A device and method for detecting stone particle gradation based on image detection

Abstract

The present invention discloses a device and method for detecting the gradation of stone particles based on image detection. First, the rockfill dam material to be detected is screened and the dust particles with smaller particle sizes are directly weighed. The remaining larger particles fall along the slope, and a video image of the falling stone is taken to identify the particle size of the stone. The weight of the dust is added to convert it into the full gradation data of the rockfill dam material and draw a gradation curve of the rockfill dam material, so that the device and method for detecting the full gradation of soil and stone based on image recognition can be put into practical application. The present invention realizes the screening and automatic weighing of tiny stone particles and optimizes the discrete degree process of stone particle imaging. On the basis of avoiding the segmentation error caused by the obstruction of small stone particles and the concentrated stone imaging, the accuracy of soil and stone gradation detection based on image recognition is greatly improved, and a feasible technical solution is provided for the realization of rapid and high-precision detection of soil and stone gradation, which has important reference value and significance for engineering applications.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a device and method for detecting stone particle gradation based on image detection. Background Art

[0002] The mechanical properties and impermeability of rockfill dams have a significant impact on the safety of the dam, and the gradation of rockfill dam materials is the main factor affecting the mechanical properties and impermeability of the filling materials after compaction, which is of great significance in the construction of the dam body. Good gradation can ensure the compaction quality of the rockfill dam project and improve the impermeability and deformation resistance of the project. Therefore, the particle size gradation detection of rockfill dam materials is an important part of the quality control of rockfill dam construction.

[0003] Traditional stone gradation detection mainly adopts screening method, through random sampling and manual screening, the gradation data is calculated. However, manual screening is time-consuming and inefficient, with high labor costs, and it is difficult to meet the needs of modern intelligent construction for fast and efficient construction. With the development of computer science and technology, image recognition, as a new detection method, has been widely used in various fields, and has also been well developed and applied in the water conservancy industry, providing a new direction for the detection of rockfill dam material gradation.

[0004] The research work on the use of image recognition technology for stone particle size detection has achieved fruitful results, and the relevant calculation theory is becoming more and more mature. However, the above research theories or methods cannot adapt to the complex environmental conditions at the construction site of rockfill dams and cannot be widely used in the gradation detection of rockfill dam materials. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a device and method for detecting the particle gradation of stone material based on image detection. First, the rockfill dam material to be detected is screened and the dust particles with smaller particle sizes are directly weighed, and the remaining larger particles fall along the slope. Video images of the falling stone materials are taken and the particle sizes of the stone materials are identified. The dust weight is added to convert it into the full gradation data of the rockfill dam material and draw a gradation curve of the rockfill dam material, so that the device and method for detecting the full gradation of soil and stone material based on image recognition can be put into practical application. The test device mainly includes: the soil and stone full gradation detection device is mainly composed of a screening component and a shooting component. The screening component is a linear vibrating screen that can screen particles below 5mm and complete automatic weighing. The shooting component is composed of freely adjustable slopes, high-speed cameras, fill lights and other equipment; secondly, combined with the video images taken by the soil and stone gradation rapid detection device, the basic process of image preprocessing includes background separation, image grayscale processing, image histogram equalization and bilateral filtering noise reduction; then, image segmentation, motion blur correction and edge detection based on the maximum inter-class variance method are carried out in turn, and the characteristic parameters of the particles are extracted and converted into equivalent particle sizes of stones; finally, the equivalent particle sizes of stones are screened according to the partitions of the frame image and are arranged in order according to the equivalent particle sizes and saved in the same set, and the gradation curve of the rockfill is drawn in combination with the weight of the screened stones below 5mm, realizing the soil and stone full gradation detection based on image recognition.

[0006] In order to solve the above problems, the technical solution of the present invention is: a stone particle gradation detection method based on image detection, comprising the following steps:

[0007] Step 1. Separate soil and stone materials using a linear vibrating screen, weigh the particles below 5 mm under the screen, and spread and flatten the stone particles above 5 mm on the screen and transport them to the inclined slope imaging background plate;

[0008] Step 2. The shooting component continuously shoots the stone particles sliding down the imaging area of ​​the slope imaging background plate to obtain a frame image;

[0009] Step 3. From all the frames, frame images of stone particles of one specification are sequentially taken at every set optimal frame-taking interval from the front to the back, and saved as valid frame images of stone particles of the specification, so that the stone particles of the specification appear only once in two valid frame images taken consecutively;

[0010] Step 4. Perform image processing on the effective frame images of stone particles of various specifications, extract characteristic parameters of the stone particles and convert them into equivalent particle size data of the stone particles;

[0011] Step 5. According to the specifications of the stone particles corresponding to the effective frame diagram, the equivalent particle size data of the stone particles that meet the specifications are taken out from the equivalent particle size data extracted from the effective frame diagram, and the equivalent mass of the stone particles is obtained according to the corresponding relationship between the equivalent particle size and the mass;

