High slope surface deformation thunder-vision and sensing integrated monitoring system and method

Abstract

The invention discloses a high slope surface deformation thunder-vision and sensing integrated monitoring system and method, and belongs to the technical field of geotechnical engineering safety monitoring. A coordinate conversion relation is established through joint calibration of a millimeter wave radar with a built-in Beidou chip and a binocular camera; radar point cloud and binocular image data are synchronously acquired by using Beidou second pulse signals; the radar point cloud is converted to a visual coordinate system through calibration external parameters, pixel-level alignment is carried out, and true color three-dimensional point cloud is generated through fusion; performing denoising filtering and time sequence analysis on the point cloud data, and extracting displacement and settlement deformation characteristics of the slope surface; and comparing the deformation feature with a preset safety threshold to trigger intelligent early warning. According to the invention, through integrated fusion of thunder vision and sensing, the defect of sensing limitation or insufficient three-dimensional modeling capability of a single sensor in severe weather is overcome, millimeter-level precision, all-weather and three-dimensional visual monitoring and accurate early warning of high slope surface deformation are realized, and the reliability of slope safety management is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of geotechnical engineering safety monitoring technology, and in particular to an integrated monitoring system and method for surface deformation monitoring of high slopes using both radar and sensors. Background Technology

[0002] The stability of high slopes is crucial for the safe operation of major infrastructure projects such as civil engineering, transportation, and water conservancy. Compared to ordinary slopes, high slopes are characterized by steepness, complex geological structures, deep burial of potential sliding surfaces, and multidimensional deformation mechanisms. Therefore, deformation monitoring technology for high slopes not only needs to have millimeter-level precision to capture initial creep, but also needs to have the ability to accurately reconstruct large-scale three-dimensional deformation fields, thereby achieving precise characterization of the spatial morphology of the sliding surface and effective early warning of disaster trends.

[0003] Traditional monitoring methods (such as total stations and inclinometers) have high measurement accuracy, but they are limited by point deployment and lack of automation, which cannot meet the urgent need for large-scale, full-surface continuous monitoring of high slopes.

[0004] With the development of remote sensing technology, non-contact monitoring technologies such as Synthetic Aperture Radar (SAR) and Ground-Based Interferometric Radar (GB-InSAR) have provided new solutions for slope monitoring. In particular, sensing technologies represented by millimeter-wave radar, with their wavelengths much longer than light waves, can effectively penetrate adverse weather conditions such as rain, fog, and dust, enabling all-weather, large-scale deformation monitoring. For example, Chinese patent CN117395622A discloses a slope monitoring solution integrating communication and sensing. Based on a 5G millimeter-wave base station, it integrates communication and sensing functions through air interface fusion processing technology, which reduces system deployment costs and expands the monitoring range to a certain extent.

[0005] However, practice has shown that the existing technology system based on millimeter-wave radar suffers from insufficient elevation angle resolution, which is particularly pronounced in high slope scenarios, thus becoming a key technical bottleneck restricting its effectiveness. This deficiency mainly stems from the physical limitations of the antenna array in the vertical direction of millimeter-wave radar (typically only 1-4 antennas), resulting in its vertical sensing capability being far lower than that in the horizontal dimension. The stability-sensitive deformations of high slopes are precisely concentrated in vertical settlement and shear displacement along the potential sliding surface, making it difficult for existing systems to directly construct high-precision, measurable 3D visualization models. Consequently, they cannot accurately capture key deformation information such as subtle vertical settlement and crack propagation on the slope surface, nor can they effectively reconstruct the 3D morphology of the sliding surface. This deficiency leads to monitoring results for high slopes mostly remaining at the macroscopic alarm level, lacking the ability to deeply analyze the deformation mechanism.

[0006] Therefore, in the face of the special and complex monitoring needs of high slopes, there is an urgent need in this field for an innovative solution that can combine the advantages of millimeter-wave radar working in all weather conditions with high-precision three-dimensional perception capabilities, so as to break through the existing technical bottlenecks, realize the leap from "deformation detection" to "mechanism analysis", and provide reliable technical support for the safety management and control of high slopes. Summary of the Invention

[0007] This invention overcomes the shortcomings of the prior art and provides an integrated monitoring system and method for surface deformation monitoring of high slopes using both radar and sensors.

