Image segmentation algorithm-based rockfill material grading detection method and device

Abstract

The invention belongs to the field of rockfill material grading detection, and relates to a rockfill material grading detection method and device based on an image segmentation algorithm, and the method comprises the steps: 1) obtaining an image after paving of stone particles; 2) preprocessing the stone image; 3) inputting the preprocessed image into an instance segmentation model to obtain mask information of stone particles; 4) extracting mask quantitative indexes and shape characteristic parameters in the mask information of the stone particles, and performing particle size and stone total volume conversion in each particle size range on the mask quantitative indexes and the shape characteristic parameters; (5) converting the obtained stone particle size and volume into three-dimensional stone particle gradation, and drawing a stone particle apparent gradation curve; 6, the obtained stone particle apparent grading curve is corrected through the data fusion model.The final stone particle three-dimensional overall grading curve is obtained.The method has the advantages of being high in adaptability, free of a special detection environment and high in detection precision.

Description

Technical Field

[0001] The invention belongs to the field of rockfill material gradation detection, and relates to a rockfill material gradation detection method and device, and in particular to a rockfill material gradation detection method and device based on an image segmentation algorithm. Background Art

[0002] Concrete face rockfill dams are widely selected as water retaining structures in the water conservancy and hydropower industry due to their low construction cost, good dam slope stability, and excellent water pressure resistance. As the main filling material for the dam body, the gradation state of the rockfill significantly affects the physical properties and mechanical behavior of the dam body, and is crucial to the construction quality and safe operation of the dam. Traditional gradation detection methods require a lot of manpower, and have poor detection stability, low sampling rate, and weak representativeness. However, some of the current gradation detection methods based on machine vision require more processes or a relatively stable detection environment, which is not convenient for use on engineering sites. In addition, when the stones are in a stacked state, the detection process has certain limitations, and only the apparent gradation of the stones exposed on the surface can be detected, and the detection results are significantly different from the screening results. Summary of the invention

[0003] In order to solve the above technical problems existing in the background technology, the present invention provides a rockfill material gradation detection method and device based on image segmentation algorithm which has strong adaptability, does not require a special detection environment and has high detection accuracy.

[0004] In order to achieve the above object, the present invention adopts the following technical solution:

[0005] A method for detecting the gradation of rockfill materials based on an image segmentation algorithm, characterized in that: the method for detecting the gradation of rockfill materials based on an image segmentation algorithm comprises the following steps:

[0006] 1) Obtain images of stone particles after paving;

[0007] 2) Preprocessing the stone image obtained in step 1);

[0008] 3) Input the preprocessed image into the instance segmentation model to obtain the mask information of the stone particles;

[0009] 4) extracting the mask quantification index and shape characteristic parameters from the mask information of the stone particles, and converting the mask quantification index and shape characteristic parameters into particle size and the total volume of the stone within each particle size range;

[0010] 5) converting the stone particle size and volume obtained in step 4) into a three-dimensional stone particle gradation and drawing a stone particle apparent gradation curve;

[0011] 6) The apparent gradation curve of the stone particles obtained in step 5) is corrected by the data fusion model to obtain the final three-dimensional overall gradation curve of the stone particles.

[0012] Preferably, the preprocessing in step 2) adopted in the present invention includes histogram equalization, bilateral filtering and grayscale conversion;

[0013] The histogram equalization can rearrange the brightness values ​​of the overexposed or dark local areas in the stone image;

[0014] The grayscale processing is to grayscale the RGB color image using a weighted average method, and the grayscale processing method is:

[0015] f(i,j)=0.3R(i,j)+0.59G(i,j)+0.11B(i,j)

[0016] in:

[0017] f(i, j) represents the gray value of the position (i, j) after graying;

[0018] R(i, j), G(i, j) and B(i, j) represent the values ​​of the three channels at the position (i, j) in the stone color image;

[0019] The bilateral filtering is to eliminate the noise mixed in when the image is digitized; the processing method of the bilateral filtering is:

[0020] The value g(i, j) at the position (i, j) in bilateral filtering depends on the weighted combination of the pixel value f(kl) at the point position (k, l) in the neighborhood and the weight coefficient ω between the corresponding point (k, l) in the neighborhood and the center point (i, j):

[0021]

[0022] The weight coefficient v(i, j, k, l) is the product of the spatial domain kernel d and the range kernel r. The expression of the weight coefficient ω(i, j, k, l) is:

[0023] ω(i,j,k,l)=d(i,j,k,l)×r(i,j,k,l)

[0024] in:

[0025] The spatial domain kernel d is expressed as the Euclidean distance between the point (k, l) in the neighborhood and the center point (i, j), where Set the number for the parameter:

[0026]

[0027] The range kernel r represents the absolute value of the difference between the gray value f(kl) of the point (k, l) in the neighborhood and the gray value f(kl) of the center point (i, j), where Set the number for the parameter:

[0028]

[0029] Preferably, the instance segmentation model in step 3) adopted in the present invention is Mask-RCNN.

