Binocular multi-point displacement monitoring equipment and method thereof

Abstract

The invention discloses binocular multi-point displacement monitoring equipment and a method thereof, aiming at temporary slope displacement monitoring and acquiring displacement based on an image recognition mode of position comparison of a reference object on a time axis, the binocular multi-point displacement monitoring equipment comprises a fixing frame and a first displacement mechanism arranged on the fixing frame, and the first displacement mechanism is provided with a mounting frame; at least two second displacement mechanisms with independent displacement control are arranged on the mounting frame, and a monitoring camera is loaded on each second displacement mechanism; the system further comprises a control module, and the control module is connected with the first displacement mechanism to control synchronous displacement of all the monitoring cameras and connected with the second displacement mechanism to control independent displacement of each monitoring camera. According to the method, two free monitoring cameras of the equipment are utilized to realize the displacement monitoring effect of monocular inspection and binocular displacement type and displacement range acquisition. The method is high in efficiency, low in cost and convenient to deploy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of geological monitoring, and particularly relates to a binocular multi-point displacement monitoring device and a method thereof. BACKGROUND

[0002] In the field of engineering construction and geological disaster prevention and control, displacement monitoring of temporary slopes is a key link to ensure construction safety and surrounding environment safety. Such slopes have the characteristics of dynamic adjustment of monitoring period and high requirement for displacement response speed, and small displacements need to be captured in time to prevent risks such as collapse and landslide. Existing slope displacement monitoring methods mainly include three categories: GNSS positioning monitoring, crack displacement meter monitoring and visual monitoring. Among them, GNSS positioning monitoring can realize long-distance point displacement measurement, but the equipment cost is high, it is easily affected by trees and buildings, and it is difficult to cover the whole slope; crack displacement meter can only carry out contact monitoring for local cracks and cannot reflect the overall deformation trend of the slope; visual monitoring has the advantages of non-contact, surface coverage and moderate cost, and has become an important choice for temporary slope monitoring. The core is to collect image or video data based on a preset reference point, and calculate the displacement by comparing the position changes of the reference points at different time axes.

[0003] The existing visual monitoring method has obvious limitations in the scene of temporary slope monitoring, and it is difficult to meet the efficient and accurate monitoring demand. The monitoring mode is fixed and the energy consumption and precision are unbalanced. Although ordinary plane visual monitoring has low energy consumption, two-dimensional displacement data can be obtained by comparing the positions of reference points, but it cannot distinguish the depth direction and actual deflection trend of displacement, and it is easy to misjudge the displacement risk. In the existing technology, binocular visual monitoring technology is used to supplement the depth information to form a three-dimensional monitoring system. Although three-dimensional displacement data can be obtained by parallax calculation to improve measurement accuracy, the binoculars need to be continuously maintained in a cooperative working state, which consumes significantly more energy than monocular monitoring. In actual application, the working mode needs to be set manually throughout the process, and it cannot be automatically switched from low-energy preliminary inspection to high-precision three-dimensional measurement according to the displacement monitoring demand, resulting in high energy consumption in daily monitoring or delayed response to abnormal displacement.

[0004] At the same time, the existing visual monitoring method cannot completely cover the monitoring area with the picture area that meets the optimal scale for monitoring, resulting in the need to collect data through scanning. However, there is a data coordination comparison process here. Because there are various types of displacement, the overall monitoring cannot determine the type of displacement, resulting in the existing visual monitoring method being able to determine the possible displacement but not the type and area of displacement, and being unable to effectively judge and accurately respond to corresponding measures. SUMMARY

[0005] In order to solve the problems existing in the prior art, the application provides a binocular multi-point displacement monitoring device and a method thereof, aiming at forming a binocular automatic calibration displacement range mode in a special low-energy consumption inspection and multi-stage early warning mode by using a low-cost visual monitoring device, and forming a new visual monitoring solution for slope displacement.

[0006] The technical scheme adopted by the application is: In a first aspect, the application provides a binocular multi-point displacement monitoring device for temporary slope displacement monitoring, which obtains a displacement amount based on image recognition of position comparison of a reference object on a time axis, and includes a fixed frame and a first displacement mechanism arranged on the fixed frame, the first displacement mechanism is provided with a mounting frame, at least two second displacement mechanisms for independent displacement control are arranged on the mounting frame, and each second displacement mechanism is loaded with a monitoring camera; the control module connected with the first displacement mechanism and the second displacement mechanism is further included, and the control module controls the first displacement mechanism to drive all monitoring cameras to perform synchronous displacement, or controls each second displacement mechanism to drive the monitoring camera to perform independent displacement.

[0007] In combination with the first aspect, the application provides a first implementation manner of the first aspect, the fixed frame is a tripod structure fixed on the ground, and the first displacement mechanism and the second displacement mechanism are both two-axis rotating movement mechanisms.

[0008] In combination with the first aspect, the application provides a second implementation manner of the first aspect, the mounting frame is a long strip-shaped fixed rod, the middle part of the fixed rod is fixed to the first displacement mechanism through a detachable mechanism, and the second displacement mechanism is arranged on the length direction of the fixed rod through the detachable mechanism; for slope surface displacement monitoring within a range of 20-100 m from the fixed frame, the spacing between adjacent second displacement mechanisms arranged on the fixed rod is 25-40 cm.

[0009] In a second aspect, the application further discloses a monitoring method, which uses the binocular multi-point displacement monitoring device according to any one of the above to monitor slope displacement, and the specific steps are as follows: Step 100, determining the range of the monitored slope surface, arranging a plurality of fixed reference objects on the slope surface, and obtaining positioning parameters of the arranged fixed reference objects through a positioning device; Step 200, selecting a stable terrain within a range of 50-200 m of the slope surface to arrange the binocular multi-point displacement monitoring device, arranging the fixed frame, the first displacement mechanism, the mounting frame, at least two second displacement mechanisms and corresponding monitoring cameras, and debugging after arranging the control module on the ground and linking the cloud for data interaction; Step 300, using the debugged binocular multi-point displacement monitoring device, a plurality of sub-regions are divided on the slope surface according to the maximum monitoring field of view of the monitoring camera, and are numbered, and then initial calibration quantities formed by image processing of all sub-region images obtained by the monitoring camera are acquired, and a two-dimensional displacement early warning threshold is set according to the initial calibration quantities; Step 400, the binocular multi-point displacement monitoring device is kept to perform monocular inspection on the slope surface according to the set period, and when monocular inspection is performed, the control module controls the monitoring camera to sequentially inspect the assigned sub-regions to acquire images, and then compares the images with the initial calibration quantities and historical detection quantities, and sends a warning to the cloud and activates the binocular inspection process when the two-dimensional displacement early warning threshold is exceeded; Step 500, after the binocular inspection process is activated, the control module controls the second displacement mechanism of the adjacent paired monitoring cameras to drive displacement synchronous calibration, and after synchronous calibration, a binocular camera unit is formed, the first displacement mechanism is controlled to drive the binocular camera unit to acquire three-dimensional data of the slope surface, and the three-dimensional data is compared with the initial calibration quantities and the historical detection quantities, and displacement information is determined and fed back to the cloud.

[0010] In combination with the second aspect, the present application provides a first implementation manner of the second aspect, and in step 100, a natural reference object is further included, the natural reference object is a fixed natural terrain feature on the slope surface; the fixed reference object includes a point reference object and a cross calibration reference object, the point reference object is a prefabricated component with a geometric center mark, and the cross calibration reference object is a prefabricated component with a cross-shaped intersection and a fixed arm length feature; when the positioning parameters are acquired by the positioning device, the geometric center three-dimensional coordinates of each point reference object, the cross-shaped intersection three-dimensional coordinates of the cross calibration reference object, and the arm length size are measured by the positioning device, the profile feature point three-dimensional coordinates of the natural reference object are recorded, and all positioning parameters and feature information are stored to the control module.

[0011] In combination with the second aspect, the present application provides a second implementation manner of the second aspect, and the debugging calibration in step 200 includes: two-dimensional calibration: the control module controls the second displacement mechanism corresponding to each monitoring camera to drive the monitoring camera to independently adjust the angle, so that each monitoring camera is aligned with a preset initial sub-region of the slope surface, an image of the calibration board in the initial sub-region is shot, a monocular intrinsic parameter of each monitoring camera is calculated and stored; three-dimensional preset calibration: based on a binocular camera unit formed by two paired monitoring cameras, a three-dimensional calibration reference is set by a preset cross calibration reference object, the control module pre-stores three-dimensional coordinates and image features of the cross calibration reference object, the second displacement mechanism is controlled to drive the binocular camera unit to move to a position aligned with the cross calibration reference object, an image of the cross calibration reference object is shot synchronously, a binocular extrinsic parameter is calculated, and the angle of the monitoring camera in the binocular camera unit is adjusted according to the stored standard binocular extrinsic parameter data to meet the binocular state requirement.

[0012] With reference to the second aspect, the present application provides a third implementation of the second aspect, in the step 300, the slope surface is divided into a plurality of sub-regions in a row-column matrix form according to the maximum monitoring field of view of the monitoring cameras, and the sub-regions are numbered in a left-to-right and top-to-bottom order, and the numbered sub-regions are divided into left-half sub-regions and right-half sub-regions, which are respectively assigned to the two monitoring cameras; the initial calibration quantity includes positioning information and size information of all the fixed reference objects and natural reference objects in each sub-region, and topographic and geomorphic spatial information of the corresponding sub-region of the slope surface, and a two-dimensional displacement early warning threshold is set according to the positioning information of the fixed reference objects and the natural reference objects in the initial calibration quantity.

