Method for grading defective cocoons of flat firewood-printed cocoons

Abstract

The invention relates to the technical field of image processing, in particular to a flat firewood-printed cocoon defective cocoon discharging grading method, which comprises the following steps of: acquiring image data of flat firewood-printed cocoons, and performing self-adaptive correction on an original image to enhance a contrast ratio and a visual effect; further enhancing the image by using a multi-scale detail enhancement algorithm to obtain a second image with rich details; performing mean shift filtering on the second image, and keeping the edge clear while removing noise to obtain a third image; constructing an eight-direction gradient operator kernel and carrying out convolution operation and binarization processing on the third image to obtain a fourth image with a prominent edge; gaussian blur and morphological processing is applied to the fourth image to smooth the boundary and optimize the structure, and a fifth image is obtained; and the region of interest is extracted based on the fifth image, the area of the defect region is accurately obtained and calculated, and the grading precision and the processing efficiency are effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, in particular to a method for grading cocoons from flat-plate firewood-printed cocoons. Background Art

[0002] As the core raw material in the silk-making process, the quality of silk cocoons directly determines the quality of the final silk products. In order to ensure the high quality of silk products, the quality assessment and grading and screening of silk cocoons have become an indispensable link. Silk cocoons can be divided into upper cocoons, lower cocoons and lower cocoons according to their use value. Upper cocoons and lower cocoons can be used for silk reeling, while lower cocoons cannot or are difficult to be used for silk reeling. Flat-plate cocoons refer to cocoons that are partially flat-shaped due to cocooning on the reed curtain cocoon foil. The local flat-shaped area is the defect of the flat-plate cocoon. At present, in the quality sorting of flat-plate cocoons in my country, the lower cocoon grading and detection are mainly carried out according to the size of the defect area. However, the current detection process is still mainly manual, and the defect detection of each batch of flat-plate cocoons depends on manual visual observation. Due to the large number of silk cocoons and the different shapes of defects, long-term manual detection can easily lead to operator fatigue, which in turn affects the detection accuracy. In addition, the process is time-consuming, labor-intensive and inefficient.

[0003] Based on this, a flat-plate Chaiyin cocoon grading method was proposed. The intelligent image processing technology was used to automatically detect and segment the defective area to improve the sorting efficiency and accuracy, and effectively solve the fatigue and precision problems caused by manual inspection. Summary of the invention

[0004] The object of the present invention is to provide a method for grading cocoons from flat-plate cocoons to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for grading cocoons from flat-plate cocoons, the method comprising the following steps:

[0006] S1: Obtain the image data of the flat plate Chaiyin cocoon and perform adaptive Gamma correction on the original image to enhance the contrast and visual effect;

[0007] S2: Apply a multi-scale detail enhancement algorithm to further enhance the image to obtain a second image with rich details;

[0008] S3: performing mean shift filtering on the second image to remove noise while keeping edges clear, thereby obtaining a third image;

[0009] S4: constructing an eight-directional gradient operator kernel and performing a convolution operation and a binarization process on the third image to obtain a fourth image with a prominent edge;

[0010] S5: applying Gaussian blur and morphological processing to the fourth image to smooth the boundary and optimize the structure to obtain a fifth image;

[0011] S6: extracting the region of interest based on the fifth image, accurately obtaining the defective area of ​​the flat-plate Chaiyin cocoon and calculating the area for grading.

[0012] Preferably, the step of obtaining the image data of the flat plate Chaiyin cocoon and performing adaptive Gamma correction on the original image to enhance the contrast and visual effect comprises:

[0013] Get the image data of the flat plate Chai Yin cocoon, perform grayscale conversion on the original image I(x,y), and obtain the grayscale image I g (x,y);

[0014] In order to enhance the contrast of the image, the grayscale image I g (x, y) is processed by adaptive histogram equalization. The equalization algorithm uses the method of cropping restriction and grid size control to adjust the local contrast of each pixel and output the equalized image I c (x, y), its operation relationship is: c (x,y)=CDF(I g (x,y))·(L-1), where the clipping limit is set to 2 and the grid size is 64×64;

[0015] Based on the equalized image I c The average gray value μ of (x, y) is used to calculate the adaptive Gamma value γ to achieve the best contrast and visual effect. The calculation formula is as follows: Where μ represents the equalized image I c The average gray value of (x,y) is defined as

[0016] Apply the calculated adaptive Gamma value γ to the equalized image I c (x, y), perform Gamma correction, and obtain the image I1(x, y) after adaptive Gamma correction. The transformation formula of Gamma correction is: Where I1(x, y) represents the corrected image, and the pixel values ​​are clipped to the range of [0, 255] to ensure the effect and accuracy of the image display.