[0030] Preferably, the specific implementation method of step 4) adopted by the present invention is:

[0031] 4.1) Extracting mask quantification indexes and shape characteristic parameters from the mask information of stone particles;

[0032] 4.2) The particle size and volume conversion algorithm is used to convert the mask quantification index and shape characteristic parameters obtained in step 4.1) into the particle size and volume of the stone.

[0033] Preferably, the mask quantization index in step 4.1) of the present invention includes the perimeter of the mask edge contour, the minimum circumscribed rectangle of the stone particles, and the convex hull of the stone particles;

[0034] The shape characteristic parameters in step 4.1) include flatness and elongation; the flatness is the ratio of the thickness of the stone particle to the width of the stone particle; the elongation is the ratio of the width of the stone particle to the length of the stone particle.

[0035] Preferably, the calculation method of the perimeter of the edge contour of the mask adopted in the present invention is: suppose there are n points constituting the edge contour, and the coordinates of the points are expressed by (x i ,y i ), then the perimeter P of the stone particle can be expressed as:

[0036]

[0037] The calculation method of the stone particle mask area S is:

[0038]

[0039] The calculation method of the minimum circumscribed rectangle of the stone particles is:

[0040] Assume that the point set of the stone contour is Q = {(x1, y1), (x2, y2), ..., (x n ,y n )}, calculate the convex hull of the stone contour, and obtain the convex hull point set Q′={(x′1, y′1), (x′2, y′2), …, (x′ n , y′ n)}, calculate the center of the convex hull point set in:

[0041] and are the average values ​​of the horizontal and vertical coordinates of all points respectively;

[0042] Transform the coordinates of each point in the convex hull point set to obtain a new point set Q″={(x″1, y″1), (x″2, y″2), …, (x″ n , y″ n )},

[0043] in:

[0044]

[0045] The new point set Q″ is constructed as a matrix M, the expression of M is:

[0046]

[0047] Use the least squares method to perform singular value decomposition on the matrix M and obtain the right singular vector matrix V:

[0048] M=UΣV T

[0049] in:

[0050] U and V are orthogonal matrices;

[0051] ∑ is a diagonal matrix;

[0052] The last column of the matrix V is the normal vector v of the minimum circumscribed rectangle of the stone outline. The rotation angle angel is calculated by the normal vector v. Assuming θ is the polar coordinate angle of v, the rotation angle is:

[0053]

[0054] Calculate the width and height of the enclosing rectangle of the convex hull point set in the rotated coordinate system, and then obtain the minimum circumscribed rectangle of the mask;

[0055] The shape characteristic parameters are obtained by dividing the stone into particle size ranges, using a vernier caliper to measure the length, width and thickness of the stone particles in different particle size ranges, and calculating the average flatness and average elongation of the stone particles in each particle size range according to the thickness, width and thickness of the stone particles in each particle size range;

[0056] Preferably, the specific method of the measurement is: find the plane where the maximum projection area of ​​the particle is located, measure the distance between the maximum two points of the plane contour as the length of the stone particle; measure the distance between the maximum two points in the vertical length direction within the contour as the width of the stone particle; perpendicular to this plane, measure the distance between the maximum two points of the particle as the thickness of the stone particle.

[0057] Preferably, the specific implementation method of the stone particle size conversion in step 4.2) of the present invention is: the second length of the stone particle is used as the particle size of the stone particle, and the second length of the stone particle is the width of the minimum circumscribed rectangle of the stone particle when the stone particle is placed in a manner with the largest horizontal projection area as the particle size of the stone particle;

[0058] The specific implementation method of the stone volume conversion in step 4.2) is:

[0059] The volume conversion formula within each particle size range is:

[0060]

[0061] in:

[0062] V i It is expressed as the volume of the i-th stone particle in the current particle size range;

[0063] n represents the total number of stone particles within the current particle size range;

[0064] S i It is expressed as the mask area of ​​the i-th stone particle in the current particle size range.

[0065] Preferably, the gradation calculation formula of the three-dimensional stone particles in step 5) adopted in the present invention is:

[0066]

[0067] in:

[0068] GIR is the ratio of the total mass of stone particles in each size range to the total mass of all stone particles in the image;

[0069] ρ is the density of stone particles;

[0070] N represents the total number of all stone particles;

[0071] n represents the total number of stone particles within the current particle size range;

[0072] V i It is expressed as the volume of the i-th stone particle in the current particle size range;

[0073] V kIt is expressed as the volume of the kth stone particle in all particle size ranges;

[0074] S i Represents the mask area of ​​the i-th stone particle in the current particle size range;

[0075] S k Represents the mask area of ​​the kth stone particle in all particle size ranges.