[0008] To achieve the above objectives, the technical solution adopted by this invention is: a method for integrated monitoring of surface deformation of high slopes using both radar and sensors, comprising:

[0009] S1. Simultaneously acquire radar point cloud data and binocular image data of the high slope surface;

[0010] S2. The radar point cloud data is fused with the binocular image data to generate color point cloud data with color information;

[0011] S3. Based on the color point cloud data in the time series, extract the deformation features of the high slope surface;

[0012] S4. Compare the deformation feature with a preset safety threshold, and generate a warning signal when the threshold is exceeded.

[0013] In a preferred embodiment of the present invention, in step S1, the synchronous acquisition is achieved through timing signals provided by the BeiDou positioning system.

[0014] In a preferred embodiment of the present invention, in step S2, the data fusion method is mainly based on a joint calibration algorithm to obtain the transformation extrinsic parameters between the radar coordinate system and the binocular vision coordinate system;

[0015] The radar point cloud data is converted to a binocular vision coordinate system using the aforementioned transformation extrinsic parameters, and then pixel-level alignment is performed.

[0016] In a preferred embodiment of the present invention, the transformation extrinsic parameters are initially solved using the PnP algorithm and further optimized using the ICP algorithm.

[0017] In a preferred embodiment of the present invention, in step S3, the extracted deformation features include at least one of displacement, settlement, or crack propagation.

[0018] In a preferred embodiment of the present invention, in step S3, the high slope is divided into multiple monitoring areas based on the geological conditions or slope structure, and the deformation features are extracted by area.

[0019] In a preferred embodiment of the present invention, before step S3, a step of performing noise reduction filtering on the color point cloud data is further included.

[0020] A high slope surface deformation integrated radar and sensor monitoring system includes:

[0021] Millimeter-wave radar is used to collect radar point cloud data of high slopes;

[0022] A binocular camera, synchronized with the millimeter-wave radar, is used to acquire binocular image data;

[0023] A data processing module, wherein the data processing module is used to perform data fusion and deformation analysis; and,

[0024] An early warning module is used to generate and send early warning signals.

[0025] In a preferred embodiment of the present invention, the millimeter-wave radar has a built-in Beidou chip; the data processing module is connected to the cloud server through the communication module built into the millimeter-wave radar to realize integrated sensing and data transmission.

[0026] In a preferred embodiment of the present invention, the early warning module is configured to: overlay the colored point cloud data with a geological information model, and perform stability assessment and early warning based on the evolution of the sliding surface morphology.

[0027] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0028] This invention innovatively integrates millimeter-wave radar, binocular vision, and the BeiDou Navigation Satellite System into a "Radar-Vision-Sensing Integrated" monitoring architecture. The core of this architecture lies in utilizing BeiDou's second pulse to achieve hardware-level synchronization, establishing a unified spatiotemporal reference for heterogeneous data. Based on this, the anti-interference distance information provided by the millimeter-wave radar and the high-resolution texture and elevation angle information provided by the binocular vision system naturally complement each other. This design directly overcomes the inherent weakness of single radar in vertical sensing capabilities and avoids the failure risk of pure vision systems under adverse weather conditions. Compared to existing technologies that rely on only a single sensor source, this invention achieves true millimeter-level precision full-field three-dimensional deformation visualization in high slope monitoring for the first time, elevating monitoring capabilities from discrete point and line perception to continuous holographic three-dimensional volume perception, thus providing unprecedented data support for precise disaster analysis.