[0013] With reference to the second implementation of the second aspect, the present application provides a fourth implementation of the second aspect, in the step 500, the specific process of the displacement synchronous calibration includes: controlling the second displacement mechanism to drive the adjacent paired monitoring cameras to move to align with the preset cross calibration reference object, and synchronously shooting the cross calibration reference object image; extracting the cross intersection pixel coordinates of the cross calibration reference object in the image, combining the preset three-dimensional coordinates, calculating the difference of the binocular external parameters, determining the displacement amount of the corresponding second displacement mechanism according to the difference and moving, and finally completing the calibration after the difference is reduced to meet the standard binocular external parameter data requirements; and then performing a reference test, shooting the cross calibration reference object in other sub-regions to obtain the current coordinates, and comparing the current coordinates with the coordinates of the cross calibration reference object in the initial calibration quantity, if the deviation exceeds the allowable range, the step of calibration is re-executed.

[0014] With reference to the fourth implementation of the second aspect, the present application provides a fifth implementation of the second aspect, in the step 500, the specific process of determining the displacement information includes: first determining the activation condition of the activated binocular inspection process, the activation condition includes that the displacement of a single fixed reference object in the same picture shot by a single monitoring camera exceeds the two-dimensional displacement early warning threshold, the relative positional relationship displacement of the fixed reference object in the front and back different pictures shot by the single monitoring camera exceeds the two-dimensional displacement early warning threshold, and the displacement of the fixed reference object in the pictures shot by different monitoring cameras exceeds the two-dimensional displacement early warning threshold; according to the activation condition, the corresponding sub-region pair of the two fixed reference objects exceeding the two-dimensional displacement early warning threshold is determined, the first displacement mechanism is controlled to drive the binocular camera unit to obtain the three-dimensional data of all the sub-regions on the straight line connecting the fixed reference objects of the sub-region pair, and then the displacement amount is calculated by expanding at least one adjacent sub-region outward, until the displacement range is determined after the displacement amount is less than the threshold; based on the displacement range and the displacement amount of the slope surface, the displacement type is judged according to the topographic and geomorphic changes in the three-dimensional data, and the displacement information is integrated and uploaded to the cloud.

[0015] The present application has the following beneficial effects: (1) The application effectively solves the problems of existing visual monitoring mode solidification, energy consumption and precision imbalance by designing an autonomous switching mechanism of monocular inspection and binocular precise measurement; in the daily monitoring stage, a monocular inspection mode is adopted, each monitoring camera covers the corresponding sub-region by independent control, which greatly reduces the energy consumption of the device; when the displacement is detected to be out of limit, the control module can autonomously drive the displacement mechanism to reset the monitoring camera to the preset calibration reference position, quickly complete the binocular state calibration and switch to the binocular precise measurement mode, which not only ensures the low energy consumption economy of daily monitoring, but also realizes the rapid high-precision response of abnormal displacement, avoiding missing the early warning opportunity; (2) The application improves the accuracy and integrity of displacement monitoring data by optimizing the reference system and sub-region division rule, sets point reference, cross calibration reference and natural reference, wherein the cross calibration reference serves as a binocular calibration reference and also assists in judging local displacement, solves the problem of single function and easy failure of existing reference points, and divides the sub-region according to the maximum field of view of the monitoring camera and allocates it to different cameras, avoiding monitoring blind area and repeated coverage, ensuring that the displacement data acquisition covers the whole slope, and providing a complete data basis for subsequent displacement analysis; (3) Based on the three-dimensional data obtained by binocular precise measurement, different types of single reference displacement, local displacement and overall displacement can be distinguished, solving the limitation that the existing visual monitoring can only output a single displacement; at the same time, a cooperative process of local data processing and cloud uploading is constructed, which can upload complete information including displacement range, displacement amount and displacement type to the cloud in real time, providing comprehensive data support for remote control and emergency decision-making, especially suitable for the needs of temporary slope for dynamic monitoring and rapid response; (4) The application further improves the measurement precision of binocular precise measurement by designing a calibration closed loop process; when the binocular mode is activated, it can not only be reset to the calibration reference position to complete the preliminary calibration, but also compare the current calibration data with the initial calibration amount through reference test, and if there is deviation, it will be recalibrated, forming a calibration closed loop, effectively correcting the camera pose deviation, ensuring the accuracy of three-dimensional displacement data, and avoiding monitoring errors caused by device deviation. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a plan view of the three-legged scheme adopted in the embodiment of the application; Figure 2 is an axonometric view of the three-legged scheme adopted in the embodiment of the application and in the binocular synchronous detection state; Figure 3 is an axonometric view of the three-legged scheme adopted in the embodiment of the application and in the single camera acquisition detection state; Figure 4 is a schematic view for slope monitoring in the embodiment of the application; Figure 5 is a schematic diagram of the rainfall experiment on the sandy soil slope by using the binocular multi-point displacement monitoring device in the embodiment of the present application.

[0017] In the figure: 1-fixed frame, 2-first steering platform, 3-fixed rod, 4-second steering platform, 5-monitoring camera, 6-fixed reference object. DETAILED DESCRIPTION

[0018] The present application will be further explained in conjunction with the accompanying drawings and specific embodiments.

[0019] Embodiment 1: The embodiment discloses a binocular multi-point displacement monitoring device, which can be used for displacement monitoring of target objects in multiple scenes. High-precision autonomous displacement monitoring is realized by using existing visual calculation algorithms. Not only local monitoring calculation can be performed, but also remote data uploading and early warning can be performed through Internet of Things technology. The device and the monitoring method based on the device provided by the embodiment are mainly used for slope surface displacement monitoring of large-area slopes formed in natural terrain or engineering construction, so as to avoid geological disasters or engineering hazards.

[0020] Specifically, the binocular multi-point displacement monitoring device of the embodiment comprises a fixed frame 1, a first displacement mechanism, a mounting frame, at least two second displacement mechanisms, at least two monitoring cameras 5 and a control module.

[0021] The fixed frame 1 is used to stably install the device in a stable area near the slope to be monitored. The first displacement mechanism is assembled on the fixed frame 1 and can drive the whole mounting frame to realize synchronous displacement.

[0022] The middle part of the mounting frame is connected with the first displacement mechanism. A plurality of second displacement mechanisms are installed on the mounting frame. Each second displacement mechanism loads a monitoring camera 5. Each second displacement mechanism can independently drive the corresponding monitoring camera 5 to adjust the pose.

[0023] The control module is connected with the first displacement mechanism, the second displacement mechanism and the monitoring camera 5. The control module can control the first displacement mechanism to drive all monitoring cameras 5 to move synchronously, can control a single second displacement mechanism to drive the corresponding monitoring camera 5 to adjust independently, and is responsible for data processing, mode switching and data interaction with the cloud. On the slope surface to be monitored, a fixed reference object 6 is pre-arranged. The fixed reference object 6 comprises a point reference object and a cross calibration reference object. The point reference object has a clear geometric center mark. The cross calibration reference object has a clear cross intersection and a fixed shape arm structure.

[0024] Meanwhile, the natural reference objects with characteristics on the slope surface are also collected as references. The natural reference objects are selected from the natural terrain, stone blocks, engineering structures and plants with stable position and shape on the slope surface. For the above reference objects, the spatial position information and size characteristics of all fixed reference objects 6, and the spatial position information of the contour feature points of the natural reference objects are obtained by the positioning device, and the information is stored in the control module as the baseline data for subsequent monitoring.

[0025] Based on the binocular multi-point displacement monitoring device, the embodiment also provides a method for monitoring the slope displacement by using the displacement monitoring device, and the specific process is as follows: The preliminary preparation and baseline establishment are as follows: firstly, the slope surface range of the temporary slope to be monitored is determined, the point reference objects and the cross calibration reference objects are uniformly arranged on the slope surface, and the natural reference objects meeting the requirements are marked; the spatial position and characteristic information of all reference objects are collected by the positioning device, the data is uploaded to the control module for storage, and the baseline database construction is completed.

[0026] The device installation is as follows: the fixed frame 1 is installed at the stable terrain with a suitable distance in the opposite direction of the slope surface of the slope to be monitored, the first displacement mechanism, the mounting frame, the second displacement mechanism and the monitoring camera 5 are sequentially assembled, the control module is arranged on the ground and the connection of the control module with each displacement mechanism and camera is established, and the communication configuration of the control module with the cloud is completed.

[0027] The debugging and calibration are as follows: on the one hand, the two-dimensional calibration is carried out, the control module drives each second displacement mechanism to independently adjust the angle of the corresponding monitoring camera 5, so that each camera is respectively aligned with the initial sub-region of the slope, the camera captures the image of the calibration plate in the sub-region, and the monocular parameters of each camera are calculated and stored; on the other hand, the three-dimensional preset calibration is carried out, the cross calibration reference object is taken as the baseline, the second displacement mechanism drives the two cameras to adjust to the binocular cooperative pose, the cross calibration reference object image is captured, the binocular cooperative parameters are calculated and stored, and the measurement accuracy in the binocular mode is ensured.

[0028] The sub-region division and early warning threshold setting are as follows: the device completed by debugging is used, the maximum monitoring field of view of a single monitoring camera 5 is taken as the unit, the entire slope surface to be monitored is divided into a plurality of sub-regions in the form of row-column matrix, all the sub-regions are numbered in a uniform order, and the sub-regions are reasonably allocated to the two monitoring cameras 5. In this embodiment, the sub-regions in the left half area are allocated to the left camera, and the sub-regions in the right half area are allocated to the right camera. The images of all the sub-regions are collected by the cameras, and the initial calibration quantity containing the position, size and terrain information of the reference objects in each sub-region is formed by processing, and the two-dimensional displacement early warning threshold is set based on the baseline position of the reference objects in the initial calibration quantity.