[0076] Preferably, the specific implementation of step 6) adopted by the present invention is:

[0077] 6.1) The mass percentage of each characteristic particle size in the three-dimensional overall state is inferred by weighted summation of the mass percentage of each characteristic particle size in the stone surface. The mass percentage calculation formula of each characteristic particle size after correction is:

[0078]

[0079] in:

[0080] Expressed as the corrected mass percentage value for the characteristic particle size d0;

[0081] l represents the number of characteristic particle sizes;

[0082] λ i is the weight coefficient of the i-th characteristic particle size;

[0083] z i is the characteristic particle size d i The mass percentage value before correction;

[0084] Among them, the weight coefficient is the estimated value that can satisfy d0 The optimal set of coefficients with the smallest difference from the true value z0, that is, At the same time, it satisfies the unbiased estimation Where Var and E are variance and expectation respectively;

[0085] 6.2) Based on the constraints of minimum variance and unbiased estimation, the Lagrange multiplier method is used, and the weight coefficient λ is i The calculation formula is:

[0086]

[0087] in:

[0088] r ij is the semivariogram function, expressed as

[0089] λ is the Lagrange multiplier;

[0090] 6.3) Assume rij The particle size difference d from each characteristic particle size ij The spherical model is used to fit this relationship to obtain the functional relationship r = r (d), then for any two points d i d j By calculating the difference between two points d ij , calculate the semivariogram function r through the functional relationship ij , through the weight coefficient λ i The calculation formula of and the corrected mass percentage calculation formula of each characteristic particle size are used to obtain the corrected three-dimensional grading curve of stone.

[0091] A rockfill material gradation detection device for realizing the rockfill material gradation detection method based on image segmentation algorithm as described above, characterized in that: the rockfill material gradation detection device comprises an image acquisition module, an image recognition module, a conversion module and a gradation curve drawing module;

[0092] The image acquisition module is used to acquire the image of the stone particles to be detected;

[0093] The image recognition module is used to perform image preprocessing operations on the photographed stone particle image and input it into the instance segmentation model, and obtain the stone particle image with mask information of the stone particles through the instance segmentation model;

[0094] The conversion module is used to extract the mask quantification index and shape characteristic parameters of the stone particle image, and convert the extracted characteristic parameters into the particle size and the total volume of the stone within each particle size range;

[0095] The grading curve drawing module converts the stone particle size and volume obtained by converting the extracted shape characteristic parameters into the stone apparent gradation, and corrects the apparent gradation through a data fusion model to obtain a three-dimensional overall grading curve of the stone.

[0096] The beneficial effects of the present invention are:

[0097] The present invention provides a method and device for detecting the gradation of rockfill materials based on an image segmentation algorithm, wherein the method comprises: 1) obtaining an image of stone particles after paving; 2) preprocessing the stone image obtained in step 1); 3) inputting the preprocessed image into an instance segmentation model to obtain mask information of stone particles; 4) extracting mask quantification indexes and shape characteristic parameters in the mask information of stone particles, and converting the mask quantification indexes and shape characteristic parameters into particle sizes and total volumes of stones within each particle size range; 5) converting the stone particle sizes and volumes obtained in step 4) into three-dimensional stone particle gradation and drawing a stone particle apparent gradation curve; 6) correcting the stone particle apparent gradation curve obtained in step 5) through a data fusion model to obtain a final three-dimensional overall gradation curve of stone particles. The rockfill gradation detection method based on image segmentation algorithm provided by the present invention can standardize the stone paving process, significantly improve the detection stability, have good adaptability and stability for stone particles with a large particle size span, and can detect stone particles with a particle size between 5-300mm. In addition, the amount of manual coordination in the entire detection process is very small, the process is simple and fast, and it can realize real-time detection of stones without affecting the stone filling process. At the same time, the device provided by the present invention can overcome the problem of stones being blocked in the stacked state. The detection process involved in the present invention has a simple process and strong adaptability, does not require a special detection environment, and can be fully applied to actual engineering sites. In summary, the rockfill gradation detection method and device based on the image segmentation algorithm provided by the present invention can realize the rapid gradation detection of rockfill. For the problem that the segmentation model cannot accurately segment large-size stone and small-size stone at the same time, the mask threshold fusion algorithm is adopted to fuse the parts of the stone images with different sampling resolutions that are accurately identified, and finally reflected in the detection results. The data fusion model is used to correct the apparent gradation of the stone to obtain the three-dimensional overall gradation of the stone. The results show that the screening test verification group has a better fit. The present invention can quickly spread the stones in the stacked state, and can quickly complete the standardization of the spreading process, and improve the stability of the recognition results of the subsequent recognition module and the gradation conversion module. In addition, the device has less manual intervention and simple procedures in the whole process from image acquisition to gradation detection, and realizes the real-time paving, detection and loading process of the stone, without affecting the filling and transportation process of the rockfill. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 It is a flow chart of the rockfill material gradation detection method based on image segmentation algorithm provided by the present invention.

