Binocular multi-point displacement monitoring device and method thereof

Abstract

The application discloses binocular multi-point displacement monitoring equipment and a method thereof, and aims at temporary slope displacement monitoring, and obtains a displacement amount based on image recognition of position comparison of a reference object on a time axis, and comprises a fixing frame and a first displacement mechanism arranged on the fixing frame, a mounting frame is arranged on the first displacement mechanism, 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 method further comprises a control module, all monitoring cameras are controlled to synchronously displace by connecting the first displacement mechanism to the control module, and each monitoring camera is controlled to independently displace by connecting the second displacement mechanism to the control module. The method in the application utilizes two free monitoring cameras of the equipment to realize the displacement monitoring effect of monocular inspection and binocular displacement type and displacement range. The application 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 the ordinary plane visual monitoring has low energy consumption, two-dimensional displacement data can be obtained by comparing the positions of the reference points, but it cannot distinguish the depth direction and actual deflection trend of the 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 the measurement accuracy, the energy consumption is significantly higher than that of monocular monitoring, and the working mode needs to be set manually throughout the actual application. 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 of daily monitoring or lag in abnormal displacement response.

[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 of displacement 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 can only determine the possible displacement, but cannot determine the type and area of displacement, and cannot 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:

[0007] 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 fixing frame and a first displacement mechanism arranged on the fixing frame, a mounting frame is arranged on the first displacement mechanism, at least two second displacement mechanisms for independent displacement control are arranged on the mounting frame, and each second displacement mechanism carries a monitoring camera; a control module connected to the first displacement mechanism and the second displacement mechanisms 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.

[0008] In combination with the first aspect, the application provides a first implementation manner of the first aspect, the fixing frame is a tripod structure fixed to the ground, and the first displacement mechanism and the second displacement mechanism are both two-axis rotating movement mechanisms.

[0009] 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 mechanisms are arranged on the length direction of the fixed rod through detachable mechanisms; for slope surface displacement monitoring within a range of 20-100 m from the fixing frame, the spacing between adjacent second displacement mechanisms arranged on the fixed rod is 25-40 cm.

[0010] In a second aspect, the application further discloses a monitoring method, which uses the binocular multi-point displacement monitoring device of any one of the above aspects for slope displacement monitoring, and the specific steps are as follows:

[0011] 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;

[0012] 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 fixing frame, the first displacement mechanism, the mounting frame, the at least two second displacement mechanisms, and the corresponding monitoring cameras, and debugging after arranging the control module on the ground and linking the cloud for data interaction;

[0013] 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 are obtained by processing images of all sub-regions obtained by the monitoring camera, and a two-dimensional displacement early warning threshold is set according to the initial calibration quantities;

[0014] 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 obtain 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;

[0015] 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 obtain 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.

[0016] 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 obtained 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 in the control module.

[0017] In combination with the second aspect, the present application provides a second implementation manner of the second aspect, and the debugging and 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 photographed, 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 synchronously photographed, 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.

[0018] In combination with 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 all 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.

[0019] In combination with 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 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; then performing a reference test, shooting the cross calibration reference object in other sub-regions to obtain the current coordinates, and comparing with the coordinates of the cross calibration reference object in the initial calibration quantity, if the deviation exceeds the allowed range, the step of calibration is re-executed.

[0020] In combination with 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 a 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 is calculated by expanding at least one adjacent sub-region outward, until the displacement range is determined after the displacement is less than the threshold; based on the displacement range and the displacement 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.

[0021] The present application has the following beneficial effects:

[0022] (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.

[0023] (2) The application improves the accuracy and integrity of displacement monitoring data by optimizing the reference system and sub-region division rule. Point reference, cross calibration reference and natural reference are set, the cross calibration reference serves as a binocular calibration reference and can also assist in judging local displacement, solving the problem of single function and easy failure of existing reference points. At the same time, the sub-regions are divided according to the maximum field of view of the monitoring camera and distributed to different cameras, avoiding monitoring blind area and repeated coverage, ensuring that the displacement data collection covers the whole slope, and providing a complete data basis for subsequent displacement analysis.

[0024] (3) Based on the three-dimensional data obtained by binocular precise measurement, the application can distinguish different types of displacement such as single reference displacement, local displacement and overall displacement, solving the limitation of existing visual monitoring that can only output a single displacement value. At the same time, a collaborative process of local data processing and cloud uploading is constructed, which can upload complete information including displacement range, displacement value and displacement type to the cloud in real time, providing comprehensive data support for remote control and emergency decision-making, especially suitable for temporary slopes that require dynamic monitoring and rapid response.

[0025] (4) The application further improves the measurement accuracy of binocular precise measurement by designing a calibration closed loop process. When the binocular mode is activated, it can not only reset to the calibration reference position to complete the preliminary calibration, but also compare the current calibration data with the initial calibration value through reference test. If there is a deviation, it will be recalibrated to form 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

[0026] Figure 1 is a plan view of the tripod scheme adopted in the embodiment of the application;

[0027] Figure 2 is an axonometric view of the tripod scheme adopted in the embodiment of the application and in the binocular synchronous detection state;

[0028] Figure 3is the axial measurement schematic diagram of the tripod scheme and in the single camera acquisition detection state in the embodiment of the application;

[0029] Figure 4 is the schematic diagram for slope monitoring in the embodiment of the application;

[0030] Figure 5 is the schematic diagram of the sand soil slope rainfall experiment by using the binocular multi-point displacement monitoring device in the embodiment of the application.

[0031] In the figure: 1-fixed frame, 2-first steering engine holder, 3-fixed rod, 4-second steering engine holder, 5-monitoring camera, 6-fixed reference object. DETAILED DESCRIPTION

[0032] The application will be further explained in conjunction with the accompanying drawings and specific embodiments.

[0033] Embodiment 1:

[0034] The embodiment discloses a binocular multi-point displacement monitoring device, which can be used for displacement monitoring of target objects in multiple scenes, realizes high-precision autonomous displacement monitoring by using existing visual calculation algorithms, can not only perform local monitoring calculation, but also can perform remote data uploading and early warning through Internet of Things technology. The device and the monitoring method based on the device provided in the embodiment are mainly used for slope displacement monitoring of large-area slopes formed in natural terrain or engineering construction, so as to avoid geological disasters or engineering dangers.

[0035] 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.

[0036] The fixed frame 1 is used for stably installing the device in a stable area near a slope to be monitored, and the first displacement mechanism is assembled on the fixed frame 1 and can drive the whole mounting frame to realize synchronous displacement.

[0037] 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, and each second displacement mechanism can independently drive the corresponding monitoring camera 5 to adjust the pose.

[0038] The control module is connected with the first displacement mechanism, the second displacement mechanism and the monitoring camera 5 respectively, can control the first displacement mechanism to drive all monitoring cameras 5 to move synchronously, can control a single second displacement mechanism to drive 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 6 is pre-laid, the fixed reference 6 includes a point reference and a cross calibration reference, the point reference has a clear geometric center mark, and the cross calibration reference has a clear cross intersection and a fixed shape arm structure.

