Method and system for monitoring stability of rock slope in real time based on three-dimensional laser scanner and deep sensor

Abstract

The invention discloses a rock slope stability real-time monitoring method and system based on a three-dimensional laser scanner and a deep sensor, and relates to the technical field of geological disaster monitoring and early warning. The method comprises the steps that multi-temporal point cloud data and a deep monitoring sequence are acquired, point cloud frames are constructed to be bound with a deep sequence window, and a point cloud frame index set and a deep abnormal event index set are output; converting abnormal space coordinates according to the abnormal measuring point depth, the drilling hole coordinates and the drilling track parameters, and projecting the abnormal space coordinates to the slope model to form an abnormal mapping relation; performing aggregation and diffusion according to the slope unit number to generate a risk hot area, performing instability mode discrimination based on the mapping weight, the trigger characteristic quantity and the normal consistency parameter, and outputting a hot area mode result set; according to the method, unified space-time correlation of deep anomaly and slope deformation, risk hot area aggregation and instability mode discrimination are realized, and the consistency and traceability of slope stability monitoring results are improved.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and early warning technology, and more specifically, to a method and system for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors. Background Technology

[0002] Rock slopes are widely used in mining, transportation engineering, water conservancy projects, and mountain infrastructure construction. Their stability directly affects the structural and operational safety of the engineering project. With the development of slope monitoring technology, the use of 3D laser scanners to periodically scan the slope surface and acquire multi-temporal point cloud data has become an important means of obtaining information on slope morphology changes, local deformation characteristics, and surface geometry. Simultaneously, deep sensors can continuously collect deep monitoring data such as displacement, strain, dip angle, and pore pressure within boreholes to reflect the internal structural response and deep evolution of the slope. In current engineering practice, 3D laser scanning and deep sensing monitoring are often deployed together to simultaneously obtain slope surface and deep information, thereby improving the completeness and reliability of slope stability monitoring.

[0003] However, in actual rock slope monitoring, 3D laser scanning data mainly reflects surface deformation, while deep sensing data mainly reflects deep evolution information such as displacement, strain, dip angle, and pore pressure within boreholes. Due to differences in time reference, indexing methods, and spatial representation, existing technologies typically employ step-by-step analysis or simple correspondence methods for comprehensive judgment, lacking a unified data link and consistent judgment criteria, making it difficult to generate continuous and consistent monitoring results. Furthermore, in existing rock slope monitoring, although surface deformation can be observed through 3D laser scanning, the deep evolution process is difficult to interpret synchronously with surface observation results. Specifically, it is difficult to determine the corresponding slope location for deep anomalies, whether they correspond to current slope changes, and to further determine the instability mode.

[0004] Therefore, there is an urgent need for a real-time monitoring method and system for rock slopes that can uniformly correlate deep anomalies with slope deformation information in time and space and achieve consistent discrimination. In view of this, this invention proposes a method and system for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors to solve the above-mentioned problems. Summary of the Invention

[0005] To overcome the aforementioned shortcomings of the prior art and achieve the above objectives, the present invention provides the following technical solution: a method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors, comprising:

[0006] Acquire multi-temporal point cloud data output by a 3D laser scanner and deep monitoring sequences output by a deep sensor, construct the binding relationship between point cloud frames and deep sequence windows, and output a set of point cloud frame indexes and a set of deep anomaly event indexes when an anomaly is triggered in the deep sequence window;

[0007] Retrieve the point cloud frame data volume corresponding to the point cloud frame index and construct the slope model; based on the depth of the abnormal measurement point corresponding to the deep abnormal event index and combined with the borehole opening coordinates and borehole trajectory parameters, calculate the abnormal spatial coordinates, project the abnormal spatial coordinates onto the slope model to obtain the slope unit number, generate an abnormal mapping table and output the abnormal mapping relationship set.

[0008] Based on the set of anomaly mapping relationships and the slope model, the deep anomaly event index is mapped and aggregated according to the slope unit number. The unit risk score is generated according to the mapping confidence parameter and the triggering feature quantity. The unit risk score is diffused and aggregated according to the slope adjacency relationship to obtain a risk hot zone set including hot zone number, hot zone unit set, hot zone risk score and hot zone associated deep anomaly event index set.

[0009] The mapping weight parameters in the anomaly mapping table are summed by slope unit number to obtain unit weight parameters. The hot zone representative normal vector is obtained by weighted summation, and the hot zone normal consistency parameter is calculated. The triggering proportion parameter is calculated by triggering channel name, and the intensity indication parameter is obtained by taking the maximum value of the ratio of the triggering feature quantity to the corresponding threshold value. The dominant parameter is generated and the tilting dominant parameter is corrected. The instability mode label and mode confidence score are determined, written into the hot zone mode result set, and output.

[0010] Furthermore, the method for obtaining the hot zone pattern result set includes:

[0011] Traverse the risk hot zone set by hot zone number to construct hot zone geometric data packets and hot zone event data packets; calculate the hot zone normal consistency parameter, channel triggering ratio parameter set, and intensity indication parameter set according to the feature aggregation rule parameters based on the anomaly mapping table summary unit weight parameters; generate displacement dominant parameters, shear dominant parameters, toppling dominant parameters, and pore pressure dominant parameters based on the dominant fusion parameter set, and correct the toppling dominant parameters based on the normal consistency threshold parameter and normal correction parameter set; determine the instability mode label based on the mode determination threshold parameter set, calculate the mode confidence score based on the mode confidence score parameter, and write it into the hot zone mode result set.

[0012] Furthermore, the construction of the hot zone geometry data packet and hot zone event data packet, as well as parameter calculation, includes:

[0013] The hot zone geometric data package is formed by reading the element normal vectors and element center coordinates corresponding to the slope element numbers from the slope model according to the hot zone element set; the hot zone event data package is formed by reading the trigger channel name, trigger characteristic quantity, threshold value and trigger timestamp from the anomaly event table according to the hot zone associated deep anomaly event index set; the element weight parameters are obtained by summing the mapping weight parameters in the anomaly mapping table according to the slope element number; the hot zone representative normal vector is calculated according to the weighted summation rule, and the hot zone normal consistency parameter is calculated based on the angle between each element normal vector and the hot zone representative normal vector; the proportion parameters of displacement change rate triggering, strain change rate triggering, tilt angle change rate triggering, and pore pressure change rate triggering are statistically analyzed according to the trigger channel name, and the hot zone displacement intensity indication parameters, hot zone strain intensity indication parameters, hot zone tilting indication parameters and hot zone pore pressure surge indication parameters are calculated according to the maximum value rule.

[0014] Furthermore, the determination of the instability mode label and mode confidence score includes:

[0015] Based on the intensity fusion weight parameter and the proportion fusion weight parameter in the dominant fusion parameter set, the displacement dominant parameter is obtained by weighted summation of the hot zone displacement intensity indicator parameter and the displacement change rate trigger proportion parameter; the shear dominant parameter is obtained by weighted summation of the hot zone strain intensity indicator parameter and the strain change rate trigger proportion parameter; the toppling dominant parameter is obtained by weighted summation of the hot zone toppling indicator parameter and the tilt angle change rate trigger proportion parameter; and the pore pressure dominant parameter is obtained by weighted summation of the hot zone pore pressure surge indicator parameter and the pore pressure change rate trigger proportion parameter. When the hot zone normal consistency parameter is greater than the normal consistency threshold parameter, the toppling dominant parameter is corrected according to the normal penalty coefficient parameter. The instability mode label is determined by comparing the pore pressure dominant parameter, toppling dominant parameter, shear dominant parameter, and displacement dominant parameter with the corresponding threshold in that order. The mode confidence score is calculated based on the first dominant value, the second dominant value, and the hot zone risk score according to the mode confidence score parameter.

[0016] Furthermore, the method for obtaining the set of risk hot zones includes:

[0017] The abnormal mapping relationship set is read from the abnormal mapping table by point cloud frame index; the trigger timestamp, trigger feature quantity, and threshold value are read from the abnormal event table by deep abnormal event index; and the unified timestamp is read from the point cloud frame index table by point cloud frame index. The maximum value of the unified timestamp in the abnormal mapping relationship set is used to determine the fusion reference timestamp. Abnormal mapping relationships whose time difference between the unified timestamp and the fusion reference timestamp is not greater than the fusion time window length parameter are selected. For the selected abnormal mapping relationships, the abnormal intensity parameter is calculated as the ratio of the trigger feature quantity to the threshold value and normalized according to the preset intensity upper limit parameter to obtain the normalized abnormal intensity parameter. The confidence weight parameter is obtained according to the mapping confidence parameter, with a value of 1 for high confidence mapping and a preset low confidence weight parameter for low confidence mapping. The normalized abnormal intensity parameter is multiplied by the confidence weight parameter to obtain the mapping weight parameter, and the mapping weight parameter is accumulated according to the slope unit number to obtain the unit risk score. At the same time, the deep abnormal event index set corresponding to the slope unit number is summarized.