[0029] This invention employs key technologies based on joint calibration and extrinsic parameter optimization to achieve deep fusion of multi-source data. To avoid the problems caused by point cloud noise and feature matching errors when directly using the PnP algorithm to solve extrinsic parameters in traditional techniques, this invention designs a two-stage optimization strategy combining PnP and ICP. First, a near-realistic initial solution is quickly obtained using the PnP algorithm. Then, the ICP algorithm is introduced, utilizing the planar geometric information of the entire point cloud rather than a few feature points to iteratively optimize the initial solution. The ICP process effectively smooths local errors by minimizing the overall distance between the radar point cloud and the visual point cloud, thereby improving the calibration accuracy to the sub-pixel level. This approach ensures accurate spatial matching between the radar point cloud and the visual image, enabling the generation of high-quality color point cloud data with both geometric accuracy and realistic RGB texture, thus perfectly combining the geometric shape and texture information of the slope surface. Compared to existing technologies where millimeter-wave radar point cloud data lacks semantic information and pure binocular vision camera data models suffer from insufficient geometric accuracy, the fused data generated by this invention combines the geometric accuracy of millimeter-wave radar with the visual detail of a binocular vision camera. This enables automated machine identification and quantitative analysis of defects such as cracks and spalling. Furthermore, this fusion algorithm significantly improves the intuitiveness and interpretability of monitoring results, allowing managers to directly observe and measure the degree of deformation on realistic 3D models. This greatly lowers the professional threshold for data analysis and improves decision-making efficiency.

[0030] This invention addresses the challenge of monitoring uneven geological conditions on high slopes, breaking through the traditional mindset of holistic monitoring and proposing a zoned monitoring concept based on geological models. By treating a large-scale slope as a complex system composed of different risk units, it independently calculates the deformation data of each area and, combined with real-time denoising filtering technology, effectively extracts early signs of local instability. This system-level solution constructs a complete closed loop from accurate perception and reliable transmission to intelligent analysis, ultimately achieving a leap from "uniform threshold alarm" to "predictive early warning based on risk zoning," elevating slope safety management from passive response to proactive early warning, and greatly enhancing the disaster prevention and mitigation capabilities of engineering projects. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart of an integrated radar and sensor monitoring system and method for surface deformation monitoring of high slopes, according to the present invention.

[0033] Figure 2 This is an architecture diagram of an integrated radar and sensor monitoring system for surface deformation of high slopes according to the present invention.

[0034] Figure 3 This is a flowchart illustrating the zonal monitoring and intelligent early warning logic of an integrated radar and sensor monitoring system and method for surface deformation of high slopes according to the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0037] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the scope of protection of this application. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0038] Application Overview:

[0039] This invention addresses a core challenge in high slope safety monitoring: achieving millimeter-level precision, visualization, and intelligent early warning of the three-dimensional deformation field of slope surfaces while ensuring all-weather reliability. Traditional optical monitoring technologies are limited by weather conditions, making reliability difficult to guarantee; while millimeter-wave radar-based monitoring schemes, although possessing strong penetration, suffer from insufficient accuracy in three-dimensional model reconstruction due to their inherent limitation in elevation angle resolution, making it difficult to support precise deformation mechanism analysis. To overcome this bottleneck, this invention proposes an integrated "radar-visual-sensing" technology approach. This involves deeply integrating a millimeter-wave radar with a built-in BeiDou chip and a binocular vision system at the hardware level, and constructing a unified perception coordinate system based on joint calibration and data fusion algorithms. This scheme fully leverages the complementary advantages of radar's anti-interference ranging and high-resolution visual imaging, and introduces the BeiDou system to solve the problems of scale reference and time synchronization. Thus, under complex environmental conditions, it achieves holographic, measurable, and intelligent monitoring of high slope surface deformation, from macroscopic displacement to minute cracks, significantly improving the predictability and reliability of slope safety management.

[0040] Example 1:

[0041] Transforming the aforementioned theoretical concept of "integrated radar-vision-sensing" into a feasible high-precision monitoring solution requires systematically addressing the core technological challenges arising from sensor heterogeneity. The essence of this challenge lies in how to establish and maintain a long-term, stable, and spatiotemporally precise collaborative sensing relationship between millimeter-wave radar and binocular cameras, both based on different physical principles (electromagnetic wave detection and optical imaging), in complex field engineering environments.

[0042] like Figure 1 As shown, a method for integrated radar and sensor monitoring of surface deformation on high slopes includes:

[0043] S1. Simultaneously acquire radar point cloud data and binocular image data of the high slope surface;

[0044] S2. The radar point cloud data is fused with the binocular image data to generate color point cloud data with color information;

[0045] S3. Based on the color point cloud data in the time series, extract the deformation features of the high slope surface;

[0046] S4. Compare the deformation feature with a preset safety threshold, and generate a warning signal when the threshold is exceeded.