[0099] Figure 2 It is a rockfill material gradation detection curve diagram based on the image segmentation algorithm provided by the present invention. DETAILED DESCRIPTION

[0100] See also Figure 1The present invention provides a method for detecting rockfill material gradation based on an image segmentation algorithm, comprising the following steps:

[0101] 1) Obtain images of stone particles after paving;

[0102] 2) Preprocessing the stone image obtained in step 1); wherein the preprocessing includes histogram equalization, bilateral filtering and grayscale conversion; through the preprocessing, the overly dark and overexposed parts of the original image are eliminated, some features of the stone are unified and the noise mixed in when the image is digitized is eliminated;

[0103] Histogram equalization can rearrange the brightness values ​​of overexposed or dark local areas in stone images; overexposure and dark phenomena are caused by the over-concentration of the brightness values ​​of pixels in the area. This operation makes the brightness distribution more uniform, and ultimately makes the regional features more obvious.

[0104] Grayscale processing In order to avoid the interference of different stone characteristics on the subsequent segmentation results, the weighted average method is used to grayscale the RGB color image. Since the human eye is most sensitive to green and least sensitive to blue, a more reasonable grayscale image can be obtained by weighted averaging the three components of RGB according to the following formula:

[0105] f(i,j)=0.3R(i,j)+0.59G(i,j)+0.11B(i,j)

[0106] in:

[0107] f(i, j) represents the gray value of the position (i, j) after graying;

[0108] R(i, j), G(i, j) and B(i, j) represent the values ​​of the three channels at the position (i, j) in the stone color image;

[0109] Bilateral filtering is to meet the requirements of image processing and eliminate the noise mixed in when the image is digitized; according to the characteristics of stone images, it is necessary to denoise the image while retaining the image edge information. Therefore, using bilateral filtering can denoise the image while retaining the important edges, textures and details in the image. The processing method of bilateral filtering is:

[0110] The value g(i, j) at the position (i, j) in bilateral filtering depends on the weighted combination of the pixel value f(kl) at the point position (k, l) in the neighborhood and the weight coefficient ω between the corresponding point (k, l) in the neighborhood and the center point (i, j):

[0111]

[0112] The weight coefficient ω(i, j, k, l) is the product of the spatial domain kernel d and the range kernel r. The expression of the weight coefficient ω(i, j, k, l) is:

[0113] ω(i,j,k,l)=d(i,j,k,l)×r(i,j,k,l)

[0114] in:

[0115] The spatial domain kernel d is expressed as the Euclidean distance between the point (k, l) in the neighborhood and the center point (i, j), where Set the number for the parameter:

[0116]

[0117] The range kernel r represents the absolute value of the difference between the gray value f(kl) of the point (k, l) in the neighborhood and the gray value f(kl) of the center point (i, j), where Set the number for the parameter:

[0118]

[0119] 3) Input the preprocessed image into the instance segmentation model to obtain the mask information of the stone particles, that is, the image of the particle contour information; exemplary, the instance segmentation model that can be used is Mask-RCNN. Due to the need for real-time detection of stone particle gradation, the speed of quickly identifying the stone contour in the stone image and generating the mask is required to be high, so the Mask-RCNN model suitable for instance segmentation is selected, which can meet the requirements in terms of speed and has high accuracy.

[0120] 4) Extracting the mask quantification index and shape characteristic parameters from the mask information of the stone particles, and converting the mask quantification index and shape characteristic parameters into particle size and the total volume of the stone within each particle size range; specifically:

[0121] 4.1) Extracting mask quantization indexes and shape characteristic parameters from the mask information of stone particles; wherein the mask quantization indexes include the perimeter of the mask edge contour, the minimum circumscribed rectangle of the stone particles, and the convex hull of the stone particles; the shape characteristic parameters include flatness and elongation; the flatness is the ratio of the thickness of the stone particles to the width of the stone particles; the elongation is the ratio of the width of the stone particles to the length of the stone particles.

[0122] The calculation method of the perimeter of the mask edge contour is: suppose there are n points constituting the edge contour, and the coordinates of the points are expressed by (x i ,y i ), then the perimeter p of the stone particle can be expressed as:

[0123]

[0124] The calculation method of the stone particle mask area S is:

[0125]

[0126] The calculation method of the minimum circumscribed rectangle of stone particles is:

[0127] Assume that the point set of the stone contour is Q = {(x1, y1), (x2, y2), ..., (x n ,y n )}, calculate the convex hull of the stone contour, and obtain the convex hull point set Q′={(x′1, y′1), (x′2, y′2), …, (x′ n , y′ n )}, calculate the center of the convex hull point set in:

[0128] and are the average values ​​of the horizontal and vertical coordinates of all points respectively;

[0129] Transform the coordinates of each point in the convex hull point set to obtain a new point set Q″={(x″1, y″1), (x″2, y″2), …, (x″ n , y″ n )},

[0130] in:

[0131]

[0132] The new point set W″ is constructed as a matrix M, and the expression of M is:

[0133]

[0134] Use the least squares method to perform singular value decomposition on the matrix M and obtain the right singular vector matrix V:

