel points in the area have been the starting point, and the area is a connected area.

[0067] A connected region is an image region consisting of pixels with the same pixel value and adjacent positions.

[0068] like Figure 2 As shown, for example, the pixel value of the starting point (1,1) is 1, and the 8-neighborhood of (1,1) is searched: (2,1) is 1, added to the region, (2,2) is 1, added to the region, and then from the neighborhoods of (2,1) and (2,2), (3,2) is 1, so the first connected region is: [(1,1), (2,1), (2,2), (3,2)]; the pixel value of the starting point (1,4) is 1, and the 8-neighborhood of (1,4) is searched: (2,5) is 1, added to the region, and then the pixel point that has not been the starting point is selected from the region, that is, (2,5), and from the neighborhood of (2,5), (3,4) is 1, so the second connected region is: [(1,4), (2,5), (3,4)].

[0069] The present invention sets the pixel values in the middle area of the texture contour on the texture binary image to 0, retains the pixel points at the edge of the texture contour, and realizes the thinning of the texture contour.

[0070] In this embodiment, the processing of the first texture skeleton image and the second texture skeleton image in S4 includes the following sub-steps:

[0071] S41: Take the pixel points with a pixel value of 1 on the texture skeleton image as texture points, connect adjacent texture points, and obtain multiple texture lines;

[0072] S42: Calculate the shape factor of the pixel points on each texture line;

[0073] S43: Divide the texture line into a curved segment and a straight segment according to the shape factor of the pixel points;

[0074] S44: Calculate the similarity of the curved segments according to the number of curved segments and the degree of curvature on each texture line;

[0075] S45: Calculate the similarity of the straight segments according to the number of straight segments and the linearity on each texture line;

[0076] S46: Add the similarity of the curved segments and the similarity of the straight segments on the same texture line to obtain a similarity value;

[0077] S47: Select the texture line with the largest similarity value as the positioning line of the tile.

[0078] The present invention connects adjacent texture points (i.e., contacting texture points) to obtain multiple texture lines, calculates the shape factor of each pixel point on each texture line, divides the texture line into a curved segment and a straight segment according to the shape factor, calculates the similarity of the curved segment and the straight segment respectively, and comprehensively screens out the positioning line of the tile.

[0079] In this embodiment, the formula for calculating the shape factor of the pixel points in S42 is: , where α i is the shape factor of the i-th pixel point, x i is the abscissa of the i-th pixel point on the texture line, y i is the ordinate of the i-th pixel point on the texture line, i and j are positive integers, x i+j is the abscissa of the (i + j)-th pixel point on the texture line, y i+j is the ordinate of the (i + j)-th pixel point on the texture line, x i+j-1 is the abscissa of the (i + j - 1)-th pixel point on the texture line, y i+j-1 is the ordinate of the (i + j - 1)-th pixel point on the texture line, x i+N is the abscissa of the (i + N)-th pixel point on the texture line, y i+Nis the ordinate of the i+Nth pixel on the texture line, the pixel (x i ,y i ) is the starting point, (x i+N , y i+N ) is the end point, and N is the number of pixels from the start point to the end point.

[0080] In the present invention, the i-th pixel point is taken as the starting point, a texture line segment with a length of N is taken, and the ratio of the actual length to the straight-line distance is calculated to obtain a shape factor for evaluating the shape of the segment.

[0081] In this embodiment, S43 includes the following sub-steps:

[0082] S431, when the shape factor of the i-th pixel is greater than a threshold, mark the area between the i-th pixel and the (i-1)-th pixel as a segmentation point, where i is a positive integer;

[0083] S432, segmenting the texture line according to each segmentation point, when a segment contains multiple pixels, the segment is marked as a straight line segment, and when a segment contains one pixel, the pixel is marked as a suspected corner point;

[0084] S433: Discard isolated suspected corner points, and connect adjacent suspected corner points to obtain a curved segment.

[0085] In this embodiment, when the shape factor of the pixel is close to 1, the segment is a straight line. When the shape factor of the pixel is larger, the segment is more curved. Therefore, the present invention sets the threshold value to 1.2.

[0086] In the present invention, an isolated suspected corner point refers to a point without other suspected corner points adjacent to it. Other adjacent suspected corner points are connected to form a curved segment, thereby realizing the segmentation of the straight segment and the curved segment.

[0087] In this embodiment, the curvature in S44 is: the average curvature of each pixel point on the curved segment;

[0088] The formula for calculating the similarity of curved segments in S44 is: , where S c is the similarity of curved segments, K is the number of curved segments on the texture line, | | is the absolute value operation, c k is the curvature of the kth curved segment on the texture line, c tar is the target curvature, k is a positive integer, and f is a normalized function.