[0047] Specifically, the technology first faces the challenge of unifying spatial coordinates, namely, how to solve the rigid body transformation parameters from the radar coordinate system to the camera coordinate system through a high-precision calibration algorithm. Any tiny calibration residual will lead to significant error accumulation in long-distance monitoring. Secondly, there is the challenge of time synchronization consistency, which requires ensuring strict temporal matching of heterogeneous data streams to avoid introducing temporal mismatch noise. Finally, there is the challenge of effective data fusion, which requires designing a fusion rule that can effectively combine the precise geometric information of millimeter-wave radar with the rich texture information of binocular cameras to generate a three-dimensional data volume that can be used for precise quantitative analysis.

[0048] Synchronous acquisition is designed to solve the problem of spatiotemporal synchronization by collecting data from multiple sensors at the same microsecond level, using a hardware synchronization mechanism based on BeiDou pulse signals.

[0049] Preferably, the millimeter-wave radar has a built-in Beidou chip, and the high-precision PPS (Pulse Per Second) signal output by the chip is directly connected to the external hardware trigger interface of the millimeter-wave radar and the binocular camera.

[0050] This hardware triggering method achieves microsecond-level or even nanosecond-level time synchronization at the physical level, thus fundamentally eliminating the uncertainty caused by software triggering or network latency, and laying an unparalleled solid foundation for the time alignment of multi-source data.

[0051] Radar point cloud data is a dataset containing information such as the three-dimensional coordinates and reflection intensity of the target point, and is acquired using millimeter-wave radar.

[0052] Binocular image data consists of two-dimensional digital images captured by a binocular camera that can be used for 3D reconstruction.

[0053] Data fusion is the precise spatial overlay of radar point cloud data and binocular image data;

[0054] It should be noted that, in order to address the challenge of coordinate system issues for radar point cloud data and binocular image data, a joint calibration method based on a specially designed large calibration board is adopted.

[0055] Specifically, during the system deployment phase, multiple (at least 5, recommended 8-12 to cover all areas of the field of view and include different depth information) large checkerboard calibration boards (recommended size not less than 2m x 2m to ensure clear identification at long distances) are moved within the shared field of view of the millimeter-wave radar and binocular camera in different poses and data is collected simultaneously.

[0056] Furthermore, the calibration plate plane in the radar point cloud is robustly fitted using the RANSAC (Random Sample Consensus) algorithm, and the three-dimensional coordinates of the virtual corner point are calculated by solving the intersection of three non-parallel planes.

[0057] Furthermore, the high-precision checkerboard corner detection function in the OpenCV (Open Source Computer Vision Library) library is used to extract sub-pixel level corner coordinates from the camera image.

[0058] The generation of color point cloud data is mainly achieved through calculation methods such as coordinate transformation and pixel projection, so that each three-dimensional point is assigned a point cloud with the corresponding pixel RGB value, thereby forming a true color three-dimensional model.

[0059] Specifically, the PnP (Perspective-n-Point) algorithm is used to solve for the initial extrinsic parameter rotation matrix. Translation vector .

[0060] Furthermore, to further improve accuracy, the ICP (Iterative Closest Point) algorithm is used for iterative optimization. The calibration board point cloud obtained from the entire millimeter-wave radar scan is registered with the dense point cloud of the calibration board reconstructed by the binocular camera vision, minimizing the distance between point pairs and outputting the optimized final extrinsic parameters (R, t).

[0061] Furthermore, the calibration accuracy was verified through reprojection error analysis, requiring the average reprojection error to be less than 0.5 pixels to ensure high accuracy of spatial alignment.

[0062] Furthermore, addressing the challenge of data fusion, the radar point cloud is transformed to the camera coordinate system using high-precision extrinsic parameters (R, t) obtained through calibration. Then, the 3D points are projected onto the 2D image plane using camera intrinsic parameters (obtained in advance through camera calibration), and their pixel coordinates are calculated. , ).

[0063] Furthermore, the RGB color information of the pixel location is extracted and assigned to the corresponding radar point cloud point.

