Three-dimensional dynamic monitoring system and method for water absorption deformation of material based on half-sphere field of view

Abstract

The application discloses a kind of material water absorption deformation three-dimensional dynamic monitoring system and method based on hemispherical field of view;The dynamic monitoring system and method can capture the external change in the water absorption process of argillaceous rock-soil body and other materials in real time, fully reveal the overall appearance of surface deformation in the water absorption process of argillaceous rock-soil body and other materials from the perspective of full-dimensional space.The dynamic monitoring system includes water absorption swelling deformation device, camera monitoring system and data acquisition device.The material water absorption deformation three-dimensional dynamic monitoring method based on hemispherical field of view includes the following steps: S1, installation and setting of monitoring system, S2, camera debugging;S3, distilled water is injected into water absorber;S4, keep the water level in the water storage cavity of water absorber unchanged;image data of sample is collected by camera at regular intervals;S5, judge that sample no longer absorbs water, complete monitoring.Using the material water absorption deformation three-dimensional dynamic monitoring system and method based on hemispherical field of view can obtain high-precision image information and improve monitoring accuracy.

Description

Technical Field

[0001] The present invention relates to the field of material water absorption deformation monitoring, and in particular to a three-dimensional dynamic monitoring system and method for material water absorption deformation based on a hemispherical field of view. Background Art

[0002] Mudstone, for example, is rich in hydrophilic minerals and is prone to expansion or contraction deformation when its water content changes, directly affecting the stability of the project.

[0003] When studying mudstone swelling due to water absorption, many researchers often rely on traditional measurement methods to collect test data using specific test equipment, such as displacement sensors, laser scanners, free dilatometers, and surface roughness meters. These instruments then plot time series curves to analyze the water absorption patterns of the mudstone after water absorption, analyzing its static characteristics, such as deformation, free expansion rate, volume change, and volume expansion coefficient, from a two-dimensional perspective. However, due to the difficulty of providing high-resolution or high-precision surface topography data and the lack of dynamic monitoring data, these traditional measurement methods are unable to capture the evolution of mudstone deformation due to water absorption. Consequently, there is a lack of in-depth research and understanding of the three-dimensional spatiotemporal development patterns of mudstone deformation due to water absorption, as well as the uneven deformation patterns.

[0004] Existing monitoring technologies have the following problems: ① Static limitations: Traditional methods rely on periodic static measurements and cannot capture millisecond-level dynamic deformation processes; ② Lack of dimensionality: Two-dimensional analysis is difficult to reflect the three-dimensional spatial deformation characteristics of materials; ③ Insufficient data accuracy: Existing equipment has low resolution and it is difficult to quantify subtle deformations.

[0005] The above defects have led to blind spots in the study of the water absorption and deformation mechanism of mudstone, restricting engineering safety assessment and disaster prevention. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a three-dimensional dynamic monitoring system and method for material water absorption deformation based on a hemispherical field of view; the dynamic monitoring system and method can capture the external changes of materials such as mudstone during the water absorption process in real time, especially comprehensively revealing the overall picture of the surface deformation of materials such as mudstone during the water absorption process from a full-dimensional spatial perspective.

[0007] The technical solution adopted by the present invention to solve the technical problem is: a three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view, including a water absorption expansion and deformation device, a camera monitoring system and a data acquisition device;

[0008] The camera monitoring system includes an annular base; a hemispherical hollow camera bracket is provided on the annular base;

[0009] The camera bracket is provided with a camera for shooting the inner side of the camera bracket; the hemispherical top formed by the camera bracket is provided with a camera;

[0010] The water absorption expansion and deformation device includes a water absorption container; the water absorption container is provided with nine calibration columns evenly distributed along the circumference;

[0011] The water absorption container has a water storage cavity; a circular permeable stone plate is provided in the middle of the upper surface of the water absorption container; a water hole is provided on the permeable stone plate and is connected to the water storage cavity; a water outlet pipe and an openable and closable water inlet pipe are provided on the water absorption container; the height of the water outlet of the water outlet pipe is flush with the lower surface of the permeable stone plate;

[0012] The data acquisition device includes a Raspberry Pi, a fill light strip, and a mounting frame; the Raspberry Pi is mounted on the mounting frame, and the fill light strip is mounted inside the mounting frame;

[0013] The Raspberry Pi corresponds to the camera one by one, and the Raspberry Pi is electrically connected to the camera;

[0014] The water absorption container is located at the center of the inner circle of the annular base; the annular base is located at the center of the bottom of the inner cavity of the installation frame; and a pressure sensor is provided at the bottom of the water absorption container.

[0015] Furthermore, the camera bracket includes a plurality of arc-shaped brackets with an arc angle of 90°; a connecting block is provided at the lower end of the arc-shaped bracket, and a vertical arc-shaped connecting block is provided at the upper end;

[0016] The annular base is provided with mounting slots evenly distributed along the circumference; the connecting block of the arc-shaped bracket is inserted into the mounting slot; the arc-shaped connecting block at the upper end of the arc-shaped bracket is spliced ​​to form a cylinder with a central mounting through hole; the cylinder is clamped by a clamp; and the clamp is locked by a locking bolt.

[0017] Furthermore, a mounting groove is provided on the arc-shaped bracket; the camera is installed in the mounting groove, and a wire collection hole is provided below the mounting groove.

[0018] Furthermore, the upper surface of the water absorption container is provided with a mounting groove matching the permeable stone slab; the bottom of the mounting groove is provided with a through hole communicating with the water storage cavity of the water absorption container.

[0019] Specifically, the camera uses a USB camera with a distortion-free lens, a maximum resolution of 2592*1944, and autofocus.

[0020] Specifically, the Raspberry Pi is electrically connected to the camera through a data cable; the data cable includes two neutral wires, a live wire, a data input wire, and a data output wire; a 5-pin piercing terminal is provided at one end of the data cable, and a Type-C female socket is provided at the other end; the piercing terminal is connected to the USB camera; a data cable connection cap is provided on the Raspberry Pi; the Type-C female socket is electrically connected to the data cable connection cap of the Raspberry Pi through a Type-C-USB data cable.

[0021] Specifically, there are 16 arc-shaped brackets; each arc-shaped bracket is provided with three cameras, and the three cameras are evenly distributed along the curvature of the arc-shaped bracket.

[0022] Specifically, the installation frame adopts an aluminum profile bracket.

[0023] Furthermore, a drainage height adjustment device is provided on the water outlet pipe; the drainage height adjustment device includes an adjustment pipe body; a connecting pipe matching the water outlet pipe is provided at one end of the adjustment pipe body; a vertical pipe is provided at the other end of the adjustment pipe body; a telescopic pipe is provided on the vertical pipe; the telescopic pipe is threaded with the vertical pipe.

[0024] The present invention also provides a monitoring method of a three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view, which uses the three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view described in the present application;

[0025] The following steps are also included:

[0026] S1. Installation and setup of monitoring system;

[0027] S11. Place the test piece in an oven at 100-105°C for 24 hours and then place it in a cooling dish for cooling;

[0028] S12. Place the sample on a circular permeable stone slab;

[0029] S13, placing the water absorption expansion deformation device in the center of the inner cavity of the camera monitoring system; at the same time, placing the data acquisition device outside the camera monitoring system so that the camera monitoring system is located in the center of the inner cavity of the data acquisition device;

[0030] S14, combining and connecting the data acquisition device and the camera monitoring system, that is, electrically connecting the camera in the camera monitoring system to the Raspberry Pi of the data acquisition device through a data cable and a Type-C-USB data cable, with a one-to-one correspondence;

[0031] S15, turning on the fill light strip, and setting the camera scheduled start time and image acquisition time interval through the Raspberry Pi, and setting the Raspberry Pi to automatically create a data folder and store image data;

[0032] S2. After the monitoring system is installed and set up, the camera is debugged through the real-time image data acquisition program;

[0033] S3. After the monitoring system is debugged, connect the water inlet pipe to the peristaltic pump through a hose, open the water inlet pipe on the water absorption container, and inject distilled water into the water storage chamber of the water absorption container through the water inlet pipe. The water level should be 1-2mm below the bottom of the permeable stone slab. At the same time, the weight of the water absorption expansion deformation device should be recorded in real time through the pressure sensor.

[0034] S4. Set the time interval for image data acquisition by the camera; after 5 minutes, start the peristaltic pump and set it to continuously inject distilled water into the water absorption container at a constant rate. During the distilled water injection process, the water outlet pipe is always kept open, and the water level in the water storage chamber of the water absorption container is kept constant during the monitoring process; and perform timed image data acquisition of the sample through the camera;

[0035] S5. Record the weight of the water absorption and deformation device through the pressure sensor. When the weight of the water absorption and deformation device no longer increases and the image information collected by the camera determines that the shape or volume increment of the test piece has not changed, stop monitoring and turn off the micro pump to complete the dynamic monitoring of the water absorption and deformation material.

[0036] Furthermore, the image data collected in step S4 is processed to extract detailed information about the entire process of deformation evolution of the mudstone mass due to water absorption, including the following steps:

[0037] S41. Using the scale-invariant feature transform (SIFT) algorithm, perform two-dimensional image feature point detection and description on the two-dimensional image data collected at the same moment by the three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view, thereby completing image feature extraction.