[0012] Step 6. The equivalent particle size and equivalent mass of the stone particles obtained in step 5 in the same valid frame are combined into the same set, and arranged in ascending order according to the equivalent particle size of the particles: let i be the arrangement number, n be the total number of detected particle sizes, then the set can be expressed as:

[0013] O i ={(d1,m1)(d2,m2)…(d i ,m i )…(d n ,m n )}

[0014] In the formula, i is the arrangement number, n is the total number of detected particle sizes, and d i is the equivalent particle size, m i is the equivalent mass;

[0015] Step 7. Calculate the particle size smaller than d i Mass percentage of all stone particles:

[0016]

[0017] In the formula, m p,i is smaller than the particle size d i The mass percentage of stone, m df is the weight of the material under the vibrating primary screen, m a is the actual total weight of the stones photographed. The particle size identified in step 6 is no larger than d i The quality of the stone, The effective mass of all stone particles identified in step 6;

[0018] Step 8. Draw a stone grading curve with the stone particle size as the horizontal coordinate and the mass percentage of stone particles smaller than each particle size as the vertical coordinate.

[0019] The result accuracy analysis model of the stone gradation curve includes two indicators: absolute accuracy and relative error. The absolute accuracy represents the degree of fit between the gradation curve obtained in step 8 and the preset gradation curve, and the relative error reflects the degree of segmental deviation between the gradation curve obtained in step 8 and the preset gradation curve.

[0020] The relative error calculation formula is as follows:

[0021] W i =[(y p,i -y p,i-1 )-(y s,i -y s,i-1 )]×100%

[0022] Where i is the sequence number, satisfying 2<<i<<M, M is the length of the grading curve data sequence obtained in step 8, and y s,i is the ordinate of the preset grading curve, y p,i is the ordinate of the matching curve obtained in step 8;

[0023] The absolute accuracy calculation formula is as follows:

[0024]

[0025] Where AAC is the absolute accuracy of the grading curve obtained in step 8 relative to the preset grading curve, <AAC<1。

[0026] The shooting frame rate of the shooting component is F, and the number of times the stone particles appear continuously in the frame image is N. t for:

[0027]

[0028] Where, v0 is the initial velocity of the stone when it leaves the screening device and enters the slope surface, ε is the slope angle influence coefficient of the slope imaging background plate, ε = g(sinα-μcosα), g is the acceleration of gravity; α is the slope inclination of the slope imaging background plate; μ is the friction coefficient of the slope imaging background plate; h0 is the displacement of the stone sliding along the slope imaging background plate to the upper edge of the imaging area, L h is the length of the imaging area.

[0029] The specifications include large, medium and small particle sizes: the large particle size range is 60 to 100 mm; the medium particle size range is 20 to 60 mm; the small particle size range is 5 to 20 mm;

[0030] The optimal frame interval N corresponding to large, medium and small particle intervals b b 、N b m 、N b s N t The values ​​obtained by multiplying the corresponding scaling factors and rounding up are as follows. When the slope angle of the slope imaging background plate is 40°, the scaling factors corresponding to the large, medium and small particle intervals are 0.34, 0.38 and 0.37, respectively. b b 、N b m 、N b s They are 14, 16 and 15 respectively.

[0031] The image processing step in step 4 includes: image preprocessing, and then sequentially performing image segmentation based on the maximum inter-class variance method and edge detection to extract particle contour features;

[0032] The image preprocessing includes: background separation, image grayscale processing, image histogram equalization and bilateral filtering noise reduction.

[0033] The bilateral filtering denoising method combines the image pixel value similarity with the spatial distance for weighted superposition, and can better preserve the image edge information while denoising.

[0034] The edge detection includes edge detection based on Roberts operator and particle size recognition based on circumscribed ellipse method.

[0035] The method of extracting the characteristic parameters of the stone particles and converting them into the equivalent particle size data of the stone particles includes: extracting the edge line scatter point coordinates of each stone object in the edge detection result image, obtaining the major axis and minor axis lengths of the graph composed of the edge line scatter point coordinates, and calculating the equivalent particle size of the stone particles. The equivalent formula for calculating the equivalent particle size of the stone particles is as follows:

[0036]

[0037] Where α is the major axis of the stone particle, and β is the minor axis of the stone particle.

