Efficient identification method and system for soil particle changes in large-scale model tests

Abstract

The present invention provides a method and system for efficiently identifying changes in soil particles in a large-scale model test. The method comprises: step 1, collecting a series of pictures of soil particle movement simulated by the test; step 2, cutting and binarizing the pictures; step 3, extracting position and angle information of rod-shaped similar soil; step 4, comparing position information of rod-shaped similar soil particles in binary images of different motion states to obtain displacement information of the rod-shaped similar soil particles; obtaining displacement information of each rod-shaped similar soil particle and outputting it as a displacement cloud map of the rod-shaped similar soil particles; step 5, identifying the longest inner diameter of the rod-shaped similar soil particle as the major axis and extracting the major axis direction; counting the average major axis direction distribution information of the particles in each grid area; comparing the average major axis direction difference in the binary images of the same grid area at different times to obtain the average major axis angle deflection information of the particles in each grid area, and further obtaining the overall rotation angle contour line distribution information.

Description

Technical Field

[0001] The present invention belongs to the technical field of geotechnical engineering microscopic image analysis, and particularly relates to a method and system for efficiently identifying soil particle changes in large-scale model tests. Background Art

[0002] In the field of geotechnical engineering, for studying topics such as soil arch effect and slope evolution, the combination of model tests and numerical simulations can often study the macroscopic and microscopic evolution laws of soil.

[0003] Discrete element numerical simulation can extract microscopic evolution information such as particle displacement, rotation angle, and porosity change. However, discrete element numerical simulation has shortcomings such as the shape and size of the established numerical models, such as soil particles and piles, not matching the actual model, and low computational efficiency. Often, when the soil model is small, numerous, and large, the computation takes too long, affecting the experimental progress. In addition, for complex working conditions such as uneven subsidence, the boundary conditions and related physical and mechanical parameters are simplified, which cannot reflect the real complex working conditions in actual engineering. Scanning with a microscopic electron microscope not only has high requirements for equipment samples, but also the processing process is more time-consuming and labor-intensive. It is only suitable for static component analysis with small samples and a small number of particles. In addition, in the prior art, there is also a method of shooting test videos and then using a computer to learn and process them frame by frame to identify the changes of all particles. This method can identify the dynamic changes of particles in a short period of time, but the processing volume is very large, and it has high requirements for the computer graphics card and memory capacity, and it is very time-consuming.

[0004] For large-scale model tests where the test process lasts for several hours and the size reaches more than 1000 mm, the above problems will be more serious and make it impossible to apply. Summary of the Invention

[0005] The present invention is developed to solve the above-mentioned problems, and its purpose is to provide an efficient method and system for identifying the changes in soil particles in large-scale model tests, which can conveniently and efficiently identify the dynamic changes of each microscopic particle during the large-scale model test and output the true constitutive information of the particles.

[0006] In order to achieve the above purpose, the present invention adopts the following scheme:

[0007] <Method>

[0008] The present invention provides a method for efficiently identifying soil particle changes in a large-scale model test, which is characterized by comprising the following steps:

[0009] Step 1: Based on the soil information to be simulated, rod structures of corresponding size, gradation, and number with elliptical cross-sectional shapes are used to simulate soil particles, and these rods are densely stacked to form rod-shaped similar soil to simulate the soil for use in the test; the elliptical side of the rod structure is used as the observation side of the model, and a high-definition camera is aimed at the observation side of the model. A series of high-resolution images of the movement of soil particles simulated by the rod structure during the test are collected at a certain interval, thereby obtaining a series of initial images containing the dynamic process of the rod-shaped similar soil particles; the interval between photographs is at least 10 seconds;

[0010] Step 2: Cut the image, select the effective area related to the rod-shaped similar soil particles in the image, and grayscale it to obtain a grayscale image; set the color division value to select an appropriate color retention range for the grayscale image, divide the grayscale image into two different grayscale values ​​of the rod-shaped similar soil particles and the pore part to obtain a preliminary distinction image; cut the preliminary distinction image to separate adjacent rod-shaped similar soil particles, and do not cut a single rod-shaped similar soil particle; perform binarization on the cut image to clearly distinguish the rod-shaped similar soil particles from the pores;