[0018] Furthermore, the method for obtaining the set of risk hot zones also includes:

[0019] A slope adjacency table is established based on the unit center coordinates corresponding to the slope unit number in the slope model and the unit adjacency distance threshold parameter. The slope unit number matched in the anomaly mapping relationship set is used as the diffusion starting point, and diffusion is carried out layer by layer along the slope adjacency table to the number of layers limited by the diffusion radius parameter. The path length parameter is taken as the number of adjacent layers, and the normalized distance is taken as the ratio of the path length parameter to the diffusion radius parameter, and the distance attenuation weight is calculated accordingly. The trigger timestamp is read according to the deep anomaly event index and the absolute difference is calculated with the unified timestamp corresponding to the diffusion starting point point cloud frame index to obtain the time difference parameter. The normalized time is taken as the ratio of the time difference parameter to the fusion time window length parameter, and the time attenuation weight is calculated accordingly.

[0020] The diffusion contribution value is obtained by multiplying the mapping weight parameter, distance decay weight, and time decay weight, and then added to the unit risk score of the candidate slope unit number. At the same time, the deep anomaly event index set corresponding to the candidate slope unit number is updated. The unit risk score is filtered according to the hot zone risk threshold parameter and clustered according to the slope adjacency list to obtain the hot zone unit set and hot zone number. The hot zone risk score is the maximum unit risk score in the hot zone unit set. The hot zone associated deep anomaly event index set is the union of the deep anomaly event index sets corresponding to the hot zone unit set, and the risk hot zone set is obtained by filtering according to the minimum hot zone unit number parameter.

[0021] Furthermore, the method for obtaining the set of abnormal mapping relationships includes:

[0022] Traverse the deep anomaly event index set and read the trigger timestamp, trigger channel name, trigger feature quantity and threshold comparison result from the anomaly event table. Determine the point cloud frame index associated with the deep anomaly event index based on the binding relationship table and obtain the deep sequence window data packet.

[0023] The data packets of the deep sequence window are grouped by the measurement point number field, and the channel value field corresponding to the trigger channel name is compared with the threshold in the threshold comparison result. The measurement point over-threshold count value is counted according to the preset continuous over-threshold count parameter, the abnormal measurement point number is determined, and the abnormal measurement point depth parameter is read from the measurement point depth field and written into the abnormal event table.

[0024] Anomaly spatial coordinates are calculated based on anomaly measurement point depth parameters, borehole opening coordinates, and borehole trajectory parameters; a slope grid cell set is obtained from the slope model based on the associated point cloud frame index, and a projection matching process is performed according to the projection method parameters to determine the matching slope cell number; mapping distance parameters, trajectory deviation upper limit parameters, and point cloud resolution parameters are calculated, and mapping confidence parameters are marked according to confidence threshold parameters; the confidence threshold parameters include distance threshold parameters and deviation threshold parameters; the distance threshold parameters are set according to multiples of the grid resolution parameters, and the deviation threshold parameters are set according to multiples of the grid resolution parameters;

[0025] Write the point cloud frame index, deep anomaly event index, anomaly spatial coordinates, matching slope unit number, and mapping confidence parameters into the anomaly mapping table and output it as a set of anomaly mapping relationships.

[0026] Furthermore, the method for obtaining the point cloud frame index set and the deep anomaly event index set includes:

[0027] Establish a device time offset table. The device time offset table records the time offset of the 3D laser scanner and the time offset of the deep sensor. The initial value of the device time offset table is written by the device registration process. The device registration process records the device identifier, communication delay measurement results and initial clock difference measurement results.

[0028] Each time a 3D laser scanner generates a frame of point cloud data, it triggers the point cloud frame storage process. The point cloud frame storage process receives the original acquisition time of the point cloud frame and generates a unified timestamp by combining it with the device time offset table. Based on the unified timestamp, a point cloud frame index is generated and written into the point cloud frame index table.

[0029] The deep sensor outputs a deep monitoring sequence according to a preset sampling period. The deep monitoring sequence includes displacement data, strain data, tilt angle data and pore pressure data inside the borehole. During the deep sequence storage process, the original deep sampling time is received and combined with the equipment time offset table to generate a unified timestamp and write it into the deep sequence index table.

[0030] Using the point cloud frame index table as the main index, a deep sequence window is generated for each point cloud frame index. The deep sequence window is limited by a preset time window length parameter. The start and end points of the window are determined based on the unified timestamp corresponding to the point cloud frame index. The deep monitoring sequence within the range from the start to the end of the window is retrieved from the deep sequence index table to form a deep sequence window data packet, which is then written to the binding relationship table. When the binding relationship table is generated or updated, an anomaly detection process is performed, and the point cloud frame index set and the deep anomaly event index set are output.

[0031] Furthermore, methods for performing anomaly detection procedures when generating or updating binding relationship tables include:

[0032] Using deep sequence window data packets as input, the set of characteristic quantities within the window is calculated based on the abnormal triggering condition parameters. The set of characteristic quantities within the window includes the displacement change rate, strain change rate, tilt angle change rate, and pore pressure change rate within the window.

[0033] The anomaly detection process compares the feature set within the window with the threshold parameter set, and determines whether the deep sequence window meets the anomaly triggering conditions based on the number of consecutive threshold exceedances. When the anomaly triggering conditions are met, a deep anomaly event index is assigned to the anomaly event, and the deep anomaly event index, trigger timestamp, trigger channel name, and comparison results of trigger features and thresholds are written into the anomaly event table. The point cloud frame index that meets the anomaly triggering conditions is added to the point cloud frame index set, and the corresponding deep anomaly event index is added to the deep anomaly event index set.

[0034] A system for real-time monitoring of rock slope stability based on a 3D laser scanner and deep sensors, used to implement the aforementioned method for real-time monitoring of rock slope stability based on a 3D laser scanner and deep sensors, includes:

[0035] The multi-source acquisition and timing module is used to acquire multi-temporal point cloud data output by the 3D laser scanner and deep monitoring sequences output by the deep sensor, construct the binding relationship between point cloud frames and deep sequence windows, and output the point cloud frame index set and the deep anomaly event index set when an anomaly is triggered in the deep sequence window.

[0036] The spatial mapping module retrieves the point cloud frame data volume corresponding to the point cloud frame index and constructs a slope model; based on the depth of the abnormal measurement point corresponding to the deep anomaly event index and combined with the borehole opening coordinates and borehole trajectory parameters, it calculates the abnormal spatial coordinates, projects the abnormal spatial coordinates onto the slope model to obtain the slope unit number, generates an anomaly mapping table and outputs the set of anomaly mapping relationships.

[0037] The fusion positioning module is used to map and aggregate the deep anomaly event index based on the anomaly mapping relationship set and the slope model, according to the slope unit number, generate unit risk scores based on the mapping confidence parameters and trigger feature quantities, and diffuse and aggregate the unit risk scores according to the slope adjacency relationship to obtain a risk hot zone set including hot zone number, hot zone unit set, hot zone risk score and hot zone associated deep anomaly event index set;

[0038] The instability mode discrimination module sums the mapping weight parameters in the anomaly mapping table by slope unit number to obtain unit weight parameters, performs weighted summation to obtain the hot zone representative normal vector, and calculates the hot zone normal consistency parameter; it counts the triggering ratio parameter by triggering channel name and takes the intensity indication parameter with the maximum value of the ratio of triggering feature quantity to the corresponding threshold value; it generates the dominant parameter and corrects the tilting dominant parameter, determines the instability mode label and mode confidence score, writes them into the hot zone mode result set, and outputs them.

[0039] Compared with existing technologies, the method and system for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors proposed in this invention have the following technical effects and advantages:

[0040] This invention addresses the core problem of visible surface deformation but difficult-to-interpret deep evolution. By incorporating multi-temporal point cloud data obtained from 3D laser scanning with deep monitoring sequences obtained from deep sensors into a unified time reference, and constructing a continuous index link using point cloud frame index and deep anomaly event index, the invention enables deep anomaly triggering results to correspond with slope observation results in the same time period. This solves the problems in existing technologies where surface observation and deep monitoring are independent, have inconsistent time series, and are difficult to interpret synchronously.

