CN107063458B - Ceramic tile coloration piecemeal detection method based on machine vision - Google Patents

Abstract

The invention discloses a tile chromaticity block detection method based on machine vision, including: step 1, image acquisition and positioning; step 2, image pose correction: geometric transformation is performed based on the oblique circumscribed moment diagonal of the ROI image to correct the tiles Pose; step 3, image segmentation: use the horizontal and vertical segmentation strategy to evenly divide the ROI image corrected in step 2 into test modules, and the test module is the smallest ROI image; step 4, composite mask: use composite logic operations to segment The masked test module is processed; step 5, color difference detection, the masked test module is converted to the HSV color space, and the color feature is generated and compared with the standard value to determine whether there is a color defect in the tile. The invention combines affine transformation, block processing, composite mask, color space, and feature extraction technology to complete the visual color difference detection of ceramic tiles. The detection performance is good, and the problem of inability to truly represent the inherent color color difference detection of ceramic tiles is solved.

Description

Chromatic block detection method of ceramic tiles based on machine vision

technical field

The invention relates to a machine vision-based tile chromaticity block detection method.

Background technique

At present, in the process of industrial production, the color difference detection of tiles is mainly completed by manual naked eye observation. This method is labor-intensive. Affected by subjective factors, the artificial cost is gradually increased, which can no longer meet the needs of modern automated production; there are also color detectors. The related methods of ceramic tile color difference detection, such as the speed detector mainly rely on the movement of the horizontal guide rail structure to scan the target, the scanning speed is relatively slow and the chromaticity detection relies too much on the manufacturer's color system, it is difficult to change and adapt to the actual product color; based on vision There are many methods for tile color difference detection, but the proposed method has great limitations, mainly in the following points:

(1) Before the tile detection, it is necessary to accurately locate the detected tiles to count the color feature information of the tiles, but it is still impossible to distinguish the pixel-level interference features of the edge neighborhood, so when performing color features, it will inevitably be brought into the environment influence of factors;

(2) Statistics are mainly based on global information for color feature statistics. This feature has little perception of local color transformation, that is, global information statistics will dilute the performance of local color difference transformation, so the accuracy and accuracy of color difference detection will be affected. influences;

(3) The selection of color space will have a great impact on the color difference detection. At present, there is no general color space used by all product color difference detection, and it is only selected according to the actual needs of the product. Based on the current situation and engineering application background of the ceramic tile color difference detection industry, this patent proposes a color difference detection algorithm based on block-based composite masks.

Contents of the invention

In order to solve the above problems, the present invention provides a tile chromaticity block detection method based on machine vision, which combines affine transformation, block processing, composite mask, color space, and feature extraction technology to complete the visual color difference detection of tiles. The detection performed well, and solved the problem that the traditional detection method could not truly represent the inherent color and color difference detection of tiles due to limitations and defects.

To achieve the above object, the technical solution adopted in the present invention is:

The tile chromaticity block detection method based on machine vision comprises the following steps:

Step 1. Image acquisition and positioning

The tile image is collected by a color camera, and then the ROI image containing only the tile is segmented through the tile positioning algorithm, and it is used as an independent processing unit for subsequent processing, and the ROI image includes the image of the boundary area of the entire tile;

Step 2. Image pose correction

Use the oblique circumscribed moment diagonal of the ROI image as the reference for geometric transformation to correct the tile pose;

Step 3, image segmentation

Use the horizontal and vertical block strategy to evenly divide the ROI image corrected in step 2 into test modules, and the test module is the smallest ROI image;

Step 4. Composite mask

Mask processing is performed on the divided test modules by using a compound logic operation method;

Step 5, color difference detection

Convert the masked test module to the HSV color space, generate color features and compare them with standard values to determine whether there are color defects in the tiles.

Described step 1 specifically comprises the following steps:

Step 11, collecting images by a high-speed linear array CCD camera;

Step 12, grayscale the image first, then perform filtering and denoising processing, segment the filtered image using an edge detection algorithm, and then extract the outline of small tiles to obtain a tile outline image;

Step 13, based on the geometric characteristics of the tile outline in step 12, corresponding primitive features are generated, and the primitive features include width, height and integrity; the corresponding HSV color space features are generated using the color characteristics of the tile itself, and the HSV color space The features include hue and saturation, and then the primitive features and color features are constructed into a positioning combination feature to jointly complete the positioning of the tiles;

Step 14, for the positioned tile image, a corresponding ROI image is generated based on the corner points of the circumscribed rectangle of the tile outline.