[0011] Step 3, extracting the position information and angle information of the stick-shaped similar soil;

[0012] Identify the position and shape of rod-like soil particles in the binary image, obtain the position coordinates and major axis angle of each rod-like soil particle, and process the series of binary images to obtain the position information of rod-like soil particles at different stages of the test;

[0013] Step 4, obtaining rod-shaped similar soil displacement;

[0014] The rod-shaped similar soil particles with the same shape and a change in centroid position and long axis direction within a certain range in the binary images adjacent in time order are regarded as the same rod-shaped similar soil particle; the position information of the same rod-shaped similar soil particle in the binary images of different motion states is compared to obtain the displacement information of the rod-shaped similar soil particle; the displacement information of each rod-shaped similar soil particle is obtained and output as a displacement cloud map of the rod-shaped similar soil particle;

[0015] Step 5, obtain the long axis direction and rotation angle of the steel rod similar to the soil;

[0016] The longest inner diameter of the rod-like similar soil particles in the binary image is identified as the long axis, and the long axis direction of the rod-like similar soil particles is extracted; the binary image is gridded, and the long axis directions of all the rod-like similar soil particles in each grid area are counted and the average long axis direction is calculated to obtain the average long axis direction distribution information of the rod-like similar soil particles in each grid area; the difference in the average long axis direction in the binary image at different times in the same grid area is compared to obtain the average long axis angle deflection information of the rod-like similar soil particles in each grid area, and then the angle contour distribution information corresponding to the time period is obtained to study the average rotation of particles in each grid area and the deflection movement of the soil as a whole.

[0017] Preferably, the method for efficiently identifying soil particle changes in a large-scale model test provided by the present invention may also have the following characteristics: the large-scale model test is a test with a model size exceeding 1000 mm and lasting for several hours.

[0018] Preferably, the method for efficiently identifying soil particle changes in a large-scale model test provided by the present invention may also have the following characteristics: in step 1, the photographing interval is 20 to 30 seconds.

[0019] Preferably, the method for efficiently identifying changes in soil particles in large-scale model tests provided by the present invention may further include: in step 2, finally deleting the parts of the image with pixel radii that are too small and too large, where a pixel radius that is too small means that it is less than 1 / 10 of the minimum rod structure cross-sectional radius, and a pixel radius that is too large means that it is more than 10 times the maximum rod structure cross-sectional radius.

[0020] Preferably, the efficient identification method for soil particle changes in large-scale model tests provided by the present invention may also have the following characteristics: in step 3, using MATLAB software, the centroid function of the regionprops function is used to capture the centroid position of the rod-shaped similar soil particles, and the Orientation function of the regionprops function is used to capture the elliptical major axis position information of the rod-shaped similar soil particles, and then the major axis angle is obtained as the major axis direction of the rod-shaped similar soil particles.

[0021] Preferably, the method for efficiently identifying changes in soil particles in large-scale model tests provided by the present invention may also have the following characteristics: in step 4, the change in the centroid position and the change in the major axis direction are within a certain range, which means that the ratio of the change in the centroid position and the change in the major axis direction to the major axis length is within 0.02.

[0022] Preferably, the efficient identification method for soil particle changes in large-scale model tests provided by the present invention may also have the following characteristics: in step 5, the grid is evenly divided into: there are no more than sixty rod-shaped similar soil particles with centroid coordinates within a grid area, so that the final overall distribution law will be more accurate.

[0023] Preferably, the method for efficiently identifying changes in soil particles in large-scale model tests provided by the present invention may also have the following features: Step 6, porosity information extraction and output: the binary image is gridded and segmented, and the RGB data of each image is read through the imread function. The white part is the rod-shaped similar soil particles R=G=B=255, and the black part is the pores R=G=B=0. The proportion of black in each grid is calculated, which is the porosity of the soil in the grid. The porosity value of each grid is calculated, and the gridded porosity information is output. A porosity change cloud map is output based on the gridded porosity information of different motion states.