[0041] Based on this, the present invention further combines the depth of the abnormal measuring point corresponding to the deep anomaly event with the borehole opening coordinates and borehole trajectory parameters to calculate the anomaly spatial coordinates, and establishes a mapping relationship with the slope unit number in the slope model. At the same time, it introduces mapping distance parameters, trajectory deviation upper limit parameters and mapping confidence parameters to constrain the mapping results, so that the occurrence time of the deep anomaly event can be determined, as well as the corresponding slope location and its confidence level. This solves the problem that it is difficult to determine the corresponding slope location of the deep anomaly event and whether it corresponds to the current slope change.

[0042] Furthermore, this invention aggregates anomaly mapping relationships along the slope unit number dimension and combines adjacency relationships, distance decay, and time decay to form risk hotspots. Then, based on the trigger channel composition, trigger feature quantities, and slope geometry information within the hotspots, instability mode discrimination is performed, outputting a hotspot risk score, instability mode label, and mode confidence score. Thus, this invention achieves a unified result chain from deep anomaly triggering to slope location, risk aggregation, and instability mode identification, solving the problems of scattered anomaly results, difficulty in forming regionalized judgments, and difficulty in further determining instability modes in existing technologies.

[0043] In summary, this invention achieves unified spatiotemporal correlation between deep anomalies and slope deformation, aggregation of risk hot zones, and identification of instability modes, thereby improving the consistency and traceability of slope stability monitoring results. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of a system for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and a deep sensor, according to Embodiment 1 of the present invention.

[0045] Figure 2 This is a flowchart of a method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors, according to Embodiment 2 of the present invention.

[0046] Figure 3 This is a flowchart of the method for obtaining the abnormal mapping relationship set in Embodiment 1 of the present invention;

[0047] Figure 4 This is a flowchart of the method for obtaining the hot zone pattern result set in Embodiment 1 of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only for better illustrating and explaining the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0049] Example 1

[0050] Please see Figure 1 As shown in the figure, this embodiment discloses a system for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and a deep sensor. It includes a multi-source acquisition and time synchronization module, a spatial mapping module, a fusion positioning module, and an instability mode discrimination module. Each module is connected by wires and / or wirelessly to realize data transmission.

[0051] The multi-source acquisition and timing module is used to acquire multi-temporal point cloud data output by the 3D laser scanner and deep monitoring sequences output by the deep sensor, establish the binding relationship between point cloud frames and deep sequence windows, and output the point cloud frame index set and the deep abnormal event index set when the deep sequence window triggers an anomaly, for subsequent spatial mapping and instability mode discrimination.

[0052] It should be noted that in existing rock slope monitoring, although surface deformation can be observed through three-dimensional laser scanning, the deep evolution process is difficult to interpret synchronously with the surface observation results. Specifically, it is difficult to determine the corresponding slope location of deep anomalies, whether they correspond to the current slope changes, and to further determine the instability mode.

[0053] The methods for obtaining the point cloud frame index set and the deep anomaly event index set include:

[0054] During the initialization phase, a device time offset table is established. This table records the time offset of the 3D laser scanner and the time offset of the depth sensor. The initial value of the device time offset table is written during the device registration process, which includes the device identifier, communication delay measurement results, and initial clock difference measurement results. Simultaneously, a data buffer is established, comprising a point cloud frame buffer and a depth sequence buffer. The point cloud frame buffer temporarily stores the point cloud frame data volume and the original acquisition time, while the depth sequence buffer temporarily stores the depth monitoring sequence data volume and the original deep sampling time.

[0055] During the point cloud frame access phase, the point cloud frame storage process is triggered every time the 3D laser scanner generates a point cloud data frame. The storage process receives the point cloud frame data body, the original acquisition time of the point cloud frame, and the 3D laser scanner device identifier. It reads the 3D laser scanner time offset from the device time offset table and converts the original acquisition time of the point cloud frame into a unified timestamp. The conversion method involves applying the 3D laser scanner time offset to the original acquisition time of the point cloud frame and rounding it according to a preset time granularity parameter. The storage process assigns a point cloud frame index to each point cloud frame. The point cloud frame index is generated in ascending order based on the unified timestamp, and the point cloud frame index, unified timestamp, and point cloud frame data body are written into the point cloud frame index table. Simultaneously, the storage process performs out-of-order correction. When the unified timestamp of a point cloud frame is earlier than the unified timestamp of already stored point cloud frames, the point cloud frame index table is reordered according to the unified timestamp, and the point cloud frame index is updated. This ensures that the subsequent binding relationship construction process relies only on the unified timestamp and is not affected by communication out-of-order issues.

[0056] During the deep sequence access phase, the deep sensors output deep monitoring sequences according to a preset sampling period. These sequences include borehole displacement, strain, tilt, and pore pressure data. The deep sequence storage process receives the deep monitoring sequence data body, the original deep sampling time, and the deep sensor device identifier. It reads the deep sensor time offset from the device time offset table, converts the original deep sampling time into a unified timestamp, and writes the unified timestamp and the deep monitoring sequence data body into the deep sequence index table. To eliminate differences in sampling frequencies among different deep sensors, the multi-source acquisition and time synchronization module reads a preset alignment step size parameter. This parameter can be set to a fixed value as needed or adaptively generated according to the deep sampling period. The resampling process generates an aligned time series using a preset alignment step size parameter. The aligned time series covers the valid time range of the deep sequence index table. For each deep monitoring sequence, the resampling process performs interpolation updates between adjacent sampling points to obtain the aligned sampling value corresponding to the aligned time series. When adjacent sampling points are missing or the time interval exceeds the preset upper limit parameter for missing interval, the resampling process marks the aligned sampling value as missing and writes it to the missing flag field, so that subsequent abnormal triggering conditions can identify incomplete data and reduce the risk of false alarms.

[0057] During the binding relationship construction phase, the point cloud frame index table is used as the main index. A deep sequence window is generated for each point cloud frame index, and the deep sequence window is limited by a preset time window length parameter, which is jointly defined by forward and backward time length parameters. The starting point of the deep sequence window is obtained by tracing back the forward time length parameter from the unified timestamp of the point cloud frame, and the ending point of the deep sequence window is obtained by extending backward the backward time length parameter from the unified timestamp of the point cloud frame. The binding process retrieves the set of aligned sampled values ​​falling into the deep sequence window from the deep sequence index table, forming a deep sequence window data packet. The point cloud frame index, the starting point of the deep sequence window, the ending point of the deep sequence window, and the deep sequence window data packet are then written into the binding relationship table. If the missing value field of the deep sequence window data packet shows a missing ratio exceeding the preset upper limit parameter, the binding process still retains the binding relationship, but a low-confidence flag is written into the binding relationship table to prompt subsequent spatial mapping and instability mode discrimination to reduce the weight of related results.

[0058] During the anomaly triggering and index output phase, an anomaly detection process is executed when the binding relationship table is generated or updated. The anomaly detection process takes the deep sequence window data package as input and calculates the set of feature quantities within the window based on preset anomaly triggering condition parameters. This set includes the displacement change rate, strain change rate, tilt angle change rate, and pore pressure change rate within the window. The anomaly detection process compares the set of feature quantities within the window with a preset threshold parameter set and, combined with a preset parameter for the number of consecutive exceedances, determines whether the deep sequence window meets the anomaly triggering conditions. When the anomaly triggering conditions are met, the anomaly detection process assigns a deep anomaly event index to the anomaly event, and writes the deep anomaly event index, trigger timestamp, trigger channel name, and the comparison result between the triggering feature quantity and the threshold into the anomaly event table. The indexing output process adds the point cloud frame index of the anomaly to the point cloud frame index set and the corresponding deep anomaly event index to the deep anomaly event index set. The point cloud frame index set and the deep anomaly event index set are then output to the subsequent spatial mapping module and instability mode discrimination module. When the same deep anomaly event is repeatedly triggered in the deep sequence window corresponding to adjacent point cloud frames, the indexing output process merges the anomaly event index according to the preset event merging time interval parameter and updates the duration field in the anomaly event table to avoid misjudgment and resource waste caused by dense and repeated alarms.

[0059] It should be noted that the triggering feature quantity refers to the value of the feature quantity in the feature quantity set within the window that actually meets the threshold and triggers the event during the anomaly detection process.