[0039] Meanwhile, natural references with characteristics on the slope surface are also collected as references. The natural references are selected from natural terrain, stones, engineering structures and plants with stable position and shape on the slope surface. For the above references, the spatial position information and size characteristics of all fixed references 6 and the contour feature point spatial position information of the natural references are obtained through a positioning device, and the information is stored in the control module as reference data for subsequent monitoring.

[0040] Based on the binocular multi-point displacement monitoring device, the embodiment also provides a method for monitoring slope displacement by using the displacement monitoring device, and the specific process is as follows:

[0041] The early preparation and reference establishment are as follows: first, the slope surface range of the temporary slope to be monitored is determined, the point reference and the cross calibration reference are uniformly laid on the slope surface, and the natural references meeting the requirements are marked; the spatial position and characteristic information of all references are collected through a positioning device, the data is uploaded to the control module for storage, and the reference database construction is completed.

[0042] The device installation is as follows: the fixing frame 1 is installed on the stable terrain at a suitable distance from the opposite direction of the slope surface to be monitored, the first displacement mechanism, the mounting frame, the second displacement mechanism and the monitoring camera 5 are assembled in sequence, the control module is laid 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.

[0043] The debugging and calibration are as follows: on the one hand, two-dimensional calibration is carried out, the control module drives each second displacement mechanism to drive corresponding monitoring camera 5 to adjust the angle independently, so that each camera is aligned with the initial sub-region of the slope respectively, the image of the calibration plate in the sub-region is shot, the monocular parameter of each camera is calculated and stored; on the other hand, three-dimensional preset calibration is carried out, taking the cross calibration reference as the reference, the second displacement mechanism drives two cameras to adjust to the binocular cooperative pose, the image of the cross calibration reference is shot, the binocular cooperative parameter is calculated and stored, and the measurement accuracy in the binocular mode is ensured.

[0044] The sub-region division and the early warning threshold setting use the device completed by debugging, divide the whole slope surface to be monitored into a plurality of sub-regions in a row-column matrix form in units of the maximum monitoring field of view of a single monitoring camera 5, number all the sub-regions in a uniform order, and reasonably allocate the sub-regions to two monitoring cameras 5. In this embodiment, the sub-regions in the left half are allocated to the left camera, and the sub-regions in the right half 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 topographic and geomorphic information of the reference objects in each sub-region is formed by processing. Based on the reference position of the reference object in the initial calibration quantity, the two-dimensional displacement early warning threshold is set.

[0045] The monocular inspection and mode switching trigger After the setting and debugging are completed, the control module controls the device to enter the monocular inspection mode, drives the two cameras to sequentially inspect the allocated sub-regions at a preset period, and collects the images of each sub-region in real time. The control module compares the collected images with the initial calibration quantity and historical monitoring data, and judges the displacement of the reference object. When any trigger condition (such as the displacement of a fixed reference object 6 in the same picture of a single camera exceeding the threshold, the relative displacement of the reference object in the front and back pictures of a single camera exceeding the threshold, or the displacement of the reference object in the pictures of different cameras exceeding the threshold) is monitored, the control module immediately sends early warning information to the cloud, and automatically triggers the binocular inspection process.

[0046] The binocular precise 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 quickly move the cameras to align with the preset cross calibration reference object position, completes the binocular cooperative calibration by shooting the image of the cross calibration reference object, and performs reference testing. After calibration is completed, 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 early warning is located, three-dimensional data of the region and the surrounding related sub-regions are collected first, then the monitoring range is gradually expanded until the displacement is less than the threshold, and the displacement range and the displacement amount are determined; the displacement type is judged in combination with the topographic and geomorphic changes in the three-dimensional data, and finally the complete displacement information is formed by integrating the displacement range, the displacement amount and the displacement type, and is uploaded to the cloud for remote control and decision-making.

[0047] It should be noted that the system and method disclosed in this embodiment are mainly used for temporary displacement monitoring of the slope surface. The core adaptive scene of this temporary geological engineering monitoring task originates from the special risk properties and dynamic monitoring needs of the temporary slope in the field of geological engineering. The necessity of the temporary slope monitoring task can be explained from the following three aspects:

[0048] The temporary slope stability risk caused by engineering construction disturbance drives the demand for temporary monitoring. For example, in traffic infrastructure, building foundation excavation, and mine stripping engineering, the temporary slope is usually formed by soil excavation or site leveling during the construction process. Such slopes are not optimized for long-term stability design, and the original stress state of the slope body is severely redistributed due to construction unloading. Unloading rebound, crack development, and other phenomena are prone to occur. 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 a short period 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.

[0049] In another case, the slope rock-soil mass of the temporary slope is mostly a fresh excavation surface, and the shear strength is low. In the geological disaster-prone period such as the rainy season and snowmelt period, rainwater infiltration or snowmelt will cause the water content of the slope body to rise sharply, further exacerbating the strength decay of the rock-soil mass, and seepage may induce piping and erosion inside the slope, increasing the risk of landslides and collapses. Meanwhile, there is an emergency monitoring demand due to extreme weather and sudden geological events, which gives rise to the task of monitoring the temporary slope. When the temporary slope is facing extreme weather such as heavy rain, typhoon, and short-time strong convective weather, or surrounding sudden geological events such as earthquakes and collapses, the slope body may have significant displacement in a short time. If the traditional monitoring means is not deployed in advance, it cannot capture the risk in time, and the long-term deployed monitoring equipment may lose its monitoring ability due to disaster damage. At this time, a temporary monitoring system needs to be quickly set up to carry out emergency monitoring on the affected temporary slope.

[0050] To solve the displacement monitoring tasks of the above several slopes, the binocular multi-point monitoring device provided in the embodiment can be quickly and temporarily set up and adjusted to adapt to the displacement monitoring of a large area of slope surface within a certain period of time.

[0051] Further, with reference to Figures 1-4 The binocular multi-point displacement monitoring device in the embodiment adopts a telescopic tripod structure as a fixing frame 1. The tripod itself has the characteristics of convenient carrying and transportation. In the displacement monitoring task of a large area of slope surface, multiple fixing frames 1 and corresponding camera assemblies can be monitored at the same time. A small number of engineering personnel can transport the equipment to the site, quickly set it up and adjust it, and can be adjusted through a terminal device with a plurality of signal ports, improving the arrangement efficiency.

[0052] In the embodiment, the first displacement mechanism and the second displacement mechanism are both a servo gimbal, which has a horizontal rotation component and a pitch angle rotation component, so that the structure loaded thereby can be rotated and adjusted in space to cover a larger area of monitoring objects.

[0053] Specifically, the fixed frame 1 is provided with a first steering platform 2, and the first steering platform 2 is connected with the control module through a wire harness. A fixed rod 3 serving as a mounting frame is fixed on the first steering platform 2 through a clamping piece. The fixed rod 3 in the embodiment is a special-shaped pipe material, the middle part has a circular cross section, and the surface has a tooth pattern, so as to be clamped and fixed by the clamping piece. The two side parts have a rectangular cross section or a structure with at least one plane, so as to be clamped and fixed by the second steering platform 4. Referring to Figure 2 and Figure 3 , two second steering platforms 4 are arranged on one fixed rod 3, and one monitoring camera 5 is arranged on each second steering platform 4. The second steering platform 4 and the monitoring camera 5 are connected with the control module through a cable. 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 and ultraviolet aging.