[0038] S42, using the K-nearest neighbor algorithm (KNN) to compare the similarity of each feature descriptor, find matching pairs of feature points in the two-dimensional images at the same time and different viewing angles, and obtain an initial matching result;

[0039] The random sampling consensus algorithm RANSAC is used to eliminate incorrect matching pairs in the initial matching results, and the optimal model parameters are obtained through parameter optimization as the final matching result;

[0040] S43. Using incremental structure-from-motion technology, a pair of optimally registered images are selected based on the feature point matching results for preliminary 3D reconstruction. Appropriate 2D images are then selected from the remaining 2D images and gradually added to the existing model to continuously restore object details to complete sparse point cloud reconstruction. Appropriate 2D images are those with high matching degree, high image quality and stability, and good adaptability to dynamic scenes; that is, in terms of time sequence: adjacent moments, and in terms of deformation: similar images.

[0041] S44, using multi-view stereo vision technology MVS, by integrating multi-view depth information, the geometric details of the complex object are recovered, the sparse point cloud is converted into dense point cloud, and dense point cloud reconstruction is completed;

[0042] S45, using statistical discrete group point removal algorithm SOR, analyzing the spatial distance distribution characteristics between each data point and its adjacent points in the dense point cloud model, identifying and removing abnormal points by using statistical principles, only retaining the mudstone sample three-dimensional point cloud model and black and white calibration points, realizing point cloud denoising, and obtaining a high-precision three-dimensional point cloud model at a certain moment;

[0043] S46, using a registration algorithm based on feature matching, the black and white calibration points are used for point cloud coarse registration of the mudstone sample three-dimensional point cloud model;

[0044] S47, using iterative closest point algorithm ICP, first, the black and white calibration points are separately subjected to point cloud fine registration, and then the corresponding conversion matrix is obtained, and then the matrix is applied to the fine registration of the three-dimensional point cloud model of the mudstone sample; according to the three-dimensional point cloud models at different moments, a time sequence three-dimensional point cloud model is constructed;

[0045] S48, based on the time sequence three-dimensional point cloud model, the capillary water absorption apparent fracture network and density development of the red bed mudstone are analyzed, the apparent fracture network profile of the mudstone sample is extracted by scalar domain calculation, and the apparent fracture density is calculated by geometric feature calculation;

[0046] S49, based on the time sequence three-dimensional point cloud model, the capillary water absorption deformation displacement development of the red bed mudstone is analyzed, the three-dimensional point cloud model deformation displacement of the mudstone sample is calculated using a multi-scale point cloud model comparison algorithm M3C2, and the axial deformation displacement at different moments is extracted according to the time sequence three-dimensional point cloud model;

[0047] S410, based on the time sequence three-dimensional point cloud model, the capillary water absorption non-uniform expansion characteristic development of the red bed mudstone is analyzed, the top surface of the three-dimensional point cloud model of the mudstone sample at different moments is subjected to plane fitting, and the fitting planes are compared to calculate the rotation angle and deflection displacement;

[0048] S411, based on the time sequence three-dimensional point cloud model, the capillary water absorption volume expansion and density development of the red bed mudstone are analyzed, the three-dimensional point cloud model of the mudstone sample is converted into a grid model with a closed boundary using Delaunay triangulation, the volume and volume density of the grid model are calculated, and the swelling volume at different moments is extracted according to the time sequence three-dimensional point cloud model;

[0049] S412, based on the time sequence three-dimensional point cloud model, the capillary water absorption surface roughness development of the red bed mudstone is analyzed, the surface roughness density of the time sequence three-dimensional point cloud model is calculated by geometric feature calculation, and the surface roughness density histogram of the time sequence three-dimensional point cloud model is extracted to analyze the surface roughness density distribution.

[0050] The beneficial effects of the present invention are as follows: the three-dimensional dynamic monitoring system and method for material water absorption deformation based on a hemispherical field of view described in the present invention can dynamically capture the external changes of materials such as mudstone and soil during the water absorption process in real time, and especially comprehensively reveal the overall picture of the surface deformation of materials such as mudstone and soil during the water absorption process from a full-dimensional spatial perspective. Through precise full-dimensional monitoring technology, this method can perform high-resolution measurements of the surface of materials such as mudstone and soil at different stages of the water absorption process, obtain subtle deformation information, and further reveal the spatial deformation characteristics of materials such as mudstone and soil under different moisture conditions. Through this dynamic real-time monitoring technology, it is possible to have a deeper understanding of the dynamic evolution process and three-dimensional spatial characteristics of water absorption deformation of materials such as mudstone and soil, provide a theoretical basis for the water absorption expansion and instability mechanism of materials such as mudstone and soil, and provide valuable practical support for the fields of geotechnical engineering, environmental protection, disaster prevention, etc.

[0051] Secondly, the three-dimensional dynamic monitoring system and method for material water absorption deformation based on a hemispherical field of view described in this application can extract detailed information about the entire evolution process of mudstone deformation due to water absorption; monitor the entire evolution process of the surface morphology of mudstone, clarify the spatial development pattern of the water absorption deformation of mudstone, and comprehensively and in real time monitor the surface morphology changes of mudstone during water absorption, thereby understanding the evolution pattern of its surface characteristics and revealing the evolution pattern of the surface morphology of mudstone during water absorption. It can also establish a three-dimensional map of the surface morphology changes of mudstone, analyze the external rock deformation pattern of mudstone, clarify the spatial development pattern of the water absorption deformation of mudstone, and help understand the water absorption deformation mechanism of mudstone. This provides theoretical guidance and data support for the engineering application of mudstone and other materials.

[0052] Finally, the three-dimensional dynamic monitoring system and method of material water absorption and deformation based on a hemispherical field of view described in this application can be applied to water absorption and deformation monitoring tests of natural polymer materials such as cellulose and starch, synthetic polymer materials such as polyacrylates and polyacrylamides, and inorganic materials and composite materials such as rocks, bentonite, silica gel, and soda lime, providing important support for material performance evaluation, research and development, and quality control, and promoting the progress of materials science in many fields such as construction, agriculture, medical care, and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 3. A top view of a three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to an embodiment of the present invention;

[0054] Figure 2 3D dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to an embodiment of the present invention;

[0055] Figure 33D dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to an embodiment of the present invention;

[0056] Figure 4 is a perspective diagram of a camera monitoring system according to an embodiment of the present invention;

[0057] Figure 5 is a three-dimensional diagram of an arc-shaped bracket according to an embodiment of the present invention;

[0058] Figure 6 is a three-dimensional diagram of a water absorption expansion and deformation device according to an embodiment of the present invention;

[0059] Figure 7 2 is a schematic structural diagram of a liquid discharge height adjustment device according to an embodiment of the present invention;

[0060] Figure 8 This is a structural diagram of the connection between the camera and the data cable in an embodiment of the present invention;

[0061] Figure 9 Flowchart of a three-dimensional dynamic monitoring method for material water absorption deformation based on a hemispherical field of view in an embodiment of the present invention;

[0062] Figure 10 This is a flow chart of automatic data collection by a camera in an embodiment of the present invention;

[0063] Figure 11 is a flow chart of a monitoring image data processing method according to an embodiment of the present invention;

[0064] Figure 12 is a two-dimensional sub-image captured by camera 1 in the application example of the present invention;

[0065] Figure 13 is a flow chart of the SIFT algorithm in an application example of the present invention;

[0066] Figure 14 This is a flow chart of the incremental motion recovery structure in an application example of the present invention;

[0067] Figure 15 This is a flow chart of the MVS algorithm based on depth map fusion in the application example of the present invention;

[0068] Figure 16 is a schematic diagram of realizing alignment of a two-dimensional image set in an application example of the present invention;

[0069] Figure 17 is a schematic diagram of sparse point cloud reconstruction in an application example of the present invention;

[0070] Figure 18 is a schematic diagram of dense point cloud reconstruction in an application example of the present invention;

[0071] Figure 19is a schematic diagram of the water absorption change process of a mudstone sample in the application example of the present application;

[0072] in the figure, Figure 19 (a) is a point cloud model of a mudstone sample before water absorption in the application example of the present application; Figure 19 (b) is a point cloud model of a mudstone sample during water absorption in the application example of the present application; Figure 19 (c) is a point cloud model of a mudstone sample after water absorption in the application example of the present application;

[0073] Figure 20 is a denoising result schematic diagram of a mudstone water absorption change sample point cloud model in the application example of the present application;

[0074] in the figure, Figure 20 (a) is a denoising result schematic diagram of a point cloud model of a mudstone sample before water absorption in the application example of the present application; Figure 20 (b) is a denoising result schematic diagram of a point cloud model of a mudstone sample after water absorption in the application example of the present application; Figure 20 (c) is a coarse registration result schematic diagram of a point cloud of a mudstone sample before and after water absorption in the application example of the present application;

[0075] Figure 21 is an ICP algorithm flowchart in the application example of the present application;

[0076] Figure 22 is a fine registration result schematic diagram of a point cloud in the application example of the present application;

[0077] Figure 23 is a fracture network diagram in the application example of the present application;