[0135] M=UΣV T

[0136] in:

[0137] U and V are orthogonal matrices;

[0138] ∑ is a diagonal matrix;

[0139] The last column of the matrix V is the normal vector v of the minimum circumscribed rectangle of the stone outline. The rotation angle angel is calculated by the normal vector v. Assuming θ is the polar coordinate angle of v, the rotation angle is:

[0140]

[0141] Calculate the width and height of the enclosing rectangle of the convex hull point set in the rotated coordinate system, and then obtain the minimum circumscribed rectangle of the mask;

[0142] The shape characteristic parameters are obtained by dividing the stone into particle size ranges of 5-10mm, 10-20mm, 20-40mm, 40-60mm, 60-80mm, 80-100mm, 100-150mm, and 150-300mm. Use a vernier caliper to measure the length, width, and thickness of stone particles in different particle size ranges, and calculate the average flatness and average elongation of stone particles in each particle size range according to the thickness, width, and thickness of stone particles in each particle size range.

[0143] Exemplarily, the specific measurement method adopted by the present invention is: find the plane where the maximum projection area of ​​the particle is located, measure the distance between the two maximum points of the plane contour as the length of the stone particle; in the vertical length direction within the contour, measure the distance between the two maximum points as the width of the stone particle; perpendicular to this plane, measure the distance between the two maximum points of the particle as the thickness of the stone particle.

[0144] 4.2) The particle size and volume conversion algorithm is used to convert the mask quantification index and shape characteristic parameters obtained in step 4.1) into the particle size and volume of the stone.

[0145] The specific implementation method of converting the particle size of the stone is: taking the second length of the stone particle as the particle size of the stone particle, and the second length of the stone particle is the width of the minimum circumscribed rectangle of the stone particle when the stone particle is placed in a manner with the largest horizontal projection area as the particle size of the stone particle;

[0146] The specific implementation method of stone volume conversion is:

[0147] The volume conversion formula within each particle size range is:

[0148]

[0149] in:

[0150] V i It is expressed as the volume of the i-th stone particle in the current particle size range;

[0151] n represents the total number of stone particles within the current particle size range;

[0152] S i It is expressed as the mask area of ​​the i-th stone particle in the current particle size range.

[0153] 5) The stone particle size and volume obtained in step 4) are converted into a three-dimensional stone particle gradation and a stone particle apparent gradation curve is drawn; wherein the gradation calculation formula of the three-dimensional stone particles is:

[0154]

[0155] in:

[0156] GIR is the ratio of the total mass of stone particles in each size range to the total mass of all stone particles in the image;

[0157] ρ is the density of stone particles;

[0158] N represents the total number of all stone particles;

[0159] n represents the total number of stone particles within the current particle size range;

[0160] V i It is expressed as the volume of the i-th stone particle in the current particle size range;

[0161] V k It is expressed as the volume of the kth stone particle in all particle size ranges;

[0162] S i Represents the mask area of ​​the i-th stone particle in the current particle size range;

[0163] S k Represents the mask area of ​​the kth stone particle in all particle size ranges.

[0164] 6) Using the Kriging interpolation algorithm as the correction model and the spherical model as the fitting model, the apparent gradation curve of the stone particles obtained in step 5) is corrected to obtain the final three-dimensional overall gradation curve of the stone particles, specifically:

[0165] 6.1) The mass percentage of each characteristic particle size in the three-dimensional overall state is inferred by weighted summation of the mass percentage of each characteristic particle size in the stone surface. The mass percentage calculation formula of each characteristic particle size after correction is:

[0166] Specifically, calculate the difference d of all characteristic particle sizes ij , this method has 9 characteristic particle sizes and 36 different differences; draw a difference-semivariance scatter plot, use the spherical model as the fitting model; calculate the semivariance function value through the fitting model, and calculate the weight coefficient through the following formula:

[0167]

[0168] After obtaining the 9 weight coefficients, the corrected passing rate of each characteristic particle size of the stone is calculated by the following formula:

[0169]

[0170] in:

[0171] Expressed as the corrected mass percentage value for the characteristic particle size d0;

[0172] l represents the number of characteristic particle sizes;

[0173] λ i is the weight coefficient of the i-th characteristic particle size;

[0174] z i is the characteristic particle size d i The mass percentage value before correction;

[0175] Among them, the weight coefficient is the estimated value that can satisfy d0 The optimal set of coefficients with the smallest difference from the true value z0, that is, At the same time, it satisfies the unbiased estimation Where Var and E are variance and expectation respectively;

[0176] 6.2) Based on the constraints of minimum variance and unbiased estimation, the Lagrange multiplier method is used, and the weight coefficient λ is i The calculation formula is:

[0177]

[0178] in:

[0179] r ij is the semivariogram function, expressed as

[0180] λ is the Lagrange multiplier;

[0181] 6.3) Assume r ij The particle size difference d from each characteristic particle size ij The spherical model is used to fit this relationship to obtain the functional relationship r = r (d), then for any two points d i d j By calculating the difference between two points d ij , calculate the semivariogram function r through the functional relationship ij , through the weight coefficient λ i The calculation formula of and the calculation formula of the mass percentage of each characteristic particle size after correction are used to obtain the corrected three-dimensional grading curve of stone.