[0017] Preferably, the step of applying a multi-scale detail enhancement algorithm to further enhance the image to obtain a second image with rich details comprises:

[0018] Use three Gaussian kernel functions G of different scales σi(x, y) performs convolution operation on image I1(x, y) to generate Gaussian blurred images B1(x, y), B2(x, y), B3(x, y) of different scales, where i is 1, 2, 3, and the standard deviation σ i They are 1, 2, and 4 respectively, and the operation expression is: The structural information of the image at different scales is obtained through these blurred images;

[0019] Calculate the difference between the original image I1(x,y) and the blurred image to obtain detail images D1(x,y), D2(x,y), and D3(x,y). The expressions are as follows: D1(x,y)=I1(x,y)-B1(x,y), D2(x,y)=B1(x,y)-B2(x,y), D3(x,y)=B2(x,y)-B3(x,y). These detail images capture the image detail information at different scales respectively.

[0020] The detail images of different scales are integrated according to the set weights to generate the final detail enhanced image I2(x,y). The integration formula is: I2(x,y)=(1-w1·sign(D1(x,y)))·D1(x,y)+w2·D2(x,y)+w3·D3(x,y)+I1(x,y), where w1, w2, and w3 are weight values, set to 0.5, 0.5, and 0.25, respectively, to control the enhancement effect of details of different scales.

[0021] Preferably, the step of constructing an eight-directional gradient operator kernel and performing a convolution operation on the third image and binarization processing to obtain a fourth image with prominent edges comprises:

[0022] Construct eight directions of gradient operator kernels to extract image gradients in different directions. The eight directions include 0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°. The operator kernel for each direction is as follows: 0° direction: 45° Orientation: 90° Direction: 135° Orientation: 180° Orientation: 225° Orientation: 270° Orientation: 315° Direction:

[0023] Use the gradient operator kernel G in each direction d Perform convolution operation on the third image I3(x,y) to calculate the gradient image G in the corresponding direction d (x,y), the convolution formula is as follows: Where d is the direction angle, G d (i, j) represents the value of the gradient operator kernel in direction d at position (i, j), and I3(x+i, y+j) is the image pixel value;

[0024] The gradient images G in eight directions d (x, y) takes the absolute value of each pixel and accumulates it to get the combined gradient image G c (x,y), the combination formula is G c (x,y)=∑ d |G d (x,y)|;

[0025] For the combined gradient image G c The pixel values ​​of (x, y) are clipped to ensure that the pixel values ​​are in the range of [0, 255]. Subsequently, the Otsu thresholding method is applied to the combined gradient image G c (x, y) is binarized to obtain the final edge-highlighted binary image I4(x, y).

[0026] The characteristics and beneficial effects of the present invention are:

[0027] (1) The present invention uses adaptive Gamma correction to correct the brightness and enhance the contrast of the image, ensuring that the details of the flat-plate Chai-printed cocoon image are fully displayed, avoiding the interference of uneven illumination on defect detection, and using multi-scale detail enhancement processing to extract image detail information by calculating the difference between images of different scales, and strengthen the image defect details in a weighted manner. Through mean shift filtering, the internal area noise of the flat-plate Chai-printed cocoon defect is effectively smoothed while the edge details of the defect are maintained;

[0028] (2) The present invention constructs an eight-directional gradient operator to perform convolution operations on the image, thereby capturing the detailed edge information of the defects in multiple directions. By combining the gradients in different directions, the edge contour of the defects is effectively highlighted, and the defects with different shapes on the surface of the flat-plate firewood printing cocoon are processed more comprehensively. By combining Gaussian blur and morphological processing, noise and small defects are eliminated, ensuring that the defect contour is clear and can be effectively distinguished from the background. Through the contour extraction technology, the defects are segmented and the defect area is calculated, thereby realizing the grading of the flat-plate firewood printing cocoon. The method is simple and efficient, with an average running time of 1.04s, and can quickly complete the defect detection and grading tasks, effectively improving the grading efficiency of the flat-plate firewood printing cocoon. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0030] Attached Figure 1 It is a schematic diagram of the method flow of the present invention;

[0031] Attached Figure 2 a is the collected flat-plate Chaiyin cocoon image;

[0032] Attached Figure 2 b is the adaptive Gamma correction image;

[0033] Attached Figure 2 c is a multi-scale detail enhanced image;