[0078] in the figure, Figure 23 (a) is a front view of a fracture network diagram in the application example of the present application; Figure 23 (b) is a top view of a fracture network diagram in the application example of the present application; Figure 23 (c) is a rear view of a fracture network diagram in the application example of the present application; Figure 23 (d) is a left view of a fracture network diagram in the application example of the present application; Figure 23 (e) is a right view of a fracture network diagram in the application example of the present application;

[0079] Figure 24 is a fracture density cloud diagram in the application example of the present application;

[0080] Figure 24 (a) is a front view of a fracture density cloud diagram in the application example of the present application; Figure 24 (b) is a top view of a fracture density cloud diagram in the application example of the present application; Figure 24 (c) is a rear view of a fracture density cloud diagram in the application example of the present application; Figure 24 (d) is a left view of a fracture density cloud diagram in the application example of the present application; Figure 24 (e) is a right view of a fracture density cloud diagram in the application example of the present application;

[0081] Figure 25 It is a deformation displacement cloud diagram in an application example of the present invention;

[0082] In the figure, Figure 25 (a) is a front view of the deformation displacement cloud map in an application example of the present invention; Figure 25 (b) is a top view of the deformation displacement cloud map in the application example of the present invention; Figure 25 (c) is a rear view of the deformation displacement cloud map in the application example of the present invention; Figure 25 (d) is a left view of the deformation displacement cloud map in the application example of the present invention; Figure 25 (e) is a right view of the deformation displacement cloud map in the application example of the present invention;

[0083] Figure 26 is an axial deformation displacement diagram in an application example of the present invention;

[0084] Figure 27 This is a schematic diagram of the deformation and tilt fitting of the model surface before and after water absorption in an application example of the present invention;

[0085] Figure 27 (a) is a schematic diagram of the top surface of the model before water absorption in an application example of the present invention; Figure 27 (b) is a schematic diagram of the top surface fitting before water absorption in an application example of the present invention; Figure 27 (c) is a schematic diagram of the top surface of the model after water absorption in an application example of the present invention; Figure 27 (d) is a fitted main view of the top surface after water absorption in an application example of the present invention;

[0086] Figure 28 is a deformation volume density diagram in an application example of the present invention; Figure 28 (a) is a front view of a deformed volume density map in an application example of the present invention; Figure 28 (b) is a top view of the deformed volume density map in an application example of the present invention; Figure 28 (c) is a rear view of the deformed volume density map in an application example of the present invention; Figure 28 (d) is a left view of the deformation volume density map in an application example of the present invention; Figure 28 (e) is a right view of the deformation volume density diagram in an application example of the present invention;

[0087] Figure 29 is a volume expansion curve diagram in an application example of the present invention;

[0088] Figure 30 is a roughness density histogram of mudstone before water absorption in an application example of the present invention;

[0089] Figure 31 Schematic diagram of mudstone roughness after water absorption in an application example of the present invention;

[0090] In the figure, Figure 31(a) is a front view of a roughness map of mudstone after water absorption in an application example of the present invention; Figure 31 (b) is a top view of the roughness map of mudstone after water absorption in an application example of the present invention; Figure 31 (c) is a rear view of the mudstone roughness map after water absorption in an application example of the present invention; Figure 31 (d) is a left view of the mudstone roughness map after water absorption in an application example of the present invention; Figure 31 (e) is a right view of the mudstone roughness map after water absorption in an application example of the present invention;

[0091] Markings in the figure: 100-camera monitoring system, 110-clamp, 120-locking bolt, 130-arc-shaped bracket; 131-connecting block, 132-arc-shaped connecting block, 133-mounting slot, 140-camera, 150-ring base, 160-wire collection hole, 200-water absorption expansion deformation device, 210-water absorption container, 220-calibration column, 230-permeable stone slab, 240-water inlet pipe, 250-water outlet pipe, 260-drainage height adjustment device, 261-adjustment pipe body, 262-connecting pipe, 263-vertical pipe, 264-telescopic pipe, 300-data acquisition device, 310-Raspberry Pi, 320-fill light strip, 330-mounting frame, 340-data cable, 350-data cable connection cap, 360-pressure sensor. DETAILED DESCRIPTION

[0092] The present invention will be further described below with reference to the accompanying drawings and examples.

[0093] like Figures 1 to 6 As shown, the three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view of the present invention includes a water absorption expansion and deformation device 200, a camera monitoring system 100 and a data acquisition device 300;

[0094] The camera monitoring system 100 includes an annular base 150 ; a hemispherical hollow camera bracket is provided on the annular base 150 ;

[0095] The camera bracket is provided with a camera 140 for photographing the inner side of the camera bracket; the hemispherical top formed by the camera bracket is provided with the camera 140;

[0096] Specifically, the camera bracket includes a plurality of arc-shaped brackets 130 with an arc angle of 90°; a connecting block 131 is provided at the lower end of the arc-shaped bracket 130, and a vertical arc-shaped connecting block 132 is provided at the upper end;

[0097] The annular base 150 is provided with mounting slots evenly distributed along the circumference; the connecting block 131 of the arc-shaped bracket 130 is inserted into the mounting slot; the arc-shaped connecting block 132 at the upper end of the arc-shaped bracket 130 is spliced ​​to form a cylinder with a central mounting hole; the cylinder is clamped by the clamp 110; the clamp 110 is locked by the locking bolt 120. The arc-shaped bracket 130 is provided with a mounting slot 133; the camera 140 is installed in the mounting slot 133, and a wire collection hole 160 is provided below the mounting slot 133. Specifically, there are 16 arc-shaped brackets 130; each arc-shaped bracket 130 is provided with three cameras 140, and the three cameras 140 are evenly distributed along the curvature of the arc-shaped bracket 130. The camera 140 uses a USB camera with a distortion-free lens, a maximum resolution of 2592*1944, and autofocus.

[0098] The water absorption expansion and deformation device 200 includes a water absorption container 210; the water absorption container 210 is provided with nine calibration columns 220 evenly distributed along the circumference;

[0099] The water absorption container 210 has a water storage cavity; a circular permeable stone plate 230 is provided in the middle position of the upper surface of the water absorption container 210; the permeable stone plate 230 is provided with a water permeable hole connected to the water storage cavity; the water absorption container 210 is provided with a water outlet pipe 250 and an openable and closable water inlet pipe 240; the height of the water outlet of the water outlet pipe 250 is flush with the lower surface of the permeable stone plate 230.

[0100] The height of the water outlet of the water outlet pipe 250 is flush with the lower surface of the permeable stone slab, which is convenient for maintaining the liquid level in the water absorption container; if the liquid level in the inner cavity of the water absorption container 210 is higher than the height of the water outlet of the water outlet pipe 250, it will be discharged from the water outlet pipe 250, thereby always keeping the liquid level in the water absorption container stable.

[0101] In order to facilitate adjustment of the height of the water outlet of the water outlet pipe 250, a drainage height adjustment device 260 is further provided on the water outlet pipe 250;

[0102] The drainage height adjustment device 260 includes an adjustment pipe body 261 ; one end of the adjustment pipe body 261 is provided with a connecting pipe 262 that matches the outlet pipe 250 ; the connecting pipe 262 is connected to the outlet pipe 250 ; and ensures that the vertical pipe 263 is distributed vertically.

[0103] A vertical tube 263 is provided at the other end of the adjustment tube body 261 ; a telescopic tube 264 is provided on the vertical tube 263 ; the telescopic tube 264 is threadedly engaged with the vertical tube 263 .

[0104] In actual application, the height of the telescopic tube 264 extending from the vertical tube 263 can be adjusted by rotating the telescopic tube 264, thereby adjusting the water outlet height of the water outlet pipe 250. Generally, the water outlet height is set slightly higher than the lower surface of the permeable stone plate to facilitate capillary water absorption by the permeable holes on the permeable stone plate 230.

[0105] Specifically, the upper surface of the water absorption container 210 is provided with a mounting groove matched with the permeable stone slab 230 ; the bottom of the mounting groove is provided with a through hole connected with the water storage cavity of the water absorption container 210 .

[0106] The data acquisition device 300 includes a Raspberry Pi 310, a fill light strip 320, and a mounting frame 330; the Raspberry Pi 310 is mounted on the mounting frame 330, and the fill light strip is mounted inside the mounting frame 330; the Raspberry Pi 310 corresponds to the camera 140 one by one, and the Raspberry Pi 310 is electrically connected to the camera 140;

[0107] The water absorption container 210 is located at the center of the inner circle of the annular base 150 ; the annular base 150 is located at the center of the bottom of the inner cavity of the mounting frame 330 ; a pressure sensor 360 is provided at the bottom of the water absorption container 210 .

[0108] In one possible embodiment, Figure 8 As shown, to facilitate the connection between the Raspberry Pi 310 and the camera 140, the Raspberry Pi 310 and the camera 140 are electrically connected via a data cable 340. The data cable 340 includes two neutral wires, a live wire, a data input wire, and a data output wire. One end of the data cable is provided with a 5-pin piercing terminal and the other end is provided with a Type-C female connector. The piercing terminal is connected to the USB camera. The Raspberry Pi 310 is provided with a data cable connection cap 350. The Type-C female connector is electrically connected to the data cable connection cap 350 of the Raspberry Pi 310 via the Type-C-USB data cable. Specifically, the mounting frame 330 is an aluminum profile bracket.