[0038] The maximum inter-class variance algorithm is used to segment the grayscale image, with the inter-class variance σ 2 The maximum threshold t is used as the binary image segmentation threshold, and then the pixel value of the background area in the grayscale image is changed to 0, and the pixel value of the foreground area is changed to 255. The grayscale mean square error of the two types of areas is:

[0039] σ 2 =ω0(μ0-μ T ) 2 +ω1(μ1-μ T ) 2

[0040] In the formula, σ 2 is the grayscale mean square error, ω0 is the probability of occurrence of pixels with grayscale levels 0 to t, μ0 is the grayscale mean of the foreground area, μ1 is the grayscale mean of the background area, and μ T is the grayscale mean of the entire image, and ω1 is the probability of occurrence of pixels from t+1 to L-1.

[0041] The edge detection based on Roberts operator determines the gradient amplitude of Roberts operator, and then determines a suitable threshold. The point with gradient amplitude greater than the threshold is the step edge point, so as to extract the particle size of stone particles. The gradient amplitude of Roberts operator is:

[0042]

[0043] Where, dx 2 (i,j),d y 2 (i, j) are the difference components of the binary image f(x, y) in the x and y directions respectively.

[0044] A stone particle grading detection system based on image detection, comprising a stone paving and conveying system composed of a linear vibrating screen and a slope imaging background plate as described above, and an image acquisition and analysis system including a shooting component as described above;

[0045] The slope imaging background board can be manually adjusted in angle, including a rotating support plate, a support frame, and a PVC pad. The upper end of the slope imaging background board is softly connected to the outlet of the linear vibration screen surface. Baffles are arranged on both sides of the slope of the rotating support plate to prevent stones from rolling down from both sides. One end of the inclined support rod is fixed to the upper end of the steel frame and can rotate around the axis, and the other end supports the slope imaging background board. An interface is arranged on the back of the slope imaging background board, and the inclined support rod is connected when adjusting the angle; the surface of the slope imaging background board is based on a steel wire mesh, and the surface layer is a red PVC pad with good wear resistance that covers the imaging area. The slope imaging background board can be freely adjusted to an angle of 60°, 55°, 50°, 45°, 40°, 35°, 30°, and 25°.

[0046] The shooting component is composed of a high-speed camera, a fill light and an image processing system;

[0047] The linear vibrating screen is a double-axis inertial vibrating screen, comprising: a screen box, a screen mesh, a motor, a transmission device, and a vibration isolation device.

[0048] The high-speed camera stand is fixed in front of the slope imaging background board so that its shooting area faces the slope imaging background board, the focal length of the high-speed camera is adjusted so that the width of the high-speed camera video recording area coincides with the width of the slope imaging background board, and the brightness and irradiation angle of the fill light are adjusted to ensure uniform light brightness on the slope imaging background board.

[0049] The image acquisition and analysis system installs and executes the programs of steps 3 to 7.

[0050] This invention realizes the screening and automatic weighing of tiny stone particles and optimizes the discreteness process of stone particle imaging. On the basis of avoiding segmentation errors caused by small stone particles being blocked and concentrated stone imaging, it greatly improves the detection accuracy of soil and stone gradation based on image recognition, provides a feasible technical solution for the realization of fast and high-precision detection of soil and stone gradation, and has important reference value and significance for engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present invention will be further described below in conjunction with the accompanying drawings:

[0052] Figure 1This is a flow chart of the soil and rock gradation detection device test and method of the present invention;

[0053] Figure 2 It is a schematic diagram of the overall three-dimensional structure of the soil and stone material gradation detection device of the present invention;

[0054] Figure 3 It is a schematic diagram of the three-dimensional structure of the adjustable falling slope of the soil and stone material gradation detection device of the present invention;

[0055] Figure 4 It is a schematic diagram of the whole object of the soil and stone material gradation detection device of the present invention;

[0056] Figure 5 This is a schematic diagram of the background separation effect of the soil and stone material gradation detection device of the present invention;

[0057] Figure 6 This is a schematic diagram of the grayscale processing effect of the soil and stone material gradation detection device of the present invention;

[0058] Figure 7 It is a schematic diagram of the image histogram equalization effect of the soil and stone material gradation detection device of the present invention;

[0059] Figure 8 It is a schematic diagram of the bilateral filtering noise reduction effect of the soil and stone material gradation detection device of the present invention;

[0060] Fig. 9 This is a schematic diagram of the effect of using the maximum inter-class variance algorithm to segment grayscale images in the soil and stone material gradation detection device of the present invention;

[0061] Fig.10 This is a schematic diagram of the edge detection effect of the soil and stone gradation detection device of the present invention based on the Roberts operator;

[0062] Fig.11 It is a schematic diagram of the instantaneous state of stone particles falling into the imaging area of ​​the slope imaging background plate in the present invention. DETAILED DESCRIPTION

[0063] The technical solutions and beneficial effects of the present invention are described below in conjunction with the accompanying drawings and specific embodiments.