[0182] See also Figure 2The screening test gradation can be regarded as the standard value. The gradation directly recognized by surface photographic image is the apparent gradation. The gradation corrected by Kriging interpolation algorithm is the data fusion corrected gradation. The corrected gradation is consistent with the screening result.

[0183] Embodiment 1:

[0184] The rockfill material gradation detection method provided by the present invention specifically comprises:

[0185] The image acquisition module of the device is used to acquire the image of stone particles; specifically, firstly, the baffle at the bottom of the chute is fixed, and the piled stone starts to slide from the top until it stably slides to the bottom of the baffle and stops, and then the remote photography program of the industrial camera is used to automatically acquire the image of the piled stone.

[0186] The acquired stone particle images are preprocessed, wherein specifically, the captured images are subjected to histogram equalization, grayscale conversion, and bilateral filtering operations to eliminate overly dark and overexposed parts of the original images, unify some features of the stones, and eliminate noise mixed in when the images are digitized.

[0187] The preprocessed stone image is input into the instance segmentation model to obtain the mask information of the stone particles, that is, the image of the particle contour information.

[0188] The contour information of the stone particles is subjected to mask quantification indexes and shape feature parameters extraction. Specifically, the shape feature parameters are measured by a vernier caliper to measure the length, width, and thickness of the stone and perform statistics. The measurement method is as follows: first, find the plane where the maximum projection area of ​​the particle is located, and measure the maximum distance between the two points of the plane contour as the particle length; in the vertical length direction within the contour, measure the maximum distance between the two points as the particle width; perpendicular to this plane, measure the maximum distance between the two points of the particle as the particle thickness. Mask quantification indexes include: the perimeter of the stone particle mask, the convex hull of the stone particle mask, the length of the minimum circumscribed rectangle of the stone particle mask, the width of the minimum circumscribed rectangle of the stone particle mask, and the area of ​​the stone particle mask. According to the calculation formula, the stone particle mask in the image is traversed to extract each mask quantification index.

[0189] The extracted mask quantification index is converted into volume. The stone particle size and volume obtained after the particle size and volume conversion by the extracted characteristic parameters are converted into the apparent gradation of the stone. Specifically, the particle size of the stone is arranged from large to small. The present invention uses 300mm, 150mm, 100mm, 80mm, 60mm, 40mm, 20mm, 10mm, and 5mm as boundaries, starting from the stone in sequence 1 and ending at each boundary, summing the volumes of all stones in the range and dividing it by the sum of the volumes of all stones, and calculating the pass rate of each particle size boundary point. The horizontal axis of the gradation curve is the stone particle size, and the vertical axis is the pass rate of the stone passing through the particle size. The position of each boundary point can be plotted, and adjacent points can be connected with straight lines.

[0190] The Kriging interpolation algorithm is used as the data fusion model, and the spherical model is used as the fitting model to correct the apparent gradation of the stone to obtain the real gradation of the stone. Specifically, the difference d of all characteristic particle sizes is calculated. ij The present invention has 9 characteristic particle sizes and 36 different differences; a difference-semivariance scatter plot is drawn, and a spherical model is used as a fitting model; the semivariance function value is calculated by the fitting model, and the weight coefficient is calculated by the following formula:

[0191]

[0192] After obtaining the 9 weight coefficients, the corrected passing rate of each characteristic particle size of the stone is calculated by the following formula:

[0193]

[0194] Embodiment 2:

[0195] The present invention also provides a rockfill material gradation detection device based on an image segmentation algorithm, the detection device comprising:

[0196] An image acquisition module, using an image acquisition device to acquire an image of stone particles to be detected;

[0197] The main body of this module consists of a slide with a slide rail and an industrial camera, i.e., its support frame. The support frame is placed on the slide rail, and images of any position inside the slide rail can be captured through the slide rail. The stone slides out through the top of the slide, and a switchable baffle is provided at the bottom. After being placed at the bottom, it forms a paving stone with certain characteristics, and the image of the stone after paving can be captured. After the shooting is completed, the stone can be collected through the bottom baffle switch. The conversion coefficient between the actual size of the stone and the pixel size of the stone image can be automatically completed by the height of the shooting rod.

[0198] The image recognition module performs image preprocessing operations on the captured stone particle image and inputs it into the instance segmentation model, and obtains the stone particle image with mask, that is, the stone particle image containing the segmentation contour through the model;

[0199] The industrial camera is connected to the computer by wire or wirelessly. According to the needs of the engineering site, the image can be quickly transferred to the computer and input into the instance segmentation model, and the contour recognition is performed to obtain the mask information of the stone particles.