[0029] Monocular inspection and mode switching trigger, after setting and debugging, the control module controls the device to enter the monocular inspection mode, and drives the two cameras to inspect the assigned sub-regions in turn according to the preset period, and real-time collects the images of each sub-region. The control module compares the collected images with the initial calibration quantity and historical monitoring data to judge the displacement of the reference object. When any trigger condition (such as the displacement of a single camera and a certain fixed reference object 6 in the same picture exceeding the threshold, the relative displacement of the reference object in the front and back pictures of a single camera exceeding the threshold, and the displacement of the reference object in the pictures of different cameras exceeding the threshold) is monitored, the control module immediately sends a warning message to the cloud and automatically triggers the binocular inspection process.

[0030] Binocular precision measurement and displacement information determination, after the binocular inspection process is activated, the control module drives the second displacement mechanism corresponding to the two cameras to drive the cameras to move quickly to the position of the preset cross calibration reference object, completes the binocular cooperative calibration by shooting the image of the cross calibration reference object, and performs reference testing. After calibration, the two cameras are synchronously activated to form a binocular camera unit, the first displacement mechanism is driven by the control module to drive the binocular camera unit to align with the sub-region where the reference object triggering the warning is located, to collect three-dimensional data of the region and the surrounding related sub-regions first, and then gradually expand the monitoring range until the displacement is less than the threshold to determine the displacement range and the displacement. Combined with the changes in the terrain and topography in the three-dimensional data, the displacement type is judged, and finally the complete displacement information is formed by integrating the displacement range, the displacement and the displacement type, and is uploaded to the cloud for remote control and decision-making.

[0031] It should be noted that the system and method disclosed in the embodiment are mainly used for temporary displacement monitoring of the slope surface. The core adaptive scene of such temporary geological engineering monitoring task originates from the special risk properties and dynamic monitoring needs of temporary slopes in the field of geological engineering. The necessity of temporary slope monitoring task can be explained from the following three aspects: The stability risk of temporary slope caused by engineering construction disturbance drives the demand for temporary monitoring. For example, in traffic infrastructure, building foundation excavation, and mine stripping engineering, temporary slopes are mostly formed by soil excavation or site leveling during construction. Such slopes are not designed for long-term stability optimization, and the original stress state of the slope body is redistributed dramatically due to construction unloading, which is prone to unloading rebound, crack development and other phenomena. In addition, personnel and equipment are concentrated during the construction period, and dynamic monitoring needs to be carried out during the entire construction process and in the short term after completion to avoid the threat of slope instability to construction safety. Traditional long-term monitoring equipment has a long deployment period and high cost, which is difficult to adapt to the temporary needs in the construction scene.

[0032] In another case, the rock and soil of temporary slopes are often freshly excavated and have low shear strength. During periods prone to geological disasters, such as the rainy season and snowmelt, rainwater infiltration or snowmelt can cause a sudden increase in the slope's moisture content, further exacerbating the weakening of the rock and soil. Furthermore, seepage can induce internal piping and erosion, increasing the risk of landslides and collapses. Temporary slope monitoring is also driven by the need for emergency monitoring due to extreme weather and sudden geological events. When temporary slopes are exposed to extreme weather such as heavy rainfall, typhoons, and short-term severe convection, or when nearby sudden geological events such as earthquakes and landslides occur, the slopes can experience significant displacement within a short period of time. Traditional monitoring methods, if not deployed in advance, cannot detect these risks in a timely manner, and long-term monitoring equipment may lose its monitoring capabilities due to damage from the disaster. In these situations, a temporary monitoring system must be quickly established to conduct emergency monitoring of the affected temporary slopes.

[0033] In order to solve the displacement monitoring tasks of the above-mentioned slopes, the binocular multi-point monitoring equipment provided in this embodiment is such a device that can be quickly temporarily built and debugged to adapt to the displacement monitoring of a large area of ​​slope within a certain period of time.

[0034] Further, refer to Figures 1-4 The binocular multi-point displacement monitoring device in this embodiment adopts a retractable tripod structure as a fixed frame 1. This tripod itself has the characteristic of being easy to carry and transport. In the displacement monitoring task of a large-area slope, multiple fixed frames 1 and corresponding camera components can be used for monitoring at the same time. A small number of engineering personnel can transport the equipment to the site for rapid installation and debugging, and can be debugged together through terminal devices with multiple signal ports, thereby improving the layout efficiency.

[0035] In this embodiment, the first displacement mechanism and the second displacement mechanism are both a servo gimbal, that is, they have a horizontal rotation component and a pitch angle rotation component, so that the structure of their load can rotate and adjust the angle in space, thereby covering a larger area of ​​the monitoring object.

[0036] Specifically, a first servo platform 2 is provided on the fixing frame 1, and the first servo platform 2 is connected to the control module via a wiring harness. A fixing rod 3 serving as a mounting frame is fixed to the first servo platform 2 via a clamping member. The fixing rod 3 in this embodiment is a heterogeneous pipe, the cross section of the middle part is circular, and the surface has a tooth pattern, which is convenient for clamping and fixing with the clamping member. The cross section of the two side parts is rectangular or has a structure with at least one plane, which is convenient for clamping and fixing the second servo platform 4. Figure 2 and Figure 3The second steering engine cluster 4 and the monitoring camera 5 are connected with the control module through cables. The control module in the embodiment adopts a wall-mounted waterproof shell, and the shell has a protection level of IP65, which can resist the influence of rainfall, dust raising and ultraviolet aging.

[0037] With reference to Figure 5 In the embodiment, simple experimental verification is carried out on the device and method. A plastic box and a sand pile in the vicinity are used to simulate a slope surface, the box is filled with dry fine sand to form a uniform sand pile, a plurality of red circular reference points are uniformly arranged on the surface of the sand pile as artificial reference points for displacement monitoring, a water pipe with a plurality of porous nozzles is arranged above the box, is connected to an adjustable speed water pump, and the water spray rate is set to 5 L / h to ensure that the water flow uniformly covers the entire surface of the sand pile to avoid local strong erosion. The fixed frame 1 of the binocular monitoring device is adjusted to a height of 1.2 m, and the two cameras are aimed at the sand pile area; the control module completes monocular calibration and binocular calibration, and is divided into two adjacent sub-areas according to the size of the sand pile. The conclusion of the experimental verification is that the monocular inspection has a small displacement recognition error, the early warning trigger response time is qualified, there is no missed judgment or misjudgment, the sensitivity of the monocular mode to the small displacement is verified, and the low-power inspection is suitable for long-term dynamic monitoring. The three-dimensional displacement data measured by the binocular unit has a small error compared with the displacement of the sand pile reference points measured directly by the vernier caliper, the displacement range has a deviation of less than 5% from the actual concave area, and the three-dimensional measurement accuracy of the binocular mode is verified. From the monocular trigger to the completion of the binocular calibration, only 30 seconds are consumed, without manual intervention, the displacement data and visual images are received in real time on the cloud, the delay of the early warning information push is less than 2 seconds, the data integrity reaches 100%, and the cooperative effect of the device autonomous switching and remote control is verified. The device stably operates in the near-distance small target and dynamic displacement scene, the recognition success rate of the red reference points is 100%, and the adaptability of the device to the simple simulation scene is verified.

[0038] Further, the object to be monitored in the embodiment is a slope in a certain area of Gannan. The slope is located in the transition zone between plateau and basin, is a typical landslide-prone bare rock mountain terrain, has a near 45° inclination as a whole, has no tall trees and vegetation coverage, only has sporadic low shrubs and weeds distributed in rock crevices, has a coverage rate of less than 10%, has a main body of gray-black sand and mudstone interbedded structure, has a rock bare area ratio of more than 90%, and has obvious joint fissures on the rock layer surface, and small weathering and peeling traces can be seen in some areas.

[0039] The slope surface has a height of about 20 m, a horizontal length of about 35 m, and a total area of about 700 m 2The top of the slope is a temporary road for construction, and the bottom is a roadbed construction area. Due to the construction excavation, the original stress of the slope body is unbalanced, and temporary displacement monitoring needs to be carried out during construction and within 3 months after completion to prevent the slope collapse from threatening the construction personnel and equipment below. In the opposite direction of the slope to be monitored, about 55 m away from the slope surface, there is a natural bedrock platform as a fixed plane for equipment erection. The area of the bedrock platform is about 1.8 m x 2.2 m, the surface is flat, there is no loose rock or floating soil, and it is far away from gullies, cracks and other easy deformation areas. The bedrock where the platform is located is confirmed by geological investigation to be a complete block structure with no obvious displacement risk, which can meet the long-term stable erection requirements of monitoring equipment.