[0109] In one possible embodiment:

[0110] The camera monitoring system 100 consists of a camera bracket, a clamp 110, and a ring base 150. The camera bracket includes multiple 90° curved brackets 130. Each curved bracket 130 has a connecting block 131 at its lower end and a vertical curved connecting block 132 at its upper end. The curved connecting blocks 132 at the upper ends of the curved brackets 130 are joined to form a cylindrical structure with a central mounting hole. This cylindrical structure is clamped by the clamp 110, which is locked with a locking bolt 120. The camera bracket is composed of sixteen curved brackets 130, forming a frame with a hemispherical field of view.

[0111] The connecting block 131 of the arc-shaped bracket 130 is inserted into the mounting slot; the arc-shaped connecting block 132 at the upper end of the arc-shaped bracket 130 is spliced ​​to form a cylinder with a central mounting hole; the cylinder is clamped by the clamp 110; the clamp 110 is locked by the locking bolt 120. The arc-shaped bracket 130 is provided with a mounting slot 133; the camera 140 is installed in the mounting slot 133, and a wire collection hole 160 is provided below the mounting slot 133. There are 16 arc-shaped brackets 130; each arc-shaped bracket 130 is provided with three cameras 140, and the three cameras 140 are evenly distributed along the curvature of the arc-shaped bracket 130. The angle between the straight lines from two adjacent cameras 140 to the center of the arc of the arc-shaped bracket is 20°. The arc-shaped connecting block 132 has an 11.25° arc angle, forming a cylindrical boss when the sixteen arc-shaped brackets 130 are assembled. Furthermore, the camera frame formed by the sixteen arc-shaped brackets 130 has a camera hole at the top for mounting the camera head 140. The annular base 150 is provided with mounting slots evenly distributed along the circumference; the connecting blocks 131 of the arc-shaped brackets 130 are inserted into the mounting slots.

[0112] The present invention also provides a three-dimensional dynamic monitoring method for material water absorption deformation based on a hemispherical field of view. Specifically, the three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view of the present invention is used; Figure 9 As shown, the following steps are also included:

[0113] S1. Installation and setup of monitoring system;

[0114] S11. Place the test piece in an oven at 100-105°C for 24 hours and then place it in a cooling dish for cooling;

[0115] S12, placing the sample on the circular permeable stone plate 230;

[0116] S13, placing the water absorption expansion deformation device 200 in the center of the inner cavity of the camera monitoring system 100; at the same time, placing the data acquisition device 300 outside the camera monitoring system 100, so that the camera monitoring system 100 is located in the center of the inner cavity of the data acquisition device 300;

[0117] S14, combining and connecting the data acquisition device 300 and the camera monitoring system 100, that is, electrically connecting the camera 140 in the camera monitoring system 100 to the Raspberry Pi 310 of the data acquisition device 300 via the data cable 340 and the Type-C-USB data cable, with a one-to-one correspondence;

[0118] S15, turning on the fill light strip 320, and setting the camera 140 predetermined start time and image acquisition time interval through the Raspberry Pi 310, and setting the Raspberry Pi 310 to automatically create a data folder and store image data;

[0119] S2. After the monitoring system is installed and set up, the camera 140 is debugged until the camera 140 image appears on the computer, indicating that the camera 140 is working properly.

[0120] S3. After the monitoring system is debugged, connect the water inlet pipe 240 to the peristaltic pump via a hose. Open the water inlet pipe 240 on the water absorption container 210 and inject distilled water into the water storage chamber of the water absorption container 210 through the water inlet pipe 240 until the water level is below the bottom of the permeable stone slab 230. Simultaneously, the pressure sensor begins to record the weight of the water absorption expansion deformation device in real time. Specifically, set the peristaltic pump at a constant rate to rapidly inject distilled water into the water absorption container.

[0121] S4. Set the image data acquisition time interval of the camera 140; after 5 minutes, start the peristaltic pump and set a slow constant rate to continuously inject distilled water into the water absorption container 210. During the process of injecting distilled water, always keep the water outlet pipe 250 open, and keep the water level in the water storage chamber of the water absorption container 210 unchanged during the monitoring process; since the water outlet height of the water outlet pipe 250 is flush with the lower surface of the permeable stone slab 230, the water level in the water storage chamber is flush with the lower surface of the permeable stone slab 230 at this time; excess water is discharged from the water outlet pipe 250; by controlling the water supply of the peristaltic pump, that is, the water supply per unit time of the peristaltic pump is greater than the water absorption amount of the mudstone through capillary action per unit time, thereby ensuring that the water level remains unchanged throughout the process.

[0122] The camera 140 is used to collect image data of the sample at regular intervals. Specifically, the camera 140 can be used to collect image data at regular intervals by running a Python program in a Raspberry Pi.

[0123] The image data collected in step S4 is processed to extract detailed information about the entire process of deformation evolution of mudstone mass due to water absorption, including the following steps:

[0124] S41. Using the scale-invariant feature transform (SIFT) algorithm, perform two-dimensional image feature point detection and description on the two-dimensional image data collected at the same moment by the three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view, thereby completing image feature extraction.

[0125] S42, using the K-nearest neighbor algorithm (KNN) to compare the similarity of each feature descriptor, find matching pairs of feature points in the two-dimensional images at the same time and different viewing angles, and obtain an initial matching result;

[0126] The random sampling consensus algorithm RANSAC is used to eliminate the obvious incorrect matching pairs in the initial matching results, and the optimal model parameters are obtained through parameter optimization as the final matching result;

[0127] Mismatches are those that deviate from the model's constrained geometric relationships. These are primarily caused by noise interference (e.g., illumination variations, sensor noise), repeated texture misassociations (e.g., similarity of mineral grains on mudstone surfaces), and dynamic deformation inconsistencies (e.g., sudden changes in crack propagation direction). To effectively eliminate mismatches, the RANSAC algorithm is used to iteratively select the optimal model through geometric consistency checks (e.g., projection error thresholds and epipolar constraints). This is combined with multi-view collaborative verification (cross-camera epipolar geometry and triangulation consistency) and temporal optical flow tracking to ensure the rationality of the spatial distribution of matching points and the coherence of their motion.

[0128] S43. Using incremental structure-from-motion technology, a pair of optimally registered images are selected based on the feature point matching results for preliminary 3D reconstruction. Appropriate 2D images are then selected from the remaining 2D images and gradually added to the existing model to continuously restore object details to complete sparse point cloud reconstruction. Appropriate 2D images are those with high matching degree, high image quality and stability, and good adaptability to dynamic scenes; that is, in terms of time sequence: adjacent moments, and in terms of deformation: similar images.

[0129] S44, using multi-view stereo vision technology MVS, by integrating multi-view depth information, recovering the geometric details of complex objects, converting sparse point clouds into dense point clouds, and completing dense point cloud reconstruction;

[0130] S45. Use the statistical discrete cluster removal algorithm (SOR) to analyze the spatial distance distribution characteristics between each data point and its neighboring points in the dense point cloud model. Utilize statistical principles to identify and remove abnormal points, retaining only the 3D point cloud model of the mudstone sample and the black and white calibration points to achieve point cloud denoising and obtain a high-precision 3D point cloud model at a certain moment.

[0131] S46, using a registration algorithm based on feature matching to perform rough point cloud registration on the three-dimensional point cloud model of the mudstone sample according to the black and white calibration points;

[0132] S47, using the iterative closest point algorithm (ICP), first performing precise point cloud registration on the black and white calibration points separately to obtain a corresponding transformation matrix, and then applying the matrix to the precise registration of the 3D point cloud model of the mudstone sample; constructing a time-series 3D point cloud model based on the 3D point cloud models at different times;

[0133] S48. Analyze the apparent fracture network and density development of red mudstone due to capillary water absorption based on a time-series 3D point cloud model. Extract the apparent fracture network contour of the mudstone sample through scalar domain calculation, and calculate the apparent fracture density through geometric features.

[0134] S49. Analyze the capillary water absorption deformation and displacement development of red-bed mudstone based on the time-series 3D point cloud model. Use the multi-scale point cloud model comparison algorithm M3C2 to calculate the deformation and displacement of the 3D point cloud model of the mudstone sample. Then extract the axial deformation and displacement at different times based on the time-series 3D point cloud model.

[0135] S410. Analyze the development of non-uniform expansion characteristics of red mudstone due to capillary water absorption based on the time-series 3D point cloud model, perform plane fitting on the top surface of the 3D point cloud model of the mudstone sample at different times, and compare the fitted planes to calculate the rotation angle and deflection displacement;

[0136] S411. Analyze the volume expansion and density development of red-bed mudstone due to capillary water absorption based on a time-series 3D point cloud model. Use Delaunay triangulation to convert the 3D point cloud model of the mudstone sample into a mesh model with a closed boundary. Calculate the volume and density of the mesh model. Extract the expansion volume at different times based on the time-series 3D point cloud model.