The ROI image is carried out to the step 2 geometric transformation, the formula is:

In the formula, θ _c is the rotation angle of the geometric transformation, Δx, Δy are the correction translations, [x, y 1] are the coordinates before transformation, and [x ₁ y ₁ 1] are the coordinates after transformation.

Described step 3 specifically comprises the following steps:

Step 31: Re-extract the contour of the rectified image to obtain the smallest ROI image that is close to the tile target, and use the origin of the image as the starting point of the block to initialize the block step size. The block step size is defined as:

Step _x and Step _y are integers after calculation, and Scale is the scale coefficient for step calculation;

Step 32 takes the origin of the corrected minimum ROI image as the starting point of block division, and then performs block division in horizontal and vertical directions with corresponding step lengths respectively, and completes the division step of the entire tile.

The step 32 is divided into:

Divide the minimum ROI image into Zone 0 in the middle, Zone 1 on the left and right sides of Zone 0, Zone 2 on the upper and lower sides, and Zone 3 intersecting Zone 1 and Zone 2. The boundary conditions are divided into 4 types of zones, and the following g( i, j) as the discriminant function of the boundary conditions, to analyze:

The discriminant function is expressed as:

If g(i,j)=1, it means that a horizontal cross-border occurs during the block traversal process, and the range of the block size width in the critical area is [i, Width];

If g(i, j) = 2, it means that a vertical cross-border occurs during the block traversal process, and the range of the block size height in the critical area is [j, Height];

If g(i, j) = 3, it means that horizontal and vertical cross-borders occur simultaneously during the block traversal process, and the ranges of the block size width and height in the critical area are [i, Width] and [j, Height] respectively;

If g(i, j) = 0, it means that there is no cross-border during the block traversal process, and the ranges of width and height are [i, i+Step _x ]; and [j, j+Step _y ];

And define: when g(i,j)={1,2,3}, the divided area is the outer block, the area when g(i,j)=0 is the inner block, and the inner block is unnecessary For the part where the pixel discrimination process is performed, only the outer blocks need to be discriminated.

Described step 4 specifically comprises the following steps:

Step 41: Initialize two images with the same size as the segmented ROI and with a pixel value of 0, respectively as the matrix of the composite tile mask, denoted as M1 and M2;

Step 42: Draw the corrected outline in the tile block step in the M1 and M2 mask images with pixel precision, and fill the inner and outer areas of the mask parent outline; that is, fill the pixels inside the M1 image outline as RGB (255, 255, 255), the outside is RGB(0,0,0); the inner and outer filling pixels of M2 are RGB(255,255,255) and RGB(255,0,0);

Step 43: Perform logical AND operations on the pose-corrected image and the composite mask images M1 and M2 respectively to obtain the processed mask ROI images I1 and I2. After processing the inner mask and the corrected image, the inside of the tile outline The images of are all preserved, and the external background colors are black and red respectively.

Step 44: When g(i, j) = {1, 2, 3} in the tile block step, the block is at a critical boundary value, since it is impossible to ensure that the edge of the tile completely coincides with the segmented ROI image, it is necessary to perform a tile The discriminative process of the inner and outer pixels, the discriminant rule is: if the values of the pixels in the boundary block in I1 and I2 satisfy I1(i, j)=RGB(0,0,0) and I1(i, j) =RGB(255,0,0), then the pixel value is a mixed background pixel in the boundary block, otherwise it is a pixel point on the tile.

The step 5 color difference detection is specifically:

Convert the ROI image after block and mask to HSV color space, and generate color features based on the weighted HSV color space components, the expression is as follows:

F _ti =(λ ₁ ×H+λ ₂ ×S)/N

Ft _i represents the color feature generated by the i-th block, H represents the hue component, and the corresponding weight is λ ₁ , S represents the saturation corresponding to the weight component λ ₂ , N represents the number of pixels traversed in the corresponding block, and λ ₁ = 0.7 or λ ₂ = 0.3;

Step 52: Calculate the generated tile color feature set F{F _ti |F _ti , i=1, 2,...N, k∈N+ and N≥1}, the standard tile color feature is F _m , and the expression of color difference detection is :

ΔE _ti =|F _ti -F _m |

The threshold threshold method is used to discriminate the positioning target, and the discriminating formula is as follows.