[0040] The fixed frame 1 selected in this embodiment has an adjustable total height when unfolded. In actual erection, the height is adjusted to 1.4 m, and the tripod bottom is fixed to the bedrock platform through expansion bolts. Then, two monitoring cameras 5 are erected, which are high-definition digital cameras with a resolution of 5 million pixels and support dynamic zoom function. The focal length can be adjusted according to the distance of the sub-monitoring area to ensure clear imaging of the reference object. They have wide dynamic and low light functions, which can adapt to environments with large differences in morning and evening light, overcast light, avoid strong light reflection or weak light caused image blur. Each camera is mounted on an independent second gimbal 4, which can be independently adjusted in horizontal and pitch angle under the drive of the control module, realizing complete coverage of the allocated sub-area. At the same time, the two monitoring cameras 5 can move synchronously under the drive of the first gimbal 2, meeting the demand of binocular collaborative monitoring. The control module has a 485 bus port and a power supply port, which are connected to the first gimbal 2, the second gimbal 4 and the monitoring camera 5 on the fixed frame 1 through waterproof cables. The cable interface is treated with waterproof sealing joint to prevent rainwater from seeping in. The control module uses a quad-core processor based on ARM Cortex-A53 architecture, which has low power consumption and data processing capability, meeting the image preprocessing in monocular inspection and disparity calculation, three-dimensional displacement vector solution in binocular mode. The processor integrates 2GB DDR4 memory and 16GB eMMC flash memory, supporting local storage of inspection images and displacement data for 30 days, and reserving a MicroSD card expansion interface for long-term continuous monitoring scenarios. The power supply is a sealed gel battery, which is composed of two 12V / 150Ah single batteries connected in series to form a 24V / 150Ah battery pack with a total energy storage capacity of 3600Wh.

[0041] Then, the specific monitoring method based on this scene is as follows: I. Equipment arrangement process (1) Preliminary site investigation and preparation On the bedrock platform 55 m away from the slope surface 55 to be monitored, the geological radar scanning confirms that there is no hidden fissure or cavity within 3 m under the platform, the total station instrument is used to measure the flatness of the platform, and the surface undulation difference is ensured to be not more than 5 cm; the center point of the platform is marked with a level instrument as a fixed frame 1 installation reference point. Then the equipment kit is unpacked and inspected on site, including a tripod fixed frame 1, a first displacement mechanism, a long strip-shaped fixed rod 3, a second displacement mechanism 2 set, 52 monitoring cameras, a control module, a 24V / 150Ah colloidal battery pack and a 300W solar panel assembly, to ensure that all parts are not damaged during transportation. Prepare auxiliary tools, carry impact drill with Φ12 mm drill bit, torque wrench, level, tape, insulating tape, waterproof sealant and other installation tools, and prepare rainproof cloth at the same time.

[0042] (II) Fixed frame 1 and displacement mechanism installation Unfold the tripod, evenly distribute the three supports around the reference point of the bedrock platform, adjust the support height to 1.4 m, observe the top mounting surface of the tripod through the level, and rotate the adjustment knob to make the mounting surface level; use the impact drill to drill holes in the corresponding bedrock position at the bottom of the support, implant expansion bolts, and use the torque wrench to tighten with a fixed torque to ensure that the tripod is rigidly connected to the bedrock and that there is no obvious shaking when applying horizontal thrust. Fix the base of the first displacement mechanism to the top mounting surface of the tripod with M10 bolts, and apply thread glue to the bolts to prevent loosening; connect the drive cable, and leave 1.5 m of redundant length for the drive cable. Test the horizontal rotation function and angle feedback accuracy. Connect the middle of the long strip-shaped fixed rod 3 to the output end of the first displacement mechanism through a detachable flange plate, and tighten the flange plate bolts in diagonal order; install one set of second displacement mechanism at a distance of 20 cm from both ends in the length direction of the fixed rod 3, to ensure a distance of 30 cm between the two mechanisms, and the center lines of the mechanism output shafts are parallel to the axis of the fixed rod 3. After installation, test the pitch angle adjustment function, the angle range is -30° to +60°, and the adjustment accuracy is 0.05°.

[0043] (III) Camera and control module deployment Two monitoring cameras 5 are mounted at the output end of the second displacement mechanism, with their lenses facing the slope. Adjust the mechanism to align the lens optical axis with the fixed rod 3. Install lens protection caps and connect waterproof cables to the cameras. Power on the cameras to test their imaging capabilities, ensuring the image is free of distortion and obstructions. Switch the resolution to 5 megapixel mode. Arrange the control module and power supply system: Install a battery protection box in a dark area of ​​the bedrock platform. Place a 24V / 150Ah battery pack inside the box and connect the positive and negative cables to the MPPT power management module. Wall-mount the control module next to the box and connect it to the first and second displacement mechanisms and the cameras via waterproof cables. Secure the solar panel bracket to an unobstructed area at the edge of the platform, adjust the inclination to 35°, connect the output cable to the MPPT module, and test the charging function. Install the 4G antenna and LoRa antenna. After powering on the control module, check the network connection. Connect the LoRa module via a handheld terminal to verify real-time image transmission.

[0044] 2. Slope reference object setting process (1) Reference object types and layout planning Point references are 30cm diameter circular precast concrete components, painted with black and white concentric rings and inlaid with a 2cm stainless steel dot in the center. Cross calibration references are cross-shaped precast concrete components with 60cm horizontal and 60cm vertical arms, 15cm wide, and 10cm thick. They are painted with red and white stripes and inlaid with a 3cm stainless steel dot at the intersection. They are arranged in three layers along the height of the slope, with one point reference placed at 5m intervals on each layer, for a total of 18 point references. One cross calibration reference is placed on the left, center, and right sides of the slope, ensuring that each cross calibration reference is no more than 8m away from surrounding point references. When selecting natural references, identify the slope's natural stability features and make circular marks around the feature points with red paint. The marking line is 5cm wide to ensure clear recognition by the camera.

[0045] (2) Installation and positioning of reference objects Fixed reference object 6 is installed. For point reference objects, holes are drilled into the slope rock, expansion bolts are inserted, and the cement structure is fixed using angle steel brackets. The cross calibration reference object is fixed in the same way. After installation, a level ruler is used to check whether the cross arm is parallel to the slope contour to ensure that the characteristic shape is not distorted. Using RTK-GPS measurement equipment, the three-dimensional coordinates of the stainless steel dot of each point reference object, the three-dimensional coordinates of the intersection of the cross calibration reference object, the coordinates of the end points of the horizontal and vertical arms, and the three-dimensional coordinates of the center of the natural reference object mark are measured in sequence. The average value of each point is recorded in the data acquisition terminal three times, and the reference object is photographed on site simultaneously. The positioning parameters are imported into the control module. The handheld terminal calls the camera to capture each reference object. The control module automatically identifies the reference object feature points, compares the measured coordinates with the image pixel coordinates, and generates an initial mapping relationship. For the cross calibration reference object, the cross arm length, angle, and other dimensional parameters are additionally stored as a reference template for binocular calibration.

[0046] (3) System linkage test First, the control module drives the second displacement mechanism, aiming the left camera at the left half of the slope and the right camera at the right half, capturing an image of the entire slope. All reference objects are checked to ensure they are within the image. If there are blind spots, the corresponding camera angles are fine-tuned and the image is retaken until all reference objects are covered. Binocular collaborative calibration is then performed, using the second displacement mechanism to adjust the two monitoring cameras 5 for synchronous calibration to form a binocular camera unit. The first displacement mechanism is controlled to rotate the fixed rod 3, causing the two cameras to synchronously align with the central cross calibration reference object to capture a stereo image pair. The control module calculates the binocular parallax and outputs the three-dimensional coordinates of the reference object, which are compared with the RTK measurement values. If there is a deviation, the second displacement mechanism is used to fine-tune the camera spacing until the accuracy requirements are met. The initial calibration quantity is uploaded to the cloud platform, which generates a digital slope model. This is compared with the actual terrain on site to confirm that the model error does not exceed 5%, completing the final verification of the equipment layout and reference object settings.

[0047] 3. Image-based Slope Area Division and Camera Responsibility Area Allocation The maximum monitoring and analysis screen of the monitoring camera 5 in this embodiment is a slope block of 30×10m. Based on the actual dimensions of the slope to be monitored, which are 20m in height and 35m in horizontal length, and the actual area size of 30m×10m corresponding to the maximum monocular monitoring field of view of the monitoring camera 5, the sub-areas are divided by combining row-column matrix and proportional segmentation to ensure that there are no blind spots on the entire slope and the data can be cross-verified.

[0048] According to the single maximum horizontal coverage range of 30 m as the benchmark along the horizontal length direction of the slope surface, a 5 m overlapping area is reserved for cross-subarea reference object correlation calibration, which is divided into two transverse units, covering the horizontal 0-30 m section and the horizontal 5-35 m section of the slope surface respectively; according to the single maximum vertical coverage range of 10 m as the benchmark along the height direction of the slope surface, it is divided into two longitudinal units without overlapping, covering the 0-10 m section of the slope surface height, i.e. the bottom area, and the 10-20 m section of the slope surface height, i.e. the top area. Four complete subareas are formed by the intersection of transverse and longitudinal units, and are named according to the rule of transverse number plus longitudinal number, i.e. the subarea of horizontal 0-30 m and height 0-10 m, the subarea of horizontal 0-30 m and height 10-20 m, the subarea of horizontal 5-35 m and height 0-10 m, and the subarea of horizontal 5-35 m and height 10-20 m. The actual area of each subarea is about 300 m 2 , and the overlapping area is about 50 m 2 .

[0049] According to the slope coverage angle difference corresponding to the subarea position and the camera installation distance of 30 cm, the allocation principle of left half area plus right half area is adopted to determine the responsible area of the two cameras. The left camera is responsible for the left half area subarea of the slope surface, i.e. the subarea of horizontal 0-30 m and height 0-10 m and the subarea of horizontal 0-30 m and height 10-20 m, covering the full height range of horizontal 0-30 m of the slope surface, focusing on monitoring 10 of the 18 point reference objects on the slope surface and 1 left cross calibration reference object, of which the 10 point reference objects include 6 in the subarea of horizontal 0-30 m and height 0-10 m and 4 in the subarea of horizontal 0-30 m and height 10-20 m, and the left cross calibration reference object is located in the subarea of horizontal 0-30 m and height 10-20 m.