[0060] The anomaly triggering condition parameters define which changes within the deep sequence window should be considered anomaly events, thereby driving the output of the point cloud frame index set and the deep anomaly event index set. These parameters consist of a monitoring channel set, feature quantity calculation caliber, threshold parameter set, continuity judgment parameters, and data integrity constraint parameters. The monitoring channel set includes in-hole displacement, strain, tilt angle, and pore pressure channels. The threshold parameter set is set separately for each monitoring channel to ensure that different physical quantities can be consistently triggered within the same judgment framework.

[0061] It should be noted that the displacement change rate, strain change rate, tilt angle change rate, and pore pressure change rate within the window are the basic discrimination parameters calculated by this invention within the deep sequence window. These parameters are used to uniformly convert deep monitoring data of different dimensions into a characterization of the rate of change at the same time scale, thus serving as the input criteria for anomaly triggering conditions. The threshold parameter set is used to limit the triggering boundaries of each rate of change channel, the consecutive threshold exceedance number parameter and the consecutive time length parameter are used to limit the continuity conditions of anomaly triggering, and the missing proportion upper limit parameter and the minimum effective point number parameter are used to limit the data integrity conditions of anomaly triggering. Through the collaborative calculation and joint constraints of the above parameters, the generation of the deep anomaly event index not only depends on the value of a single sampling point, but is also controlled by the rate of change, the degree of persistence, and the data integrity, thereby ensuring that the outputs of the point cloud frame index set and the deep anomaly event index set have a unified judgment criterion.

[0062] The feature quantity calculation method is used to obtain the set of features within a window for anomaly detection from the deep sequence window data package. This invention uniformly adopts a rate-of-change method for anomaly detection. The set of features within a window includes the rate of change of displacement, the rate of change of strain, the rate of change of tilt angle, and the rate of change of pore pressure. The rate of change of displacement within a window is obtained by subtracting the starting value of displacement from the ending value of displacement within the deep sequence window and then dividing by the window time span. The rate of change of strain within a window is obtained by subtracting the starting value of strain from the ending value of strain within the deep sequence window and then dividing by the window time span. The rate of change of tilt angle within a window is obtained by subtracting the starting value of tilt angle from the ending value of tilt angle within the deep sequence window and then dividing by the window time span. The rate of change of pore pressure within a window is obtained by subtracting the starting value of pore pressure from the ending value of pore pressure within the deep sequence window and then dividing by the window time span. The trigger channel names in the anomaly event table correspond to the above four types of rate-of-change channels, and the threshold parameter set is used to determine whether the set of features within the window exceeds the threshold. The threshold parameter set includes the displacement rate of change threshold, strain rate of change threshold, tilt angle rate of change threshold, and pore pressure rate of change threshold.

[0063] The continuity determination parameters are used to avoid single-point spike triggering. These parameters include the consecutive threshold exceedance count parameter and the consecutive duration parameter. The consecutive threshold exceedance count parameter specifies that triggering is only allowed when at least a certain number of aligned sampling points within the window meet the threshold exceedance condition for the same channel. The consecutive duration parameter specifies the minimum duration that the aforementioned threshold exceedance points must cover in time. If the number of sampling points meeting the threshold exceedance condition within the deep sequence window reaches the consecutive threshold exceedance count parameter and the coverage duration reaches the consecutive duration parameter, then the corresponding channel is determined to have been triggered. If only one condition is met, the abnormal candidate event is marked as pending confirmation, and no deep abnormal event index is output, or only a low-confidence abnormal event index is output for subsequent modules to use with reduced weighting.

[0064] Data integrity constraints are used to control trigger reliability in the event of data loss. These constraints include at least an upper limit parameter for the missing proportion and a minimum number of valid points parameter. The upper limit parameter for the missing proportion limits the proportion of missing data indicated by the missing marker field within the deep sequence window to no more than the upper limit; otherwise, alarm-level anomaly candidate events are not allowed to be generated directly. The minimum number of valid points parameter limits the number of valid aligned sampling points used for judgment within the window to no less than the lower limit; otherwise, anomaly candidate events are marked as low-confidence. Data integrity constraints may also include a channel consistency threshold parameter. This threshold specifies that in multi-channel joint triggering mode, at least two different channels must trigger simultaneously or sequentially within a preset time deviation range before outputting the deep anomaly event index, further reducing false alarms caused by single sensor drift or occasional disturbances.

[0065] In an example of data in this invention, the preset time granularity parameter is set to 0.001 seconds, the preset sampling period is set to 0.5 seconds, the preset alignment step size parameter is set to 1 second, the preset time window length parameter is composed of a forward time length parameter and a backward time length parameter, wherein the forward time length parameter is set to 1 second and the backward time length parameter is set to 1 second and 200 milliseconds, the preset missing interval upper limit parameter is set to 0.4, and the preset consecutive threshold number parameter is set to two. The preset threshold parameter set is set as follows: displacement change rate threshold 0.05 mm / s, strain change rate threshold 8 microstrain / s, tilt angle change rate threshold 0.02 degrees / s, and pore pressure change rate threshold 0.6 kPa / s.

[0066] Create a device time offset table and record it as shown in the example below. The device identifiers are taken as scanner device number one and sensor device number two. The communication delay measurement results are taken as 0.04 seconds and 0.06 seconds respectively. The initial clock difference measurement results are recorded as scanner device clock 0.12 seconds fast and sensor device clock 0.08 seconds slow. The device time offset table records the scanner time offset as -0.12 seconds and the sensor time offset as +0.08 seconds. The time offset is defined as the sum of the original device time and the time offset to obtain a unified timestamp.

[0067] Each time a 3D laser scanner generates a frame of point cloud data, it triggers the point cloud frame storage process. This process involves applying a scanner time offset to the original acquisition time of the point cloud frame to obtain a unified timestamp, and then generating a point cloud frame index by incrementing the unified timestamp and writing it into the point cloud frame index table. An example point cloud frame index table contains three records: the first record has an original acquisition time of 10:00:00.120 seconds, which is offset by -0.12 seconds to obtain a unified timestamp of 10:00:00.000 seconds, and is written to point cloud frame index one, along with the number of points in the point cloud frame data body; the second record has an original acquisition time of 10:00:02.120 seconds, which is replaced by a unified timestamp of 10:00:02.000 seconds, and is written to point cloud frame index two; the third record has an original acquisition time of 10:00:04.120 seconds, which is replaced by a unified timestamp of 10:00:04.000 seconds, and is written to point cloud frame index three.

[0068] In the deep sequence access phase, the deep sensor outputs the deep monitoring sequence according to the preset sampling period and triggers the deep sequence storage process. The deep sequence storage process applies a sensor time offset to the original deep sampling time to obtain a unified timestamp and writes it into the deep sequence index table. In the example data of the deep sequence access phase, the deep sensor continuously outputs the monitoring values ​​of the same measuring point number at multiple unified timestamps according to the preset sampling period. In the example, measuring point number 1 corresponds to a borehole displacement of 0.00 mm, a strain of 10 microstrain, an inclination angle of 0.02 degrees, and a borehole pressure of 120 kPa at the unified timestamp of 10:00:00:00. The unified timestamp of 10:00:00:0500 millimeters corresponds to a borehole displacement of 0.01 mm, a strain of 11 microstrain, an inclination angle of 0.021 degrees, and a borehole pressure of 120.2 kPa. The unified timestamp of 10:00:01:00 corresponds to a borehole displacement of 0.02 mm, a strain of 12 microstrain, an inclination angle of 0.023 degrees, and a borehole pressure of 120.5 kPa. In the example above, “strain value ten microstrain, strain value eleven microstrain, strain value twelve microstrain” are the numerical records of the strain channel at adjacent sampling times, which are used to calculate the strain change rate according to a unified timestamp.

[0069] An aligned time series is generated according to a preset alignment step size parameter. An example aligned time series is taken as 10:00:00, 10:00:01, 10:00:02, 10:00:03, and 10:00:04. For each deep monitoring sequence, aligned sample values ​​are obtained by interpolation between adjacent sampling points. Since 500 milliseconds are missing at 10:00:01, but the time interval between adjacent sampling points does not exceed the preset upper limit of the missing interval parameter, the aligned sample value can still be obtained by interpolation between 10:00:01 and 10:00:02. The aligned sample value for 10:00:01 is directly taken from the original record, and the aligned sample value for 10:00:02 is also directly taken from the original record.