T is the upper and lower limits of the threshold, g(ΔE _ti )=1, which means that there is no color deviation in the block position of the tile, and only when the g() function corresponding to all blocks of the entire tile is 1, it is judged that there is no color in the tile Deviation, if any one of them corresponds to g()=0, it is judged that there is a chromaticity deviation in the tile.

After adopting the technical scheme, the present invention has the following advantages:

1. The horizontal and vertical block strategy is used to divide the tile target into a uniform detection module, which overcomes the defect that the traditional global information feature cannot perceive small color difference changes, and can highlight the change information of local color difference;

2. Adopt the ceramic tile pose correction method based on the geometric transformation of the diagonal external moment of the tile outline to ensure the effect of the horizontal and vertical block strategy;

3. In view of the problem that the block unit will overflow the tile boundary and affect the color feature statistics, the composite mask logic operation is used to prevent the non-tile pixels in the cross-border block from participating in the color feature statistics, thereby ensuring the accuracy of color expression features;

4. Use the HSV color space to detect the color difference of the segmented blocks, generate color features and compare them with the standard values, identify the color difference defects of tiles, and accurately locate the defects.

Description of drawings

The accompanying drawings described here are used to provide a further understanding of the present invention, and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute improper limitations to the present invention. In the attached picture:

Fig. 1 is a schematic flow chart of the tile chromaticity separation detection method based on machine vision in the present invention;

Fig. 2 is a schematic diagram of the geometry of tile image correction in the present invention;

Fig. 3 is a block diagram of the present invention;

Fig. 4 is a diagram of color difference detection of ceramic tiles according to the present invention.

Detailed ways

In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention clearer and clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

As shown in Figure 1, the tile chromaticity block detection method based on machine vision disclosed by the present invention comprises the following steps:

Step 1, image acquisition and positioning, its step 1 specifically includes the following steps:

Step 11, collecting images by a high-speed linear array CCD camera;

Step 12, grayscale the image first, then perform filtering and denoising processing, segment the filtered image using an edge detection algorithm, and then extract the outline of small tiles to obtain a tile outline image;

Step 13, based on the geometric characteristics of the tile outline in step 12, corresponding primitive features are generated, and the primitive features include width, height and integrity; the corresponding HSV color space features are generated using the color characteristics of the tile itself, and the HSV color space The features include hue and saturation, and then the primitive features and color features are constructed into a positioning combination feature to jointly complete the positioning of the tiles;

Step 14, for the positioned tile image, a corresponding ROI image is generated based on the corner points of the circumscribed rectangle of the tile outline.

The ROI image is carried out to the step 2 geometric transformation, the formula is:

In the formula, θ _c is the rotation angle of the geometric transformation, Δx, Δy are the correction translations, [x, y 1] are the coordinates before transformation, and [x ₁ y ₁ 1] are the coordinates after transformation.

Step 2. Image pose correction

The geometric transformation based on the oblique circumscribed moment diagonal of the ROI image is used as the reference to correct the tile pose; the geometric diagram of the correction is shown in Figure 2.

Let the oblique circumscribed rectangle of the tile at the standard position be ABCD, the intersection of the diagonals be p(x _p , y _p ), the angle between the diagonal BD and the pixel coordinates in the W direction is θ, and the coordinates of points B and D are respectively set to (x _B ,y _B ), (x _D ,y _D ), similarly, let the oblique circumscribed rectangle of any target tile be A ₁ B ₁ C ₁ D ₁ , and the diagonal intersection point p ₁ (x _p1 ,y _p1 ), B ₁ D ₁ on the diagonal, and the angle between the pixel coordinates in the W direction is θ ₁ , and the coordinates of B ₁ and D ₁ are respectively set to (xB ₁ , y _B1 ), (x _D1 , y _D1 ) , W and H respectively represent the horizontal and vertical axes of the pixel coordinates, the correction calculation is based on the pixel dimension, the correction translation is set to Δx, Δy, the rotation angle is set to θ _c , and the geometric transformation formula is:

In the formula, θ _c is the rotation angle of the geometric transformation, Δx, Δy are the correction translations, [x, y 1] are the coordinates before transformation, and [x ₁ y ₁ 1] are the coordinates after transformation.