[0034] Attached Figure 2 d is the mean shift filtered image;

[0035] Attached Figure 2 e is the gradient combination image;

[0036] Attached Figure 2 f is the gradient combined binary image;

[0037] Attached Figure 2 g is the Gaussian blurred image;

[0038] Attached Figure 2 h is the morphologically processed image;

[0039] Attached Figure 2 i is the defect area image;

[0040] Attached Figure 2 j is the defect marking image;

[0041] Attached Figure 3 This is a data diagram of the effect of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Example 1

[0044] Please see attached Figure 1The present invention provides a technical solution for a flat-plate firewood-printed cocoon grading method: a flat-plate firewood-printed cocoon grading method, the method comprising the following steps:

[0045] S1: Obtain the image data of the flat plate Chaiyin cocoon and perform adaptive Gamma correction on the original image to enhance the contrast and visual effect;

[0046] Get the image data of the flat plate Chaiyin cocoon as follows Figure 2 a, convert the original image into grayscale to obtain grayscale image I g (x,y);

[0047] In order to enhance the contrast of the image, the grayscale image I g (x, y) is processed by adaptive histogram equalization. The equalization algorithm uses the method of cropping restriction and grid size control to adjust the local contrast of each pixel and output the equalized image I c (x, y), its operation relationship is: c (x,y)=CDF(I g (x,y))·(L-1), where the clipping limit is set to 2 and the grid size is 64×64;

[0048] Based on the equalized image I c The average gray value μ of (x, y) is used to calculate the adaptive Gamma value γ to achieve the best contrast and visual effect. The calculation formula is as follows: Where μ represents the equalized image I c The average gray value of (x,y) is defined as

[0049] Apply the calculated adaptive Gamma value γ to the equalized image I c (x, y), perform Gamma correction, and obtain the image I1(x, y) after adaptive Gamma correction as shown in Figure 2 b. The transformation formula for Gamma correction is: Where I1(x, y) represents the corrected image, and the pixel values ​​are clipped to the range of [0, 255] to ensure the effect and accuracy of the image display.

[0050] S2: Apply a multi-scale detail enhancement algorithm to further enhance the image to obtain a second image with rich details;

[0051] Use three Gaussian kernel functions of different scales Convolution operations are performed on the image I1(x,y) to generate Gaussian blurred images B1(x,y), B2(x,y), and B3(x,y) of different scales. When i is 1, 2, or 3, the standard deviation σ iThey are 1, 2, and 4 respectively, and the operation expression is: The structural information of the image at different scales is obtained through these blurred images;

[0052] Calculate the difference between the original image I1(x,y) and the blurred image to obtain detail images D1(x,y), D2(x,y), and D3(x,y). The expressions are as follows: D1(x,y)=I1(x,y)-B1(x,y), D2(x,y)=B1(x,y)-B2(x,y), D3(x,y)=B2(x,y)-B3(x,y). These detail images capture the image detail information at different scales respectively.

[0053] The detail images of different scales are integrated according to the set weights to generate the final detail enhanced image I2(x,y) as shown in Figure 2 c. The integration formula is: I2(x,y)=(1-w1·sign(D1(x,y)))·D1(x,y)+w2·D2(x,y)+w3·D3(x,y)+I1(x,y), where w1, w2, and w3 are weight values, set to 0.5, 0.5, and 0.25, respectively, to control the enhancement effect of details at different scales.

[0054] S3: performing mean shift filtering on the second image to remove noise while keeping edges clear, thereby obtaining a third image;

[0055] For each pixel, its color value (r, g, b) and spatial coordinates (x, y) form a five-dimensional vector x0 = [r, g, b, x, y];

[0056] In the neighborhood of the current pixel x0, the mean offset vector m(x0) is calculated according to the Gaussian kernel function K to determine the amount by which the pixel should be offset toward the average color and spatial center of the neighborhood. The specific calculation is as follows: Among them, the Gaussian kernel function is defined as: The parameter h controls the influence of the search neighborhood, and N(x0) represents the set of pixels in the vicinity.

[0057] According to the mean offset vector m(x0), update the color and spatial position of the pixel, that is, update the current pixel position to: x i =x0+m(x0);

[0058] Calculation and updating are repeated until the amplitude of the mean offset vector approaches zero, indicating convergence to the average center of local color and space;

[0059] After multiple iterations, all pixels will be moved to the smooth center of their area, achieving denoising while retaining edge clarity. The final output image I3(x, y) presents a smooth and detail-preserving effect, as shown in the figure below. Figure 2 d.