[0064] like Figure 2 to Figure 4An indoor test device for soil and stone grading detection considering the discreteness of stone particle imaging is mainly composed of a linear vibrating screen 1, a manually adjustable angle slope imaging background plate 2 and a shooting component 3, wherein the linear vibrating screen 1 is mainly used to filter and automatically weigh particles below 5 mm, and to vibrate and feed stone particles in batches to the slope imaging background plate 2, and to disperse and spread the stones as evenly as possible through vibration. The manually adjustable angle slope imaging background plate 2 mainly comprises a rotating support plate 21, a support frame (23, 24, 25), a PVC pad 21, etc., which can realize the fixed multi-angle adjustment of the slope imaging background plate 2 and the optimal adjustment of the discreteness of stone particle imaging; the shooting component 3 is composed of a high-speed camera, a fill light and an image processing system. On the basis of shooting video images, background separation, image grayscale processing, image histogram equalization and bilateral filtering noise reduction and other image preprocessing are performed, and image segmentation and edge detection based on the maximum inter-class variance method are performed in turn, and the characteristic parameters of the particles are extracted and converted into equivalent stone particle sizes; finally, the equivalent particle sizes of the stones are screened according to the partitions of the frame images and are arranged in order according to the equivalent particle sizes and saved in the same set, and the gradation curve of the rockfill material is drawn in combination with the weight of the screened stones below 5 mm, so as to realize the soil and stone gradation detection considering the discreteness of stone particle imaging.

[0065] In the preferred solution, the linear vibrating screen is also called a double-axis inertial vibrating screen, and its main components include: screen box, screen, motor, transmission device, vibration isolation device, etc. The screen surface of the linear vibrating screen adopts a standard screen with a width of 700mm, a length of 1700mm, a height of 900mm, and a screen diameter of 5mm. The lower end of the screen is connected to the storage container, and an electronic scale is arranged under the container.

[0066] The characteristic parameters of the linear vibrating screen mainly include the screen surface inclination angle, vibration direction angle, throwing intensity, amplitude and frequency, material movement speed and screening efficiency.

[0067] In the preferred solution, the upper end of the manually adjustable angle slope imaging background plate 2 is softly connected to the outlet of the linear vibration screen surface, baffles 22 are arranged on both sides of the slope surface to prevent stones from rolling down from both sides, one end of the slope support rod 23 is fixed to the upper end of the steel frame 24 and can rotate around the axis, and the other end supports the slope imaging background plate 2, and an interface is arranged on the back of the slope imaging background plate 2, so that the slope support rod 23 can be quickly connected when adjusting the angle, such as Figure 3 shown.

[0068] In the preferred solution, the manually adjustable angle slope imaging background plate 2 is based on a steel mesh, and the surface layer is a red PVC mat 21 with good wear resistance. The color of the imaging background is obviously contrasted with the color of the stone particles, and the imaging effect is better. The size of the PVC mat 21 is the same as the size of the imaging area and covers the imaging area.

[0069] In a preferred embodiment, the manually adjustable angle slope imaging background plate 2 can be freely adjusted to an angle of 60°, 55°, 50°, 45°, 40°, 35°, 30°, and 25°.

[0070] In the preferred scheme, the high-speed camera stand is fixed in front of the manually adjustable angle slope imaging background board 2 so that its shooting area is facing the slope imaging background board 2, the focal length of the high-speed camera is adjusted so that the width of the high-speed camera video recording area coincides with the width of the slope imaging background board 2, and the brightness and irradiation angle of the fill light are adjusted to ensure uniform lighting on the slope imaging background board 2.

[0071] In a preferred solution, the particle size range of the stone samples is 0 to 100 mm in combination with the detection device, the mass percentage of 5 mm is controlled below 5%, and the total weight of the two groups of stone samples is 25 kg.

[0072] In a preferred embodiment, the gradation detection experimental step comprises the following steps:

[0073] Step 1: Install the detection device and prepare stone samples;

[0074] Step 2: Set up the high-speed camera 3 on the horizontal ground so that its camera lens faces the center of the slope imaging background plate 2. Fix the height and adjust the focal length of the high-speed camera so that the width of the high-speed camera video recording area is equal to the width of the slope imaging background plate 2. Turn on the fill light.

[0075] Step 3: Pour the stone sample onto the screen surface of the linear vibrating screen;

[0076] Step 4: Turn on the high-speed camera to start recording, and turn on the linear vibrating screen motor;

[0077] Step 5: After all the stones in the group have fallen, the high-speed camera stops recording, and then turns off the motor to record the weight of the stones in the container under the screen and the weight of the stones after falling along the slope;

[0078] Step 6: Export the video image in the high-speed camera, and the indoor rockfill dam material screening and shooting process is completed;

[0079] Step 7: Conduct video image analysis and full gradation conversion of rockfill dam materials.

[0080] In a preferred solution, the rockfill dam material video image analysis and full gradation conversion includes establishing a comprehensive extraction model for effective frame images, image preprocessing, extracting particle contour features, full gradation conversion of rockfill dam materials and drawing stone gradation curves.