[0024] Preferably, the method for efficiently identifying soil particle changes in large-scale model tests provided by the present invention may also have the following feature: in step 6, after cropping and segmentation, each grid block contains hundreds of rod-shaped similar soil particles.

[0025] <system>

[0026] Furthermore, the present invention also provides an efficient identification system for soil particle changes in large-scale model tests, which is characterized by comprising:

[0027] The image acquisition unit simulates soil particles using rod structures of corresponding size, gradation, and number and elliptical cross-sectional shape according to the soil information to be simulated, and densely stacks the rod structures to form rod-shaped similar soil to simulate the soil for use in the test; the elliptical side of the rod structures is used as the observation side of the model, and a high-definition camera is aimed at the observation side of the model. A series of high-resolution images of the movement of soil particles simulated by the rod structures during the test are collected at a certain interval, thereby obtaining a series of initial images containing the dynamic process of the rod-shaped similar soil particles; the interval between photos is at least 10 seconds;

[0028] The image processing unit cuts the image, selects the effective area related to the rod-like similar soil particles in the image, and grayscales it to obtain a grayscale image; sets the color division value to select an appropriate color retention range for the grayscale image, divides the grayscale image into two different grayscale values ​​of the rod-like similar soil particles and the pore part, and obtains a preliminary distinction image; cuts the preliminary distinction image to separate adjacent rod-like similar soil particles, while not cutting a single rod-like similar soil particle; and performs binarization on the cut image to clearly distinguish the rod-like similar soil particles from the pores;

[0029] Extraction unit: extracts the position and angle information of rod-shaped similar soil particles; identifies the position and shape of rod-shaped similar soil particles in the binary image, obtains the position coordinates and major axis angle of each rod-shaped similar soil particle, and processes the series of binary images to obtain the position information of rod-shaped similar soil particles at different stages of the test;

[0030] The displacement output unit obtains the displacement of rod-shaped similar soil particles: rod-shaped similar soil particles with the same shape in the binary images adjacent in time order and with changes in centroid position and major axis direction within a certain range are regarded as the same rod-shaped similar soil particle; the position information of the same rod-shaped similar soil particle in the binary images of different motion states is compared to obtain the displacement information of the rod-shaped similar soil particle; the displacement information of each rod-shaped similar soil particle is obtained and output as a displacement cloud map of the rod-shaped similar soil particle;

[0031] The rotation information acquisition unit obtains the major axis direction and rotation angle of the steel rod-like soil: the longest inner diameter of the rod-like soil particles in the binary image is identified as the major axis, and the major axis direction of the rod-like soil particles is extracted; the binary image is gridded, and the major axis directions of all the rod-like soil particles in each grid area are counted and the average major axis direction is calculated for each grid area, thereby obtaining the average major axis direction distribution information of the rod-like soil particles in each grid area; the difference in the average major axis direction in the binary image at different times in the same grid area is compared to obtain the average major axis angle deflection information of the rod-like soil particles in each grid area, and then obtain the rotation angle contour line distribution information corresponding to the time period, so as to study the average rotation of the particles in each grid area and the deflection movement of the soil as a whole;

[0032] The control unit is connected to the image acquisition unit, the image processing unit, the extraction unit, the displacement output unit, and the rotation information acquisition unit, and controls their operations.