Step 3, image segmentation, its step 3 specifically includes the following steps:

Step 31: Re-extract the contour of the rectified image, obtain the smallest ROI image close to the tile target (this image is considered to be a tile image during processing), take the origin of the image as the starting point of the block, initialize the block step size, and divide The block step size is defined as:

Step _x and Step _y are integers taken after calculation, and Scale is the scale coefficient for step calculation; the smaller the value, the higher the accuracy of the block, and the time complexity of the algorithm will also increase, and the general value is 0.1;

Step 32, taking the origin of the corrected minimum ROI image as the starting point of the block, and then performing block division in the horizontal direction and the vertical direction with corresponding step lengths respectively, and completing the division step of the entire tile.

However, in the actual division project, it is difficult to ensure that the image length and width are integer multiples of the block step size. In order to prevent processing exceptions from occurring when the block crosses the boundary, the following discrimination conditions are added:

As shown in Figure 3, the minimum ROI image is divided into Zone 0 in the middle, Zone 1 on the left and right sides of Zone 0, Zone 2 on the upper and lower sides, and Zone 3 where Zone 1 and Zone 2 intersect. The boundary conditions are divided into four types of regions. , using the following g(i, j) as the discriminant function of the boundary conditions for analysis:

The discriminant function is expressed as:

If g(i,j)=1, it means that a horizontal cross-border occurs during the block traversal process, and the range of the block size width in the critical area is [i, Width];

If g(i, j) = 2, it means that a vertical cross-border occurs during the block traversal process, and the range of the block size height in the critical area is [j, Height];

If g(i, j) = 3, it means that horizontal and vertical cross-borders occur simultaneously during the block traversal process, and the ranges of the block size width and height in the critical area are [i, Width] and [j, Height] respectively;

If g(i, j) = 0, it means that there is no cross-border during the block traversal process, and the ranges of width and height are [i, i+Step _x ]; and [j, j+Step _y ];

And define: when g(i,j)={1,2,3}, the divided area is the outer block, the area when g(i,j)=0 is the inner block, and the inner block is unnecessary For the part where the pixel discrimination process is performed, only the outer blocks need to be discriminated.

Through the above algorithm process, the uniform division of tiles is completed, the scale of the division and the abnormal problem of crossing the border are successfully solved, and a new strategy is provided for highlighting the local color difference change;

Step 4, compound mask, it specifically comprises the following steps:

Step 41: Initialize two images with the same size as the segmented ROI and with a pixel value of 0, respectively as the matrix of the composite tile mask, denoted as M1 and M2;

Step 42: Draw the corrected outline in the tile block step in the M1 and M2 mask images with pixel precision, and fill the inner and outer areas of the mask parent outline; that is, fill the pixels inside the M1 image outline as RGB (255, 255, 255), the outside is RGB(0,0,0); the inner and outer filling pixels of M2 are RGB(255,255,255) and RGB(255,0,0);

Step 43: Perform logical AND operations on the image after external pose correction and the composite mask images M1 and M2 to obtain the processed mask ROI images I1 and I2, and the inner mask and the corrected image are processed to make the tile outline The internal images are fully preserved, and the external background colors are black and red.

Step 44: When g(i, j) = {1, 2, 3} in the tile block step, the block is at a critical boundary value, since it is impossible to ensure that the edge of the tile completely coincides with the segmented ROI image, it is necessary to perform a tile The discriminative process of the inner and outer pixels, the discriminant rule is: if the values of the pixels in the boundary block in I1 and I2 satisfy I1(i, j)=RGB(0,0,0) and I1(i, j) =RGB(255,0,0), then the pixel value is a mixed background pixel in the boundary block, otherwise it is a pixel point on the tile. The above judgment conditions only perform pixel discrimination processing on the boundary blocks, so the complexity of the entire algorithm will not be increased. Through the above-mentioned logical operation method of the composite mask, the goal of non-target pixels strictly not participating in the generation of color features can be fully realized .

Only geometrically correcting the tiles is not enough to ensure that the non-target pixels that cross the border are not strictly involved in the generation of color features, because the accuracy loss in the image processing process is inevitable, and the boundaries of the tiles cannot completely coincide with the segmented ROI image. Membrane is a very practical technology in image topology processing. The biggest feature is that it can control the region of interest of any shape, and completely shield the features that are not related to interest, so that we only care about the image target to be processed. In the present invention, not only Only the shielding features of the mask are used, and more importantly, the proposed "composite mask" logic operation method is used to identify the non-target pixels of the boundary blocks, so that they are strictly not involved in the generation of color features, thus affecting the accuracy of tile color difference detection Spend;

Step 5, color difference detection

Convert the masked test module to the HSV color space, generate color features and compare them with standard values to determine whether there are color defects in the tiles.