[0081] A comprehensive extraction model for effective frame images is established to sequentially extract frame images of stone particles of one specification from all frame images at every set optimal frame interval and save them as effective frame images of stone particles of said specification, so that stone particles of said specification appear only once in two consecutively extracted effective frame images.

[0082] In a preferred embodiment, the basic process of image preprocessing includes background separation, grayscale processing, histogram equalization and bilateral filtering noise reduction, such as Figures 5 to 8 shown.

[0083] In a preferred solution, the bilateral filtering denoising performs weighted superposition based on the consideration of spatial distance and the similarity of image pixel values, which can better preserve image edge information while denoising.

[0084] In a preferred embodiment, the extraction of particle contour features includes image segmentation based on the maximum inter-class variance method, edge detection based on the Roberts operator, and particle size recognition based on the circumscribed ellipse method, such as Figures 9-10 shown.

[0085] In a preferred embodiment, the basic process of converting the particle profile characteristics into the equivalent particle size of stone includes:

[0086] Step 1: Use the maximum inter-class variance algorithm to segment the grayscale image, with the inter-class variance σ 2 The maximum threshold t is used as the binary image segmentation threshold, and the pixel values ​​of the background area in the grayscale image are changed to 0 (black), and the pixel values ​​of the foreground area are changed to 255 (white). Variance is a measure of the uniformity of grayscale distribution. The larger the inter-class variance between the background and foreground, the greater the difference between the two parts of the image. When part of the foreground is mistakenly classified as the background or part of the background is mistakenly classified as the foreground, the difference between the two parts will become smaller.

[0087] The grayscale mean square error of the two types of regions is: 2 =ω0(μ0-μ T ) 2 +ω1(μ1-μ T ) 2

[0088] In the formula, σ 2 is the grayscale mean square error, ω0 is the probability of occurrence of pixels with grayscale levels 0 to t, μ0 is the grayscale mean of the foreground area, μ1 is the grayscale mean of the background area, and μ T is the grayscale mean of the entire image, ω1 is the probability of occurrence of pixels with grayscale levels t+1 to L-1, and L is the total grayscale level of the grayscale image.

[0089] Step 2: After determining the gradient amplitude of the Roberts operator, determine a suitable threshold. The point with a gradient amplitude greater than the threshold is the step edge point, thereby extracting the stone particle size. The gradient amplitude of the Roberts operator is:

[0090]

[0091] Where, d x 2 (i,j),d y 2 (i, j) are the difference components of the binary image f(x, y) in the x and y directions respectively.

[0092] Step 3: In the edge detection result image, extract the edge line scatter point coordinates of each stone object, obtain the major axis and minor axis lengths of the graph composed of the edge line scatter point coordinates, and calculate the equivalent particle size of the stone particles. Then, based on the corresponding relationship between the equivalent particle size and the mass, obtain the equivalent mass of the stone particles. The equivalent formula for calculating the equivalent particle size of stone particles is as follows:

[0093]

[0094] Where α is the major axis of the stone particle, and β is the minor axis of the stone particle.

[0095] In the preferred solution, theoretically, if the camera shooting frame rate is F, then the number of photos taken continuously by the camera, that is, the number of times the spherical stone appears continuously in the frame image, is N. t for:

[0096]

[0097] Where, v0 is the initial velocity of the stone when it leaves the screening device and enters the slope surface, ε is the slope angle influence coefficient of the slope imaging background plate, ε = g(sinα-μcosα), g is the acceleration of gravity; α is the slope inclination of the slope imaging background plate; μ is the friction coefficient of the slope imaging background plate; h0 is the displacement of the stone sliding along the slope imaging background plate to the upper edge of the imaging area, L h is the length of the imaging area.

[0098] like Fig.11 , when the spherical stone P1 theoretically enters the shooting area (that is, slides to point B), the displacement on the slope imaging background plate is h0, and the time taken for the spherical stone to slide from point A to point B is:

[0099]

[0100]

[0101] The speed of the spherical stone when it rolls to point B is:

[0102]

[0103] The total displacement of the spherical stone on the inclined plane at the moment it passes through the shooting area (that is, when it slides to point C) is:

[0104] s c =L h +h0

[0105] Where, L h Indicates the height of the camera's view window on the background.

[0106] The corresponding time spent is:

[0107]

[0108] At this time, the maximum speed of the spherical stone is:

[0109]

[0110] Based on the analysis of the motion state of the stone in the previous text, the number of frames that the spherical stone may appear in under ideal conditions is calculated. The time it takes for the spherical stone to fall from the beginning to enter the shooting area is t b , the time taken for the spherical stone to fall and leave the shooting area is t c Assume that the camera frame rate is F, then the number of photos taken continuously by the camera, that is, the number of times the spherical stone appears continuously in the frame image, is N. t for:

[0111]

[0112] Where F is the camera frame rate; t c The time it takes for the spherical stone to fall and leave the shooting area; t b It is the time it takes for the spherical stone to enter the shooting area from the beginning of its fall.