[0054] Referring to Figure 5 , simple experimental verification is carried out for the device and method in the embodiment. A plastic box and a sand pile 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, the water spraying rate is set to 5L / h, and the water flow is uniformly distributed on 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 1.2m, and the two cameras are aligned with the sand pile area; the control module completes monocular calibration and binocular calibration, and divides the sand pile into two adjacent sub-areas according to the size. The experimental verification conclusion confirms 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 recessed 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 is 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 red reference point recognition success rate is 100%, and the adaptability of the device to the simple simulation scene is verified.

[0055] 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, belongs to typical landslide-prone bare rock mountain terrain, the overall slope surface is at a near 45° angle, there are no tall trees for vegetation coverage, only scattered low shrubs and weeds are distributed in the rock crevices, the coverage rate is less than 10%, the main part of the slope surface is gray-black sandstone and mudstone interbedded structure, the exposed rock area accounts for more than 90%, and there are obvious joint fissures in the rock layer surface, and small weathering and peeling marks can be seen in some areas.

[0056] The slope surface height is about 20 m, the horizontal length of the slope surface is about 35 m, and the total area of the slope surface is about 700 m 2 The top of the slope surface is a temporary construction road, and the bottom is a highway roadbed construction area. Due to 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 surface to be monitored, about 55 m away from the slope surface, there is a natural bedrock platform as a device erection fixing plane. 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 deformation-prone areas such as gullies and cracks. According to the geological survey, the bedrock where the platform is located is a complete block structure, and there is no obvious displacement risk, which can meet the long-term stable erection requirements of the monitoring equipment.

[0057] The total height of the selected fixed frame 1 in this embodiment can be adjusted after unfolding, and is adjusted to 1.4 m in height during actual erection, and the tripod base is fixed to the bedrock platform by 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 monitoring sub-area to ensure that the reference object is clearly imaged. It has wide dynamic and low light function, which can adapt to the environment with large difference in light between morning and evening, overcast light, avoid strong light reflection or weak light caused image blur. Each camera is mounted on an independent second servo cloud 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 be moved synchronously under the drive of the first servo cloud 2, meeting the demand of binocular collaborative monitoring. The port of the control module is 485 bus port, and it also has a power supply port. It is connected with the first servo cloud 2, the second servo cloud 4 and the monitoring camera 5 on the fixed frame 1 through waterproof cable. The cable interface is treated with waterproof sealing joint to avoid rainwater infiltration. The control module selects a four-core processor based on ARM Cortex-A53 architecture. The processor has low power consumption and data processing capacity, which can meet the image preprocessing in monocular inspection and the disparity calculation and three-dimensional displacement vector solution in binocular mode. The processor integrates 2GB DDR4 memory and 16GB eMMC flash memory, which can support local storage of inspection images and displacement data for 30 days. At the same time, a MicroSD card expansion interface is reserved, which can adapt to long-term continuous monitoring scene. The power supply is a sealed gel battery, which is composed of two 12V / 150Ah single batteries in series, forming a 24V / 150Ah battery pack, with a total energy storage capacity of 3600Wh.

[0058] Then, the specific monitoring method based on the scene is as follows:

[0059] I. Device arrangement process

[0060] (1) Preliminary site survey and preparation

[0061] On the bedrock platform 55 m away from the slope surface to be monitored, use the geological radar to scan to confirm that there is no hidden fissure or cavity within 3 m below the platform, use the total station instrument to measure the flatness of the platform to ensure that the surface undulation difference is not greater than 5 cm; use the level to mark the center point of the platform as the installation reference point of the fixed frame 1. Then, the equipment kit is unpacked and inspected on site, including tripod fixed frame 1, first displacement mechanism, long strip fixed rod 3, second displacement mechanism 2 set, monitoring camera 52, control module, 24V / 150Ah gel battery pack and 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, etc. installation tools, and prepare rainproof cloth.

[0062] (II) Mounting of the fixed frame 1 and displacement mechanism

[0063] The tripod is unfolded so that the three legs are evenly distributed around the reference points on the bedrock platform. The height of the legs is adjusted to 1.4 m. The top mounting surface of the tripod is observed using a level. The mounting surface is leveled by rotating the adjustment knobs. Holes are drilled in the bedrock at the corresponding positions at the bottom of the legs using a percussion drill. Expansion bolts are inserted into the holes. The bolts are tightened using a torque wrench to a fixed torque. The tripod is rigidly connected to the bedrock. A horizontal pushing force is applied to test for any significant movement. The base of the first displacement mechanism is fixed to the top mounting surface of the tripod using M10 bolts. Threaded glue is applied to the bolts to prevent loosening. The drive cable of the connecting mechanism is connected. The drive cable is pre-emptively provided with 1.5 m of excess length. The horizontal rotation function and angle feedback accuracy are tested by applying power.

[0064] (III) Deployment of the camera and control module

[0065] Two monitoring cameras 5 are fixed to the output ends of the second displacement mechanisms. The lenses face the slope surface. The lens optical axes are perpendicular to the fixed rod 3 by adjusting the mechanisms. Lens protection covers are installed. Waterproof camera cables are connected. The camera imaging function is tested by applying power. The image is ensured to be free of distortion and obstruction. The resolution is switched to the 5 million pixel mode. The control module and power supply system are arranged. A battery protection box is installed in a light-protected area on the bedrock platform. A 24V / 150Ah battery pack is placed in the box. The positive and negative cables are connected to the MPPT power management module. The control module is wall-mounted next to the protection box. Waterproof cables are connected to the first displacement mechanism, the second displacement mechanism, and the camera. A solar panel support is fixed to the edge of the platform where there is no obstruction. The inclination angle is adjusted to 35°. The output cable is connected to the MPPT module. The charging function is tested. A 4G antenna and a LoRa antenna are installed. The network connection status is checked after the control module is powered on. The real-time image transmission function is verified by connecting the LoRa module using a handheld terminal.

[0066] II. Slope reference object setting process

[0067] (I) Reference object type and layout planning

[0068] The point reference adopts a circular prefabricated cement member with a diameter of 30 cm, the surface of which is sprayed with black and white concentric rings, and a Φ2 cm stainless steel dot is embedded in the center; the cross calibration reference adopts a cross-shaped prefabricated cement member, the horizontal arm of which is 60 cm long, the vertical arm is 60 cm long, the arm width is 15 cm, the thickness is 10 cm, the surface is sprayed with red and white stripes, and a Φ3 cm stainless steel dot is embedded at the intersection of the cross. Along the height direction of the slope surface, 3 layers are arranged, and 1 point reference is arranged every 5 m horizontally in each layer, a total of 18 point references; 1 cross calibration reference is arranged in the left, middle and right areas of the slope surface, respectively, to ensure that the distance between each cross calibration reference and the surrounding point references does not exceed 8 m. When selecting a natural reference, identify the natural stability features of the slope surface, and make a circular mark around the feature point with red paint, the mark line being 5 cm wide, to ensure that the camera can clearly identify it.