[0050] The right camera is responsible for the right half area subarea of the slope surface, i.e. the subarea of horizontal 5-35 m and height 0-10 m and the subarea of horizontal 5-35 m and height 10-20 m, covering the full height range of horizontal 5-35 m of the slope surface, focusing on monitoring 8 of the 18 point reference objects on the slope surface and 2 cross calibration reference objects, of which the 8 point reference objects include 4 in the subarea of horizontal 5-35 m and height 0-10 m and 4 in the subarea of horizontal 5-35 m and height 10-20 m, and the 2 cross calibration reference objects are respectively located in the middle of the subarea of horizontal 5-35 m and height 0-10 m and the top of the subarea of horizontal 5-35 m and height 10-20 m.

[0051] For the overlapping area of the horizontal 0-30m and the height 0-10m sub-region and the horizontal 5-35m and the height 0-10m sub-region, the horizontal 0-30m and the height 10-20m sub-region and the horizontal 5-35m and the height 10-20m sub-region, the images are synchronously collected by two cameras, and the control module verifies the reference displacement data of the overlapping area by cross matching of feature points, so as to avoid misjudgment caused by the visual angle deviation of a single camera.

[0052] IV. Monocular monitoring debugging and function verification process The monocular monitoring debugging is carried out from the aspects of intrinsic parameter calibration and field of view matching, so as to ensure the monocular displacement calculation accuracy and the region coverage integrity. In the intrinsic parameter calibration stage, the control module drives the left camera to align with the preset 1m*1m size checkerboard calibration plate in the horizontal 0-30m and the height 0-10m sub-region, the checkerboard grid spacing is 10cm, five groups of calibration images at different angles are shot, and the angle adjustment range includes pitch and horizontal; the intrinsic parameters of the left camera are calculated by OpenCV algorithm, wherein the focal length is 12mm, and the intrinsic parameter data is stored in the local parameter library. The same process is used to complete the intrinsic parameter calibration of the right camera, and the right camera aligns with the checkerboard calibration plate in the horizontal 5-35m and the height 0-10m sub-region, and the focal length in the finally obtained intrinsic parameter is 12mm.

[0053] In the field of view matching debugging stage, the control module drives the two cameras to independently traverse the responsible region in the order of sub-regions, and the sub-region traversal order is the horizontal 0-30m and the height 0-10m sub-region, the horizontal 0-30m and the height 10-20m sub-region, the horizontal 5-35m and the height 0-10m sub-region, and the horizontal 5-35m and the height 10-20m sub-region. After shooting a sub-region image, the control module automatically identifies all references in the image, including point references, cross calibration references and natural references, and records the pixel coordinates of each reference; if the pixel ratio of a reference is less than 30 pixels, the pitch angle of the corresponding camera is adjusted by the sub-displacement mechanism, and the adjustment range is controlled within ±2°, until the pixel ratio of all references reaches 30 pixels or more, so as to ensure the subsequent identification accuracy.

[0054] The monocular monitoring function test verifies the system stability through three tests of full coverage, displacement simulation and periodic inspection. In the full coverage test, the two cameras continuously shoot three rounds of full slope images according to the allocated regions, the control module automatically splices the images to form a complete slope view, checks that there is no missed shooting in the four sub-regions and the overlapping area, the identification success rate of 18 point references, 3 cross calibration references and 5 natural references reaches 98% or above, and the scattered weeds are removed by the image preprocessing algorithm to eliminate interference, and there is no identification failure caused by shrub obstruction.

[0055] In the displacement simulation test, a point reference object is selected in the sub-region of 0-30m in horizontal and 0-10m in height. The reference object is moved by 2mm, 4mm and 6mm in horizontal direction respectively by a micro displacement platform with an accuracy of 0.1mm, corresponding to the non-overrun, critical overrun and overrun states. The left camera captures the image after moving, and the control module calculates the pixel coordinate change of the point reference object and converts it into the actual displacement. The actual displacements of the three tests are 2.05mm, 4.1mm and 6.08mm respectively, with an error within 0.1mm, verifying that the monocular displacement calculation accuracy meets the requirements.

[0056] In the periodic inspection test, the inspection period is set to 2 hours each time, and the control module automatically wakes up the camera to complete sub-region shooting, data processing, local storage and cloud uploading at the preset time. All inspection data are lossless, and the cloud platform can real-time view the displacement curve of each sub-region reference object. The displacement fluctuation of a single reference object within 24 hours is less than 0.5mm, and the stability of the monocular monitoring system meets the requirements.

[0057] V. Activation conditions of binocular monitoring process Based on the monocular monitoring data and combined with the risk level of temporary monitoring of rock slope, four types of binocular activation conditions are set. Any condition can trigger the binocular monitoring process. Each type of condition has clear determination criteria and applicable scenarios to ensure that no key risk points are missed.

[0058] The first type is single displacement overrun of a single reference object, including point reference object, cross calibration reference object and natural reference object. When the point reference object is triggered, the two-dimensional displacement of any single point reference object in the same sub-region image captured by a single camera reaches 3mm or more, and the displacement direction is consistent with the potential sliding direction of the slope surface. The potential sliding direction of the slope surface in this scenario is 315°. For example, the horizontal displacement of a point reference object in the image of the sub-region of 0-30m in horizontal and 10-20m in height captured by the left camera is -2.5mm, the vertical displacement is -1.8mm, the combined displacement is 3.08mm, which reaches the threshold of 3mm and the direction is consistent with the potential sliding direction of the slope surface, triggering the binocular monitoring process.

[0059] When the cross calibration reference object is triggered, the two-dimensional displacement of the cross intersection in the image of the cross calibration reference object captured by a single camera reaches 2.5mm or more, or the deflection angle of the cross arm reaches 1° or more. For example, the deflection angle of the cross horizontal arm of the cross calibration reference object in the sub-region of 5-35m in horizontal and 0-10m in height captured by the right camera is 1.2°, which is determined as local torsional displacement, triggering the binocular monitoring process.

[0060] When the natural reference object triggers, three feature points of the natural reference object are selected, the average two-dimensional displacement of the three feature points reaches 3.5 mm or above, and the displacement directions of the three feature points are consistent to exclude image noise interference, triggering the binocular monitoring process.

[0061] The second type is that the relative displacement of the reference object exceeds the limit, including the same area relative displacement and cross-area relative displacement. When the same area relative displacement triggers, the interval between the two images of the same sub-area taken by a single camera is 2 hours, and the relative displacement difference between the adjacent two point reference objects reaches 2 mm or above; for example, the left camera takes images of the 0-30 m horizontal and 0-10 m high sub-area at two different times, and the relative displacement of the two adjacent point reference objects increases from 0.5 mm to 2.6 mm, with a difference of 2 mm threshold, triggering the binocular monitoring process. When the cross-area relative displacement triggers, the displacement difference of the same reference object in the overlapping area of the front and rear images taken by the two cameras reaches 1.5 mm or above; for example, the left camera measures the displacement of a point reference object in the overlapping area as 2.2 mm, and the right camera measures the displacement of the point reference object as 3.8 mm, with a difference of 1.6 mm, which is determined as displacement data anomaly caused by single camera view angle deviation, triggering the binocular calibration and precision measurement process.

[0062] The third type is that the cumulative displacement of the reference object exceeds the limit, which is divided into short-term accumulation and long-term accumulation. When the short-term accumulation triggers, the cumulative displacement of the same reference object reaches 5 mm or above in the interval of 2 hours for three consecutive inspections within 6 hours; for example, the displacement of a point reference object in the 5-35 m horizontal and 10-20 m high sub-area in three inspections is 1.8 mm, 1.7 mm and 1.6 mm, respectively, with a cumulative amount of 5.1 mm, reaching the 5 mm threshold and the displacement direction continuously along the slope, which is determined as an acceleration deformation precursor, triggering the binocular monitoring process. When the long-term accumulation triggers, the average cumulative displacement of 3 or more reference objects in the same sub-area within 24 hours reaches 8 mm or above; for example, the 24-hour cumulative displacement of three point reference objects in the 0-30 m horizontal and 10-20 m high sub-area is 7.8 mm, 8.2 mm and 8.5 mm, respectively, with an average cumulative displacement of 8.17 mm, reaching the 8 mm threshold, which is determined as regional overall deformation, triggering the binocular monitoring process.

[0063] The fourth type is an abnormal displacement trend trigger, including displacement rate mutation and displacement direction mutation. When the displacement rate mutation trigger occurs, the displacement rate of a single reference object, i.e., the ratio of the displacement difference between two adjacent inspections and the interval time, suddenly increases from a stable value of 0.5 mm per hour and below to 1.5 mm per hour and above; for example, the displacement rate of a certain point reference object suddenly increases from 0.4 mm per hour to 1.6 mm per hour, which is determined as deformation acceleration, triggering the binocular monitoring process. When the displacement direction mutation trigger occurs, the displacement direction of the same reference object deviates from the random fluctuation of 15° and below to the fixed direction, the direction deviation is 5° and below in two consecutive inspections, and the direction is consistent with the slope joint crack trend, which is 45° in this scenario. It is determined as potential slip controlled by the structure surface, triggering the binocular monitoring process.

[0064] Six, displacement type definition In combination with the geological background of the sand and mud interbedded rock slope, the defined four types of typical geological displacements are based on the displacement direction, range and correlation data of the reference objects obtained by monocular monitoring. Each type of displacement has a clear geological origin, displacement performance and risk level, which meets the actual needs of temporary monitoring of rock slopes.