[0064] It should be noted that for points whose coordinates are not integers after projection, bilinear interpolation is used to obtain smooth color values, ultimately generating a true-color 3D point cloud that combines radar millimeter-level geometric accuracy with real texture information from the camera.

[0065] The color point cloud data based on time series mainly refers to multi-period point cloud data arranged in chronological order. The system automatically executes the above acquisition and fusion process according to a preset cycle (which can be set to 1 hour, 6 hours or 24 hours according to the stability of the slope) to generate a time series color point cloud dataset.

[0066] Extracting deformation features mainly refers to aligning the current frame point cloud with the selected historical reference point cloud using a point cloud registration algorithm (such as ICP or its improved algorithm), and calculating the overall displacement field and rigidity transformation parameters of the point cloud.

[0067] Deformation characteristics mainly refer to parameters used to quantify slope changes, such as displacement, settlement, and crack propagation.

[0068] Furthermore, abnormal deformations in local areas can be identified through change detection techniques (such as distance threshold method and statistical analysis method), such as the extent of subsidence basins, the extension length and width of cracks, etc.

[0069] The safety threshold is a critical value set according to slope design specifications and stability analysis to trigger an alarm;

[0070] The warning signal is generated and issued by the system when the data exceeds the safety threshold.

[0071] Specifically, the warning signal is triggered when the extracted deformation feature values ​​(such as maximum displacement, average settlement rate, and crack propagation area) exceed the safety threshold set comprehensively based on slope design specifications, historical monitoring data, and geological survey results.

[0072] Furthermore, early warning information can be sent to relevant management personnel through various means (such as audible and visual alarms at the monitoring center, SMS messages, and message push notifications from dedicated platforms), and a monitoring report containing deformation cloud maps, change curves, and brief analysis conclusions can be automatically generated.

[0073] The basic framework provided in this embodiment successfully solves the core challenges in radar-visual fusion, realizing automated and three-dimensional monitoring of surface deformation on high slopes, and laying the foundation for more refined monitoring solutions in the future.

[0074] Example 2:

[0075] In the basic joint calibration process described in Example 1, the extrinsic parameters solved directly using the PnP algorithm are easily affected by point cloud noise, corner extraction errors, and inaccurate matching. To solve this problem, the core of this embodiment lies in adopting a two-stage optimization strategy consisting of a preliminary solution using the PnP algorithm and iterative optimization using the ICP algorithm. The technical idea is as follows:

[0076] The PnP algorithm is used to quickly obtain an initial solution that is close to reality and can be used as the starting point for optimization. Then, the ICP algorithm is used to perform fine iteration based on this initial solution, which does not depend on specific feature points (but uses information from the entire point cloud surface). This effectively eliminates noise and local errors and achieves the best spatial alignment at the point cloud level.

[0077] Based on the geological survey report and topographic survey data of the high slope, the deployment location of the monitoring unit is selected. The location must be stable (such as where bedrock is exposed) and have a wide field of vision, covering the entire potential deformation area.

[0078] Preferably, the monitoring unit adopts an integrated design, encapsulating the millimeter-wave radar, binocular camera, and BeiDou chip within a housing with an IP67 protection rating. The BeiDou chip is directly connected to the millimeter-wave radar's main control chip via a high-speed SPI bus. Its centimeter-level precision positioning data serves as the absolute spatial reference for the radar point cloud, while its high-precision PPS (pulse per second) signal is directly connected to the external hardware trigger interfaces of the radar and camera, achieving microsecond-level time synchronization at the hardware level. This lays a solid foundation for subsequent data spatiotemporal alignment.

[0079] After deployment, joint calibration is performed. Specifically, a large checkerboard calibration board (e.g., 2m×2m) is set up on-site and moved to multiple different postures (at least 5 groups, covering all areas of the field of view) within the shared field of view of the radar and camera. For each posture, radar point cloud data and binocular camera images are collected simultaneously.

[0080] Regarding the extraction of radar virtual corner points:

[0081] For each set of radar point cloud data acquired synchronously with the calibration board attitude, the RANSAC (Random Sample Consensus) algorithm is used for robust plane fitting. This algorithm can effectively remove stray noise points in the point cloud, accurately fit the plane where the calibration board is located, and obtain its plane equation:

[0082]

[0083] in,( , , ) is the plane normal vector. This is the intercept.