[0033] Functions and effects of the invention

[0034] The present invention provides an efficient method and system for identifying changes in soil particles in large-scale model tests. First, based on the soil information to be simulated, rod structures with corresponding sizes, gradations, and quantities and elliptical cross-sectional shapes are used to simulate soil particles densely stacked to form rod-like similar soil to simulate the soil; a series of high-resolution images containing particle movement conditions in the test are collected at a certain interval; the images are then processed to obtain a binary image; then, the positions and shapes of the rod-like similar soil particles in the binary image are identified to obtain the position coordinates and major axis angles of the rod-like similar soil particles; and then, rod-like similar soil particles with the same shape in adjacent binary images and changes in centroid position and major axis direction within a certain range are regarded as the same rod-like similar soil particle, and the positions of the same rod-like similar soil particle in binary images of different motion states are compared. The information is obtained to obtain the displacement information of the rod-like similar soil particles, and then a rod-like similar soil particle displacement cloud map showing the displacement information of all rod-like similar soil particles is obtained; then, the longest inner diameter of the rod-like similar soil particles in the binary map is identified as the major axis, and the major axis direction is extracted; for each grid area, the major axis directions of all rod-like similar soil particles in the area are counted and the average major axis direction is calculated, thereby obtaining the average major axis direction distribution information of the rod-like similar soil particles in each grid area; the difference in the average major axis direction in the binary map at different times in the same grid area is compared to obtain the average major axis angle deflection information of the rod-like similar soil particles in each grid area, and then the angle contour distribution information corresponding to the time period is obtained to study the average rotation of particles in each grid area and the deflection motion of the soil as a whole. Through the above processing, the dynamic changes of each mesoscopic particle during the large-scale model test can be conveniently, accurately and efficiently identified, and the mesoscopic constitutive information of each particle is accurate. The calculation process is simple and does not involve complex calculation and learning processes. The effect is more obvious, especially for sample processes with a large analysis range, a large number of particles and a long time. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of a method for efficiently identifying soil particle changes in a large-scale model test according to the first embodiment of the present invention;

[0036] Figure 2 Schematic diagram of the medium steel bar similar soil movable door test involved in Example 1 of the present invention (analysis size: 1800mm×700mm);

[0037] Figure 3 Schematic diagram of the image binarization process involved in Example 1 of the present invention, where (a) is the grayscale image and (b) is the binarized image;

[0038] Figure 4: This is a graph of the steel bar similar soil position and displacement information extraction involved in Example 1 of the present invention, wherein (a) is the extracted steel bar similar soil position information, and (b) is the output steel bar similar soil displacement cloud map;

[0039] Figure 5 : This is a diagram of the extraction of the long axis direction and rotation angle information of the steel bar similar soil involved in Example 1 of the present invention, wherein (a) is the extracted long axis direction information of the steel bar similar soil, (b) is the output long axis direction distribution diagram of the steel bar similar soil, and (c) is the output contour distribution diagram of the rotation angle of the steel bar similar soil;

[0040] Figure 6 This is a porosity information extraction diagram involved in Example 1 of the present invention, wherein (a) is the image grid segmentation diagram, (b) is the extracted image grid porosity information diagram, and (c) is the output porosity change cloud diagram. DETAILED DESCRIPTION

[0041] The specific implementation scheme of the method and system for efficiently identifying soil particle changes in a large-scale model test according to the present invention will be described in detail below with reference to the accompanying drawings.

[0042] <Example 1>

[0043] In this embodiment, the sliding door model test using steel rods to simulate soil is taken as an example. It should be understood that the test method is also applicable to other experimental models.

[0044] Step 1: Collect high-definition pictures:

[0045] According to the soil information to be simulated, rod structures of corresponding size, gradation, and number with elliptical cross-sections are used to simulate soil particles. These rods are densely stacked to form rod-shaped similar soil to simulate the soil for use in the test. The elliptical side of the rod structure (where the elliptical cross-section can be seen) is used as the observation side of the model. A high-definition camera is aimed at the observation side of the model and photographs are taken at a certain interval (25 seconds) in front of the test box. During the sliding door test, photos of the soil at different stages of the sliding door sinking are collected, such as Figure 2 shown.

[0046] In this example, steel rods were stacked densely to form a steel rod-like soil to simulate soil. The minimum size (minor axis * major axis) of a single rod structure was 3*6 mm, and the maximum size was 5*10 mm. The test lasted for more than 6 hours.

[0047] Step 2: Image binarization:

[0048] Use image processing software to cut the image, select the effective parts, and grayscale them to get the following Figure 3(a) is the grayscale image shown. After all the grayscale photos are imported using image processing software, the appropriate color retention range is selected for the grayscale image by setting the color division value. The value range is usually between 100 and 200. At this time, the image is divided into two colors: the steel rod similar soil part and the pore part (the part within the color division range is the steel rod similar soil part, and the part outside the range is the pore part), ensuring that each color block is independent of each other; using image processing software, according to the steel rod similar soil and pore part divided by the color division value, select the appropriate cutting intensity and cut. After cutting, the adjacent steel rod similar soils must be cut apart and the single steel rod similar soil is not over-cut; finally, the cut image is binarized and the parts with too small and too large pixel radius are removed, and the following is obtained: Figure 3 (b) shows the binary image.