[0065] The first type is block slip, also known as rock block slip type. Its geological origin is controlled by the slope surface structure surface, including bedding and joint. A single or multiple block rock blocks with a size of 2 m and above in the slope body slip along the structure surface, and the slip surface is mostly flat rock bedding. In this scenario, the sand and mud interbedding surface is 35°. In terms of displacement performance, 3-5 adjacent point references in the sub-region occur synchronous displacement in the same direction, the displacement direction is consistent with the structure surface trend, which is 315° in this scenario, and the displacement range is 5-20 mm. The relative displacement difference between adjacent reference objects is controlled to be 2 mm and below. The cross calibration reference object has no obvious deflection, and the deflection angle is less than 0.5°. The angular displacement of the block rock in the natural reference object is consistent with that of the point reference object, and there is no local protrusion or depression phenomenon. The risk level of this type of displacement is medium, and the sliding rock block is easy to cause local collapse, and the depth of the sliding surface needs to be determined by binocular precise measurement to evaluate the subsequent risk.

[0066] The second is wedge slip, also known as double structural plane control type. The geological origin is that two groups of intersecting structural planes cut to form a wedge. The structural planes include joints. In this scenario, the strike of the two groups of joints is 30° and 330°, and the dip angle is 60° and 55°, respectively. The wedge slips along the intersection line direction of the two groups of structural planes. It is commonly seen in the middle and upper sub-regions of sand and mud interbedded slope, i.e. the sub-regions with a horizontal distance of 0-30 m and a height of 10-20 m and the sub-regions with a horizontal distance of 5-35 m and a height of 10-20 m. The displacement is the displacement of the reference object in the wedge along the intersection line direction. In this scenario, the intersection line direction is 285°, and the displacement range is 8-30 mm. The closer to the front edge of the wedge, i.e. the lower part of the slope surface, the greater the displacement. For example, when the displacement of the reference object at the rear edge of the wedge is 10 mm, the displacement of the reference object at the front edge can reach 25 mm. The cross calibration reference object appears slight deflection, and the deflection angle range is 1-2°. The edge of the wedge in the natural reference object is concave in the displacement direction, which is consistent with the wedge slip direction. The risk level of this type of displacement is high. The wedge slip is easy to cause medium-scale landslide, so emergency monitoring and timely warning are needed.

[0067] The third is tilting deformation, also known as rock layer tilting type. The geological origin is that the rock layer direction is opposite to the slope direction. In this scenario, the rock layer direction is 135°, and the slope direction is 315°. Under the influence of gravity or rainwater infiltration, the rock layer tilts along the layer surface. It is commonly seen in the thin layer of sand and mud in the top sub-region of the slope, i.e. the sub-regions with a horizontal distance of 0-30 m and a height of 10-20 m and the sub-regions with a horizontal distance of 5-35 m and a height of 10-20 m. The displacement is the outward tilting displacement of the reference object in the top sub-region. The horizontal displacement away from the slope is 3-8 mm, and the vertical displacement upward is 1-3 mm. The transverse arm of the cross calibration reference object deflects by an angle of 2-3°, showing the characteristics of the outer side upward. The rock outcrop in the natural reference object shows the characteristics of forward inclination, and the displacement direction is perpendicular to the rock layer strike. The risk level of this type of displacement is medium. The tilting rock layer is easy to cause surface peeling. If not monitored in time, it may gradually develop into overall slip.

[0068] The fourth type is surface peeling, also known as weathering peeling type, and the geological origin is that the slope surface rock is affected by weathering, including temperature difference and rainwater erosion, and the surface rock block with a thickness of less than 0.5 m is peeled from the parent rock, which mainly occurs on the surface of weathered sandstone and mudstone in the middle and lower parts of the slope, that is, the sub-regions with a horizontal distance of 0-30 m and a height of 0-10 m and the sub-regions with a horizontal distance of 5-35 m and a height of 0-10 m, and there is no obvious sliding surface. The displacement is characterized by independent displacement of single or scattered point references, and the displacement range is 2-5 mm. The displacement directions of adjacent references are irrelevant, for example, when a point reference is horizontally displaced by +1.5 mm, the horizontal displacement of the adjacent point reference may be -2 mm. The cross calibration reference has no displacement, and the deflection angle is less than 0.5°. The displacement of the surface weathered rock in the natural reference is less than 3 mm, which is mainly point displacement caused by local peeling. The risk level of this type of displacement is low, only the surface rock peeling occurs, and there is no overall landslide risk. Regular tracking and monitoring are needed to observe whether there is a development trend.

[0069] Seven, determination process of displacement area, displacement amount and displacement type when switching binoculars When switching the binocular monitoring process, the current deformation state of the slope is determined through four steps of core overrun area positioning, displacement range determination, displacement amount calculation and displacement type determination, which provides accurate direction for subsequent binocular precision measurement. The whole process is based on monocular monitoring data and does not need to rely on three-dimensional terrain model.

[0070] In the core overrun area positioning stage, the control module extracts the overrun references that trigger the activation conditions from the monocular monitoring data, such as the left camera monitoring a point reference in the horizontal 0-30 m and height 0-10 m sub-region with a displacement of 6.08 mm, and the right camera monitoring a point reference in the horizontal 5-35 m and height 0-10 m sub-region with a displacement of 5.2 mm, and locking these two point references as core overrun references. According to the sub-regions where the core overrun references are located, the initial core monitoring area is determined as the horizontal 0-30 m and height 0-10 m sub-region and the horizontal 5-35 m and height 0-10 m sub-region where the two core overrun references are located, that is, the area with a horizontal distance of 0-35 m and a height of 0-10 m at the bottom of the slope, covering the two sub-regions, and the initial estimated area is about 600 m 2 .

[0071] The displacement range, i.e. the displacement area determination stage, adopts a combination of boundary expansion detection and boundary verification. In the boundary expansion detection, the control module drives the two cameras to expand an adjacent sub-region outside the core region, i.e. a sub-region of 0-30 m horizontally and 10-20 m high and a sub-region of 5-35 m horizontally and 10-20 m high, to shoot the images of the expanded region and calculate the displacement of the reference objects in the region. If the displacement of the reference objects in the expanded region is less than 3 mm threshold, for example, the displacement of a reference object at a point in the sub-region of 0-30 m horizontally and 10-20 m high is 2.1 mm, and the displacement of a reference object at a point in the sub-region of 5-35 m horizontally and 10-20 m high is 2.8 mm, it is determined that the displacement boundary is the initial core region, and the displacement area is calculated by subtracting the overlapping area from the areas of the two sub-regions in the core region, i.e. 300 m2+ 300 m2- 50 m2= 550 m2. In the boundary verification, the reference objects at the displacement boundary are detected, for example, the point reference objects at the intersection of the sub-region of 0-30 m horizontally and 0-10 m high and the sub-region of 0-30 m horizontally and 10-20 m high, and it is confirmed that the displacement is 2.2 mm, which is less than the 3 mm threshold, and there is no cross-boundary overrun phenomenon, and finally the displacement area is determined to be 550 m 2 .

[0072] In the displacement calculation stage, representative reference objects and boundary reference objects in the core region are selected to ensure that the data covers the entire displacement region. Five representative reference objects are selected in the core region, including three point reference objects, one cross calibration reference object and one natural reference object. The displacement values of these reference objects in the monocular monitoring data are extracted, for example, the displacements of the three point reference objects are 6.08 mm, 5.2 mm and 4.8 mm, the displacement of the cross calibration reference object is 3.5 mm, and the displacement of the natural reference object is 4.2 mm. The average displacement of these reference objects is calculated to be 4.756 mm. At the same time, the displacement of the boundary reference object is detected, for example, the displacements of the two point reference objects at the displacement boundary are 2.2 mm and 2.8 mm, and the maximum displacement of the boundary reference object is taken as the boundary displacement of the displacement region, which is 2.8 mm. Finally, the core displacement range of the displacement region is determined to be 4.756 mm±0.5 mm, and the boundary displacement is 2.8 mm.

[0073] Displacement type determination stage, combined with displacement performance and geological background analysis. First, analyze the displacement direction and correlation of the reference objects in the core area. The displacement direction of the three point references and one cross calibration reference in the core area is 315°, which is consistent with the potential sliding direction of the slope surface, and the relative displacement difference between adjacent references is less than 2mm, showing synchronous displacement characteristics. The angular displacement direction of the blocky rock in the natural reference is consistent with that of the point reference, without local protrusions or depressions, which is consistent with the displacement performance of block sliding. Combined with the geological background of the slope, the core area is located in the middle and lower part of the slope surface, corresponding to the flat rock layer of sand and mudstone interbedding, without the characteristics of wedge-shaped body formed by the intersection of two groups of structural planes, and without the opposite conditions of rock layer inclination and slope direction required for rock layer dumping. The three types of wedge sliding, dumping deformation and surface peeling are excluded. Finally, it is determined that the displacement type is block sliding, the risk level is medium, and the depth of the sliding surface and the sliding amount need to be further determined by subsequent binocular precise measurement to provide basis for slope safety control.

[0074] Eight, under monocular state, activate the rear camera fast reset and reference alignment After activating the binocular monitoring process, the control module first starts the reference alignment program, taking the pre-stored cross calibration reference as the core positioning reference, to realize the fast reset of the two monitoring cameras 5 at any patrol angle. The control module retrieves the preset cross calibration reference information from the local parameter library, including its three-dimensional coordinates, image features, and reads the real-time angle data of the current two cameras. No matter what angle the camera is at during the monocular patrol stage, the control module will calculate the angle adjustment amount of each camera from the current angle to the alignment cross calibration reference.