[0069] (II) Reference installation and positioning

[0070] The fixed reference 6 is installed. For the point reference, a hole is drilled in the rock on the slope surface, an expansion bolt is implanted, and the cement member is fixed through an angle steel support; the cross calibration reference is also fixed in the same way, and after installation, a level is used to check whether the cross arms are parallel to the contour lines of the slope surface, to ensure that the feature form is not distorted. An RTK-GPS measurement device is used to measure the three-dimensional coordinates of the stainless steel dot of each point reference, the three-dimensional coordinates of the intersection point of the cross calibration reference, the end point coordinates of the horizontal / vertical arm, and the three-dimensional coordinates of the marked center of the natural reference; each point is measured 3 times to take the average value, which is recorded to the data collection terminal, and the on-site photos of the references are taken synchronously. The positioning parameters are imported into the control module, the camera is called through the handheld terminal to shoot each reference, the control module automatically identifies the feature points of the reference, compares the measurement coordinates with the image pixel coordinates, and generates an initial mapping relationship; for the cross calibration reference, additional storage of its cross arm length, angle and other size parameters is used as a binocular calibration benchmark template.

[0071] (III) System linkage test

[0072] The first control module drives the second displacement mechanism to make the left camera align with the left half of the slope surface and the right camera align with the right half of the slope surface, and to shoot the image of the whole slope surface; check whether all reference objects are located in the image, if there is a visual blind area, fine-tune the corresponding camera angle and re-shoot until all reference objects are covered. Then perform binocular collaborative calibration, adjust the two monitoring cameras 5 through the second displacement mechanism to form a binocular camera unit. Control the first displacement mechanism to drive the fixed rod 3 to rotate, so that the two cameras are aligned with the central cross calibration reference object at the same time, and shoot a stereo image pair; the control module calculates the binocular parallax and outputs the three-dimensional coordinates of the reference object, which is compared with the RTK measurement value, if the difference is too large, the second displacement mechanism is used to fine-tune the distance between the cameras until the accuracy requirement is met. Upload the initial calibration quantity to the cloud platform, and the cloud generates a digital model of the slope surface, which is compared with the actual terrain on site to confirm that the model error is not more than 5%, and the final verification of the device arrangement and the reference object setting is completed.

[0073] III. Image-based slope surface region division and camera responsible region allocation

[0074] The maximum monitoring and analysis picture of the monitoring camera 5 in the embodiment is a slope block of 30*10m, based on the actual size of the slope surface to be monitored, which is 20m in height and 35m in horizontal length, combined with the actual area size corresponding to the maximum monitoring field of view of the monitoring camera 5, which is 30m*10m, the sub-regions are divided by using the combination of row-column matrix and equal proportion division, to ensure that there is no monitoring blind area on the whole slope surface and the data can be cross-validated.

[0075] Along the horizontal length direction of the slope surface, take 30m as the benchmark of the maximum horizontal coverage range of monocular, reserve 5m overlapping area for cross-sub-region reference object correlation calibration, divide it into two horizontal units, which cover the horizontal 0-30m section and the horizontal 5-35m section of the slope surface respectively; along the height direction of the slope surface, take 10m as the benchmark of the maximum vertical coverage range of monocular, without overlapping, divide it into two vertical units, which cover the bottom area of 0-10m in height and the top area of 10-20m in height respectively. Four complete sub-regions are formed by crossing the horizontal and vertical units, which are named according to the rule of horizontal number plus vertical number, respectively, the sub-region of horizontal 0-30m and height 0-10m, the sub-region of horizontal 0-30m and height 10-20m, the sub-region of horizontal 5-35m and height 0-10m, and the sub-region of horizontal 5-35m and height 10-20m, each sub-region has an actual area of about 300m 2 , and the overlapping area is about 50m 2 .

[0076] According to the slope surface coverage angle difference corresponding to the sub-region 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 sub-region of the slope surface, that is, the sub-region of horizontal 0-30 m and height 0-10 m and the sub-region of horizontal 0-30 m and height 10-20 m, covering the full height range of horizontal 0-30 m of the slope surface, and monitoring 10 of the 18 point references on the slope surface and 1 left cross calibration reference, of which the 10 point references include 6 in the sub-region of horizontal 0-30 m and height 0-10 m and 4 in the sub-region of horizontal 0-30 m and height 10-20 m, and the left cross calibration reference is located in the sub-region of horizontal 0-30 m and height 10-20 m.

[0077] The right camera is responsible for the right half area sub-region of the slope surface, that is, the sub-region of horizontal 5-35 m and height 0-10 m and the sub-region of horizontal 5-35 m and height 10-20 m, covering the full height range of horizontal 5-35 m of the slope surface, and monitoring 8 of the 18 point references on the slope surface and 2 cross calibration references, of which the 8 point references include 4 in the sub-region of horizontal 5-35 m and height 0-10 m and 4 in the sub-region of horizontal 5-35 m and height 10-20 m, and the 2 cross calibration references are respectively located in the middle of the sub-region of horizontal 5-35 m and height 0-10 m and the top of the sub-region of horizontal 5-35 m and height 10-20 m.

[0078] For the overlapping areas of the sub-region of horizontal 0-30 m and height 0-10 m and the sub-region of horizontal 5-35 m and height 0-10 m, the sub-region of horizontal 0-30 m and height 10-20 m and the sub-region of horizontal 5-35 m and height 10-20 m, the images are synchronously collected by the two cameras, and the control module verifies the displacement data of the reference in the overlapping area by cross matching of feature points, so as to avoid misjudgment caused by the visual angle deviation of a single camera.

[0079] Four, monocular monitoring debugging and function verification process

[0080] The monocular monitoring debugging is carried out from two aspects of intrinsic parameter calibration and field of view matching, 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 sub-region of horizontal 0-30 m and height 0-10 m, the checkerboard grid spacing is 10 cm, five groups of calibration images with 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, in which the focal length is 12 mm, 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 sub-region of horizontal 5-35 m and height 0-10 m, and the focal length of the finally obtained intrinsic parameter is 12 mm.

[0081] In the field of view matching debugging stage, the control module drives two cameras to independently traverse the responsible area in the order of sub-regions. The sub-region traversal order is the sub-region of horizontal 0-30m and height 0-10m, the sub-region of horizontal 0-30m and height 10-20m, the sub-region of horizontal 5-35m and height 0-10m, and the sub-region of horizontal 5-35m and height 10-20m. Every time a sub-region image is taken, the control module automatically identifies all reference objects in the image, including point reference objects, cross calibration reference objects and natural reference objects, and records the pixel coordinates of each reference object. If the pixel ratio of a certain reference object is less than 30 pixels, the pitch angle of the corresponding camera is fine-tuned through the sub-displacement mechanism, and the fine-tuning range is controlled within ±2°, until the pixel ratio of all reference objects reaches 30 pixels or more, to ensure the subsequent recognition accuracy.

[0082] 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, two cameras continuously take three rounds of full slope images according to the allocated regions, and the control module automatically splices the images to form a complete slope view, checks that there is no missing shot in the four sub-regions and the overlapping area, and the recognition success rate of 18 point reference objects, 3 cross calibration reference objects and 5 natural reference objects reaches 98% or above. The sporadic weeds are removed by the image preprocessing algorithm to eliminate interference, and there is no recognition failure caused by shrub obstruction.

[0083] In the displacement simulation test, a point reference object is selected in the horizontal 0-30m and height 0-10m sub-region. The reference object is moved by 2mm, 4mm and 6mm in the horizontal direction respectively by a micro-displacement table with an accuracy of 0.1mm, corresponding to the non-overrun, critical overrun and overrun states. The left camera takes 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 of less than 0.1mm, verifying that the monocular displacement calculation accuracy meets the requirements.