[0070] Using the point cloud frame index table as the primary index, a deep sequence window is generated for each point cloud frame index and written to the binding relationship table. Point cloud frame index one corresponds to a unified timestamp of 10:00:00, with the window starting at 10:00:00 and ending at 10:00:01. The window data packet contains aligned sample values ​​for 10:00:00 and 10:00:01 in the aligned time sequence. Point cloud frame index two corresponds to a unified timestamp of 10:00:02, with the window starting at 10:00:01 and ending at 10:00:03. The window data packet contains aligned sample values ​​for 10:00:01, 10:00:02, and 10:00:03 in the aligned time sequence. Point cloud frame index three corresponds to a unified timestamp of 10:00:04, with the window starting at 10:00:03 and ending at 10:00:04. The window data packet contains aligned sample values ​​for 10:00:03 and 10:00:04 in the aligned time sequence. The missing flag field of the example window data packet is calculated according to the missing ratio within the window. No alignment time missing occurs within the window range of point cloud frame index two, and the missing ratio is zero. The low confidence flag is not written to the window data packet corresponding to point cloud frame index two.

[0071] An anomaly detection process is performed when the binding relationship table is generated or updated. The anomaly detection process takes the deep sequence window data packet as input, calculates the set of features within the window based on preset anomaly triggering condition parameters, and compares it with a preset threshold parameter set. Anomaly event index allocation is triggered when the preset consecutive threshold exceedance number parameter is continuously met. For example, for the window start time of 10:00:01 and the window end time of 10:00:03 corresponding to point cloud frame index two, first, the displacement change rate from 10:00:01 to 10:00:02 is calculated, resulting in 0.06 mm / s; the strain change rate from 10:00:01 to 10:00:02 is calculated, resulting in 8 microstrain / s; the tilt change rate from 10:00:01 to 10:00:02 is calculated, resulting in 0.03 degrees / s; and the pore pressure change rate from 10:00:01 to 10:00:02 is calculated, resulting in 0.8 kPa / s. All four of these parameters satisfy the threshold exceedance condition when compared with their corresponding thresholds. Subsequently, the displacement change rate from 10:00:02 to 10:00:03 was calculated, and the result was 0.08 mm / s. The strain change rate was 16 microstrains / s, the tilt angle change rate was 0.04 degrees / s, and the pore pressure change rate was 1.6 kPa / s. All four of these items met the threshold conditions when compared with their corresponding thresholds.

[0072] The anomaly detection process counts the threshold exceedance results twice for two consecutive time intervals, satisfying the preset consecutive threshold exceedance number parameter. This generates a deep anomaly event index one, which is then written into the anomaly event table. The anomaly event table is then written with a trigger timestamp of 10:00:30, the trigger channel names (bore displacement channel, strain channel, tilt angle channel, and pore pressure channel) and the trigger characteristic quantities (displacement change rate of 0.08 mm / s, strain change rate of 16 microstrains / s, tilt angle change rate of 0.04 degrees / s, and pore pressure change rate of 1.6 kPa / s). Finally, the threshold comparison result is written as exceeding the threshold.

[0073] Add the point cloud frame indexes that meet the abnormal triggering conditions to the point cloud frame index set, and add the corresponding deep abnormal event indexes to the deep abnormal event index set. Example output shows that the point cloud frame index set contains point cloud frame index two, and the deep abnormal event index set contains deep abnormal event index one. If the window corresponding to point cloud frame index three is subsequently triggered again, and the time difference between the trigger timestamp and 10:00:03 is less than the preset event merging time interval parameter, then update the duration field in the abnormal event table and merge the deep abnormal event indexes; no new deep abnormal event indexes will be added.

[0074] The spatial mapping module, based on the point cloud frame index set and the deep anomaly event index set, retrieves the point cloud frame data volume corresponding to the point cloud frame index and constructs a slope model; based on the depth of the abnormal measurement points corresponding to the deep anomaly event index and combined with the borehole opening coordinates and borehole trajectory parameters, it calculates the abnormal spatial coordinates, projects the abnormal spatial coordinates onto the slope model to obtain the slope unit number, generates an anomaly mapping table and outputs an anomaly mapping relationship set, which includes the point cloud frame index, the deep anomaly event index, the abnormal spatial coordinates, the matching slope unit number, and the mapping confidence parameters.

[0075] The method for constructing the slope model includes:

[0076] Read the point cloud frame index table, point cloud frame data body storage location, binding relationship table, abnormal event table, deep sequence index table, borehole opening coordinates and borehole trajectory parameters, establish a coordinate reference parameter set and write the point cloud registration parameters into the registration field of the point cloud frame index table; traverse the point cloud frame index set and retrieve the point cloud frame data body;

[0077] A point cloud preprocessing process is performed on the point cloud frame data volume. The point cloud preprocessing process includes distance clipping parameters, reflection intensity filtering parameters, and outlier removal parameters. Based on the preprocessed point set, a slope grid cell set is generated according to the grid resolution parameters and the slope cell number is assigned. The cell center coordinates, cell normal vector, cell slope value, and cell aspect value are calculated to form a slope model and record the point cloud frame index and the cell information.

[0078] like Figure 3 As shown, the method for obtaining the set of abnormal mapping relationships includes:

[0079] The spatial mapping module reads the point cloud frame index table, the point cloud frame data volume storage location, the binding relationship table, the abnormal event table, the deep sequence index table, and the borehole opening coordinates and borehole trajectory parameters. It establishes a coordinate reference parameter set, which includes the scanning station coordinates, point cloud registration parameters, and the definition of the origin of the unified coordinate system. The point cloud registration parameters are calculated from the control point coordinates and the corresponding point set of the control points in the point cloud frame. These parameters are written to the registration field of the point cloud frame index table to ensure that different point cloud frames are in the same unified coordinate system.

[0080] The point cloud frame index set is traversed, and a unified timestamp and point cloud frame data volume are retrieved from the point cloud frame index table according to the point cloud frame index. A point cloud preprocessing process is performed on the point cloud frame data volume. This process includes distance clipping parameters, reflection intensity filtering parameters, and outlier removal parameters. The distance clipping parameters limit the point set retained in the slope monitoring area, the reflection intensity filtering parameters remove low-confidence echo points, and the outlier removal parameters remove isolated points and sparse noise points. After point cloud preprocessing, a slope model construction process is executed. This process projects the preprocessed point set onto a raster plane defined by the grid resolution parameters and generates a slope raster cell set. For each cell in the slope raster cell set, the cell center coordinates, cell normal vector, cell slope value, and cell aspect value are calculated, and each cell is assigned a slope cell number. A slope model is formed and the point cloud frame index and the unit information are recorded. The unit information includes the slope unit number, unit center coordinates, unit normal vector, unit slope value and unit aspect value. The slope model establishes an index relationship according to the point cloud frame index.

[0081] The spatial mapping module traverses the deep anomaly event index set, reads the trigger timestamp, trigger channel name, and comparison results of trigger feature quantity and threshold from the anomaly event table according to the deep anomaly event index, and determines the point cloud frame index associated with the deep anomaly event index based on the binding relationship table. It retrieves the deep sequence window data packet from the deep sequence index table, which contains the measurement point number field, measurement point depth field, and channel value field.

[0082] The abnormal measurement point determination process involves grouping the deep sequence window data packets by measurement point number field to obtain a set of measurement point data sub-packets. For each measurement point data sub-packet set, the process iterates through the channel value fields of each aligned sampling time within the window, comparing the channel value fields with the threshold values ​​of the corresponding channels in the threshold comparison results to obtain the threshold ex-threshold marker for that measurement point at that aligned sampling time. A continuous count is then performed on the threshold ex-threshold marker according to a preset continuous threshold ex-threshold count parameter. The number of times the continuous count satisfies the preset continuous threshold ex-threshold count parameter is determined as the measurement point threshold ex-threshold count value. The abnormal measurement point determination process identifies the measurement point number with the largest threshold ex-threshold count value as the abnormal measurement point number. The abnormal measurement point depth parameter is then read from the measurement point depth field using the abnormal measurement point number. Finally, the abnormal measurement point number and abnormal measurement point depth parameter are written into the abnormal measurement point number field and abnormal measurement point depth field of the abnormal event table.

[0083] It should be noted that in the example of preset parameters related to abnormal triggering conditions, the number of consecutive exceedances can be preset to twice, the alignment sampling time of the deep sequence window data packet can be set to 10:00:01, 10:00:02, and 10:00:03, and the threshold corresponding to the trigger channel name in the preset threshold parameter set can be set to 0.05 mm / s. The threshold comparison results recorded in the abnormal event table include a displacement change rate threshold of 0.05 mm / s, a trigger channel name of "in-hole displacement change rate channel", a deep abnormal event index of "deep abnormal event index one", and a trigger timestamp of 10:00:03.