[0075] Subsequently, the control module drives the second displacement mechanism corresponding to each of the two cameras to independently adjust the angle in the order of "horizontal first and then pitch": first, through horizontal adjustment, the lens optical axes of the two cameras are preliminarily pointed to the sub-area where the cross calibration reference is located; then through pitch adjustment, combined with real-time image feedback - the control module acquires an image taken by the camera every 0.5 seconds, automatically identifies the features of the cross calibration reference in the image, if it is not identified, continue to adjust the angle, if it is identified, further optimize the angle, until the cross calibration reference is located in the center area of the image of the two cameras, completing the preliminary reset. The entire reset process does not require manual intervention, based on the linkage of pre-stored reference and real-time image recognition, ensuring that the alignment can be completed within 30 seconds at any initial angle, laying the foundation for subsequent calibration.

[0076] Nine, precise calibration process of binocular unit After preliminary reset, enter the calibration stage of the binocular unit, the core of which is to make the two cameras meet the pose requirements of standard binocular vision through external parameter calibration and pose fine adjustment, which is specifically three steps: The first step is external parameter calibration. The control module drives the two cameras to synchronously capture the aligned cross calibration reference image, extracts the pixel coordinates of the cross intersection in the two images, and calculates the external parameters of the two cameras based on the known three-dimensional coordinates of the cross calibration reference, including the relative rotation matrix and translation vector. The actual external parameters calculated are compared with the pre-stored standard binocular external parameters to obtain the external parameter deviation, wherein the rotation deviation needs to be controlled within 0.5°, and the translation deviation needs to be controlled within 1 mm.

[0077] The second step is pose fine-tuning. If the external parameter deviation exceeds the allowed range, the control module drives the second displacement mechanism to make targeted adjustment according to the deviation type: if there is a rotation deviation, the pitch angle or horizontal angle of the corresponding camera is fine-tuned, and the image is recaptured and the external parameters are recalculated after each adjustment of 0.1°, until the rotation deviation is not more than 0.5°; if there is a translation deviation, the horizontal position of the camera on the mounting frame is fine-tuned through the second displacement mechanism, and the calibration is repeated after each adjustment of 0.5 mm, until the translation deviation is not more than 1 mm.

[0078] The third step is calibration verification. After fine-tuning, the control module drives the binocular unit to turn to another preset cross calibration reference, and repeats the above external parameter calibration step to verify whether the external parameter deviations of the two calibrations are within the allowed range. If both calibration deviations meet the standards, it is determined that the binocular unit calibration is complete, forming a stable binocular measurement unit; if the second calibration deviation is out of standard, the first step to the second step is re-executed until the two calibration results are consistent, ensuring the calibration accuracy of the binocular unit and avoiding the three-dimensional measurement deviation caused by single calibration error.

[0079] Ten, binocular three-dimensional terrain acquisition process After the binocular unit calibration is completed, the binocular three-dimensional terrain acquisition stage is entered, the core of which is to acquire the three-dimensional terrain data of the target displacement area through stereo image acquisition, disparity calculation and three-dimensional coordinate solution, which is divided into four steps: The first step is target area positioning and scanning path planning. The control module determines the displacement-related sub-area based on the monocular monitoring in the previous stage, including a single sub-area and the case of two sub-areas. Then the scanning path of the binocular unit is planned, taking the geometric center of the displacement sub-area as the starting point, and dividing the scanning sub-area according to the snake-shaped traversal rule to avoid monitoring blind area. The control module drives the first displacement mechanism to drive the binocular unit to move to each scanning sub-area according to the planned path, and keeps the pose of the binocular unit stable after each movement to ensure that the two cameras are synchronized and aligned with the current scanning sub-area.

[0080] The second step is stereo image pair acquisition. In each scanning sub-area, the control module sends a synchronous shooting instruction to the two cameras to ensure that the shooting time difference of the two images is less than or equal to 1 ms, and a set of stereo image pairs is acquired. For the key area in the displacement area, 2-3 additional shootings are performed to obtain multiple sets of stereo image pairs for subsequent data verification and to improve the reliability of three-dimensional data.

[0081] The third step is parallax calculation and three-dimensional coordinate calculation. The control module pre-processes each set of stereo image pairs: first, the corresponding feature points of the two images are constrained on the same horizontal line through the epipolar correction algorithm to eliminate matching interference caused by image distortion and angle difference; then, a stereo matching algorithm is used to perform pixel-level feature matching on the corrected images to find the corresponding pixel positions of the same terrain feature points in the two images, and the pixel difference, i.e., the parallax, is calculated.

[0082] Combined with the internal and external parameters obtained through the previous calibration, the three-dimensional measurement principle of binocular vision is used to convert the parallax of each feature point into three-dimensional space coordinates according to the mapping relationship between parallax and actual distance, i.e., the larger the parallax, the closer the feature point to the camera; the smaller the parallax, the farther the feature point to the camera. A set of discrete three-dimensional feature point coordinates in the displacement area is obtained.

[0083] The fourth step is three-dimensional terrain data integration. The control module splices the three-dimensional feature point coordinates of all scanning sub-areas according to the spatial position, removes duplicate points and abnormal points, and forms a three-dimensional point cloud model of the displacement area. Then, through point cloud gridding processing, the discrete points are connected into a continuous three-dimensional terrain surface, and the three-dimensional coordinate information of all reference objects is retained. Finally, a complete binocular three-dimensional terrain data is formed, which contains not only the overall relief form of the terrain but also the accurate three-dimensional displacement information of the key reference objects, providing a basis for subsequent displacement depth analysis and risk assessment.

[0084] In one implementation process, after the calibration of the binocular unit is completed, the control module first retrieves the information of the two interval sub-areas locked in the monocular monitoring stage. The two sub-areas are non-adjacent sub-areas with displacement exceeding the standard in the monocular monitoring stage. The geometric center coordinates of the two sub-areas are extracted, and the initial monitoring band is determined with the line connecting the two points as the axis. The monitoring band includes a range of half the width of each sub-area on both sides of the line, covering 3-4 consecutive sub-areas to ensure complete coverage of the potential associated deformation area of the two interval sub-areas.

[0085] The control module drives the first displacement mechanism to drive the binocular unit to perform three-dimensional scanning on the sub-areas in the initial monitoring band in order from one end sub-area to the other end sub-area.

[0086] Synchronously collect 2-3 sets of stereo image pairs for each sub-region scanned, and obtain the three-dimensional coordinates of all feature points in the sub-region through epipolar rectification and stereo matching. Compare the three-dimensional coordinates with those in the initial calibration, and calculate the displacement of each feature point. If the displacement of more than 80% of the feature points in the initial monitoring zone exceeds 3mm, and the displacement direction deviates from the displacement direction of the two interval sub-regions by no more than 15°, it is determined that there is continuous displacement in the connecting region, and the subsequent range expansion stage is entered; if the proportion of feature points that meet the displacement requirement is less than 80%, it is determined that the two interval sub-regions are independent displacements, and the range expansion is carried out for each sub-region respectively.

[0087] In the displacement range expansion scanning and boundary locking process, the gradient expansion method is used to expand the scanning from the initial monitoring zone to each sub-region in turn. The expansion sequence is to preferentially expand along the dominant direction of the displacement of the two interval sub-regions, and then to expand vertically. The specific process is as follows: One-way expansion verification takes a boundary sub-region of the initial monitoring zone as the starting point, drives the binocular unit to move to the adjacent sub-region, scans and calculates the displacement of the feature points in the sub-region. If more than 60% of the feature points in the sub-region have a displacement greater than 3mm, it is determined that the displacement range extends in that direction, and the next sub-region is expanded; if the proportion of feature points that meet the displacement requirement is less than 60%, it is determined that the direction is the displacement boundary, and the expansion in that direction is stopped.

[0088] Full-direction boundary locking completes the expansion scanning in the horizontal left, horizontal right, vertical up and vertical down directions of the initial monitoring zone in turn according to the above logic, until all directions meet the boundary sub-region with a proportion of feature points that meet the displacement requirement less than 60%. All sub-regions that meet the displacement requirement are integrated to form a closed displacement range. The geometric boundary of the displacement range is calculated through the pre-stored coordinates of the sub-regions, and the actual area is converted. For example, in the present scenario, the final locked displacement range covers 5 sub-regions, and the actual area is about 650m 2 At the same time, the geometric center of the displacement range is marked as the displacement core region.

[0089] Based on the three-dimensional feature point data of all sub-regions within the locked displacement range, quantitative analysis of the displacement direction and displacement is carried out.

[0090] First, the displacement direction is determined. The three-dimensional displacement vectors of all feature points in the displacement range are extracted, and the actual displacement direction of each feature point is calculated through vector synthesis. The displacement direction of all feature points is counted, and the direction with the highest frequency is taken as the dominant displacement direction. If the direction consistency of the feature points in the displacement core region is greater than 90%, and the direction of the boundary region diverges slightly outward, it is determined to be a one-way displacement; if there are two dominant directions, it is determined whether it is a composite displacement according to the geological structure.

[0091] The displacement amount is graded and calculated, and the displacement range is divided into three parts: the core area, which is the geometric center of the displacement range and has an area of 1 / 3 of the total area, the average value of the displacement amount of all feature points is calculated as the core displacement amount; the transition area, which is 1 / 3 of the area from the outside of the core area to the inside of the boundary area, the median of the displacement amount is calculated as the transition displacement amount; the boundary area, which is 1 / 3 of the area of the edge of the displacement range, the maximum and minimum values of the displacement amount are taken as the boundary displacement amount range; and the position of the feature point with the largest displacement amount is recorded as the subsequent key tracking object.