[0084] In the periodic inspection test, the inspection period is set to 2 hours each time, and the system runs continuously for 24 hours. 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 not lost, 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.

[0085] Five, activation conditions of binocular monitoring process

[0086] Based on monocular monitoring data, combined with the risk level of temporary monitoring of rock slope, four types of binocular activation conditions are set, and any condition can trigger the binocular monitoring process. Each type of condition has clear determination criteria and application scenarios to ensure that no key risk points are missed.

[0087] The first type is single reference single displacement exceeding the limit, including point reference, cross calibration reference and natural reference. When the point reference is triggered, the two-dimensional displacement of any single point reference in the same sub-region image captured by a single camera reaches 3 mm 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 is 315° in this scenario. For example, the horizontal displacement of a point reference is -2.5 mm and the vertical displacement is -1.8 mm in the sub-region image captured by the left camera at a horizontal distance of 0-30 m and a height of 10-20 m. The combined displacement is 3.08 mm, which reaches the threshold of 3 mm and is consistent with the potential sliding direction of the slope surface, triggering the binocular monitoring process.

[0088] When the cross calibration reference is triggered, the two-dimensional displacement of the cross intersection in the cross calibration reference image captured by a single camera reaches 2.5 mm or more, or the cross arm deflection angle reaches 1° or more. For example, the cross arm deflection angle of the cross calibration reference in the sub-region captured by the right camera at a horizontal distance of 5-35 m and a height of 0-10 m is 1.2°, which is determined as local torsion displacement, triggering the binocular monitoring process.

[0089] When the natural reference is triggered, three feature points are selected from the profile of the natural reference, the average two-dimensional displacement of the three feature points reaches 3.5 mm or more, and the displacement directions of the three feature points are consistent to exclude image noise interference, triggering the binocular monitoring process.

[0090] The second type is reference relative displacement exceeding the limit, including same region relative displacement and cross region relative displacement. When the same region relative displacement is triggered, the interval between the two images of the same sub-region captured by a single camera is 2 hours, and the relative displacement difference between two adjacent point references reaches 2 mm or more. For example, the relative displacement of two adjacent point references increases from 0.5 mm to 2.6 mm in the images of the same sub-region captured by the left camera at a horizontal distance of 0-30 m and a height of 0-10 m at different times, and the difference reaches the threshold of 2 mm, triggering the binocular monitoring process. When the cross region relative displacement is triggered, the displacement difference of the same reference in the overlapping region of the front and rear frames of the images captured by the two cameras reaches 1.5 mm or more. For example, the displacement of a point reference in the overlapping region measured by the left camera is 2.2 mm, and the displacement of the same point reference measured by the right camera is 3.8 mm, with a difference of 1.6 mm. It is determined that the displacement data is abnormal due to the viewing angle deviation of the single camera, triggering the binocular calibration and precision measurement process.

[0091] The third type is a reference object cumulative displacement exceeding the limit, which is divided into short-term accumulation and long-term accumulation triggering situations. When the short-term accumulation triggers, the displacement of the same reference object accumulates to 5 mm or more in 6 hours, which is the interval of 3 consecutive inspections; for example, the displacement of a reference object in a certain point in a sub-region with a horizontal range of 5-35 m and a height range of 10-20 m is 1.8 mm, 1.7 mm and 1.6 mm in three inspections, and the cumulative displacement is 5.1 mm, which reaches the threshold of 5 mm and the displacement direction is continuously along the slope surface, which is determined as an accelerated 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-region within 24 hours reaches 8 mm or more; for example, the 24-hour cumulative displacement of three reference objects in a sub-region with a horizontal range of 0-30 m and a height range of 10-20 m is 7.8 mm, 8.2 mm and 8.5 mm, and the average cumulative displacement is 8.17 mm, which reaches the threshold of 8 mm, which is determined as a regional overall deformation, triggering the binocular monitoring process.

[0092] The fourth type is triggered by abnormal displacement trend, including displacement rate mutation and displacement direction mutation. When the displacement rate mutation triggers, the displacement rate of a single reference object, i.e. the ratio of the displacement difference between two adjacent inspections to the interval time, increases from a stable value of 0.5 mm per hour or less to 1.5 mm per hour or more; for example, the displacement rate of a certain point reference object increases from 0.4 mm per hour to 1.6 mm per hour, which is determined as accelerated deformation, triggering the binocular monitoring process. When the displacement direction mutation triggers, the displacement direction of the same reference object changes from random fluctuation deviation of 15° or less to fixed direction, the direction deviation is 5° or less in two consecutive inspections, and the direction is consistent with the slope joint fissure trend, which is 45° in this scenario, which is determined as potential slip controlled by structural plane, triggering the binocular monitoring process.

[0093] Six, displacement type definition

[0094] Based on the geological background of sand and mud interbedding and joint fissure development in bare rock slope, four types of typical geological displacement are defined based on the displacement direction, range and correlation data of reference objects obtained by monocular monitoring. Each type of displacement has clear geological origin, displacement performance and risk level, which meets the actual needs of temporary monitoring of rock slope.

[0095] The first is block sliding, also known as rock block sliding type, which is controlled by the slope surface dominant structure plane, including bedding plane and joint. Single or multiple block rock with size of 2m or above in the slope body slides along the structure plane. The sliding surface is mostly flat rock layer plane. In this scenario, the inclination of sand and mud interbedding plane is 35°. In terms of displacement performance, 3-5 adjacent points of reference in the sub-region synchronously displace in the same direction, and the displacement direction is consistent with the inclination of the structure plane, which is 315° in this scenario. The displacement amount ranges from 5mm to 20mm, and the relative displacement difference of adjacent reference objects is controlled to be 2mm or 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. The depth of the sliding surface needs to be determined by binocular precise measurement to evaluate the subsequent risk.

[0096] The second is wedge sliding, also known as double structure plane control type. The wedge is cut by two groups of intersecting structure planes, and the structure plane includes joint. In this scenario, the strike of the two groups of joints is 30° and 330°, and the inclination is 60° and 55°, respectively. The wedge slides along the intersection direction of the two structure planes. It is commonly seen in the middle and upper sub-regions of sand and mud interbedding slope, i.e. the sub-regions with horizontal distance of 0-30m and height of 10-20m, and the sub-regions with horizontal distance of 5-35m and height of 10-20m. The displacement of the reference objects in the wedge range is in the intersection direction, which is 285° in this scenario. The displacement amount ranges from 8mm to 30mm, and the closer to the front edge of the wedge, i.e. the lower part of the slope surface, the larger the displacement amount. For example, when the displacement of the reference object at the rear edge of the wedge is 10mm, the displacement of the reference object at the front edge can reach 25mm. The cross calibration reference object has slight deflection, and the deflection angle ranges from 1° to 2°. The edge depression direction of the wedge in the natural reference object is consistent with the sliding direction of the wedge. The risk level of this type of displacement is high, and the wedge sliding is easy to cause medium-scale landslide. Emergency monitoring needs to be started and timely warning needs to be given.