Step 5 color difference detection is specifically:

Convert the ROI image after block and mask to HSV color space, and generate color features based on the weighted HSV color space components, the expression is as follows:

F _ti =(λ ₁ ×H+λ ₂ ×S)/N

Ft _i represents the color feature generated by the i-th block, H represents the hue component, and the corresponding weight is λ ₁ , S represents the saturation corresponding to the weight component λ ₂ , N represents the number of pixels traversed in the corresponding block, and λ ₁ = 0.7 or λ ₂ = 0.3;

Step 52: Calculate the generated tile color feature set F{F _ti |F _ti , i=1, 2,...N, k∈N+ and N≥1}, the standard tile color feature is F _m , and the expression of color difference detection is :

ΔE _ti =|F _ti -F _m |

The threshold threshold method is used to discriminate the positioning target, and the discriminating formula is as follows.

T is the upper and lower limits of the threshold, g(ΔE _ti )=1, which means that there is no color deviation in the block position of the tile, and only when the g() function corresponding to all blocks of the entire tile is 1, it is judged that there is no color in the tile Deviation, if any one of them corresponds to g()=0, it is judged that there is a chromaticity deviation in the tile.

Adopt the detection method of the present invention to carry out experimental detection to ceramic tile and find that the detection method of the present invention can effectively detect the position of tile color mutation, as shown in Figure 4, the abscissa of statistical diagram is the number of blocks, and the ordinate is The normalized value of the feature color value F _ti .

While the above description shows and describes preferred embodiments of the present invention, it should be understood that the present invention is not limited to the forms disclosed herein and should not be viewed as excluding other embodiments, but can be used in various other combinations, modifications and environments , and can be modified within the scope of the inventive concept herein through the above teachings or techniques or knowledge in related fields. However, changes and changes made by those skilled in the art do not depart from the spirit and scope of the present invention, and should all be within the protection scope of the appended claims of the present invention.

Claims (7)