[0084] In the field example of the deep sequence window data packet, the measurement point number field includes measurement point number 1, measurement point number 2, and measurement point number 3; the measurement point depth field corresponds to 5 meters, 10 meters, and 15 meters respectively; the channel value field records the sequence of judgment values ​​corresponding to the trigger channel name, as shown in the following examples. The channel value field corresponding to measurement point number 1 is 0.04 mm / s at 10:00:01, 0.06 mm / s at 10:00:02, and 0.07 mm / s at 10:00:03. The channel value field corresponding to measurement point number 2 is 0.06 mm / s at 10:00:01, 0.07 mm / s at 10:00:02, and 0.08 mm / s at 10:00:03. The channel value field corresponding to measurement point number 3 is 0.03 mm / s at 10:00:01, 0.04 mm / s at 10:00:02, and 0.06 mm / s at 10:00:03.

[0085] In the execution example of the abnormal measurement point determination process, the spatial mapping module extracts the channel value field corresponding to the displacement change rate channel in the borehole from the deep sequence window data packet based on the comparison result between the trigger channel name and the threshold in the abnormal event table, and groups the deep sequence window data packet according to the measurement point number field to obtain a set of measurement point data sub-packets. For measurement point number one, the 0.04 mm / s at 10:00:01 is compared with the threshold of 0.05 mm / s, indicating no exceedance; the 0.06 mm / s at 10:00:02 is compared with the threshold, indicating yes exceedance; the 0.07 mm / s at 10:00:03 is compared with the threshold, indicating yes exceedance. Subsequently, a continuous count is performed twice according to the preset continuous exceedance number parameter. Measurement point number one forms a continuous exceedance at 10:00:02 and 10:00:03, and the measurement point exceedance count value is determined to be one. For measuring point number two, the times 10:00:01, 10:00:02, and 10:00:03 are all marked as exceeding the threshold. Specifically, 10:00:01 and 10:00:02 form one consecutive threshold exceedance, and 10:00:02 and 10:00:03 form another consecutive threshold exceedance. The threshold exceedance count for this measuring point is determined to be two times. For measuring point number three, only 10:00:03 is marked as exceeding the threshold, and no consecutive threshold exceedance occurs. The threshold exceedance count for this measuring point is determined to be zero times. In the abnormal measuring point determination process, measuring point number two with the largest threshold exceedance count is identified as the abnormal measuring point number. Ten meters are read from the measuring point depth field of abnormal measuring point number two as the abnormal measuring point depth parameter. The abnormal measuring point number and abnormal measuring point depth parameter are then written into the abnormal measuring point number field and abnormal measuring point depth field of the abnormal event table.

[0086] The abnormal measuring point depth parameter is the depth value corresponding to the abnormal measuring point number in the measuring point depth field. This parameter characterizes the axial position of the abnormal measuring point within the borehole. The measurement caliber of the abnormal measuring point depth parameter is the cumulative distance from the borehole opening along the borehole centerline to the abnormal measuring point location, and the unit can be meters or millimeters. The subsequent spatial mapping module uses the abnormal measuring point depth parameter as input, combines the borehole opening coordinates and borehole trajectory parameters to convert the abnormal measuring point depth parameter into abnormal spatial coordinates, and then projects these abnormal spatial coordinates onto the slope model to obtain the slope element number.

[0087] It should be noted that the abnormal measuring point depth parameter is a depth positioning parameter calculated during the abnormal measuring point determination process of this invention. The abnormal measuring point depth parameter is generated by the one-to-one correspondence between the measuring point number field and the measuring point depth field, and is jointly determined with the over-threshold count result corresponding to the trigger channel name. The function of the abnormal measuring point depth parameter is to further convert the deep abnormal event index from the time-triggered result into the borehole axial position result, providing a clear input for subsequent conversion of abnormal spatial coordinates. If there is only a deep abnormal event index without the abnormal measuring point depth parameter, the deep abnormal event cannot establish a positional conversion relationship with the borehole opening coordinates and borehole trajectory parameters, making it difficult to further map to the slope unit number. Therefore, the abnormal measuring point depth parameter is an intermediate parameter connecting deep abnormal triggering and slope spatial mapping. Its setting and calculation process is used to support the technical problem solved by this invention of the difficulty in corresponding deep abnormal information to slope positions.

[0088] Anomaly spatial coordinates are calculated based on anomaly measurement point depth parameters, borehole opening coordinates, and borehole trajectory parameters. The borehole trajectory parameters include borehole inclination parameters, borehole direction parameters, and segment length parameters. The borehole inclination and direction parameters determine the borehole centerline direction vector, and the segment length parameter determines the cumulative distance along the depth direction of the borehole centerline. Starting from the borehole opening coordinates, the anomaly spatial coordinates are accumulated along the borehole centerline direction vector according to the anomaly measurement point depth parameters. These anomaly spatial coordinates are then represented as three-dimensional coordinate values ​​in a unified coordinate system, and written into the anomaly spatial coordinate field of the anomaly mapping table.

[0089] The slope grid cell set is retrieved from the slope model using the associated point cloud frame index. A projection matching process is then performed, controlled by the projection mode parameters. When the projection mode parameter is normal projection, the intersection coordinates of the anomaly spatial coordinates and the slope grid surface are calculated along the normal vector direction of the slope grid cells. If the intersection coordinates fall within the slope grid coverage area, the slope cell number corresponding to the intersection coordinates is determined as the matching slope cell number. When the projection mode parameter is shortest distance projection, the center coordinates of the cell with the smallest Euclidean distance to the anomaly spatial coordinates are searched in the slope grid cell set. The slope cell number corresponding to the cell with the smallest distance is determined as the matching slope cell number. If the intersection coordinates calculated by normal projection exceed the slope grid coverage area, the projection mode parameter is switched to shortest distance projection, and the matching slope cell number determination is completed.

[0090] During the confidence parameter calculation phase, the spatial mapping module calculates the mapping confidence parameters and writes them into the anomaly mapping table. The mapping confidence parameters are calculated from the mapping distance parameter, the trajectory deviation upper limit parameter, and the point cloud resolution parameter. The mapping distance parameter is the three-dimensional distance between the anomaly spatial coordinates and the center coordinates of the matching slope unit. The point cloud resolution parameter is the mesh resolution parameter.

[0091] It should be noted that the mapping distance parameter characterizes the geometric deviation between the anomalous spatial coordinates and the center coordinates of the matching slope unit, while the trajectory deviation upper limit parameter characterizes the upper limit of the spatial conversion deviation introduced by the error between the borehole opening coordinates and the borehole trajectory parameters at the depth position of the anomalous measuring point. These two parameters correspond to the two error sources: point cloud projection deviation and borehole conversion deviation, respectively. The mapping confidence parameter is jointly determined by the above parameters and the confidence threshold parameter and is written into the anomalous mapping table to characterize the confidence level of a single anomalous mapping relationship. By calculating the mapping distance parameter and the trajectory deviation upper limit parameter separately and uniformly labeling the mapping confidence parameter, the subsequent fusion positioning process can distinguish between high-confidence mappings and low-confidence mappings when using the set of anomalous mapping relationships.

[0092] The trajectory deviation upper limit parameter characterizes the upper limit of spatial positioning deviation introduced by the borehole opening coordinate measurement deviation and the borehole trajectory measurement deviation at the abnormal measuring point depth. The trajectory deviation upper limit parameter is determined based on the calibration records of the borehole trajectory parameters. These records include the upper limit of the borehole opening coordinate deviation, the upper limit of the borehole inclination deviation, and the upper limit of the borehole direction deviation. The upper limit of the borehole opening coordinate deviation is given by repeated measurements of the borehole opening coordinates, while the upper limits of the borehole inclination deviation and borehole direction deviation are given by the instrument accuracy and repeated measurement results of the inclination measurement records. When calculating the trajectory deviation upper limit parameter during the spatial mapping process, the angle values ​​of the upper limits of the borehole inclination deviation and borehole direction deviation are first converted to radians. Then, the corresponding lateral offset upper limit is calculated based on the abnormal measuring point depth parameter. The upper limit of the lateral offset in the borehole inclination direction is taken by multiplying the abnormal measuring point depth parameter by the upper limit of the borehole inclination deviation in radians, and the upper limit of the lateral offset in the borehole direction is taken by multiplying the abnormal measuring point depth parameter by the upper limit of the borehole direction deviation in radians. The trajectory deviation upper limit parameter is the sum of the upper limits of the borehole opening coordinate deviation, the upper limit of the lateral offset in the borehole inclination direction, and the upper limit of the lateral offset in the borehole direction. The trajectory deviation upper limit parameter, together with the mapping distance parameter, is used for mapping confidence parameter marking.