[0060] S4: constructing an eight-directional gradient operator kernel and performing a convolution operation and a binarization process on the third image to obtain a fourth image with a prominent edge;

[0061] Construct eight directions of gradient operator kernels to extract image gradients in different directions. The eight directions include 0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°. The operator kernel for each direction is as follows: 0° direction: 45° Orientation: 90° Direction: 135° Orientation: 180° Orientation: 225° Orientation: 270° Orientation: 315° Direction:

[0062] Use the gradient operator kernel G in each direction d Perform convolution operation on the third image I3(x,y) to calculate the gradient image G in the corresponding direction d (x,y), the convolution formula is as follows: Where d is the direction angle, G d (i, j) represents the value of the gradient operator kernel in direction d at position (i, j), and I3(x+i, y+j) is the image pixel value;

[0063] The gradient images G in eight directions d (x, y) takes the absolute value of each pixel and accumulates it to get the combined gradient image G c (x,y), such as Figure 2 e, the combination formula is G c (x,y)=∑ d |G d (x,y)|;

[0064] For the combined gradient image G c The pixel values ​​of (x, y) are clipped to ensure that the pixel values ​​are in the range of [0, 255]. Subsequently, the Otsu thresholding method is applied to the combined gradient image G c (x, y) is binarized to obtain the final edge-highlighted binary image I4(x, y), as shown in Figure 2 f.

[0065] S5: applying Gaussian blur and morphological processing to the fourth image to smooth the boundary and optimize the structure to obtain a fifth image;

[0066] Use a Gaussian blur with a kernel size of 51×51 to achieve a smoothing effect, such as Figure 2 g;

[0067] Define an elliptical structure element of size 21×21, and use morphological closing operation to fill the holes between the boundaries of the image and connect adjacent areas, thereby improving the overall structure of the image;

[0068] The Otsu threshold method is used to binarize the Gaussian blurred image to ensure that the generated image I5(x,y) is clearer in terms of boundaries and structures. Figure 2 h.

[0069] S6: extracting the ROI region based on the fifth image, accurately obtaining the defective area of ​​the flat-plate Chaiyin cocoon and calculating the area to achieve grading;

[0070] A pixel-by-pixel traversal method is used for the fifth image I5(x, y) to detect the boundary change points between the target area and the background in the image, and the connected boundary lines in the image are extracted by calculating the gray value changes of the neighborhood to generate a closed contour set of the target area;

[0071] Mark the pixels inside each closed contour, set the pixel value inside the contour to a predetermined value of 255, and generate a mask image I m (x,y);

[0072] The mask image I m (x, y) is logically operated with the fifth image I5(x, y), and the non-target area is eliminated through the XOR operation to obtain the ROI area, such as Figure 2 i;

[0073] Perform geometric analysis on the ROI area and count the number of pixels to determine the defect area value, as shown in the attached figure. Figure 3 The defect area of ​​the middle flat plate Chaiyin cocoon 1 is calculated by the number of pixels and compared with the preset grading standard of 0.5 cm 2 Compare and realize the classification of second-rate cocoons;

[0074] Extract the contour of the ROI area and mark it on the original image, such as Figure 2 j;

[0075] By using adaptive Gamma correction and multi-scale detail enhancement algorithm to process the image, combined with mean shift filtering, the image quality is effectively improved. By constructing an eight-directional gradient operator kernel for convolution operation, integrating multi-directional gradient information, and combining Gaussian blur and morphological processing, the defect area of ​​the flat-plate Chaiyin cocoon is accurately extracted, the defect segmentation quality is significantly improved, and the automatic and accurate detection of defects on the flat-plate Chaiyin cocoon and area calculation are realized, thereby effectively improving the efficiency of grading detection of flat-plate Chaiyin cocoons.

[0076] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example.

[0077] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for grading cocoons from flat-plate cocoons, characterized in that: The following steps are involved: S1: Obtain the image data of the flat plate Chaiyin cocoon and perform adaptive Gamma correction on the original image to enhance the contrast and visual effect; S2: Apply a multi-scale detail enhancement algorithm to further enhance the image to obtain a second image with rich details; S3: performing mean shift filtering on the second image to remove noise while keeping edges clear, thereby obtaining a third image; S4: constructing an eight-directional gradient operator kernel and performing a convolution operation and a binarization process on the third image to obtain a fourth image with a prominent edge; S5: applying Gaussian blur and morphological processing to the fourth image to smooth the boundary and optimize the structure to obtain a fifth image; S6: extracting the region of interest based on the fifth image, accurately obtaining and calculating the area of ​​the defective region, and realizing the grading of the flat-plate Chaiyin cocoon.