[0097] The third is the dumping deformation, also known as the rock layer dumping type, and 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°. Affected by gravity or rainwater infiltration, the rock layer along the layer surface occurs dumping, which is mostly seen in the thin sand and mudstone at the top 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 that the reference object at the top of the sub-region appears outward dumping displacement, the horizontal displacement range is 3-8 mm away from the slope, and the vertical displacement range is 1-3 mm upward; the transverse arm of the cross calibration reference object deflects by an angle range of 2-3°, showing the characteristics of the outer side upward, and the rock outcrop in the natural reference object shows the characteristics of forward inclination. The risk level of this type of displacement is medium, and the dumping rock layer is prone to surface peeling. If not monitored in time, it may gradually develop into overall sliding.

[0098] The fourth is the surface peeling, also known as the weathering peeling type, and the geological origin is that the slope 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 mostly occurs on the surface of the weathered sand and mudstone in the middle and lower parts of the slope, i.e., 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 that single or scattered point reference objects independently displace, the displacement range is 2-5 mm, and the displacement direction of adjacent reference objects is irrelevant, for example, when a point reference object horizontally displaces by +1.5 mm, the adjacent point reference object may horizontally displace by -2 mm; the cross calibration reference object has no displacement, and the deflection angle is less than 0.5°; the displacement of the surface weathered rock in the natural reference object is less than 3 mm, which is mostly point displacement caused by local peeling. The risk level of this type of displacement is low, only the surface rock block peels off, and there is no overall landslide risk, which needs to be regularly tracked and monitored to observe whether there is a development trend.

[0099] Seven, determination process of displacement area, displacement amount and displacement type when switching binoculars

[0100] When switching the binocular monitoring process, the current deformation state of the slope is determined through four steps of core out-of-limit area positioning, displacement range determination, displacement amount calculation and displacement type determination, which provides an accurate direction for subsequent binocular precision measurement. The whole process is based on monocular monitoring data and does not need to rely on a three-dimensional terrain model.

[0101] In the core overrun area positioning stage, the control module extracts the overrun reference object that triggers the activation condition from the monocular monitoring data, for example, the left camera monitors that the displacement of a certain point reference object in the 0-30m horizontal and 0-10m height sub-area reaches 6.08mm, and the right camera monitors that the displacement of a certain point reference object in the 5-35m horizontal and 0-10m height sub-area reaches 5.2mm, and locks these two point reference objects as the core overrun reference objects. According to the sub-area where the core overrun reference objects are located, the initial core monitoring area is determined as the 0-30m horizontal and 0-10m height sub-area and the 5-35m horizontal and 0-10m height sub-area where the two core overrun reference objects are located, i.e. the 0-35m horizontal and 0-10m height area at the bottom of the slope, covering the two sub-areas, and the preliminary estimated area is about 600m 2 .

[0102] In the displacement range, i.e. displacement area determination stage, the boundary expansion detection and boundary verification combination method is adopted. In the boundary expansion detection, the control module drives the two cameras to expand one adjacent sub-area, i.e. the 0-30m horizontal and 10-20m height sub-area and the 5-35m horizontal and 10-20m height sub-area, based on the core area, to shoot the images of the expanded area and calculate the displacement of the reference objects in the area. If the displacement of the reference objects in the expanded area is less than the 3mm threshold, for example, the displacement of a certain point reference object in the 0-30m horizontal and 10-20m height sub-area is 2.1mm, and the displacement of a certain point reference object in the 5-35m horizontal and 10-20m height sub-area is 2.8mm, then it is determined that the displacement boundary is the initial core area, and the displacement area is calculated by subtracting the overlapping area from the areas of the two sub-areas in the core area, i.e. 300m2+300m2-50m2=550m2. In the boundary verification, the reference objects at the displacement boundary are detected, for example, the point reference object at the junction of the 0-30m horizontal and 0-10m height sub-area and the 0-30m horizontal and 10-20m height sub-area, and it is confirmed that the displacement is 2.2mm, which is less than the 3mm threshold, and there is no cross-boundary overrun phenomenon, and finally the displacement area is determined as 550m 2 .

[0103] In the displacement amount calculation stage, representative reference objects in the core area and boundary reference objects are selected to ensure that the data covers the entire displacement area. In the core area, five representative reference objects are selected, 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, respectively, 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 boundary are 2.2 mm and 2.8 mm, respectively. The maximum displacement of the boundary reference object is taken as the boundary displacement of the displacement area, and the final core displacement range of the displacement area is determined to be 4.756 mm ± 0.5 mm, and the boundary displacement is 2.8 mm.

[0104] In the displacement type determination stage, the displacement performance and geological background are comprehensively analyzed. First, the displacement direction and correlation of the reference objects in the core area are analyzed. The displacement directions of the three point reference objects and the cross calibration reference object in the core area are all 315°, which is consistent with the potential sliding direction of the slope surface, and the relative displacement differences between adjacent reference objects are all less than 2 mm, showing synchronous displacement characteristics. The displacement direction of the angular block rock in the natural reference object is consistent with that of the point reference object, and there is no local protrusion or depression, 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 conditions of opposite rock layer inclination and slope direction required for rock layer dumping, excluding wedge sliding, dumping deformation, and surface peeling. Finally, it is determined that the displacement type is block sliding, the risk level is medium, and the depth and displacement of the sliding surface need to be further determined through subsequent binocular precise measurement to provide basis for slope safety control.

[0105] Eight, in the monocular state, activate the rear camera to quickly reset and align the reference

[0106] After activating the binocular monitoring process, the control module first starts the reference alignment program, taking the pre-stored cross calibration reference object as the core positioning reference, to quickly reset the two monitoring cameras 5 at any patrol angle. The control module retrieves the pre-set cross calibration reference object information from the local parameter library, including its three-dimensional coordinates, image features, and real-time angle data of the current two cameras. Regardless of the angle of the camera in 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 object.

[0107] Subsequently, the control module drives the second displacement mechanism corresponding to each camera to independently adjust the angle in the order of "horizontal first and then pitch": first, through horizontal adjustment, the lens optical axis of the two cameras is preliminarily pointed to the sub-region where the cross calibration reference object is located; then, through pitch adjustment, combined with real-time image feedback - the control module acquires an image shot by the camera every 0.5 seconds, automatically identifies the features of the cross calibration reference object in the image, if it is not identified, it continues to fine-tune the angle, if it is identified, it further optimizes the angle, until the cross calibration reference object is located in the center region of the image of the two cameras, and the preliminary reset is completed. The entire reset process does not require manual intervention, and is 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.

[0108] Nine, precise calibration process of binocular unit

[0109] After preliminary reset, enter the binocular unit calibration stage, 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-tuning, which is specifically three steps:

[0110] The first step is external parameter calibration. The control module drives the two cameras to synchronously shoot the images of the aligned cross calibration reference object, extracts the pixel coordinates of the cross intersection in the two images, and calculates the external parameters of the two cameras, including the relative rotation matrix and translation vector, based on the known three-dimensional coordinates of the cross calibration reference object and using the PnP algorithm. Compare the calculated actual external parameters 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.