[0049] The above cutting process is implemented using the Separate Objects function of the avizo software. Overcutting means that one particle is cut into two.

[0050] Step 3: Extract the steel bar similar soil position information and long axis direction information:

[0051] Perform image analysis on the binary image obtained in step 2. Each steel bar similar soil is represented as a white independent area block in the binary image. Using Matlab software, the Centroid function of the regionprops function is used to capture the centroid position of the white independent area block, and the Orientation function of the regionprops function is used to capture the ellipse major axis position information of the white independent area block, that is, the major axis direction of the steel bar similar soil. Figure 4 (a) is a screenshot of the extracted steel bar similar soil position information. Figure 5 (a) is a screenshot of the extracted long axis direction information of the steel bar similar soil.

[0052] Step 4: Output of steel bar similar soil displacement:

[0053] Through step 3, the position information of the steel rod-like soil particles at different stages of the test is obtained. Using MATLAB software, a logical operation program is written to locate the same steel rod-like soil particles in different photos (using adjacent photos, according to the same shape, the centroid position changes within a certain range, and the long axis direction changes within a certain range to locate the same steel rod-like soil particles). The position information of the steel rod-like soil particles is compared and processed to obtain the displacement information of the steel rod-like soil particles, and the imshow function is used to output it as follows Figure 4 (b) shows the displacement contour of the steel bar in similar soil. The ratio of the position change and the change in the long axis direction (x-direction offset, y-direction offset) to the long axis length is within 0.02.

[0054] Step 5: Output of long axis direction and rotation angle:

[0055] The long axis direction information obtained in step 3 is used to identify the longest inner diameter of the steel rod-like soil particles as the long axis using MATLAB software, and the long axis direction of the steel rod-like soil particles is extracted. The image is gridded. In this embodiment, the size of each grid is 40mm*40mm, and there are about 50 particles in it. Then, the average long axis direction distribution map of all steel rod-like soils in each grid is calculated and output. At the same time, the long axis direction information in different motion states in the same grid is compared, and the long axis angle deflection of the steel rod-like soil in each grid is calculated and output as follows: Figure 5 The major axis rotation angle distribution diagram of the steel rod similar soil shown in (b) and the rotation angle contour distribution diagram shown in 5(c) are used to study the average rotation of particles in each grid and the deflection motion law of the soil as a whole.

[0056] Step 6: Extract and output porosity information:

[0057] Use the slicing tool of the image processing software to select a suitable grid size. In this embodiment, the grid is about 80*80mm, and there are about 200 particles in it. The binary image obtained in step 2 is grid-segmented to obtain the following: Figure 6 The grid segmentation diagram of the picture shown in a is removed after the irrelevant areas are removed (that is, the irrelevant areas are colored non-black and white). The RGB data of each picture is read by the imread function using the Matlab software. The white part is the steel bar-like soil (R=G=B=255), and the black part is the pore (R=G=B=0). The black proportion in each grid is calculated, which is the porosity of the soil in the grid. After outputting the porosity value, the numerical analysis software can be used to draw the following Figure 6 The image shown in b is gridded into a porosity information diagram, and its porosity is output using numerical analysis software as follows: Figure 6 c is the porosity change cloud diagram.

[0058] In the above process, the calculation method after gridding in step 6 is different from that in step 5. After gridding in step 5, for each particle within a grid, the mean value along the corresponding long axis of all particles whose centroid coordinates are within the grid is calculated (step 5 only uses the grid range for particle screening). Step 6 completely crops the image into small pieces according to the grid, and then calculates the porosity by the proportion of black and white colors in each small piece (step 6 is to cut the original image with the grid).

[0059] In addition, in the above process, the images can be processed using image processing software such as Photoshop, Avizo, ImageJ, etc. The image information extraction is calculated using numerical analysis software such as MATLAB, and output is performed using MATLAB, Origin, Excel, etc.