[0113] Since the stone particles are irregular ellipsoids in reality, the speeds at which stone particles of different sizes slide down the slope are not equal. Therefore, stone particles of different sizes may appear twice or not at all in adjacent frames taken at the same optimal frame interval.

[0114] In the preferred scheme, a comprehensive extraction model of effective frame images is established to extract three effective frame image groups according to the different specification ranges of stones, respectively, by setting the best frame interval division intervals, for the gradation identification of large, medium and small particle size intervals.

[0115] The specifications include large, medium and small particle size ranges: the large particle size range is 60 to 100 mm; the medium particle size range is 20 to 60 mm; the small particle size range is 5 to 20 mm;

[0116] The optimal frame interval N corresponding to large, medium and small particle intervals b b 、N b m 、N b s N t The values ​​are obtained by multiplying them with the corresponding scaling factors and rounding them up. The large, medium and small particle intervals correspond to a scaling factor respectively.

[0117] According to the analysis and demonstration of the stone frame number in the previous article, a comprehensive extraction model of effective frame images is established for the video image processing of stones sliding down the slope.

[0118] First, according to the working conditions of the video shooting, the theoretical optimal frame interval N is calculated. t .

[0119] The optimal frame interval N t Multiply by the scaling factor of large, medium and small particle intervals respectively and round up to get the optimal frame interval N in each interval b b 、N b m 、N b s .

[0120] Starting from the first frame of the captured video, every N b b One image is saved as a valid frame image for large particle detection, and the rest of the images are kept in the database without processing until the last frame image, then all the detection frames of large particle size particles can be selected. b m 、N b s The above process is repeated for each image and saved as valid frames for medium and small particle detection.

[0121] Image recognition is performed using these three sets of frames, and the equivalent particle sizes and equivalent volumes of large, medium and small particle size stone objects are extracted according to their particle size partitions. That is, only particle data with an equivalent particle size of 60-100mm is saved in the large particle size partition, only particle data with an equivalent particle size of 20-60mm is saved in the medium particle size partition, and only particle data with an equivalent particle size of 5-20mm is saved in the small particle size partition.

[0122] The effective frame image extraction model can ensure that no data is missed or redundant when conducting particle size detection on stone objects, thereby improving the accuracy of the entire detection system.

[0123] When the slope angle of the slope imaging background plate is 40°, the scaling factors corresponding to the large, medium and small particle size ranges are 0.34, 0.38 and 0.37 respectively, N b b 、N b m 、N b s They are 14, 16 and 15 respectively.

[0124] In a preferred solution, the three sets of frame images are used for image recognition, and the equivalent particle sizes and equivalent volumes of large, medium and small particle size stone objects are extracted according to their particle size partitions.

[0125] In a preferred solution, the three valid frame groups are retained to obtain the equivalent particle size data of particles that meet the particle size specification range of the partition, and the data are merged into the same set and arranged in ascending order according to the equivalent particle size of the particles: let i be the arrangement number, and n be the total number of detected particle sizes, then the set can be expressed as:

[0126] O i ={(d1,m1)(d2,m2)…(d i ,m i )…(d n ,m n )}

[0127] In the formula, i is the arrangement number, n is the total number of detected particle sizes, d is the equivalent particle size, and m is the equivalent mass.

[0128] In the preferred embodiment, the particle size d is calculated to be smaller than i Mass percentage of all stone particles:

[0129]

[0130] In the formula, m p,i is smaller than the particle size d i The mass percentage of stone, m df is the weight of the material under the vibrating primary screen, m a is the actual total weight of the stones photographed. The particle size identified in step 6 is no larger than d i The quality of the stone, The effective mass of all stone particles identified in step 6;

[0131] Among them, when di>5mm, the equivalent mass of the stone particles below 5mm screened out can be written as:

[0132] The total equivalent mass of all stone particles can be written as:

[0133] The weight of stone particles with a particle size smaller than di can be expressed as (the equivalent mass of all stone particles with a particle size smaller than di is):

[0134] In the preferred scheme, the stone grading data below 5 mm is obtained by dividing the mass percentage of stone particles below 5 mm by particle size. This section appears as a straight line on the grading curve. After obtaining the mass percentage of stone particles smaller than a certain particle size, the stone grading curve is drawn with the stone particle size as the horizontal coordinate and the mass percentage of particles smaller than a certain particle size as the vertical coordinate.