[0084] Each interior corner of the calibration plate can be considered as the intersection of three non-parallel planes. The three-dimensional coordinates (x, y, z) of this corner in the radar coordinate system can be calculated by solving the following system of linear equations:

[0085]

[0086] Write in matrix form ,in:

[0087] , ,

[0088] Solution .

[0089] This method yields the three-dimensional coordinates of the virtual corner point in the radar coordinate system. It also allows for the acquisition of the set of three-dimensional coordinates for all interior corner points on the calibration board. ,in =1,2,…,N, where N is the total number of corner points.

[0090] Camera corner extraction:

[0091] For synchronously acquired stereo camera images, functions from the OpenCV library are used to detect checkerboard corner points and obtain their sub-pixel coordinates. Given the physical dimensions of the calibration plate, the 3D coordinates of each corner point in the world coordinate system can be calculated. .

[0092] Calculation and optimization of extrinsic parameters:

[0093] Corner points extracted by radar Corner points detected by the camera Match in sequence. Use the cv2.solvePnP function in OpenCV to solve for the initial rotation matrix. Translation vector The transformation relationship solved by this function is:

[0094]

[0095] To improve accuracy, the ICP (Iterative Closest Point) algorithm was further optimized. This was applied to the radar point cloud. Projected onto the camera coordinate system using initial extrinsic parameters:

[0096]

[0097] Next, we'll work with the camera's dense point cloud. Perform iterative alignment of ICP (Iterative Closest Point):

[0098]

[0099] in, , It is a nearest neighbor pair.

[0100] Output optimized extrinsic parameters .

[0101] Calibration and verification:

[0102] Verify calibration accuracy by checking reprojection error. This involves using 3D points in the point cloud. Transform to the camera coordinate system using extrinsic parameters:

[0103]

[0104] Then use the camera's intrinsic parameters (focal length) Main point Project it onto the image plane:

[0105]

[0106]

[0107] Calculate projection points With true corner Euclidean distance error:

[0108]

[0109] The average reprojection error is required to be less than 0.5 pixels.

[0110] Regarding data fusion and color point cloud generation:

[0111] External parameters obtained from calibration Transform the radar point cloud from the radar coordinate system to the camera coordinate system:

[0112]

[0113] Project the transformed point cloud onto the image plane. For points... Calculate its pixel coordinates :

[0114]

[0115]

[0116] Extract the RGB color information of the pixel and assign it to the corresponding point cloud point to generate a true-color 3D point cloud.

[0117] Example 3:

[0118] Based on Example 1, this embodiment optimizes the data processing flow and monitoring strategy, focusing on point cloud denoising and filtering and slope regional monitoring methods, aiming to improve data reliability and the pertinence of early warning.

[0119] Preferably, point cloud data denoising and filtering processing is performed:

[0120] In the data acquisition and fusion process described in Example 1, the obtained point cloud data often contains outliers caused by environmental interference (such as birds and vegetation), sensor noise, etc., as well as random noise caused by measurement accuracy limitations. These noises can seriously affect the accuracy of deformation feature extraction.

[0121] Therefore, introducing denoising filtering before step S3 (deformation feature extraction) is crucial. Denoising refers to eliminating or reducing interference components in the data through algorithms, preserving the true and valid signal to improve monitoring accuracy and reliability. This embodiment employs a denoising strategy combining spatial and temporal domains, as detailed below:

[0122] Spatial statistical filtering (for outlier noise): This method aims to filter out obvious isolated noise points.

[0123] First, for each point in the point cloud, calculate the average distance to its fifty nearest neighbor points.

[0124] Furthermore, the mean and standard deviation of all average distances in the entire point cloud are calculated.

[0125] Furthermore, points whose average distance is far above or far below the mean (e.g., beyond the mean plus one and a half standard deviations) are removed, and only points whose average distance is within a reasonable statistical range are retained, thereby effectively eliminating outliers and noise.

[0126] Preferably, time-domain averaging filtering (for random noise): This method aims to smooth the data and suppress high-frequency random fluctuations.