1. the ceramic tile chroma block detection method based on machine vision, it is characterized in that: comprise the following steps: Step 1. Image acquisition and positioning The tile image is collected by a color camera, and then the ROI image containing only the tile is segmented through the tile positioning algorithm, and it is used as an independent processing unit for subsequent processing, and the ROI image includes the image of the boundary area of the entire tile; Step 2. Image pose correction Use the oblique circumscribed moment diagonal of the ROI image as the reference for geometric transformation to correct the tile pose; Step 3, image segmentation Use the horizontal and vertical block strategy to evenly divide the ROI image corrected in step 2 into test modules, and the test module is the smallest ROI image; Step 4. Composite mask Mask processing is performed on the divided test modules by using a compound logic operation method; Step 5, color difference detection Convert the masked test module to the HSV color space, generate color features and compare them with standard values to determine whether there are color defects in the tiles. 2. The machine vision-based ceramic tile chromaticity block detection method as claimed in claim 1, characterized in that: said step 1 specifically comprises the following steps: Step 11, collecting images by a high-speed linear array CCD camera; Step 12, grayscale the image first, then perform filtering and denoising processing, segment the filtered image using an edge detection algorithm, and then extract the outline of small tiles to obtain a tile outline image; Step 13, based on the geometric characteristics of the tile outline in step 12, corresponding primitive features are generated, and the primitive features include width, height and integrity; the corresponding HSV color space features are generated using the color characteristics of the tile itself, and the HSV color space The features include hue and saturation, and then the primitive features and color features are constructed into a positioning combination feature to jointly complete the positioning of the tiles; Step 14, for the positioned tile image, a corresponding ROI image is generated based on the corner points of the circumscribed rectangle of the tile outline. 3. the tile chromaticity detection method based on machine vision as claimed in claim 1, is characterized in that: ROI image is carried out described step 2 geometric transformations, and formula is: In the formula, θ _c is the rotation angle of the geometric transformation, Δx, Δy are the correction translations, [x, y 1] are the coordinates before transformation, and [x ₁ y ₁ 1] are the coordinates after transformation. 4. The machine vision-based ceramic tile chromaticity block detection method as claimed in claim 1, characterized in that: said step 3 specifically comprises the following steps: Step 31: Re-extract the contour of the rectified image to obtain the smallest ROI image that is close to the tile target, and use the origin of the image as the starting point of the block to initialize the block step size. The block step size is defined as: Step _x and Step _y are integers after calculation, and Scale is the scale coefficient for step calculation; Step 32 takes the origin of the corrected minimum ROI image as the starting point of block division, and then performs block division in horizontal and vertical directions with corresponding step lengths respectively, and completes the division step of the entire tile. 5. The tile chromaticity block detection method based on machine vision as claimed in claim 4, characterized in that: said step 32 is divided into: Divide the minimum ROI image into Zone 0 in the middle, Zone 1 on the left and right sides of Zone 0, Zone 2 on the upper and lower sides, and Zone 3 intersecting Zone 1 and Zone 2. The boundary conditions are divided into 4 types of zones, and the following g( i, j) as the discriminant function of the boundary conditions, to analyze: The discriminant function is expressed as: If g(i,j)=1, it means that a horizontal cross-border occurs during the block traversal process, and the range of the block size width in the critical area is [i, Width]; If g(i, j) = 2, it means that a vertical cross-border occurs during the block traversal process, and the range of the block size height in the critical area is [j, Height]; If g(i, j) = 3, it means that horizontal and vertical cross-borders occur simultaneously during the block traversal process, and the ranges of the block size width and height in the critical area are [i, Width] and [j, Height] respectively; If g(i, j) = 0, it means that there is no cross-border during the block traversal process, and the ranges of width and height are [i, i+Step _x ]; and [j, j+Step _y ]; And define: when g(i,j)={1,2,3}, the divided area is the outer block, the area when g(i,j)=0 is the inner block, and the inner block is unnecessary For the part where the pixel discrimination process is performed, only the external blocks need to be discriminated. 6. The tile chromaticity block detection method based on machine vision as claimed in claim 5, characterized in that: said step 4 specifically comprises the following steps: Step 41: Initialize two images with the same size as the segmented ROI and with a pixel value of 0, respectively as the matrix of the composite tile mask, denoted as M1 and M2; Step 42: Draw the corrected outline in the tile block step in the M1 and M2 mask images with pixel precision, and fill the inner and outer areas of the mask parent outline; that is, fill the pixels inside the M1 image outline as RGB (255, 255, 255), the outside is RGB(0,0,0); the inner and outer filling pixels of M2 are RGB(255,255,255) and RGB(255,0,0); Step 43: Perform logical AND operations on the pose-corrected image and the composite mask images M1 and M2 respectively to obtain the processed mask ROI images I1 and I2, and the inner mask and the corrected image are processed so that the tile outline The images of all images are preserved, and the external background colors are black and red respectively; Step 44: When g(i, j) = {1, 2, 3} in the tile block step, the block is at a critical boundary value, since it is impossible to ensure that the edge of the tile completely coincides with the segmented ROI image, it is necessary to perform a tile The discriminative process of the inner and outer pixels, the discriminant rule is: if the values of the pixels in the boundary block in I1 and I2 satisfy I1(i, j)=RGB(0,0,0) and I1(i, j) =RGB(255,0,0), then the pixel value is a mixed background pixel in the boundary block, otherwise it is a pixel point on the tile. 7. The tile chromaticity block detection method based on machine vision as claimed in claim 1, characterized in that: said step 5 color difference detection is specifically: Convert the ROI image after block and mask to HSV color space, and generate color features based on the weighted HSV color space components, the expression is as follows: F _ti =(λ ₁ ×H+λ ₂ ×S)/N Ft _i represents the color feature generated by the i-th block, H represents the hue component, and the corresponding weight is λ ₁ , S represents the saturation corresponding to the weight component λ ₂ , N represents the number of pixels traversed in the corresponding block, and λ ₁ = 0.7 or λ ₂ = 0.3; Step 52: Calculate the generated tile color feature set F{F _ti |F _ti , i=1, 2,...N, k∈N+ and N≥1}, the standard tile color feature is F _m , and the expression of color difference detection is : ΔE _ti =|F _ti -F _m | The method of threshold threshold is used to discriminate the positioning target, and the discriminating formula is as follows: T is the upper and lower limits of the threshold, g(ΔE _ti )=1, which means that there is no color deviation in the block position of the tile, and only when the g() function corresponding to all blocks of the entire tile is 1, it is judged that there is no color in the tile Deviation, if any one of them corresponds to g()=0, it is judged that there is a chromaticity deviation in the tile.