[0092] In combination with the quantitative data of the displacement range, direction and amount, and the geological background of the slope, the displacement type is determined by a multi-dimensional matching method: block sliding determination, if the displacement range is a regular rectangle / block, the dominant displacement direction is consistent with the inclination of the dominant structural plane of the slope surface, the relative displacement difference of the feature points in the core area is less than or equal to 2mm, and there is no obvious gradient change of large front edge displacement and small rear edge displacement, it is determined as block sliding. Wedge sliding determination, if the displacement range is a triangle, the dominant displacement direction is consistent with the intersection direction of the two joints, and the displacement amount presents a gradient difference of large front edge and small rear edge, and the cross calibration reference appears 1-2° deflection, it is determined as wedge sliding. Dump deformation determination, if the displacement range is concentrated in the top sub-area of the slope surface, the dominant displacement direction is the horizontal direction away from the slope surface, the feature points are vertically displaced upward, and the cross calibration reference is tilted upward on the outside of the horizontal arm, it is determined as dump deformation. Surface peeling determination, if the displacement range is scattered, the displacement direction of the feature points has no relevance, the displacement amount is generally less than 5mm, and is concentrated in the weathered rock layer area in the middle and lower parts of the slope surface, it is determined as surface peeling. The control module integrates all monitoring results in the form of structured data and visualized data, and uploads to the cloud platform through a preset wireless communication protocol.

[0093] The present application is not limited to the above-mentioned optional embodiments, and anyone can derive other various forms of products under the inspiration of the present application. The above specific embodiments should not be understood as limiting the protection scope of the present application, and the protection scope of the present application should be defined by the claims, and the specification can be used to explain the claims.

Claims

1. Binocular multi-point displacement monitoring equipment for temporary slope displacement monitoring, based on image recognition of reference object position comparison on time axis to obtain displacement, characterized by: The device comprises a fixed frame (1) and a first displacement mechanism arranged on the fixed frame (1), and a mounting frame is arranged on the first displacement mechanism, and at least two second displacement mechanisms independently controlled are arranged on the mounting frame, and each second displacement mechanism is loaded with a monitoring camera (5); The device further comprises a control module connected with the first displacement mechanism and the second displacement mechanisms, and the control module controls the first displacement mechanism to drive all the monitoring cameras (5) to perform synchronous displacement, or controls each second displacement mechanism to drive the monitoring camera (5) to perform independent displacement.

2. The binocular multi-point displacement monitoring device according to claim 1, characterized in that: The fixed frame (1) is a tripod structure fixed on the ground, and the first displacement mechanism and the second displacement mechanisms are both two-axis rotating displacement mechanisms.

3. The binocular multi-point displacement monitoring device according to claim 1, wherein: The mounting frame is a fixed rod (3) in a strip shape, and the middle part of the fixed rod (3) is fixed to the first displacement mechanism through a detachable mechanism, and the second displacement mechanisms are arranged on the fixed rod (3) in the length direction through detachable mechanisms. The binocular multi-point displacement monitoring device is used for monitoring the displacement of a slope surface within a range of 20-100 m from the fixed frame (1), and the spacing between adjacent second displacement mechanisms arranged on the fixed rod (3) is 25-40 cm.

4. A binocular multi-point displacement monitoring method, characterized in that: The binocular multi-point displacement monitoring device according to any one of the above claims 1-3 is used for monitoring the displacement of a slope, and the specific steps are as follows: Step 100, the range of the slope surface to be monitored is determined, a plurality of fixed reference objects (6) are arranged on the slope surface, and positioning parameters of the arranged fixed reference objects (6) are obtained through a positioning device; Step 200, the binocular multi-point displacement monitoring device is arranged on a stable terrain within a range of 50-200 m of the slope surface, the fixed frame (1), the first displacement mechanism, the mounting frame, at least two second displacement mechanisms and corresponding monitoring cameras (5) are arranged, and after a control module is arranged on the ground, debugging is performed and data interaction with the cloud is linked; Step 300, the binocular multi-point displacement monitoring device is used to divide a plurality of sub-regions of the slope surface and number them according to the maximum monitoring field of view of the monitoring cameras (5), and then initial calibration quantities are obtained by processing images of all the sub-regions by the monitoring cameras (5), and two-dimensional displacement early warning thresholds are set according to the initial calibration quantities; Step 400, the binocular multi-point displacement monitoring device is kept to perform monocular inspection on the slope surface according to a set period, when the monocular inspection is performed, the control module controls the monitoring cameras to perform inspection on the assigned sub-regions in sequence to obtain images, and then the images are compared with the initial calibration quantities and historical detection quantities, and when the two-dimensional displacement early warning thresholds are exceeded, an early warning is sent to the cloud and a binocular inspection process is activated; Step 500, after the binocular inspection process is activated, the control module controls the second displacement mechanisms of the adjacent paired monitoring cameras to perform displacement synchronous calibration, and after the synchronous calibration, a binocular camera unit is formed, the first displacement mechanism is controlled to drive the binocular camera unit to obtain three-dimensional data of the slope surface, and the three-dimensional data are compared with the initial calibration quantities and the historical detection quantities, and displacement information is determined and fed back to the cloud.

5. The binocular multi-point displacement monitoring method according to claim 4, characterized in that: In step 100, a natural reference object is further included, and the natural reference object is a natural terrain feature fixed on the slope surface. The fixed reference (6) includes a point reference and a cross calibration reference, the point reference is a prefabricated component with a geometric center mark, and the cross calibration reference is a prefabricated component with a cross-shaped intersection and a fixed arm length feature; When obtaining the positioning parameters through the positioning device, the geometric center three-dimensional coordinates of each point reference, the cross intersection three-dimensional coordinates of the cross calibration reference, and the arm length size are measured through the positioning device, the profile feature point three-dimensional coordinates of the natural reference are recorded, and all the positioning parameters and feature information are stored to the control module.

6. The binocular multi-point displacement monitoring method according to claim 4, characterized in that: The debugging after the control module is arranged on the ground in step 200 includes: Two-dimensional calibration: the control module controls the second displacement mechanism corresponding to each monitoring camera to independently adjust the angle of the monitoring camera, so that each monitoring camera is aligned with a preset initial sub-region of the slope surface, an image of the calibration plate in the initial sub-region is shot, the monocular intrinsic parameter of each monitoring camera is calculated and stored; Three-dimensional preset calibration: based on a binocular camera unit formed by two paired monitoring cameras, the three-dimensional calibration reference is preset through the cross calibration reference, the control module pre-stores the three-dimensional coordinates and image features of the cross calibration reference, the second displacement mechanism is controlled to move the binocular camera unit to a position aligned with the cross calibration reference, the image of the cross calibration reference is shot synchronously, the binocular extrinsic parameter is calculated, and the angle of the monitoring camera in the binocular camera unit is adjusted according to the stored standard binocular extrinsic parameter data to meet the binocular state requirement.

7. The binocular multi-point displacement monitoring method according to claim 4, characterized in that: In step 300, when the sub-regions are divided and numbered, the slope surface is divided into a plurality of sub-regions in a row-column matrix form according to the maximum monitoring field of view of the monitoring camera, all the sub-regions are numbered in the order from left to right and from top to bottom, the numbered sub-regions are divided into left half sub-regions and right half sub-regions, and the left half sub-regions and the right half sub-regions are respectively assigned to the two monitoring cameras; The initial calibration quantity includes the positioning information, size information of all the fixed references (6) and natural references in each sub-region, and the topographic and geomorphic spatial information of the corresponding sub-region of the slope surface, and the two-dimensional displacement early warning threshold is set according to the positioning information of the fixed references (6) and the natural references in the initial calibration quantity.

8. The binocular multi-point displacement monitoring method according to claim 6, characterized in that: In step 500, the specific process of displacement synchronous calibration includes: The second displacement mechanism is controlled to move the adjacent paired monitoring cameras to align with the preset cross calibration reference, and the image of the cross calibration reference is shot synchronously; The cross intersection pixel coordinates of the cross calibration reference in the image are extracted, the binocular extrinsic parameter difference is calculated by combining the preset three-dimensional coordinates, the displacement amount of the corresponding second displacement mechanism is determined according to the difference value and moved, and finally the calibration is completed after the difference value is reduced to meet the standard binocular extrinsic parameter data requirement; Reference test is performed again, the cross calibration reference in other sub-regions is shot to obtain the current coordinates, which are compared with the coordinates of the cross calibration reference in the initial calibration quantity, if the deviation exceeds the allowed range, the calibration is re-executed.

9. The binocular multi-point displacement monitoring method according to claim 8, characterized in that: In step 500, the specific process of determining the displacement information includes: The activation condition of activating the binocular inspection process is determined first, and the activation condition includes that the displacement of a single fixed reference object (6) in the same picture taken by a single monitoring camera exceeds a two-dimensional displacement warning threshold, the displacement of the relative position relationship of the fixed reference object (6) in the front and back different pictures taken by the single monitoring camera exceeds the two-dimensional displacement warning threshold, and the displacement of the fixed reference object (6) in the pictures taken by different monitoring cameras exceeds the two-dimensional displacement warning threshold; According to the activation condition, two fixed reference objects (6) exceeding the two-dimensional displacement warning threshold are determined to determine the corresponding sub-region pair, the first displacement mechanism drives the binocular camera unit to obtain the three-dimensional data of all sub-regions on the straight line connecting the fixed reference objects (6) of the sub-region pair, and then at least one adjacent sub-region is expanded outward to calculate the displacement, and the displacement range is determined after the displacement is less than the threshold value; Based on the displacement range and the displacement of the slope surface, the displacement type is judged combined with the topographic and geomorphic changes in the three-dimensional data, and the displacement information is integrated and uploaded to the cloud.

Images