[0137] S412. Analyze the surface roughness development of red mudstone due to capillary water absorption based on the time series three-dimensional point cloud model. Calculate the surface roughness density of the time series three-dimensional point cloud model through geometric features, extract the surface roughness density histogram of the time series three-dimensional point cloud model, and analyze the surface roughness density distribution.

[0138] S5. The weight of the water-absorbing, swelling, and deformation device is recorded via pressure sensor 360. When the weight of the water-absorbing, swelling, and deformation device ceases to increase and stabilizes, and the image information captured by camera 140 indicates that the specimen's shape or volume increment has not changed, monitoring is stopped and the micropump is turned off, completing dynamic monitoring of the water-absorbing, deformable material. In this example, after 24 hours of water absorption, the mudstone specimen's swelling stabilizes and the deformation is minimal, indicating the end of the test. This results in real-time, all-encompassing image data of the mudstone's swelling and deformation process.

[0139] In order to facilitate the realization of automatic control, Python programming is used to realize automatic control. The control flow chart is as follows Figure 10 To ensure that all cameras 140 can work synchronously, specifically, a local area network is used to adjust the time of all Raspberry Pi 310, and then the same data collection time interval is set for each Raspberry Pi 310 to collect data.

[0140] Application Examples

[0141] The three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view described in the present invention is used to monitor image information in a mudstone water absorption deformation test; and the three-dimensional dynamic monitoring method for material water absorption deformation based on a hemispherical field of view described in the present invention is used to perform technical optimization on the image feature extraction, matching, and reconstruction problems in the mudstone water absorption deformation test.

[0142] Specifically, first, the scale-invariant feature transform (SIFT) algorithm is used to stably extract feature points under different perspectives, and the K-nearest neighbor (KNN) algorithm and the random sampling consensus algorithm (RANSAC) are combined to effectively deal with texture blur and mismatching problems, ensuring high-precision feature matching.

[0143] Then, the incremental structure from motion (incremental SFM) technique is used to gradually fuse multi-view images for efficient reconstruction of sparse point clouds.

[0144] Finally, multi-view stereo (MVS) technology is used to transform the sparse point cloud into a high-density dense point cloud, accurately restoring the object's geometric details. The hemispherical field-of-view, three-dimensional dynamic monitoring system and method for material water absorption deformation innovatively integrates image feature extraction, feature matching, and three-dimensional reconstruction techniques, overcoming challenges such as deformation and noise during the experiment. This provides a stable and accurate three-dimensional modeling solution with broad application prospects.

[0145] The system and method for monitoring the water absorption deformation of mudstone using the hemispherical field of view-based three-dimensional dynamic monitoring system and method for monitoring the water absorption deformation of mudstone, as well as three-dimensional reconstruction and point cloud processing, include the following steps:

[0146] 1. 3D reconstruction;

[0147] 1.1 Image acquisition and processing;

[0148] Through indoor experimental research, a new test device was designed independently to conduct mudstone water absorption deformation monitoring tests, and 49 cameras were used to collect two-dimensional sub-images of corresponding viewing angles. The two-dimensional sub-image captured by one camera is as follows: Figure 12 As shown. Due to the presence of various uncertainties in the experimental process, these factors will directly affect the image quality, thereby indirectly affecting the accuracy of feature extraction and three-dimensional model reconstruction. Therefore, before performing three-dimensional reconstruction, systematic image preprocessing must be performed to restore image quality and enhance key features. To this end, the three-dimensional dynamic monitoring method of material water absorption deformation based on hemispherical field of view described in the present invention adopts preprocessing technologies such as image filtering, sharpening and edge detection to effectively remove noise and enhance detail information, which not only significantly improves the image quality, but also optimizes the feature matching accuracy between multi-view images; providing a solid image data foundation for subsequent three-dimensional reconstruction.

[0149] 1.2 Image alignment;

[0150] In the image alignment process, SIFT algorithm is first used to detect and describe feature points, such as Figure 13As shown in the figure, the specific implementation steps and technical process of the algorithm are shown in detail. Due to its large computational load, it may lead to resource consumption and long processing time. Therefore, the KNN (K-NearestNeighbors) algorithm is used for feature matching on this basis. Although KNN can quickly find potential matching points, due to factors such as noise interference and perspective changes in the image, the preliminary matching results may have mismatches. Therefore, the RANSAC (RandomSampleConsensus) algorithm is combined to further optimize the matching results. Through the combined application of this series of algorithms, the feature matching accuracy can be effectively improved, and the accuracy and robustness of image alignment can be ensured. The alignment results are shown in Figure 1. Figure 16 shown.

[0151] 1.3 Sparse point cloud reconstruction;

[0152] In the sparse point cloud reconstruction process, the incremental motion recovery structure (SFM) technology is used to reconstruct the 3D geometric structure of the target object from the multi-view 2D image sequence and simultaneously estimate the camera pose. This process achieves the accurate mapping and conversion of feature points from 2D images to 3D space, and preliminarily reconstructs the 3D geometric shape of the mudstone sample. The sparse point reconstruction process is as follows: Figure 14 The reconstruction results are shown in Figure 17 shown.

[0153] 1.4 Dense point cloud reconstruction;

[0154] In the process of dense point cloud reconstruction, multi-view stereo vision (MVS) technology, such as Figure 15 As shown in , the depth value of each pixel is further calculated to generate a high-precision depth map, and the three-dimensional spatial coordinates corresponding to each pixel are determined, thereby generating denser point cloud data with higher density and richer details, further restoring the three-dimensional surface structure details of the mudstone sample to generate a more accurate and detailed three-dimensional point cloud model, as shown in Figure 18 shown.

[0155] In order to accurately characterize the dynamic changes in the surface morphology of mudstone samples during water absorption, this embodiment adopts a time-series three-dimensional point cloud modeling method. Based on the independently proposed two-dimensional image processing process, the two-dimensional image data set at each moment is reconstructed in three dimensions. Figure 19 As shown in the figure, as water absorption time increases, water penetrates upward from the bottom of the specimen, resulting in a distinct color gradient on the surface. Initially, the dark area is concentrated at the bottom of the specimen, then gradually expands upward to cover the entire surface. Notably, water migration under capillary absorption not only causes color changes but also leads to the continuous development of surface cracks. These features are reflected in the 3D point cloud model as a distinct dark crack network.

[0156] 2. Point cloud processing;

[0157] 2.1 Point cloud denoising;

[0158] The statistical outlier removal (SOR) algorithm was used to denoise the dense point cloud model. By analyzing the spatial distance distribution between each data point in the point cloud and its neighboring points, combined with statistical principles, abnormal points were identified and removed. This process effectively removed noise points, only retaining the true three-dimensional point cloud data of the mudstone sample and the black and white calibration points, thereby obtaining a high-precision three-dimensional point cloud model, ensuring the accuracy and stability of subsequent three-dimensional reconstruction. The results are shown in Figure 20 .

[0159] 2.2 Point cloud registration

[0160] (1) Point cloud coarse registration

[0161] The artificial auxiliary marking method was used to perform point cloud coarse calibration based on the black and white calibration points on the water absorption container in the millisecond-level dynamic three-dimensional reconstruction deformation monitoring system with full-dimensional stereoscopic hemispherical view. According to the source point cloud and target point cloud, eight black and white calibration points were artificially selected as key marker points for extraction, description, translation, and rotation matrix calculation. The transformation matrix was applied to the source point cloud for translation and rotation, which roughly aligned the source point cloud to the position of the target point cloud, achieving point cloud coarse registration. The point cloud coarse registration results are shown in Figure 24 , most of the point cloud model of the mudstone sample before water absorption is wrapped inside the point cloud model of the mudstone sample after water absorption, and the eight black and white calibration points are basically completely coincident, with good coarse registration effect.

[0162] (2) Point cloud fine registration

[0163] Based on coarse registration, the iterative closest point (ICP) algorithm was used to further optimize the registration results. The specific process of the ICP algorithm is shown in Figure 21 .

[0164] Since the indoor test research device was designed with eight black and white calibration points in the water absorption container. After point cloud denoising, only the mudstone sample model and calibration point data were retained. After completing point cloud coarse registration, the marker points were first separated from the mudstone sample. Then a step-by-step registration strategy was adopted: first, the eight black and white marker points were fine registered to obtain an accurate transformation matrix; then the transformation matrix was applied to the fine registration of the mudstone sample, and the final point cloud registration results are shown in Figure 22 . According to the proposed three-dimensional point cloud model processing method, the point cloud of the mudstone sample model was processed in a continuous time sequence, and the dynamic evolution process of the apparent morphology during the capillary water absorption process of the mudstone sample was reconstructed.

[0165] In summary, the technical solution described in the present invention innovatively proposes a two-dimensional image processing method based on oblique photography modeling and point cloud processing technology. After image processing, sparse point cloud reconstruction, dense point cloud reconstruction, point cloud denoising, and point cloud registration, three-dimensional point cloud models of mudstone samples at different times are extracted, and a series of three-dimensional point cloud models with time series are constructed to characterize the apparent morphological development of red-bed mudstone during capillary water absorption.