[0093] The spatial mapping module marks mapping relationships based on confidence threshold parameters, which include distance threshold parameters and deviation threshold parameters. The distance threshold parameter is set as a multiple of the grid resolution parameter, and the deviation threshold parameter is also set as a multiple of the grid resolution parameter. If the mapped distance parameter is not greater than the distance threshold parameter and the trajectory deviation upper limit parameter is not greater than the deviation threshold parameter, the mapping relationship is marked as a high-confidence mapping and written to the abnormal mapping table. If the mapped distance parameter is greater than the distance threshold parameter or the trajectory deviation upper limit parameter is greater than the deviation threshold parameter, the mapping relationship is marked as a low-confidence mapping and written to the abnormal mapping table. Simultaneously, the mapped distance parameter, trajectory deviation upper limit parameter, and corresponding threshold parameter are recorded in the abnormal mapping table.

[0094] For example, in the technical solution of the present invention, the distance threshold parameter can be twice the grid resolution parameter, and the deviation threshold parameter can be three times the grid resolution parameter.

[0095] It should be noted that the distance threshold parameter is set to twice the grid resolution parameter primarily to allow for margin in the discretization error of the slope grid. The slope model discretizes a continuous slope into grid cells. After anomaly spatial coordinates are projected onto the slope, there is an inherent discretization error between the actual landing point and the coordinates of the center of the matching slope cell. Even without considering any measurement noise, the distance between the cell center and any point within the cell can be close to the order of one grid resolution. Adding point cloud noise and point cloud registration residuals, the mapped distance parameter may exceed one grid resolution even without anomalies. Setting the distance threshold to twice the grid resolution allows the projected landing point to deviate from several adjacent cells without being directly classified as low confidence, thus avoiding misclassifying normal deviations caused by grid discretization as low confidence.

[0096] The deviation threshold parameter is set to three times the grid resolution parameter, primarily to allow for a margin in the "cumulative upper limit of borehole positioning deviation." The upper limit of trajectory deviation is obtained by accumulating the upper limit of borehole coordinate deviation and the upper limit of inclination deviation along the depth of the anomalous measurement point. The greater the depth, the greater the cumulative deviation, and this deviation is manifested as a lateral drift of the anomalous spatial coordinates in the slope projection direction. If the deviation threshold is only on the same order of magnitude as the grid resolution, anomalous measurement points at deeper locations are easily marked as low confidence due to the cumulative deviation, resulting in overly conservative mapping results. Setting the deviation threshold to three times the grid resolution is equivalent to allowing the anomalous spatial coordinates to maintain a high confidence judgment even when crossing a limited number of grid cells on the slope, thus taking into account usability in deep borehole scenarios.

[0097] Two or three times is an example default value. Whether it is better depends on the data quality and engineering constraints on site. For example, when the grid resolution is finer, the threshold multiplier can be increased or decreased accordingly; when the point cloud registration residual is large, the distance threshold multiplier should be increased appropriately; when the borehole depth is greater or the inclination accuracy is lower, the deviation threshold multiplier should be increased appropriately. The multiplier reflects how much margin is willing to be left for error, and the goal is to make the upper bound of the error corresponding to high confidence mapping controllable, rather than pursuing the optimal of a certain fixed multiplier.

[0098] An anomaly mapping relationship set is formed by combining the point cloud frame index, deep anomaly event index, anomaly spatial coordinates, matching slope unit number, and mapping confidence parameters. The anomaly mapping relationship set is grouped by the point cloud frame index to obtain a frame-level anomaly mapping result set, which is then output to subsequent modules for performing fusion localization and instability mode discrimination along the slope unit number dimension.

[0099] The fusion positioning module is used to map and aggregate deep anomaly event indexes based on the anomaly mapping relationship set and the slope model, according to the slope unit number. It generates unit risk scores based on the mapping confidence parameters and triggering feature quantities, and diffuses and aggregates the unit risk scores according to the slope adjacency relationship to obtain a risk hot zone set including hot zone number, hot zone unit set, hot zone risk score and hot zone associated deep anomaly event index set.

[0100] The method for obtaining the set of risk hot zones includes:

[0101] The fusion localization module reads the anomaly mapping table to obtain the set of anomaly mapping relationships and mapping confidence parameters, and reads the slope model to obtain the coordinates of the cell center corresponding to the slope cell number. It reads the anomaly event table to obtain the trigger timestamp, trigger feature quantity, and threshold value corresponding to the deep anomaly event index. It reads the point cloud frame index table to obtain the unified timestamp corresponding to the point cloud frame index in the anomaly mapping relationship set. The point cloud frame index table records the point cloud frame index, unified timestamp, and the storage location of the point cloud frame data volume.

[0102] The following parameters are established: fusion time window length, unit adjacency distance threshold, diffusion radius, distance decay, time decay, preset intensity upper limit, confidence threshold, preset low confidence weight, hot zone risk threshold, and minimum number of hot zone units. The fusion time window length parameter is used to filter abnormal mapping relationship records participating in the fusion and is used for time decay normalization. The unit adjacency distance threshold parameter is used to determine the adjacency relationship between slope unit numbers. The diffusion radius parameter is used to limit the maximum number of adjacent layers for diffusion. The distance decay and time decay parameters are used to calculate the decay weight of diffusion contribution. The preset intensity upper limit, confidence threshold, and preset low confidence weight parameters are used to convert abnormal intensity parameters and mapping confidence parameters into mapping weight parameters.

[0103] It should be noted that all parameters used by the fusion positioning module are loaded from the parameter storage area. The parameter storage area is written by the project configuration file during the system deployment phase, and includes the parameter version number and effective timestamp. After the point cloud grid resolution setting, scanning cycle setting and deep sensor noise statistics are completed on site, the corresponding parameter values ​​are solidified into the parameter storage area. The fusion positioning module reads the parameter once at the beginning of each fusion cycle and keeps it unchanged within the cycle.

[0104] The fusion time window length parameter is derived from the update cycle of the unified timestamp of point cloud frames in the point cloud frame index table. The fusion time window length parameter is taken as an integer multiple of the point cloud frame update cycle to cover several consecutive frames. The cell adjacency distance threshold parameter is derived from the grid resolution parameter used when constructing the slope model. The cell adjacency distance threshold parameter is taken as twice the grid resolution parameter to ensure that adjacent raster cells can be determined as adjacent. The diffusion radius parameter is derived from the slope raster discrete scale and diffusion range settings. The diffusion radius parameter is measured by the number of adjacent layers and takes a fixed integer value to limit the maximum number of layers for layer-by-layer diffusion. The distance decay parameter and time decay parameter are derived from the diffusion contribution decay strategy configuration. Both are real numbers greater than zero and used for linear decay calculation.

[0105] The preset intensity upper limit parameter is derived from the normalization scale configuration of the abnormal intensity parameter. This parameter is a real number greater than one and is used to truncate the abnormal intensity parameter to the range of zero to one. The preset low confidence weight parameter is derived from the mapping confidence reduction strategy configuration. This parameter is a real number greater than zero and less than one. The confidence threshold parameter is derived from the mapping confidence field generation caliber in the abnormal mapping table. This field is written by the spatial mapping module based on the distance threshold and deviation threshold parameters. These parameters are also configured in the parameter storage area and solidified into mapping confidence fields after use by the spatial mapping module. The fusion positioning module does not recalculate the mapping confidence field; it only uses this field to obtain either a high-confidence mapping or a low-confidence mapping and maps it to the confidence weight parameter. The hot zone risk threshold parameter and the minimum hot zone unit number parameter are derived from the hot zone output constraint configuration. The hot zone risk threshold parameter is used for unit risk scoring and filtering, while the minimum hot zone unit number parameter is used to remove connected clustering results that are too small.