[0135] In a preferred solution, the accuracy analysis model of the gradation test result includes two indicators: absolute accuracy and relative error. The absolute accuracy represents the degree of fit between the two curves, and the relative error reflects the degree of segmental deviation between the two curves.

[0136] In the preferred solution, the relative error calculation formula is as follows:

[0137] W i =[(y p,i -y p,i-1 )-(y s,i -y s,i-1 )]×100%

[0138] Where i is the sequence number, satisfying 2<<i<<M, M is the length of the screening distribution curve data sequence, y s,i is the ordinate of the preset grading curve, y p,i It is the ordinate of the screening distribution curve.

[0139] In the preferred solution, the absolute accuracy calculation formula is as follows:

[0140]

[0141] Where AAC is the absolute accuracy of the gradation curve detected by the video image relative to the preset gradation curve, <AAC<1。

[0142] The above-mentioned embodiments are only preferred technical solutions of the present invention and should not be regarded as limitations of the present invention. The embodiments and features in the embodiments of the present application can be arbitrarily combined with each other without conflict. The protection scope of the present invention shall be the technical solutions recorded in the claims, including the equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for detecting stone particle gradation based on image detection, characterized in that: The following steps are involved: Step 1. Separate soil and stone materials using a linear vibrating screen, weigh the particles under the screen that are less than 5 mm, and spread and flatten the stone particles on the screen that are larger than 5 mm, and transport them to the inclined slope imaging background plate; Step 2. The shooting component continuously shoots the stone particles sliding down the imaging area of ​​the slope imaging background plate to obtain a frame image; Step 3. From all the frames, frame images of stone particles of one specification are sequentially taken at every set optimal frame-taking interval from the front to the back, and saved as valid frame images of stone particles of the specification, so that the stone particles of the specification appear only once in two valid frame images taken consecutively; The shooting frame rate of the shooting component is F, and the number of times the stone particles appear continuously in the frame image is N. t for: Where 𝑣0 is the initial velocity of the stone when it leaves the screening device and enters the slope surface; 𝜀 is the slope angle influence coefficient of the slope imaging background plate, 𝜀=𝑔(sin𝛼-μcos𝛼), 𝑔 is the acceleration of gravity; 𝛼 is the slope inclination of the slope imaging background plate; 𝜇 is the friction coefficient of the slope imaging background plate; ℎ0 is the displacement of the stone sliding along the slope imaging background plate to the upper edge of the imaging area, 𝐿 ℎ is the length of the imaging area; The specifications include large, medium and small particle sizes: the large particle size range is 60 to 100 mm; the medium particle size range is 20 to 60 mm; the small particle size range is 5 to 20 mm; The optimal frame interval N corresponding to large, medium and small particle intervals b b 、N b m 、N b s N t The values ​​obtained by multiplying the corresponding scaling factors and rounding up are as follows. When the slope angle of the slope imaging background plate is 40°, the scaling factors corresponding to the large, medium and small particle intervals are 0.34, 0.38 and 0.37, respectively. b b 、N b m 、N b s They are 14, 16, and 15 respectively; Step 4. Perform image processing on the effective frame images of stone particles of various specifications, extract characteristic parameters of the stone particles and convert them into equivalent particle size data of the stone particles; The image processing steps include: image preprocessing, and then sequentially performing image segmentation based on the maximum inter-class variance method, particle edge detection, and particle size recognition; the image preprocessing includes: background separation, image grayscale processing, image histogram equalization, and bilateral filtering noise reduction; The image segmentation based on the maximum inter-class variance method is to segment the grayscale image using the maximum inter-class variance algorithm, with the inter-class variance σ 2 The maximum threshold t is used as the binary image segmentation threshold, and then the pixel value of the background area in the grayscale image is changed to 0, and the pixel value of the foreground area is changed to 255; the grayscale mean square error of the two types of areas is: s 2 =ω0(μ0-μ T ) 2 +ω1(μ1-μ T ) 2 In the formula, σ 2 is the grayscale mean square error, ω0 is the probability of occurrence of pixels with grayscale levels 0 to t, μ0 is the grayscale mean of the foreground area, μ1 is the grayscale mean of the background area, and μ T is the grayscale mean of the entire image, ω1 is the probability of occurrence of pixels with grayscale levels t+1 to L-1, and L is the total grayscale level of the grayscale image; Based on the particle edge detection of Roberts operator, after determining the gradient amplitude of Roberts operator, a suitable threshold is determined. The point with gradient amplitude greater than the threshold is the step edge point, so as to extract the particle size of stone particles. The gradient amplitude of Roberts operator is: Where, d x 2 (i,j),d y 2 (i, j) are the difference components of the binary image f(x, y) in the x and y directions respectively; The method of extracting the characteristic parameters of the stone particles and converting them into the equivalent particle size data of the stone particles includes: extracting the edge line scatter point coordinates of each stone object in the particle edge detection result image, obtaining the major axis and minor axis lengths of the graph composed of the edge line scatter point coordinates, and calculating the equivalent particle size of the stone particles. The equivalent formula for calculating the equivalent particle size of the stone particles is as follows: In the formula, 𝛼 is the major axis of the stone particle, and 𝛽 is the minor axis of the stone particle; Step 5. According to the specifications of the stone particles corresponding to the effective frame image, the equivalent particle size data of the stone particles that meet the specifications are taken out from the equivalent particle size data extracted from the effective frame image, and the equivalent mass of the stone particles is obtained according to the corresponding relationship between the equivalent particle size and the mass; Step 6. The equivalent particle size and equivalent mass of the stone particles obtained in step 5 in the same valid frame are combined into the same set, and arranged in ascending order according to the equivalent particle size of the particles: let i be the arrangement number, n be the total number of detected particle sizes, then the set can be expressed as: O i ={(d1,m1)(d2,m2)…(d i ,m i )…(d n ,m n )} In the formula, i is the arrangement number, n is the total number of detected particle sizes, and d i is the equivalent particle size, m i is the equivalent mass; Step 7. Calculate the particle size smaller than d i Mass percentage of all stone particles: In the formula, 𝑚 𝑝,𝑖 is smaller than the particle size d i The mass percentage of stone, d𝑓 is the weight of the material under the vibrating primary screen, 𝑚 𝑎 is the actual total weight of the stones photographed. The particle size identified in step 6 is no larger than d i The quality of the stone, The effective mass of all stone particles identified in step 6; Among them, d i When the particle size is >5mm, the equivalent mass of the sieved stone particles below 5mm can be written as: The total equivalent mass of all stone particles can be written as: Particle size smaller than d i The equivalent mass of all stone particles is: Step 8. Screen the equivalent particle size of the stone according to the partition of the frame image and arrange them in order according to the equivalent particle size and save them in the same set. Draw the gradation curve of the rockfill material based on the weight of the screened stones below 5 mm: draw the stone gradation curve with the particle size of the stone as the horizontal coordinate and the mass percentage of the stone particles smaller than each particle size as the vertical coordinate.