2. A method for grading cocoons from flat-plate cocoons according to claim 1, characterized in that: Step S1 includes the following steps: S11: Obtain the image data of the flat plate Chai Yin cocoon, perform grayscale conversion on the original image I(x, y), and obtain a grayscale image I g (x,y); S12: Grayscale image I g (x, y) is subjected to adaptive histogram equalization to enhance the image contrast. The equalization algorithm uses the method of cropping restriction and grid size control to adjust the local contrast of each pixel and output the equalized image I c (x,y); S13: Based on the equalized image I c The average gray value μ of (x, y) is used to calculate the adaptive Gamma value γ to achieve the best contrast and visual effect. The calculation formula is as follows: Where μ represents the equalized image I c The average gray value of (x,y) is defined as S14: Apply the calculated adaptive Gamma value γ to the equalized image I c (x, y), perform Gamma correction, and obtain the image I1(x, y) after adaptive Gamma correction. The transformation formula of Gamma correction is: Where I1(x, y) represents the corrected image, and the pixel values ​​are clipped to the range of [0, 255] to ensure the effect and accuracy of the image display.

3. A method for grading cocoons from flat-plate cocoons according to claim 1, characterized in that: Step S2 includes the following steps: S21: Using three Gaussian kernel functions of different scales Perform convolution operations on the image I1(x,y) to generate Gaussian blurred images B1(x,y), B2(x,y), B3(x,y) of different scales, and obtain the structural information of the image at different scales through these blurred images; S22: Calculate the difference between the original image I1(x,y) and the blurred image to obtain detail images D1(x,y), D2(x,y), and D3(x,y). The expressions are as follows: D1(x,y)=I1(x,y)-B1(x,y), D2(x,y)=B1(x,y)-B2(x,y), D3(x,y)=B2(x,y)-B3(x,y). These detail images capture the image detail information at different scales respectively. S23: Integrate detail images of different scales according to the set weights to generate the final detail enhanced image I2(x,y). The integration formula is: I2(x,y)=(1-w1·sign(D1(x,y)))·D1(x,y)+w2·D2(x,y)+w3·D3(x,y)+I1(x,y), where w1, w2, w3 are weight values ​​respectively.

4. A method for grading cocoons from flat-plate cocoons according to claim 1, characterized in that: Step S4 includes the following steps: S41: construct eight-directional gradient operator kernels to extract image gradients in different directions; S42: Use the gradient operator kernel G in each direction d Perform convolution operation on the third image I3(x,y) to calculate the gradient image G in the corresponding direction d (x,y), the convolution formula is as follows: Where d is the direction angle, G d (i, j) represents the value of the gradient operator kernel in direction d at position (i, j), and I3(x+i, y+j) is the image pixel value; S43: The gradient images G in eight directions d (x, y) takes the absolute value of each pixel and accumulates it to get the combined gradient image G c (x,y), the combination formula is G c (x,y)=∑ d |G d (x,y)|; S44: Combined gradient image G c The pixel values ​​of (x, y) are clipped to ensure that the pixel values ​​are in the range of [0, 255]. Subsequently, the Otsu thresholding method is applied to the combined gradient image G c (x, y) is binarized to obtain the final edge-highlighted binary image I4(x, y).

5. A method for grading cocoons from flat-plate cocoons according to claim 2, characterized in that: For grayscale image I g (x, y) performs adaptive histogram equalization, and the operation relationship is: I c (x,y)=CDF(I g (x,y))·(L-1), where the clipping limit is set to 2 and the grid size is 64×64.

6. A method for grading cocoons from flat-plate cocoons according to claim 3, characterized in that: The Gaussian kernel function When i is 1, 2, or 3, the standard deviation σ i They are 1, 2, and 4 respectively, and the operation expression is:

7. A method for grading cocoons from flat-plate cocoons according to claim 6, characterized in that: The weight values ​​w1, w2, and w3 are 0.5, 0.5, and 0.25, respectively, which are used to control the enhancement effect of details at different scales.

8. A method for grading cocoons from flat-plate cocoons according to claim 4, characterized in that: The eight directions are 0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°, and the operator kernel of each direction is as follows: 0° direction: 45° Orientation: 90° Direction: 135° Orientation: 180° Orientation: 225° Orientation: 270° Orientation: 315° Direction:

Images