[0127] First, the system maintains a time window containing the most recent ten consecutive frames of data. The point cloud of the current frame is registered and aligned with the point clouds of historical frames within the time window to eliminate the overall displacement effect caused by slope deformation.

[0128] Furthermore, for points that coincide or are adjacent in spatial location, their three-dimensional coordinates are arithmetically averaged within a time window, and the coordinates of that point in the current frame are updated with this average value. This effectively suppresses random noise and highlights the true deformation trend.

[0129] For high slopes with complex geological conditions and a wide area, it is difficult to accurately reflect the risk of local instability using overall monitoring and single-threshold early warning modes. This embodiment introduces the concept of zoned monitoring to achieve differentiated and precise control.

[0130] Preferably, the slope is monitored by region:

[0131] a. Divide the monitoring area: During the system initialization phase, based on the engineering geological survey report, topographic map and initial 3D model of the high slope, it is divided into multiple monitoring areas.

[0132] Specifically, the zoning is based on the properties of the soil and rock mass (such as strongly weathered areas and weakly weathered areas), the slope structure (such as stepped areas and overall slopes), historical deformation patterns, and risk levels.

[0133] b. Feature extraction by partition: In step S3, the extraction of deformation features is performed independently by partition.

[0134] Specifically, the system automatically calculates deformation indices for each region, such as the region's average displacement vector, the region's maximum displacement, and the zone's average displacement rate. For identified cracks, it records their zone, length, width, and propagation rate.

[0135] c. Threshold determination: In step S4, an independent warning threshold is set for each partition.

[0136] Specifically, these thresholds are set differently based on the region's geological conditions, design parameters, and historical stability analysis results. When the deformation index of a certain zone exceeds its specific threshold, the system generates a clear zone warning signal, clearly indicating the specific zone that was triggered, making management measures more targeted.

[0137] d. Generate early warning information: Generate regional early warning information and comprehensive monitoring reports.

[0138] Method for calculating displacement of high slopes by region:

[0139] Dividing a large-scale slope into multiple characteristic sub-regions and independently calculating deformation data for each region can accurately capture the relationship between local instability signs and overall coordinated deformation.

[0140] This method overcomes the shortcomings of traditional overall monitoring in responding slowly to local abrupt changes. By comparing and analyzing displacement fields between regions, it can identify potential sliding surface locations and cascading failure risks, providing spatial targeting basis for differentiated reinforcement decisions. It is particularly suitable for complex scenarios such as stepped slopes or heterogeneous geological bodies, significantly improving the precision of large-area slope monitoring and the effectiveness of disaster early warning.

[0141] This embodiment significantly improves the purity of data and the accuracy of early warning by introducing noise reduction filtering and zone monitoring, thus realizing differentiated and refined intelligent management and control of high slopes.

[0142] Example 4:

[0143] A high slope surface deformation integrated radar and sensor monitoring system, whose core hardware lies in the deep integration of millimeter-wave radar, binocular camera and Beidou chip, as well as the integrated sensor data transmission design.

[0144] Preferably, the millimeter-wave radar is responsible for transmitting frequency-modulated continuous waves and receiving echoes, and for obtaining the target's range, velocity, and azimuth information through signal processing.

[0145] Preferably, the binocular camera consists of two precisely calibrated cameras used to acquire high-resolution stereo image pairs and reconstruct the three-dimensional geometry and texture information of the scene through stereo vision algorithms or photogrammetry software.

[0146] Preferably, the BeiDou chip, serving as the core of the system's spatiotemporal reference, is directly connected to the millimeter-wave radar main control chip via a high-speed SPI bus. It provides three key functions:

[0147] Preferably, absolute geographic coordinates with centimeter-level accuracy are used as the absolute spatial reference for the radar point cloud;

[0148] Preferably, a high-precision timestamp is used for data labeling;

[0149] Preferably, the PPS second pulse signal is directly connected to the external hardware trigger interface of the radar and camera to achieve microsecond-level data acquisition synchronization.