[0089] In the present invention, the normalization function adopts the tanh hyperbolic tangent function.

[0090] In this embodiment, the linearity in S45 is: the average value of the shape factor of each pixel point on the straight line segment;

[0091] The formula for calculating the similarity of straight line segments in S45 is: , where S r is the straight line segment similarity, f is the normalization function, L is the number of straight line segments on the texture line, r k is the linearity of the kth straight line segment on the texture line, r tar is the target linearity, || is the absolute value operation, and k is a positive integer.

[0092] The present invention uses the tile outline as the positioning line, such as Figure 3 As shown, therefore, there are 4 straight segments and 4 curved segments. The present invention calculates the difference between the linearity of each straight segment and the target linearity, calculates the difference between the curvature of each curved segment and the target curvature, and uses the difference in the number of straight segments and the difference in the number of curved segments for enhancement. When the number of straight segments is 4 and the linearity is similar, the straight segment similarity can reach the maximum; when the number of curved segments is 4 and the curvature is similar, the curved segment similarity can reach the maximum.

[0093] In the present invention, the target linearity and the target curvature are pre-stored target values.

[0094] In this embodiment, S5 includes the following sub-steps:

[0095] S51, extracting a geometric center point from the positioning line on the first texture skeleton image to obtain a first center point;

[0096] S52, extracting a geometric center point from the positioning line on the second texture skeleton image to obtain a second center point;

[0097] S53, taking the average of the first center point and the second center point to obtain a positioning point: , , where x p is the horizontal coordinate of the positioning point, y p is the ordinate of the positioning point, x o,1 is the horizontal coordinate of the first center point, y o,1 is the ordinate of the first center point, x o,2 is the horizontal coordinate of the second center point, y o,2 is the ordinate of the second center point;

[0098] S54, subtracting the positioning point from the stored positioning point to obtain the deviation position of the marble tile.

[0099] The present invention obtains positioning points through positioning lines on two texture skeleton images, improves the accuracy of obtaining positioning points, subtracts the obtained positioning points from the stored positioning points, and obtains the deviation position of the marble tile. After the deviation position of the marble tile is known, the position of the inkjet port can be adjusted for position calibration.

[0100] The present invention constructs the first and second texture images by shooting the tile images and performing adjacent row and column operations, and then extracts clear texture features through binarization processing. Then, the binarized image is skeletonized to obtain a more simplified texture skeleton image. By analyzing the similarity between the curved segments and the straight segments of the texture line, the positioning line is screened out, and finally the positioning point of the tile is determined and its deviation position is identified. The present invention finds the corresponding positioning line from both texture skeleton images, combines the two positioning lines, obtains the positioning point, realizes the re-positioning of the tile, accurately identifies the deviation position of the tile, and reduces the texture incoherence problem caused by position deviation.

[0101] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent inkjet positioning method for marble tiles based on image recognition, characterized in that: The following steps are involved: S1, capturing an image of the placed marble tiles, performing adjacent row operations and adjacent column operations on the image, and constructing a first texture image and a second texture image; S2, binarizing the first texture image and the second texture image respectively to obtain a first texture binarized image and a second texture binarized image; S3, skeletonizing the textures in the first texture binary image and the second texture binary image to obtain a first texture skeleton image and a second texture skeleton image; S4, selecting a positioning line from each texture line according to the similarity of curved segments and straight segments of texture lines in the first texture skeleton image and the second texture skeleton image; S5, determining a positioning point according to the positioning lines on the first texture skeleton image and the second texture skeleton image, and obtaining a deviation position of the marble tile based on the positioning point; The processing of the first texture skeleton image and the second texture skeleton image in S4 comprises the following sub-steps: S41, taking pixel points with a pixel value of 1 on the texture skeleton image as texture points, and connecting adjacent texture points to obtain a plurality of texture lines; S42, calculating the shape factor of the pixel point on each texture line; S43, dividing the texture line into curved segments and straight segments according to the shape factor of the pixel point; S44, calculating the similarity of the curved segments according to the number and curvature of the curved segments on each texture line; S45, calculating the straight line segment similarity according to the number and linearity of the straight line segments on each texture line; S46, adding the similarity of the curved segment and the similarity of the straight segment on the same texture line to obtain a similarity value; S47, selecting the texture line with the largest similarity value on the texture skeleton image as the positioning line of the tile; The formula for calculating the shape factor of the pixel in S42 is: , where α i is the shape factor of the i-th pixel, x i is the horizontal coordinate of the i-th pixel on the texture line, y i is the ordinate of the i-th pixel on the texture line, i and j are positive integers, x i+j is the horizontal coordinate of the i+jth pixel on the texture line, y i+j is the ordinate of the i+jth pixel on the texture line, x i+j-1 is the horizontal coordinate of the i+j-1th pixel on the texture line, y i+j-1 is the ordinate of the i+j-1th pixel on the texture line, x i+N is the horizontal coordinate of the i+Nth pixel on the texture line, y i+N is the ordinate of the i+Nth pixel on the texture line, the pixel (x i ,y i ) is the starting point, (x i+N, y i+N ) is the end point, and N is the number of pixels from the start point to the end point.

2. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The S1 comprises the following sub-steps: S11, taking an image of the placed marble tiles, taking the pixel value of each row in the image as a row vector, and taking the pixel value of each column as a column vector; S12, subtract the nth row vector from the n+1th row vector bit by bit, and take the absolute value of each subtraction result to obtain the nth row texture vector, and form a first texture image with the texture vectors of each row, where n is a positive integer; S13. Subtract the m column vector from the m+1 column vector bit by bit, and take the absolute value of each subtraction result to obtain the mth column texture vector, and use the column texture vectors to form a second texture image, where m is a positive integer.

3. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The S2 comprises the following sub-steps: S21, extracting the pixel value of each pixel point on the first texture image, calculating the average value of each pixel value to obtain a first average value, setting the pixel values ​​greater than the first average value to 1, and setting the pixel values ​​of other pixels to 0, to obtain a first texture binary image; S22, extracting the pixel value of each pixel point on the second texture image, calculating the average value of each pixel value to obtain a second average value, setting the pixel values ​​greater than the second average value to 1, and setting the pixel values ​​of other pixels to 0, to obtain a second texture binary image.

4. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The S3 comprises the following sub-steps: S31, taking each connected area formed by the pixel points with a pixel value of 1 on the first texture binary image as a texture contour, retaining the pixel values ​​of the edge pixels in each texture contour, and setting the pixel values ​​of other pixels to 0, to obtain a first texture skeleton image; S32, taking each connected area formed by the pixel points with a pixel value of 1 on the second texture binary image as a texture contour, retaining the pixel values ​​of the edge pixels in each texture contour, and setting the pixel values ​​of other pixels to 0, to obtain a second texture skeleton image.

5. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The S43 comprises the following sub-steps: S431, when the shape factor of the i-th pixel is greater than a threshold, mark the area between the i-th pixel and the (i-1)-th pixel as a segmentation point, where i is a positive integer; S432, segmenting the texture line according to each segmentation point, when a segment contains multiple pixels, the segment is marked as a straight line segment, and when a segment contains one pixel, the pixel is marked as a suspected corner point; S433: Discard isolated suspected corner points, and connect adjacent suspected corner points to obtain a curved segment.

6. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The curvature in S44 is: the average curvature of each pixel point on the curved segment; The formula for calculating the similarity of the curved segments in S44 is: , where S c is the similarity of curved segments, K is the number of curved segments on the texture line, | | is the absolute value operation, c k is the curvature of the kth curved segment on the texture line, c tar is the target curvature, k is a positive integer, and f is a normalized function.

7. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The linearity in S45 is: the average value of the shape factor of each pixel point on the straight line segment; The formula for calculating the straight line segment similarity in S45 is: , where S r is the straight line segment similarity, f is the normalization function, L is the number of straight line segments on the texture line, r k is the linearity of the kth straight line segment on the texture line, r tar is the target linearity, || is the absolute value operation, and k is a positive integer.

8. The method for intelligent inkjet positioning of marble tiles based on image recognition according to claim 1 is characterized in that: The S5 comprises the following sub-steps: S51, extracting a geometric center point from the positioning line on the first texture skeleton image to obtain a first center point; S52, extracting a geometric center point from the positioning line on the second texture skeleton image to obtain a second center point; S53, taking the average of the first center point and the second center point to obtain a positioning point: , , where x p is the horizontal coordinate of the positioning point, y p is the ordinate of the positioning point, x o,1 is the horizontal coordinate of the first center point, y o,1 is the ordinate of the first center point, x o,2 is the horizontal coordinate of the second center point, y o,2 is the ordinate of the second center point; S54, subtracting the positioning point from the stored positioning point to obtain the deviation position of the marble tile.

Images