[0060] If the existing technology for studying microscopic structures is used to identify and process the dynamic changes of particles during the above experiments, on the one hand, due to the long time, large number of images, and excessive processing volume, it cannot be processed at all and cannot be applied; on the other hand, if the time interval between adjacent photos analyzed is greater than 10s, tracking loss is likely to occur.

[0061] The large-size particle movement microscopic image analysis method of this embodiment, due to the above steps, can directly perform image analysis through the process pictures of the steel rod similar soil model test to obtain information such as particle center of mass, displacement, major axis direction, rotation angle, porosity, etc., which can greatly improve the test efficiency. Moreover, since the relevant information obtained comes from the model test, its boundary conditions and related parameter settings are closer to the actual working conditions than the numerical parameters artificially simplified in traditional discrete element numerical simulation. The test results are more accurate and reliable, and a lot of processing time can also be saved.

[0062] The above embodiments are merely illustrative of the method of the present invention. The method for analyzing the microscopic image of large-size particle motion involved in the present invention is not limited to the model tests described in the above embodiments. In the present invention, different model tests can be designed according to different engineering contexts to meet actual engineering requirements, and soil motion microscopic image analysis can be performed on the model tests.

[0063] <Example 2>

[0064] In the second embodiment, a highly efficient identification system for soil particle changes in large-scale model tests that can automatically implement the method of the present invention is provided. The system includes an image acquisition unit, an image processing unit, an extraction unit, a displacement output unit, a rotation information acquisition unit, an input display unit, and a control unit.

[0065] The image acquisition unit simulates soil particles using rod structures of corresponding size, gradation, and number with elliptical cross-sectional shapes based on the soil information to be simulated. These rod structures are densely stacked to form rod-shaped similar soil to simulate the soil for use in the test. The elliptical side of the rod structure serves as the observation side of the model. A high-definition camera is aimed at the observation side of the model. A series of high-resolution images of the movement of soil particles simulated by the rod structure during the test are collected at a certain interval, obtaining a series of initial images containing the dynamic process of the rod-shaped similar soil particles. The interval between photographs is at least 10 seconds.

[0066] The image processing unit cuts the image, selects the effective area related to the rod-like similar soil particles in the image, and grayscales it to obtain a grayscale image; sets the color division value to select the appropriate color retention range for the grayscale image, divides the grayscale image into two different grayscale values ​​of the rod-like similar soil particles and the pore part, and obtains a preliminary distinction image; cuts the preliminary distinction image to separate adjacent rod-like similar soil particles, while not cutting a single rod-like similar soil particle; and performs binarization on the cut image to clearly distinguish the rod-like similar soil particles from the pores;

[0067] Extract the position information and angle information of the stick-like soil particles; identify the position and shape of the stick-like soil particles in the binary image, obtain the position coordinates and major axis angle of each stick-like soil particle, and process the series of binary images to obtain the position information of the stick-like soil particles at different stages of the test;

[0068] The displacement output unit obtains the displacement of the rod-shaped similar soil particles: rod-shaped similar soil particles with the same shape in the binary images adjacent in time sequence and with changes in centroid position and major axis direction within a certain range are regarded as the same rod-shaped similar soil particle; the position information of the same rod-shaped similar soil particle in the binary images of different motion states is compared to obtain the displacement information of the rod-shaped similar soil particle; the displacement information of each rod-shaped similar soil particle is obtained and output as a displacement cloud map of the rod-shaped similar soil particle;

[0069] The rotation information acquisition unit obtains the major axis direction and rotation angle of the steel rod-like soil: identifies the longest inner diameter of the rod-like similar soil particles in the binary image as the major axis, and extracts it as the major axis direction of the rod-like similar soil particles; grids the binary image, and for each grid area, counts the major axis directions of all the rod-like similar soil particles in the area and calculates the average major axis direction, thereby obtaining the average major axis direction distribution information of the rod-like similar soil particles in each grid area; compares the difference in the average major axis direction in the binary image at different times in the same grid area, obtains the average major axis angle deflection information of the rod-like similar soil particles in each grid area, and then obtains the rotation angle contour line distribution information corresponding to the time period, so as to study the average rotation of particles in each grid area and the deflection movement of the soil as a whole;

[0070] The input display unit is used to allow the user to input operation instructions, and display the input, output, and intermediate processing information of each unit according to the operation instructions. Figures 2 to 6 The content shown.