[0111] 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 adjustments according to the deviation type: if there is a rotation deviation, fine-tune the pitch angle or horizontal angle of the corresponding camera, and after each adjustment of 0.1°, re-shoot the image and calculate the external parameters until the rotation deviation is not more than 0.5°; if there is a translation deviation, fine-tune the lateral position of the camera on the mounting bracket through the second displacement mechanism, and after each adjustment of 0.5 mm, re-calibrate until the translation deviation is not more than 1 mm.

[0112] The third step is calibration verification. After fine-tuning, the control module drives the binocular unit to turn to another pre-set cross calibration reference object, 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 the deviations of the two calibrations are up to standard, it is determined that the binocular unit calibration is completed, 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.

[0113] X. Binocular three-dimensional terrain acquisition process

[0114] After the binocular unit calibration is completed, the binocular three-dimensional terrain acquisition stage is entered, the core of which is to obtain the three-dimensional terrain data of the target displacement region through stereo image acquisition, disparity calculation and three-dimensional coordinate solution, the process is divided into four steps:

[0115] The first step is target region positioning and scanning path planning. The control module determines the displacement-related sub-regions based on the monocular monitoring in the early stage, including a single sub-region and two sub-regions. Then the scanning path of the binocular unit is planned, taking the geometric center of the displacement sub-region as the starting point, and dividing the scanning sub-region according to the snake-shaped traversal rule to avoid monitoring blind area. The control module drives the first displacement mechanism to move the binocular unit to each scanning sub-region according to the planned path, and keeps the pose of the binocular unit stable after each movement to ensure that the two cameras are aligned with the current scanning sub-region at the same time.

[0116] The second step is stereo image pair acquisition. In each scanning sub-region, the control module sends a synchronous shooting instruction to the two cameras to ensure that the shooting time difference of the two images is ≤1ms, and a set of stereo image pairs is acquired. For key areas in the displacement region, an additional 2-3 times of shooting is added to obtain multiple sets of stereo image pairs for subsequent data verification and improve the reliability of three-dimensional data.

[0117] The third step is disparity calculation and three-dimensional coordinate solution. 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, the 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 calculate the pixel difference, i.e. the disparity.

[0118] Combined with the binocular internal and external parameters obtained in the early calibration, the three-dimensional measurement principle of binocular vision is used to convert the disparity of each feature point into three-dimensional space coordinates according to the mapping relationship between disparity and actual distance, i.e. the larger the disparity, the closer the feature point to the camera; the smaller the disparity, the farther the feature point to the camera. The three-dimensional feature point coordinate set in the displacement region is obtained.

[0119] The fourth step is three-dimensional terrain data integration. The control module splices the three-dimensional feature point coordinates of all scanning sub-regions according to the spatial position, removes the repeated points and abnormal points, and forms a three-dimensional point cloud model of the displacement region; 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, and 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.

[0120] In one implementation process, after the binocular unit completes calibration, the control module first calls the two interval sub-region information locked in the monocular monitoring stage. The two sub-regions are non-adjacent sub-regions with displacement exceeding the standard in the monocular monitoring. The geometric center coordinates of the two sub-regions are extracted first, and the initial monitoring band is determined by taking the line connecting the two points as the axis. The monitoring band contains a range of half the width of the sub-region on both sides of the line, covering 3-4 continuous sub-regions, ensuring complete coverage of the potential associated deformation area of the two interval sub-regions.

[0121] The control module drives the first displacement mechanism to drive the binocular unit to perform three-dimensional scanning on the sub-regions in the initial monitoring band in sequence from one end sub-region to the other end sub-region.

[0122] Every time a sub-region is scanned, 2-3 groups of stereo image pairs are synchronously collected, and the three-dimensional coordinates of all feature points in the sub-region are obtained through epipolar rectification and stereo matching. The displacement of each feature point is calculated by comparing the three-dimensional coordinates with the initial calibration. If the displacement of more than 80% of the feature points in the initial monitoring band 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 line connecting region, and the subsequent range expansion stage is entered. If the proportion of feature points meeting the displacement standard is less than 80%, it is determined that the two interval sub-regions are independent displacements, and the range expansion is carried out for the individual sub-regions.

[0123] In the displacement range expansion scanning and boundary locking process, the gradient expansion method is used to expand scanning from the initial monitoring band to each sub-region in sequence. The expansion sequence is to expand along the dominant direction of the displacement of the two interval sub-regions first, and then to expand vertically. The specific process is as follows:

[0124] One-way expansion verification takes a boundary sub-region of the initial monitoring band 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 meeting the displacement standard is less than 60%, it is determined that the direction is the displacement boundary, and the expansion in that direction is stopped.

[0125] All-direction boundary locking completes the expansion scanning in the horizontal left, horizontal right, vertical up, and vertical down directions of the initial monitoring band in sequence according to the above logic until all directions meet the boundary sub-region with a proportion of feature points meeting the displacement standard less than 60%. All sub-regions meeting the displacement standard 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 this 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 area.

[0126] Based on the three-dimensional feature point data of all sub-regions in the displacement range, quantitative analysis of displacement direction and displacement amount is carried out.

[0127] Firstly, the displacement direction is determined, the three-dimensional displacement vector of all feature points in the displacement range is 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 area is greater than 90%, and the direction of the boundary area diverges slightly to the outside, it is determined as single-direction displacement; if there are two dominant directions, it is determined whether it is composite displacement combined with the geological structure.

[0128] The displacement amount is calculated in stages. The displacement range is divided into core area, transition area and boundary area. The average value of the displacement amount of all feature points in the core area, which is 1 / 3 of the area around the geometric center of the displacement range, is taken as the core displacement amount. The median of the displacement amount in the transition area, which is 1 / 3 of the area from the outside of the core area to the inside of the boundary area, is taken as the transition displacement amount. The maximum and minimum values of the displacement amount in the boundary area, which is 1 / 3 of the area at the edge of the displacement range, are taken as the boundary displacement amount range. The position of the feature point with the largest displacement amount is recorded as the subsequent key tracking object.

[0129] Combined with the quantitative data of displacement range, direction and amount, and the geological background of the slope, the displacement type is determined by multi-dimensional matching method: block sliding is determined if the displacement range is regular rectangular / block, the dominant displacement direction is consistent with the inclination of the dominant structure surface of the slope, 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 displacement and small rear displacement; wedge sliding is determined if the displacement range is triangular, 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 displacement and small rear displacement, and the cross calibration reference appears 1-2° deflection; dump deformation is determined if the displacement range is concentrated in the sub-area at the top of the slope, the dominant displacement direction is horizontal direction away from the slope, the feature points are vertically displaced upward, and the cross calibration reference is tilted upward on the outside of the horizontal arm; surface peeling is determined if the displacement range is scattered, the displacement direction of the feature points has no relevance, and the displacement amount is generally less than 5mm, and is concentrated in the weathered rock layer area at the middle and lower parts of the slope; the control module integrates all monitoring results in the form of structured data and visual data, and uploads to the cloud platform through the preset wireless communication protocol.