[0166] 3. Analysis of temporal and spatial evolution of mudstone deformation due to water absorption;

[0167] 3.1 Spatiotemporal evolution of apparent crack development;

[0168] 3.1.1 Apparent fracture network;

[0169] Based on oblique photography modeling and point cloud processing, the apparent crack network diagram of the sample is obtained as follows: Figure 23 As shown in the figure, combined with analysis of the mudstone water absorption deformation monitoring test presented by two-dimensional real-time images, it was found that unevenly distributed, relatively small cracks formed primarily in the lower region of the specimen, with a generally upward expansion angle. Although the cracks developed primarily horizontally, they maintained a certain oblique expansion trend, which also led to the more obvious long and wide cracks in the upper and middle regions. During this process, the fracture network morphology gradually expanded and developed from the initial scattered cracks, eventually forming a grid-like fracture network.

[0170] Through 3D modeling, point cloud processing, and image analysis, it was found that the fracture network of the mudstone specimens exhibited significant temporal and spatial heterogeneity. In the early stages of the experiment, water migrated upward from the bottom of the specimen through capillary action, and the wetting front spread from bottom to top. However, due to the complex pore structure and uneven distribution of hydrophilic minerals within the mudstone, the permeability of the mudstone was limited, the propagation of the wetting front was slow, and the migration speed varied between different regions. Therefore, the development of the mudstone capillary water absorption fracture network can be divided into three stages:

[0171] (1) Scattered crack network development stage (0-148 minutes): The wetting front gradually migrates upward from the bottom of the specimen to 25 mm. In the early stage (67 minutes), short, narrow, non-directional scattered cracks are formed in the lower part of the specimen, mainly in the horizontal direction. In the middle stage (92 minutes), the cracks expand along the upward slope (0-25 degrees), forming a small network with local interlacing. In the late stage (148 minutes), the cracks in the middle gradually widen due to the accumulation of pore pressure, and the overall distribution characteristics are "dense at the bottom and sparse in the middle". This stage is mainly driven by the cementation failure caused by the water absorption and expansion of clay minerals (adsorption-wedge failure mechanism). The water migration rate at the bottom is high, but the expansion stress has not fully accumulated, so the cracks are short and scattered.

[0172] (2) Grid-like crack network development stage (148-399 minutes): The wetting front expanded to 50 mm. In the early stage (214 minutes), a wide main crack with a size of 25 mm developed rapidly. In the middle stage (289 minutes), the original cracks extended upward to form new cracks and widened. In the late stage (399 minutes), the main cracks expanded in a nearly horizontal direction (angle > 55°), and secondary cracks were generated vertically, forming a grid-like structure composed of wide main cracks and short and wide secondary cracks. In this stage, under the synergistic action of radial compressive stress and axial tensile stress, the main cracks expanded along the direction of least resistance, while the secondary cracks were driven by the strain energy release of the expansion core.

[0173] (3) Stable development stage of the fracture network (399-501 minutes): After the wetting front reaches the top of the specimen, the free boundary conditions cause stress to be released mainly in the radial direction, forming three radial main fractures and "branch"-like secondary fractures at the top. The crack width decreases from 0.7-1.2 mm at the circumferential edge to 0.1-0.3 mm at the center, reflecting the spatial attenuation characteristics of the stress concentration. The local area causes radial convex deformation due to the tensile stress concentration at the crack tip, accompanied by accelerated expansion of volume. In this stage, the anisotropic constraints of the red mudstone layer structure and the hard interlayer limit the vertical deformation, forcing the strain energy to be released in the radial direction, ultimately forming a self-organized fracture morphology.

[0174] 3.1.2 Apparent crack density;

[0175] Through 3D point cloud modeling, the crack density map is extracted according to the crack width distribution. Figure 24 According to the analysis of the crack density map, it was found that in the axial dimension of the specimen, the crack density showed a non-uniform distribution characteristic of "sparse at the bottom, dense in the middle and upper parts, and less dense at the top"; while in the radial dimension, it showed a gradient decreasing pattern of "high density at the edge and low density in the center".

[0176] The distribution of crack density in the axial dimension (0-50 mm) of the red bed mudstone sample shows an obvious "low-high-slowly decreasing" staged development trend, and its peak migration trajectory has a certain effect on the spread of the wetting front: the crack density and peak value gradually increase as the wetting front spreads from bottom to top along the axial direction of the sample, reaching a maximum value in the upper and middle part of about 40 mm. However, thereafter, as the wetting front spreads upward, the crack density shows a slow downward trend, with a small decline.

[0177] (1) Scattered fracture network development stage (0-148 minutes): The wetting front migrates upward from the bottom to 25 mm. Cracks mainly appear in the lower area, showing uneven distribution of small cracks. They expand in a nearly horizontal direction, but gradually develop in an oblique direction. The fracture density is low, and the closer to the bottom, the smaller the fracture density.

[0178] (2) Grid-like crack network development stage (148-399 minutes): The wetting front spread to 50 mm, the crack density increased, and the distribution was relatively uniform, reaching about 76.54%. With the accumulation of expansion stress, the crack width increased, and the secondary cracks expanded upward at an angle greater than 55°, eventually forming a grid-like structure. The crack density cloud map shows a relatively high density, especially at 214 minutes, when the wetting front reached 40 mm and the crack density reached its peak. The crack expansion was restricted by the free boundary at the top, resulting in a gradual decrease in crack width and a slow decrease in crack density.

[0179] (3) Stable development stage of the fracture network (399-501 minutes): The top is affected by the free boundary, and the expansion stress is mainly radial, resulting in the formation of radial fractures at the top. The fracture density gradually decreases from the outside to the inside at the top, and the high-density area shifts to the 40 mm area. The fracture density shows a "low-high-gradually decreasing" trend.

[0180] Combined with the development of the fracture network, an analysis from the perspective of time evolution revealed that the fracture density gradually increases from bottom to top, reaching a maximum peak at around 40 mm, and forming a high-density zone around this area. Subsequently, as fractures in the upper middle and top continue to develop, the high-density zone continues to approach the 30-50 mm region, while the low-density zone continues to gather towards the lower region of the specimen and the top center area. Ultimately, in the axial dimension, a non-uniform distribution characteristic of "sparse at the bottom, dense in the upper middle, and less dense at the top" is presented, with fracture area densities of 0.008, 0.184, and 0.097 cm² / cm², respectively, showing a clear "low-high-gradually declining" staged development trend. In the radial dimension, a gradient decreasing pattern of "high density at the edge, low density at the center" is presented, with fracture area densities of 0.062 and 0.035 cm² / cm², respectively.

[0181] 3.2 Analysis of deformation and displacement of mudstone due to water absorption;

[0182] 3.2.1 Spatial differentiation characteristics of deformation and displacement;

[0183] Displacement cloud map Figure 25 The displacement cloud map is shown in Figure 1. The color bar of the displacement cloud map is mainly blue, green, and red, and the values ​​represented increase from negative to positive. The blue area represents negative displacement, the green area represents displacement of 0mm or close to 0mm, and the red area represents positive displacement.

[0184] Analysis of displacement contours reveals spatial heterogeneity in the deformation of the cylindrical mudstone specimen. The specimen's lateral surfaces, primarily characterized by blue and green areas, exhibit minimal displacement, indicating primarily radial, horizontal displacement from the inside out. The top region, dominated by red, exhibits significant deformation, primarily manifesting as axial, vertical displacement from the bottom up. The lower portion of the specimen exhibits minimal deformation, with displacement variations approaching 0 mm, indicating no significant deformation. In contrast, the top region exhibits displacement variations ranging from 4.51 mm to 6.78 mm, indicating significant axial expansion. The upper-middle portion of the specimen exhibits blue and yellow-green areas, representing negative and positive values ​​slightly above 0 mm, respectively, and are symmetrically distributed about the specimen's axis. Displacement variations in the upper-middle region range from 0 to 2.64 mm, demonstrating an opposite trend, indicating that the specimen did not shrink during capillary water absorption, but rather experienced a slight tilt due to complex deformation. Analysis of the experimental phenomena suggests that this variation is related to the radial expansion and horizontal displacement of the specimen as a result of water absorption.

[0185] 3.2.2 Apparent crack density;

[0186] like Figure 26 The figure shows the axial deformation displacement diagram used in this application example. By calculating the axial displacement of the specimen's 3D point cloud model at different times, an axial displacement-time curve is plotted. The curve is roughly S-shaped, expanding and extending from bottom to top at varying rates, with a maximum axial displacement of 6.78 mm. Because the development and width of cracks directly affect the axial extension displacement of the mudstone specimen to a certain extent, this application example conducts a detailed analysis of the axial displacement based on the temporal and spatial evolution of crack development:

[0187] (1) Scattered crack network development stage: Water gradually enters the pore network inside the specimen, and with the continuous generation of scattered small cracks on the surface of the specimen, the specimen begins to gradually extend from bottom to top along the axial direction, and the extension rate also increases continuously. When the water absorption time reaches 116 minutes, the displacement development enters the acceleration stage, and the extension rate rises sharply from 0.01 mm / min to 0.05 mm / min. At this time, the wetting front spreads to the middle of the specimen. The expansion stress generated by the water entering the specimen exceeds the tensile strength, resulting in enhanced connectivity of the internal pore network, making it easier for water to migrate inside the specimen, thereby generating grid-like cracks, and the crack width gradually increases.