[0200] The particle size and volume conversion module performs mask quantification index and shape feature parameter extraction on the contour information of stone particles, and converts the extracted feature parameters into particle size and total volume of stone within each particle size range;

[0201] The stone mask image is exported, and the particle size and volume of each stone in the image are calculated by running the program that completes the above business.

[0202] The grading curve drawing module converts the stone particle size and volume obtained by converting the particle size and volume through the extracted characteristic parameters into the stone apparent gradation, and corrects the apparent gradation through the data fusion model to obtain the three-dimensional overall grading curve of the stone.

Claims

1. A method for detecting rockfill material gradation based on image segmentation algorithm, characterized in that: The rockfill material gradation detection method based on image segmentation algorithm comprises the following steps: 1) Obtain images of stone particles after paving; 2) Preprocessing the stone image obtained in step 1); 3) Input the preprocessed image into the instance segmentation model to obtain the mask information of the stone particles; 4) extracting the mask quantification index and shape characteristic parameters from the mask information of the stone particles, and converting the mask quantification index and shape characteristic parameters into particle size and the total volume of the stone within each particle size range; 5) converting the stone particle size and volume obtained in step 4) into a three-dimensional stone particle gradation and drawing a stone particle apparent gradation curve; 6) The apparent gradation curve of the stone particles obtained in step 5) is corrected by the data fusion model to obtain the final three-dimensional overall gradation curve of the stone particles.

2. The rockfill material gradation detection method based on image segmentation algorithm according to claim 1 is characterized in that: The preprocessing in step 2) includes histogram equalization, bilateral filtering and grayscale conversion; The histogram equalization can rearrange the brightness values ​​of the overexposed or dark local areas in the stone image; The grayscale processing is to grayscale the RGB color image using a weighted average method, and the grayscale processing method is: f(i,j)=0.3R(i,j)+0.59G(i,j)+0.11B(i,j) in: f(i, j) represents the gray value of the position (i, j) after graying; R(i, j), G(i, j) and B(i, j) represent the values ​​of the three channels at the position (i, j) in the stone color image; The bilateral filtering is to eliminate the noise mixed in when the image is digitized; the processing method of the bilateral filtering is: The value g(i, j) at the position (i, j) in bilateral filtering depends on the weighted combination of the pixel value f(kl) at the point position (k, l) in the neighborhood and the weight coefficient ω between the corresponding point (k, l) in the neighborhood and the center point (i, j): The weight coefficient ω(i, j, k, l) is the product of the spatial domain kernel d and the range kernel r. The expression of the weight coefficient ω(i, j, k, l) is: ω(i,j,k,l)=d(i,j,k,l)×r(i,j,k,l) in: The spatial domain kernel d is expressed as the Euclidean distance between the point (k, l) in the neighborhood and the center point (i, j), where Set the number for the parameter: The range kernel r represents the absolute value of the difference between the gray value f(kl) of the point (k, l) in the neighborhood and the gray value f(kl) of the center point (i, j), where Set the number for the parameter:

3. The rockfill material gradation detection method based on image segmentation algorithm according to claim 2 is characterized in that: The instance segmentation model in step 3) is Mask-RCNN.

4. The rockfill material gradation detection method based on image segmentation algorithm according to claim 3 is characterized in that: The specific implementation method of step 4) is: 4.1) Extracting mask quantification indexes and shape characteristic parameters from the mask information of stone particles; 4.2) The particle size and volume conversion algorithm is used to convert the mask quantification index and shape characteristic parameters obtained in step 4.1) into the particle size and volume of the stone.

5. The rockfill material gradation detection method based on image segmentation algorithm according to claim 4 is characterized in that: The mask quantization index in step 4.1) includes the perimeter of the mask edge contour, the minimum circumscribed rectangle of the stone particles, and the convex hull of the stone particles; The shape characteristic parameters in step 4.1) include flatness and elongation; the flatness is the ratio of the thickness of the stone particle to the width of the stone particle; the elongation is the ratio of the width of the stone particle to the length of the stone particle.