2. The stone particle gradation detection method based on image detection according to claim 1 is characterized in that: The result accuracy analysis model of the stone gradation curve includes two indicators: absolute accuracy and relative error. The absolute accuracy represents the degree of fit between the gradation curve obtained in step 8 and the preset gradation curve, and the relative error reflects the degree of segmental deviation between the gradation curve obtained in step 8 and the preset gradation curve. The relative error calculation formula is as follows: W i =[(and p,i -and p,i-1 )-(and s,i -and s,i-1 )]×100% Where i is the sequence number, satisfying 2<<i<<M, M is the length of the grading curve data sequence obtained in step 8, and y s,i is the ordinate of the preset grading curve, y p,i is the ordinate of the matching curve obtained in step 8; The absolute accuracy calculation formula is as follows: Where AAC is the absolute accuracy of the grading curve obtained in step 8 relative to the preset grading curve, <AAC<1。 3. A stone particle gradation detection system used in the stone particle gradation detection method based on image detection as described in any one of claims 1 to 2, characterized in that: It includes a stone paving and conveying system consisting of a linear vibrating screen, a slope imaging background plate, and an image acquisition and analysis system for shooting components; The slope imaging background board can be manually adjusted in angle, including a rotating support plate, a support frame, and a PVC pad. The upper end of the slope imaging background board is softly connected to the outlet of the linear vibration screen surface. Baffles are arranged on both sides of the slope of the rotating support plate to prevent stones from rolling down from both sides. One end of the inclined support rod is fixed to the upper end of the steel frame and can rotate around the axis, and the other end supports the slope imaging background board. An interface is arranged on the back of the slope imaging background board, and the inclined support rod is connected when adjusting the angle; the surface of the slope imaging background board is based on a steel wire mesh, and the surface layer is a red PVC pad with good wear resistance that covers the imaging area. The slope imaging background board can be freely adjusted to an angle of 60°, 55°, 50°, 45°, 40°, 35°, 30°, and 25°. The shooting component is composed of a high-speed camera, a fill light and an image processing system; The linear vibrating screen is a double-axis inertial vibrating screen, including: a screen box, a screen mesh, a motor, a transmission device, and a vibration isolation device; The image acquisition and analysis system installs and executes programs.

4. The stone particle gradation detection system according to claim 3 is characterized in that: The high-speed camera stand is fixed in front of the slope imaging background board so that its shooting area faces the slope imaging background board, the focal length of the high-speed camera is adjusted so that the width of the high-speed camera video recording area coincides with the width of the slope imaging background board, and the brightness and irradiation angle of the fill light are adjusted to ensure uniform light brightness on the slope imaging background board.

Images