[0071] The control unit is connected to the image acquisition unit, the image processing unit, the extraction unit, the displacement output unit, the rotation information acquisition unit, and the input and display unit for communication, and controls the operation of these units.

[0072] The above embodiments are merely illustrative of the technical solutions of the present invention. The method and system for efficiently identifying soil particle changes in large-scale model tests involved in the present invention are not limited solely to those described in the above embodiments, but are subject to the scope defined by the claims. Any modifications, supplements, or equivalent substitutions made by those skilled in the art based on these embodiments are within the scope of protection claimed by the claims of the present invention.

Claims

1. An efficient method for identifying soil particle changes in large-scale model tests, characterized by: The following steps are involved: Step 1: Based on the soil information to be simulated, rod structures of corresponding size, gradation, and number with elliptical cross-sectional shapes are used to simulate soil particles, and these rods are densely stacked to form rod-shaped similar soil to simulate the soil for use in the test; the elliptical side of the rod structure is used as the observation side of the model, and a high-definition camera is aimed at the observation side of the model. A series of high-resolution images of the movement of soil particles simulated by the rod structure during the test are collected at a certain interval, thereby obtaining a series of initial images containing the dynamic process of the rod-shaped similar soil particles; the interval between photographs is at least 10 seconds; Step 2: Cut the image, select the valid area related to the rod-shaped similar soil particles in the image, and grayscale it to obtain a grayscale image; set the color division value to select an appropriate color retention range for the grayscale image, divide the grayscale image into two different grayscale values ​​of the rod-shaped similar soil particles part and the pore part to obtain a preliminary distinction image; cut the preliminary distinction image to separate adjacent rod-shaped similar soil particles, and do not cut a single rod-shaped similar soil particle; The cut images were binarized to clearly distinguish between stick-like soil particles and pores. Step 3, extracting the position information and angle information of the stick-shaped similar soil; Identify the position and shape of rod-like soil particles in the binary image, obtain the position coordinates and major axis angle of each rod-like soil particle, and process the series of binary images to obtain the position information of rod-like soil particles at different stages of the test; Step 4, obtaining rod-shaped similar soil displacement; The rod-shaped similar soil particles with the same shape and a change in centroid position and long axis direction within a certain range in the binary images adjacent in time order are regarded as the same rod-shaped similar soil particle; the position information of the same rod-shaped similar soil particle in the binary images of different motion states is compared to obtain the displacement information of the rod-shaped similar soil particle; the displacement information of each rod-shaped similar soil particle is obtained and output as a displacement cloud map of the rod-shaped similar soil particle; Step 5, obtain the long axis direction and rotation angle of the steel rod similar to the soil; The longest inner diameter of the rod-like similar soil particles in the binary image is identified as the long axis, and the long axis direction of the rod-like similar soil particles is extracted; the binary image is gridded, and the long axis directions of all the rod-like similar soil particles in each grid area are counted and the average long axis direction is calculated to obtain the average long axis direction distribution information of the rod-like similar soil particles in each grid area; the difference in the average long axis direction in the binary image at different times in the same grid area is compared to obtain the average long axis angle deflection information of the rod-like similar soil particles in each grid area, and then the angle contour distribution information corresponding to the time period is obtained to study the average rotation of particles in each grid area and the deflection movement of the soil as a whole.

2. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized by: in, Large model tests are tests with a model size exceeding 1000 mm and a duration of several hours.

3. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized by: in, In step 1, the photo taking interval is 20 to 30 seconds.

4. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized by: in, In step 2, the parts of the image with pixel radius that are too small or too large are finally deleted. A pixel radius that is too small means that it is less than 1 / 10 of the minimum rod structure cross-sectional radius, and a pixel radius that is too large means that it is more than 10 times the maximum rod structure cross-sectional radius.

5. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized by: in, In step 3, Matlab software is used to capture the centroid position of the rod-shaped similar soil particles through the Centroid function of the regionprops function, and the Orientation function of the regionprops function is used to capture the elliptical major axis position information of the rod-shaped similar soil particles, and then the major axis angle is obtained as the major axis direction of the rod-shaped similar soil particles.

6. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized by: in, In step 4, the change in the centroid position and the change in the major axis direction being within a certain range means that the ratio of the change in the centroid position and the change in the major axis direction to the major axis length is within 0.

02.

7. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized by: in, In step 5, the grid is evenly divided so that there are no more than sixty rod-like soil particles whose centroid coordinates are within a grid area.

8. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 1 is characterized in that: Also includes: Step 6: Extract and output porosity information: The binary image is gridded and segmented, and the RGB data of each image is read through the imread function. The white part is the rod-shaped similar soil particles R=G=B=255, and the black part is the pore R=G=B=0. The black proportion in each grid is calculated, which is the porosity of the soil in the grid. The porosity value of each grid is calculated, and the gridded porosity information is output. Based on the gridded porosity information of different motion states, a porosity change cloud map is output.

9. The method for efficiently identifying soil particle changes in large-scale model tests according to claim 8 is characterized by: in, In step 6, after cropping and segmentation, each grid block contains hundreds of rod-like soil particles.

10. An efficient identification system for soil particle changes in large-scale model tests, characterized by: include: The image acquisition unit simulates soil particles using rod structures of corresponding size, gradation, and number and elliptical cross-sectional shape according to the soil information to be simulated, and densely stacks the rod structures to form rod-shaped similar soil to simulate the soil for use in the test; the elliptical side of the rod structures is used as the observation side of the model, and a high-definition camera is aimed at the observation side of the model. A series of high-resolution images of the movement of soil particles simulated by the rod structures during the test are collected at a certain interval, thereby obtaining a series of initial images containing the dynamic process of the rod-shaped similar soil particles; the interval between photos is at least 10 seconds; The image processing unit cuts the image, selects the effective area related to the rod-like soil particles in the image, and grayscales it to obtain a grayscale image; sets the color division value to select an appropriate color retention range for the grayscale image, divides the grayscale image into two different grayscale values ​​of the rod-like soil particles part and the pore part, and obtains a preliminary distinction image; cuts the preliminary distinction image to separate adjacent rod-like soil particles, while not cutting a single rod-like soil particle; The cut images were binarized to clearly distinguish between stick-like soil particles and pores. Extraction unit: extracts the position and angle information of rod-shaped similar soil particles; identifies the position and shape of rod-shaped similar soil particles in the binary image, obtains the position coordinates and major axis angle of each rod-shaped similar soil particle, and processes the series of binary images to obtain the position information of rod-shaped similar soil particles at different stages of the test; The displacement output unit obtains the displacement of rod-shaped similar soil particles: rod-shaped similar soil particles with the same shape in the binary images adjacent in time order and with changes in centroid position and major axis direction within a certain range are regarded as the same rod-shaped similar soil particle; the position information of the same rod-shaped similar soil particle in the binary images of different motion states is compared to obtain the displacement information of the rod-shaped similar soil particle; the displacement information of each rod-shaped similar soil particle is obtained and output as a displacement cloud map of the rod-shaped similar soil particle; The rotation information acquisition unit obtains the major axis direction and rotation angle of the steel rod-like soil: the longest inner diameter of the rod-like soil particles in the binary image is identified as the major axis, and the major axis direction of the rod-like soil particles is extracted; the binary image is gridded, and the major axis directions of all the rod-like soil particles in each grid area are counted and the average major axis direction is calculated for each grid area, thereby obtaining the average major axis direction distribution information of the rod-like soil particles in each grid area; the difference in the average major axis direction in the binary image at different times in the same grid area is compared to obtain the average major axis angle deflection information of the rod-like soil particles in each grid area, and then obtain the rotation angle contour line distribution information corresponding to the time period, so as to study the average rotation of the particles in each grid area and the deflection movement of the soil as a whole; The control unit is connected to the image acquisition unit, the image processing unit, the extraction unit, the displacement output unit, and the rotation information acquisition unit, and controls their operations.

Images