6. The method for detecting rockfill material gradation based on image segmentation algorithm according to claim 5, characterized in that: The calculation method of the perimeter of the edge contour of the mask is: suppose there are n points constituting the edge contour, and the coordinates of the points are expressed by (x i ,y i ), then the perimeter P of the stone particle can be expressed as: The calculation method of the stone particle mask area S is: The calculation method of the minimum circumscribed rectangle of the stone particles is: Assume that the point set of the stone contour is Q = {(x1, y1), (x2, y2), ..., (x n ,y n )}, calculate the convex hull of the stone contour, and obtain the convex hull point set Q′={(x′1,y1′),(x′2,y2′),…,(x′ n , y′ n )}, calculate the center of the convex hull point set in: and are the average values ​​of the horizontal and vertical coordinates of all points respectively; Transform the coordinates of each point in the convex hull point set to obtain a new point set Q″={(x″1, y″1), (x″2, y″2), …, (x″ n , y″ n )}, in: The new point set Q″ is constructed as a matrix M, the expression of M is: Use the least squares method to perform singular value decomposition on the matrix M and obtain the right singular vector matrix V: M=UΣV T in: U and V are orthogonal matrices; ∑ is a diagonal matrix; The last column of the matrix V is the normal vector v of the minimum circumscribed rectangle of the stone outline. The rotation angle angel is calculated by the normal vector v. Assuming θ is the polar coordinate angle of v, the rotation angle is: Calculate the width and height of the enclosing rectangle of the convex hull point set in the rotated coordinate system, and then obtain the minimum circumscribed rectangle of the mask; The shape characteristic parameters are obtained by dividing the stone into particle size ranges, using a vernier caliper to measure the length, width and thickness of the stone particles in different particle size ranges, and calculating the average flatness and average elongation of the stone particles in each particle size range according to the thickness, width and thickness of the stone particles in each particle size range; Preferably, the specific method of the measurement is: find the plane where the maximum projection area of ​​the particle is located, measure the distance between the maximum two points of the plane contour as the length of the stone particle; measure the distance between the maximum two points in the vertical length direction within the contour as the width of the stone particle; perpendicular to this plane, measure the distance between the maximum two points of the particle as the thickness of the stone particle.

7. The method for detecting rockfill material gradation based on image segmentation algorithm according to claim 6, characterized in that: The specific implementation method of the stone particle size conversion in step 4.2) is: the second length of the stone particle is used as the particle size of the stone particle, and the second length of the stone particle is the width of the smallest circumscribed rectangle of the stone particle when the stone particle is placed in a manner with the largest horizontal projection area as the particle size of the stone particle; The specific implementation method of the stone volume conversion in step 4.2) is: The volume conversion formula within each particle size range is: in: V i It is expressed as the volume of the i-th stone particle in the current particle size range; n represents the total number of stone particles within the current particle size range; S i It is expressed as the mask area of ​​the i-th stone particle in the current particle size range.

8. The method for detecting rockfill material gradation based on image segmentation algorithm according to claim 7, characterized in that: The gradation calculation formula of the three-dimensional stone particles in step 5) is: in: GIR is the ratio of the total mass of stone particles in each size range to the total mass of all stone particles in the image; ρ is the density of stone particles; N represents the total number of all stone particles; n represents the total number of stone particles within the current particle size range; V k It is expressed as the volume of the kth stone particle in all particle size ranges; S i Represents the mask area of ​​the i-th stone particle in the current particle size range; S k Represents the mask area of ​​the kth stone particle in all particle size ranges.

9. The method for detecting rockfill material gradation based on image segmentation algorithm according to claim 8, characterized in that: The specific implementation of step 6) is: 6.1) The mass percentage of each characteristic particle size in the three-dimensional overall state is inferred by weighted summation of the mass percentage of each characteristic particle size in the stone surface. The mass percentage calculation formula of each characteristic particle size after correction is: in: Expressed as the corrected mass percentage value for the characteristic particle size d0; l represents the number of characteristic particle sizes; λ i is the weight coefficient of the i-th characteristic particle size; z is the characteristic particle size d i The mass percentage value before correction; Among them, the weight coefficient is the estimated value that can satisfy d0 The optimal set of coefficients with the smallest difference from the true value z0, that is, At the same time, it satisfies the unbiased estimation Where Var and E are variance and expectation respectively; 6.2) Based on the constraints of minimum variance and unbiased estimation, the Lagrange multiplier method is used, and the weight coefficient λ is i The calculation formula is: in: r ij is the semivariogram function, expressed as λ is the Lagrange multiplier; 6.3) Assume r ij The particle size difference d from each characteristic particle size ij The spherical model is used to fit this relationship to obtain the functional relationship r = r (d), then for any two points d i ,d j By calculating the difference between two points d ij , calculate the semivariogram function r through the functional relationship ij , through the weight coefficient λ i The calculation formula of and the calculation formula of the mass percentage of each characteristic particle size after correction are used to obtain the corrected three-dimensional grading curve of stone.

10. A rockfill material gradation detection device for implementing the rockfill material gradation detection method based on image segmentation algorithm as claimed in any one of claims 1 to 9, characterized in that: The rockfill material gradation detection device comprises an image acquisition module, an image recognition module, a conversion module and a gradation curve drawing module; The image acquisition module is used to acquire the image of the stone particles to be detected; The image recognition module is used to perform image preprocessing operations on the photographed stone particle image and input it into the instance segmentation model, and obtain the stone particle image with mask information of the stone particles through the instance segmentation model; The conversion module is used to extract the mask quantification index and shape characteristic parameters of the stone particle image, and convert the extracted characteristic parameters into the particle size and the total volume of the stone within each particle size range; The grading curve drawing module converts the stone particle size and volume obtained by converting the extracted shape characteristic parameters into the stone apparent gradation, and corrects the apparent gradation through a data fusion model to obtain a three-dimensional overall grading curve of the stone.

Images