[0188] (2) Grid-like crack development stage: wide grid-like cracks began to appear in the middle and upper part of the specimen. Although the crack width continued to develop and expand during this stage, the internal pore network became more complex as the cracks continued to increase and expand, and the migration path of water in the specimen became more tortuous, resulting in a decrease in the propagation rate of the wetting front, slow development of the crack width, and the occurrence of "local displacement hysteresis" in the middle and lower regions, that is, the small cracks in the middle and lower regions were compressed. Therefore, the axial displacement extension rate decreased to 0.02 mm / min at this time.

[0189] (3) Stable development stage of crack network: As capillary water absorption continues to increase, the crack development time further increases due to the weakening of capillary action, causing the axial displacement extension rate of the specimen to decrease again to 0.007 mm / min.

[0190] 3.3 Analysis of heterogeneous expansion of mudstone due to water absorption;

[0191] like Figure 27 As shown in the figure, the top surface of the point cloud model before and after water absorption was intercepted and plane fitted. The top surface of the point cloud model before water absorption was relatively flat, and the fitting plane and the top surface of the model alternately penetrated each other. However, after water absorption, the top surface of the point cloud model and the fitting plane penetrated each other in a single manner, which indicated that the sample column was tilted.

[0192] Through plane fitting and comparison, it was found that the rotation angle of the top surface of the three-dimensional point cloud model of the two samples before and after water absorption was 2.01926°, and the inclination shifted from 36° east-northeast to 33° east-northeast, indicating that the top surface rotated slightly during the capillary water absorption process. In addition, the offset of the top surface of the three-dimensional point cloud model of the two samples before and after water absorption was approximately 5.473459 mm. Therefore, the expansion deformation of the red-bed mudstone sample during capillary water absorption is not a simple vertical expansion deformation, but a non-uniform expansion deformation with a certain inclination. This composite deformation pattern confirms that the red-bed mudstone not only undergoes axial expansion during water migration, but is also accompanied by obvious shear and torsional deformation, reflecting the spatial heterogeneity of the fracture network within the rock mass.

[0193] 3.4 Analysis of mudstone volume expansion due to water absorption;

[0194] The volume calculation shows that the volume of the red mudstone sample before water absorption is 98.6855cm3, while the volume of the red mudstone sample after water absorption is 114.1415cm3, the volume expansion is about 15.456cm3, and the volume expansion rate is about 15.66%. According to the proposed volume expansion analysis method based on the three-dimensional point cloud model, the volume density of the mudstone samples before and after water absorption is calculated, as shown in Figure 2. Figure 28 shown.

[0195] From a spatial dimension analysis, there are obvious density peaks in the circumferential area of ​​the top surface of the specimen and the wide crack development area on the side. The axial deformation shows a nonlinear increasing characteristic from bottom to top. The volume expansion in the 0-25mm area from bottom to top of the specimen accounts for 36.17%, while the volume expansion in the 25-50mm area accounts for 63.82%. Therefore, the volume expansion of the red mudstone under capillary action shows obvious axial gradient and radial heterogeneity:

[0196] (1) Axial gradient: The volume density cloud map shows that the volume expansion of the mudstone sample gradually increases from bottom to top along the axial direction. The volume expansion of the lower part of the sample is relatively small, mainly because when water migrates to this area, the hydrophilic minerals form a dense gel layer, resulting in the development of small cracks and limited expansion effect. As water migrates to the middle and upper parts, expansion stress accumulates, the crack width increases, and local areas expand or even peel off, resulting in increased volume expansion.

[0197] (2) Radial heterogeneity: During the development of the grid fissures, the radial compressive stress generated by water migration and the axial tensile stress act together to promote the horizontal expansion of large primary fissures and form vertical secondary fissures, reducing the interaction between minerals and causing radial expansion of the specimen. This effect is particularly pronounced in the wide fissure area, resulting in abnormal volume density in this area.

[0198] In order to analyze the volume expansion of red mudstone due to capillary water absorption from the time dimension, the volume of the point cloud model of the mudstone sample at different times was calculated and extracted. The results are shown in the figure below. Figure 29 shown.

[0199] Volume expansion generally increases with capillary absorption time, but the growth rate varies. Combining the analysis of apparent crack development and deformation displacement development in red-bed mudstone, it was found that when wide cracks developed, the volume expansion rates were 0.051 cm³ / min and 0.057 cm³ / min, respectively. Therefore, there is a certain positive correlation between volume expansion and crack and deformation development:

[0200] (1) Scattered crack network development stage: The average expansion rate is 0.037 cm³ / min. Initially, water migrates mainly from bottom to top along the axial direction, and crack development is not obvious. As the expansion effect intensifies, small cracks and some wide cracks gradually appear in the specimen, resulting in an increase in the expansion rate.

[0201] (2) Grid-like crack development stage: The average expansion rate is 0.025 cm³ / min. As wide cracks form, the axial extension rate of the specimen increases, and local radial stress appears in some areas, causing local bulges in the specimen, further increasing the expansion rate.

[0202] (3) Fracture network development stage: The average expansion rate is 0.037 cm³ / min. Initial expansion is slow, water migration is slow, and the expansion of small cracks is the main process. At about 418 minutes, the wide cracks penetrate and release radial compressive stress, causing the expansion rate to rise sharply, and then enter a stable stage. In addition, at 300 minutes, a slow expansion rate stage occurs, and the wide cracks compress the small cracks below and cause the sample to tilt, resulting in a decrease in the expansion rate.

[0203] 3.5 Analysis of surface roughness of mudstone water absorption;

[0204] The roughness density of the point cloud model of the mudstone sample before and after water absorption is calculated as follows: Figure 30 、 Figure 31 shown.

[0205] The roughness bars of the point cloud model of the mudstone sample before water absorption are similar in color, and the histogram shows a left-skewed distribution with a peak at zero. This indicates that in its dry state, due to its mineralogy and bedding structure, the mudstone has no obvious apparent cracks or structural defects, and its surface is relatively flat and smooth with low roughness. However, due to mudstone's particular sensitivity to moisture, water gradually migrates upward through the internal pore network under capillary action, affecting the physical properties of the mudstone both internally and on its surface.

[0206] The roughness density cloud of the point cloud model of the mudstone sample after water absorption shows that the blue low-density area is mainly distributed in the 10-40mm area of ​​the mudstone sample, with the green medium-density area filling the remaining area; while the 0-10mm and 40-50mm areas of the sample are composed of green medium-density areas and red high-density areas. The color stripe distribution on the side of the cylindrical sample is roughly symmetrical with "high at both ends and low in the middle" centered at a height of 25mm. The color stripe distribution on the top surface of the sample is distributed in a "concentric circle" shape, gradually spreading from the center of the circle to the outside, from the blue high-density area to the red low-density area, showing a clear radial annular gradient increase feature:

[0207] (1) Axial dimension:

[0208] When water migrates to the 0-10mm region, the cracks are primarily fine, with bottom constraints causing local bulges at the crack edges. As the water migrates further upward, the crack and pore network becomes saturated, and some areas begin to peel, increasing surface roughness.

[0209] When moisture migrates to the 10-40mm area, wide cracks develop, extending axially overall, resulting in a relatively uniform surface with low roughness. Changes in localized fine cracks lead to larger fluctuations in surface roughness, especially cracking, which manifests as localized radial expansion and increased surface roughness.

[0210] When moisture continues to migrate toward the 40-50 mm area, the top surface is constrained by the circumferential free boundary, and radial expansion triggers the expansion of radial cracks, resulting in local bulges on the circumferential edge, extreme roughness, and red color bars.

[0211] (2) Radial dimension:

[0212] When moisture migrates to the top surface of the specimen, due to the weakening of capillary action, only three main cracks develop in the top area. Combined with the real-time image analysis of the experiment, it is found that these cracks serve as boundaries to divide the top surface into regions. The local small cracks in the sub-regions are mainly close to the circumferential edge, and the surface is relatively smooth. Therefore, the surface roughness density in the top area increases in a circular gradient from the inside to the outside along the radial direction.