[0130] 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. A binocular multi-point displacement monitoring device, designed for temporary slope displacement monitoring, obtains displacement data based on image recognition of reference object positions on a time axis, characterized by: It includes a fixed frame (1) and a first displacement mechanism set on the fixed frame (1). The first displacement mechanism is provided with a mounting frame, and the mounting frame is provided with at least two independent displacement control second displacement mechanisms. Each second displacement mechanism is loaded with a monitoring camera (5). It also includes a control module, which is connected to the first displacement mechanism to control the synchronous displacement of all monitoring cameras (5) and connected to the second displacement mechanism to control the independent displacement of each monitoring camera (5); The control module controls each second displacement mechanism to drive the corresponding monitoring camera (5) to perform monocular inspection of several sub-regions of the slope surface according to a set cycle, acquire images of each sub-region and compare them with the preset initial calibration quantity; When the displacement of the reference object in the image exceeds the two-dimensional displacement warning threshold set according to the initial calibration quantity, the control module controls the second displacement mechanism corresponding to the adjacent paired monitoring camera (5) to perform displacement synchronous calibration, so that the adjacent paired monitoring camera (5) forms a binocular camera unit, and controls the first displacement mechanism to drive the binocular camera unit to acquire the three-dimensional data of the slope surface, and compares it with the initial calibration quantity and historical detection quantity to determine the displacement information.

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

3. The binocular multi-point displacement monitoring device according to claim 1, characterized in that: The mounting bracket is a long strip-shaped fixed rod (3). The middle part of the fixed rod (3) is fixed to the first displacement mechanism through a detachable mechanism. The second displacement mechanism is set in the length direction of the fixed rod (3) through a detachable mechanism. For slope surface displacement monitoring within a range of 20-100m from the fixed frame (1), the spacing between adjacent second displacement mechanisms set on the fixed rod (3) is 25-40cm.

4. A monitoring method, characterized in that: The specific steps for slope displacement monitoring using the binocular multi-point displacement monitoring equipment described in any one of claims 1-3 are as follows: Step 100: Determine the range of the slope surface to be monitored, set up several fixed reference objects (6) on the slope surface, and obtain positioning parameters from the fixed reference objects (6) through the positioning device; Step 200: Select stable terrain within 50-200m of the slope surface to set up binocular multi-point displacement monitoring equipment, set up the fixed frame (1), the first displacement mechanism, the mounting frame, at least two second displacement machines and the corresponding monitoring cameras (5), set up the control module on the ground, debug it and connect to the cloud for data interaction; Step 300: Using the debugged binocular multi-point displacement monitoring equipment, divide the slope surface into several sub-regions and number them with the maximum monitoring field of the monitoring camera (5). Then, obtain the initial calibration value formed by image processing of all sub-regions through the monitoring camera (5) and set the two-dimensional displacement early warning threshold according to the initial calibration value. Step 400: The binocular multi-point displacement monitoring device performs monocular inspections of the slope surface according to the set cycle. During monocular inspection, the control module controls the monitoring camera to sequentially inspect several sub-areas assigned to it. After acquiring images, the images are compared with the initial calibration quantity and historical detection quantity. If the two-dimensional displacement warning threshold is exceeded, an warning is sent to the cloud and the binocular inspection process is activated. Step 500: After activating the binocular inspection process, the control module controls the second displacement mechanism of the adjacent paired monitoring cameras to perform synchronous displacement calibration. 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 compare it with the initial calibration quantity and historical detection quantity. After determining the displacement information, it is fed back to the cloud.

5. The monitoring method according to claim 4, characterized in that: Step 100 also includes natural reference points, which are fixed natural terrain features on the slope surface; The fixed reference object (6) 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. When obtaining positioning parameters through the positioning device, the positioning device measures the three-dimensional coordinates of the geometric center of each reference object, the three-dimensional coordinates of the intersection point of the crosshair calibration reference object, and the arm length, records the three-dimensional coordinates of the contour feature points of the natural reference object, and stores all positioning parameters and feature information in the control module.

6. The monitoring method according to claim 4, characterized in that: The debugging and calibration in step 200 includes: Two-dimensional calibration: The control module controls the second displacement mechanism corresponding to each monitoring camera, which drives the monitoring camera to adjust its angle independently, so that each monitoring camera is aligned with the preset initial sub-region of the slope surface, takes a picture of the calibration plate in the initial sub-region, and calculates and stores the monocular intrinsic parameters of each monitoring camera. 3D Preset Calibration: Based on a binocular camera unit formed by two paired monitoring cameras, a preset crosshair calibration reference is used as the 3D calibration benchmark. The control module pre-stores the 3D coordinates and image features of the crosshair calibration reference, controls the second displacement mechanism to move the binocular camera unit to the position aligned with the crosshair calibration reference, synchronously captures the image of the crosshair calibration reference, calculates the binocular extrinsic parameters, and adjusts the angle of the monitoring camera in the binocular camera unit according to the stored standard binocular extrinsic parameter data to meet the binocular state requirements.

7. The monitoring method according to claim 4, characterized in that: In step 300, when dividing and numbering sub-regions, the slope surface is divided into several sub-regions in the form of a row and column matrix, with the maximum monitoring field of view of the monitoring camera as the unit. All 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, which are respectively assigned to two monitoring cameras. The initial calibration quantity includes: the positioning information and size information of all fixed reference objects (6) and natural reference objects in each sub-region, as well as the topographic and geomorphological spatial information of the corresponding sub-region of the slope. A two-dimensional displacement early warning threshold is set based on the positioning information of the fixed reference objects (6) and natural reference objects in the initial calibration quantity.

8. The monitoring method according to claim 6, characterized in that: In step 500, the specific process of displacement synchronization calibration includes: The second displacement mechanism is controlled to move the adjacent paired monitoring cameras to align with the preset crosshair calibration reference object, and simultaneously capture images of the crosshair calibration reference object; Extract the pixel coordinates of the cross intersection of the cross calibration reference in the image, combine them with its preset three-dimensional coordinates, calculate the difference between the binocular extrinsic parameters, determine the displacement of the corresponding second displacement mechanism based on the difference, and move it. Finally, reduce the difference to meet the standard binocular extrinsic parameter data requirements to complete the calibration. Then, a reference test is performed. The current coordinates of the cross calibration reference in other sub-regions are photographed and compared with the coordinates of the cross calibration reference in the initial calibration. If the deviation exceeds the allowable range, the calibration step is repeated.

9. The monitoring method according to claim 8, characterized in that: In step 500, the specific process for determining the displacement information includes: First, determine the activation conditions for activating the binocular inspection process. The activation conditions include the displacement of a single fixed reference object (6) in the same frame captured by a single monitoring camera exceeding the two-dimensional displacement warning threshold, the relative positional displacement of the fixed reference object (6) in different frames captured by a single monitoring camera exceeding the two-dimensional displacement warning threshold, and the displacement of the fixed reference object (6) in frames captured by different monitoring cameras exceeding the two-dimensional displacement warning threshold. Based on the activation conditions, two fixed reference objects (6) that exceed the two-dimensional displacement warning threshold are determined to identify the corresponding sub-region pair. The first displacement mechanism is controlled to drive the binocular camera unit to acquire the three-dimensional data of all sub-regions on the straight line connecting the fixed reference objects (6) of the sub-region pair. Then, the displacement is calculated by expanding at least one adjacent sub-region to the outside until the displacement is less than the threshold and the displacement range is determined. Based on the displacement range and displacement amount of the slope surface, combined with the topographic changes in the three-dimensional data, the displacement type is determined, and the displacement information is integrated and uploaded to the cloud.

Images