[0150] Preferably, the integrated sensing transmission unit:

[0151] The millimeter-wave radar integrates a 4G / 5G communication module. This module not only transmits the raw or pre-processed fused data to the cloud in real time, but also achieves an integrated sensing and communication design. This means that by utilizing the physical layer characteristics of the communication link, partial or complete data fusion of multiple sensor sources is performed directly in the wireless communication channel during data transmission, rather than through centralized post-processing. This significantly reduces end-to-end latency and improves the real-time performance of monitoring.

[0152] Preferably, a cloud processing and early warning unit:

[0153] Data processing module: Located on the cloud server, it is responsible for receiving data transmitted back by the integrated sensor and executing core algorithms such as data decoding, point cloud fusion, noise reduction filtering (as described in Example 3), time series analysis, and deformation feature extraction.

[0154] Early warning module: Also located in the cloud, it incorporates zone monitoring and threshold comparison logic (as described in Example 3). When deformation exceeds the limit, it automatically generates an early warning message and sends it to the user terminal through multiple channels such as API interface, SMS, and email. Simultaneously, the system automatically generates a comprehensive monitoring report including a deformation distribution map, historical data curves, and evaluation results.

[0155] The system architecture described in this embodiment, through deep hardware integration and integrated sensor design, seamlessly connects perception, communication and intelligent decision-making, forming a complete and efficient closed-loop system for slope safety monitoring, which is the physical basis for realizing the aforementioned method innovation.

[0156] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for integrated radar and sensor monitoring of surface deformation on high slopes, characterized in that, include: S1. Simultaneously acquire radar point cloud data and binocular image data of the high slope surface; S2. The radar point cloud data is fused with the binocular image data to generate color point cloud data with color information; S3. Based on the color point cloud data in the time series, extract the deformation features of the high slope surface; S4. Compare the deformation feature with a preset safety threshold, and generate a warning signal when the threshold is exceeded.

2. The integrated radar and sensor monitoring method for surface deformation of high slopes according to claim 1, characterized in that, In step S1, the synchronous acquisition is achieved through timing signals provided by the BeiDou positioning system.

3. The integrated radar and sensor monitoring method for surface deformation of high slopes according to claim 1, characterized in that, In step S2, the data fusion method is mainly based on a joint calibration algorithm to obtain the transformation extrinsic parameters between the radar coordinate system and the binocular vision coordinate system; The radar point cloud data is converted to a binocular vision coordinate system using the aforementioned transformation extrinsic parameters, and then pixel-level alignment is performed.

4. The integrated radar and sensor monitoring method for surface deformation of high slopes according to claim 3, characterized in that, The transformation extrinsic parameters are initially solved using the PnP algorithm and further optimized using the ICP algorithm.

5. The integrated radar and sensor monitoring method for surface deformation of high slopes according to claim 1, characterized in that, In step S3, the extracted deformation features include at least one of displacement, settlement, or crack propagation.

6. The integrated radar and sensor monitoring method for surface deformation of high slopes according to claim 5, characterized in that, In step S3, the high slope is divided into multiple monitoring areas based on its geological conditions or slope structure, and the deformation features are extracted from each area.

7. The integrated radar and sensor monitoring method for surface deformation of high slopes according to claim 1, characterized in that, Before step S3, the method further includes a step of performing noise reduction filtering on the color point cloud data.

8. A high slope surface deformation integrated radar and sensor monitoring system, used to implement the method according to any one of claims 1 to 7, characterized in that, include: Millimeter-wave radar is used to collect radar point cloud data of high slopes; A binocular camera, synchronized with millimeter-wave radar, is used to acquire binocular image data; The data processing module is used to perform data fusion and deformation analysis; as well as, An early warning module is used to generate and send early warning signals.

9. The integrated radar-visual-sensor monitoring system for surface deformation of high slopes according to claim 8, characterized in that, The millimeter-wave radar has a built-in Beidou chip; the data processing module connects to the cloud server through the communication module built into the millimeter-wave radar to realize integrated sensing and data transmission.

10. The integrated radar-visual-sensor monitoring system for surface deformation of high slopes according to claim 9, characterized in that, The early warning module is configured to overlay the colored point cloud data with the geological information model and perform stability assessment and early warning based on the evolution of the sliding surface morphology.

Images