Claims

1. A three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view, characterized by: It comprises a water absorption expansion and deformation device (200), a camera monitoring system (100), and a data acquisition device (300); The camera monitoring system (100) comprises an annular base (150); a hemispherical hollow camera bracket is provided on the annular base (150); The hollow camera bracket is provided with a camera (140) for photographing the inner side of the camera bracket; the hemispherical top formed by the camera bracket is provided with the camera (140); The water absorption expansion deformation device (200) comprises a water absorption container (210); the water absorption container (210) is provided with nine calibration columns (220) evenly distributed along the circumference; The water absorption container (210) has a water storage cavity; a circular permeable stone plate (230) is provided in the middle of the upper surface of the water absorption container (210); a permeable hole communicating with the water storage cavity is provided on the permeable stone plate (230); a water outlet pipe (250) and an openable and closable water inlet pipe (240) are provided on the water absorption container (210); the height of the water outlet of the water outlet pipe (250) is flush with the lower surface of the permeable stone plate (230); and a liquid discharge height adjustment device (260) is provided on the water outlet pipe (250); The drainage height adjustment device (260) comprises an adjustment pipe body (261); a connecting pipe (262) matching the water outlet pipe (250) is provided at one end of the adjustment pipe body (261); a vertical pipe (263) is provided at the other end of the adjustment pipe body (261); a telescopic pipe (264) is provided on the vertical pipe (263); the telescopic pipe (264) is threadedly matched with the vertical pipe (263); The data acquisition device (300) comprises a Raspberry Pi (310), a fill light strip (320), and a mounting frame (330); the Raspberry Pi (310) is mounted on the mounting frame (330), and the fill light strip is mounted inside the mounting frame (330); the Raspberry Pi (310) corresponds to the camera (140) in a one-to-one manner, and the Raspberry Pi (310) is electrically connected to the camera (140); The water absorption container (210) is located at the exact center of the inner circle of the annular base (150); the annular base (150) is located at the exact center of the bottom of the inner cavity of the mounting frame (330); a pressure sensor (360) is provided at the bottom of the water absorption container (210); The hollow camera bracket comprises a plurality of arc-shaped brackets (130) with an arc angle of 90°; a connecting block (131) is provided at the lower end of the arc-shaped bracket (130), and a vertical arc-shaped connecting block (132) is provided at the upper end; The annular base (150) is provided with mounting slots evenly distributed along the circumference; the connecting block (131) of the arc-shaped bracket (130) is inserted into the mounting slot; the arc-shaped connecting block (132) at the upper end of the arc-shaped bracket (130) is spliced ​​to form a cylinder with a central mounting through hole; the cylinder is clamped by a clamping hoop (110); and the clamping hoop (110) is locked by a locking bolt (120).

2. The three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to claim 1, characterized in that: The arc-shaped bracket (130) is provided with a mounting groove (133); the camera (140) is installed in the mounting groove (133), and a wire collection hole (160) is provided below the mounting groove (133).

3. The three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to claim 1, characterized in that: The upper surface of the water absorption container (210) is provided with a mounting groove matching the permeable stone plate (230); the bottom of the mounting groove is provided with a through hole communicating with the water storage cavity of the water absorption container (210).

4. The three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to claim 1, characterized in that: The camera (140) adopts a USB camera with a distortion-free lens, a maximum resolution of 2592*1944, and automatic focus.

5. The three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to claim 4, characterized in that: The Raspberry Pi (310) is electrically connected to the camera (140) via a data line (340); the data line (340) comprises two neutral lines, a live line, a data input line, and a data output line; one end of the data line is provided with a 5-pin piercing terminal, and the other end is provided with a Type-C female socket; the piercing terminal is connected to the USB camera; a data line connection cap (350) is provided on the Raspberry Pi (310); the Type-C female socket is electrically connected to the data line connection cap (350) of the Raspberry Pi (310) via a Type-C-USB data line.

6. The three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view according to claim 2, characterized in that: There are 16 arc-shaped brackets (130); three cameras (140) are provided on each arc-shaped bracket (130), and the three cameras (140) are evenly distributed along the curvature of the arc-shaped bracket (130).

7. A three-dimensional dynamic monitoring method for material water absorption deformation based on a hemispherical field of view, characterized by: A three-dimensional dynamic monitoring system for water absorption and deformation of materials based on a hemispherical field of view as described in any one of claims 1 to 6 is used; The following steps are also included: S1. Installation and setup of monitoring system; S11. Place the test piece in an oven at 100-105°C for 24 hours and then place it in a cooling dish for cooling; S12, placing the sample on a circular permeable stone plate (230); S13, placing the water absorption expansion deformation device (200) in the exact center of the inner cavity of the camera monitoring system (100); and simultaneously placing the data acquisition device (300) outside the camera monitoring system (100), so that the camera monitoring system (100) is located in the exact center of the inner cavity of the data acquisition device (300); S14, combining and connecting the data acquisition device (300) and the camera monitoring system (100), that is, electrically connecting the camera (140) in the camera monitoring system (100) to the Raspberry Pi (310) of the data acquisition device (300) via a data cable (340) and a Type-C-USB data cable, with a one-to-one correspondence; S15, turning on the fill light strip (320), and setting the camera (140) predetermined start time and image acquisition time interval through the Raspberry Pi (310), and setting the Raspberry Pi (310) to automatically create a data folder and store image data; S2. After the monitoring system is installed and set up, the camera (140) is debugged until the camera (140) screen appears on the computer; S3. After the monitoring system is debugged, the water inlet pipe (240) is connected to the peristaltic pump via a hose, the water inlet pipe (240) on the water absorption container (210) is opened, and distilled water is injected into the water storage chamber of the water absorption container (210) through the water inlet pipe (240) until the water level is 1-2 mm below the bottom of the capillary permeable stone plate (230); at the same time, the weight of the water absorption expansion deformation device (200) is recorded in real time via the pressure sensor (360); S4, setting the image data acquisition time interval of the camera (140); after 5 minutes, starting the peristaltic pump and setting a constant rate to continuously inject distilled water into the water absorption container (210); during the injection of distilled water, always keeping the water outlet pipe (250) open, and keeping the water level in the water storage chamber of the water absorption container (210) unchanged during the monitoring process; and the water level in the water storage chamber is flush with the lower surface of the permeable stone slab (230); performing timed image data acquisition on the sample through the camera (140); S5. The weight of the water absorption expansion and deformation device is recorded by the pressure sensor (360). When the weight of the water absorption expansion and deformation device no longer increases and the image information collected by the camera (140) shows that the shape or volume increment of the test piece has not changed, the monitoring is stopped and the micro pump is turned off; thus, the dynamic monitoring of the water absorption and deformation material is completed.

8. The three-dimensional dynamic monitoring method for material water absorption deformation based on a hemispherical field of view according to claim 7, characterized in that: The image data collected in step S4 is processed to extract detailed information about the entire process of deformation evolution of mudstone mass due to water absorption, including the following steps: S41. Using the scale-invariant feature transform (SIFT) algorithm, perform two-dimensional image feature point detection and description on the two-dimensional image data collected at the same moment by the three-dimensional dynamic monitoring system for material water absorption deformation based on a hemispherical field of view, thereby completing image feature extraction. S42, using the K-nearest neighbor algorithm (KNN) to compare the similarity of each feature descriptor, find matching pairs of feature points in the two-dimensional images at the same time and different viewing angles, and obtain an initial matching result; The random sampling consensus algorithm RANSAC is used to eliminate incorrect matching pairs in the initial matching results, and the optimal model parameters are obtained as the final matching results through parameter optimization; S43. Using incremental structure-from-motion technology, a pair of optimally registered images is selected based on the feature point matching results for preliminary 3D reconstruction. Appropriate 2D images are then selected from the remaining 2D images and gradually added to the existing model, continuously restoring object details to complete sparse point cloud reconstruction. S44, using multi-view stereo vision technology MVS, by integrating multi-view depth information, recovering the geometric details of complex objects, converting sparse point clouds into dense point clouds, and completing dense point cloud reconstruction; S45. Use the statistical discrete cluster removal algorithm (SOR) to analyze the spatial distance distribution characteristics between each data point and its neighboring points in the dense point cloud model. Utilize statistical principles to identify and remove abnormal points, retaining only the 3D point cloud model of the mudstone sample and the black and white calibration points to achieve point cloud denoising and obtain a high-precision 3D point cloud model at a certain moment. S46, using a registration algorithm based on feature matching to perform rough point cloud registration on the three-dimensional point cloud model of the mudstone sample according to the black and white calibration points; S47, using the iterative closest point algorithm (ICP), first performing precise point cloud registration on the black and white calibration points separately to obtain a corresponding transformation matrix, and then applying the matrix to the precise registration of the 3D point cloud model of the mudstone sample; constructing a time-series 3D point cloud model based on the 3D point cloud models at different times; S48. Analyze the apparent fracture network and density development of red mudstone due to capillary water absorption based on a time-series 3D point cloud model. Extract the apparent fracture network contour of the mudstone sample through scalar domain calculation, and calculate the apparent fracture density through geometric features. S49. Analyze the capillary water absorption deformation and displacement development of red-bed mudstone based on the time-series 3D point cloud model. Use the multi-scale point cloud model comparison algorithm M3C2 to calculate the deformation and displacement of the 3D point cloud model of the mudstone sample. Then extract the axial deformation and displacement at different times based on the time-series 3D point cloud model. S410. Analyze the development of non-uniform expansion characteristics of red mudstone due to capillary water absorption based on the time-series 3D point cloud model, perform plane fitting on the top surface of the 3D point cloud model of the mudstone sample at different times, and compare the fitted planes to calculate the rotation angle and deflection displacement; S411. Analyze the volume expansion and density development of red-bed mudstone due to capillary water absorption based on a time-series 3D point cloud model. Use Delaunay triangulation to convert the 3D point cloud model of the mudstone sample into a mesh model with a closed boundary. Calculate the volume and density of the mesh model. Extract the expansion volume at different times based on the time-series 3D point cloud model. S412. Analyze the surface roughness development of red mudstone due to capillary water absorption based on the time series three-dimensional point cloud model. Calculate the surface roughness density of the time series three-dimensional point cloud model through geometric features, extract the surface roughness density histogram of the time series three-dimensional point cloud model, and analyze the surface roughness density distribution.

Images