[0106] For ease of understanding, a set of example values ​​consistent with the above sources are provided. The grid resolution parameter is set to 0.5 meters, and the cell adjacency distance threshold parameter is set to 1 meter. The point cloud frame update cycle is set to 10 seconds, and the fusion time window length parameter is set to 120 seconds. The diffusion radius parameter is set to 3, indicating a maximum diffusion of three layers of adjacent cells. The distance decay parameter is set to 1, and the time decay parameter is set to 1. The preset intensity upper limit parameter is set to 2, and the preset low confidence weight parameter is set to 0.5. The distance threshold parameter is set to 1.5 meters, and the deviation threshold parameter is set to 2.25 meters. Based on these values, the spatial mapping module marks the mapping confidence field as a high-confidence mapping or a low-confidence mapping. The hot zone risk threshold parameter is set to 1.2, and the minimum number of hot zone cells is set to 3.

[0107] Here is another set of example data corresponding to this module. The example set includes three records of anomaly mapping relationships, with point cloud frame indices of 101, 102, and 103, corresponding to unified timestamps of 13:05:00, 13:05:10, and 13:05:20 respectively. The fusion baseline timestamp is 13:05:20, and the fusion time window length parameter is 120 seconds. Therefore, all three records meet the time window filtering requirement. In the anomaly event table, the trigger channel name for deep anomaly event index 21 is the pore pressure change rate channel, the trigger feature is 1.2 kPa / s, and the threshold value is 0.6 kPa / s. Therefore, the anomaly intensity parameter is 2, and the normalized anomaly intensity parameter is 1. Since this record's mapping confidence field is a high-confidence mapping, the confidence weight parameter is 1, and the mapping weight parameter is 1.

[0108] The trigger channel name for Deep Anomaly Index 22 is the Tilt Change Rate Channel, the trigger feature is 0.04 degrees per second, the threshold value is 0.02 degrees per second, so the anomaly intensity parameter is 2, and the normalized anomaly intensity parameter is 1. The record's mapping confidence field is a low-confidence mapping, so the confidence weight parameter is 0.5, and the mapping weight parameter is 0.5. During diffusion updates, the path length parameter from the diffusion start point to a candidate diffusion unit number is 2, and the diffusion radius parameter is 3, so the normalized distance is 2 divided by 3. The distance decay weight is one-third obtained by subtracting the normalized distance multiplied by the distance decay parameter from one. The time difference parameter between the trigger timestamp and the unified timestamp of the diffusion start point is 20 seconds, and the fusion time window length parameter is 120 seconds, so the normalized time is one-sixth, and the time decay weight is six-fifths obtained by subtracting the normalized time multiplied by the time decay parameter from one. The diffusion contribution value is the mapping weight parameter multiplied by the distance decay weight, then multiplied by the time decay weight, and accumulated in the unit risk score of the candidate diffusion unit number. During the threshold screening stage, slope units with a risk score of not less than 1.2 are entered into connected clustering. When the number of hot zone units output by connected clustering is not less than three, hot zone numbers are assigned and risk hot zone sets are output.

[0109] A slope adjacency table is established based on the center coordinates of each slope element number in the slope model. The table is constructed by iterating through all other slope element numbers for each element number, calculating the distance between their center coordinates, and adding the slope element number to its adjacency set when the distance is less than a threshold parameter. This establishes an index relationship between slope element numbers and their adjacency sets. The adjacency distance threshold parameter is set to a multiple of the grid resolution parameter to ensure that the adjacency set is consistent with the discrete scale of the slope grid.

[0110] During the data fusion screening phase, a fusion baseline timestamp is determined based on the point cloud frame index table. The fusion baseline timestamp is taken as the maximum value of the unified timestamp corresponding to the point cloud frame index in the set of abnormal mapping relationships being processed. A time window screening is performed on the set of abnormal mapping relationships. For each abnormal mapping relationship, the point cloud frame index is read, and the unified timestamp is read from the point cloud frame index table. The time difference between the unified timestamp and the fusion baseline timestamp is calculated. If the time difference is not greater than the fusion time window length parameter, the abnormal mapping relationship is added to the fusion set; if the time difference is greater than the fusion time window length parameter, it is not included in this fusion process.

[0111] During the unit risk table generation phase, the participating fusion sets are traversed, and the trigger feature quantity and corresponding threshold value are read from the anomaly event table according to the deep anomaly event index to calculate the anomaly intensity parameter. The anomaly intensity parameter is calculated by dividing the trigger feature quantity by the corresponding threshold value to obtain the feature ratio parameter. The threshold value is taken from the threshold value field of the anomaly event table. When the threshold value is less than or equal to zero, the anomaly mapping relationship is marked as an invalid record and skipped in subsequent weight calculation. The anomaly intensity parameter is taken from the feature ratio parameter. Normalization processing is performed on the anomaly intensity parameter to obtain the normalized anomaly intensity parameter. The normalization processing method is to first subtract one from the anomaly intensity parameter, and then divide the result by the preset intensity upper limit parameter minus one. When the normalization result is less than zero, it is taken as zero; when the normalization result is greater than one, it is taken as one.

[0112] The fusion positioning module calculates confidence weight parameters based on the mapping confidence parameters. These mapping confidence parameters originate from the mapping confidence field of the abnormal mapping table. This field is marked and written into the abnormal mapping table by the preceding spatial mapping process based on the mapping distance parameter, trajectory deviation upper limit parameter, distance threshold parameter, and deviation threshold parameter. The fusion positioning module does not repeatedly perform mapping confidence determination; it only performs weight transformation based on the mapping confidence fields already written in the abnormal mapping table. When the mapping confidence parameter is marked as a high-confidence mapping, the confidence weight parameter is set to one; when the mapping confidence parameter is marked as a low-confidence mapping, the confidence weight parameter is a preset low-confidence weight parameter.

[0113] The normalized anomaly intensity parameter is multiplied by the confidence weight parameter to obtain the mapping weight parameter, which is then written into the mapping weight field of the anomaly mapping table. A unit risk table is created using the slope unit number as the key, recording the slope unit number, unit risk score, and associated event set. The unit risk score is updated by accumulating the mapping weight parameter into the unit risk score corresponding to the matching slope unit number, and adding the deep anomaly event index to the associated event set corresponding to that slope unit number.

[0114] During the diffusion update phase, the unit risk table is updated. The diffusion update uses the matching slope unit number with an abnormal mapping relationship as the diffusion starting point, and diffuses layer by layer in the slope adjacency table according to adjacency relationships to obtain a set of candidate diffusion unit numbers. For each diffusion path from the diffusion starting point to a candidate diffusion unit number, a path length parameter is calculated, which is the number of adjacent layers traversed along the diffusion path. A normalized distance is calculated based on the path length parameter, which is the path length parameter divided by the diffusion radius parameter. If the normalized distance is greater than one, no further diffusion contribution is added to that candidate diffusion unit number.

[0115] The distance attenuation weight is calculated based on the normalized distance. The calculation method is to subtract the normalized distance from one and multiply it by the distance attenuation parameter. The weight is zero when it is less than zero and one when it is greater than one. The point cloud frame index corresponding to the anomaly mapping relationship of the diffusion starting point is read, and the unified timestamp of the diffusion starting point is read from the point cloud frame index table. Simultaneously, the trigger timestamp is read from the anomaly event table according to the deep anomaly event index. The time difference parameter is calculated as the absolute difference between the trigger timestamp and the unified timestamp of the diffusion starting point. The normalized time is calculated based on the time difference parameter, which is the time difference parameter divided by the fusion time window length parameter. The time attenuation weight is reset to zero when the normalized time is greater than one. The time attenuation weight is calculated based on the normalized time. The calculation method is to subtract the normalized time from one and multiply it by the time attenuation parameter. The weight is zero when it is less than zero and one when it is greater than one.

[0116] The diffusion contribution value is obtained by multiplying the mapping weight parameter, the distance decay weight, and the time decay weight. The diffusion contribution value is accumulated into the unit risk score corresponding to the candidate diffusion unit number, and the deep anomaly event index is added to the associated event set corresponding to the candidate diffusion unit number.

[0117] It should be noted that the path length parameter, distance decay weight, time difference parameter, time decay weight, and diffusion contribution value are diffusion control parameters calculated by this invention during the diffusion process of unit risk scoring. The path length parameter is used to characterize the adjacency level distance between the candidate diffusion slope unit number and the diffusion starting point; the distance decay weight is used to limit the risk contribution from decreasing as the spatial adjacency level increases. The time difference parameter is used to characterize the temporal deviation between the trigger timestamp and the unified timestamp corresponding to the diffusion starting point; the time decay weight is used to limit the risk contribution from decreasing as the time deviation increases. The diffusion contribution value is jointly determined by the mapping weight parameter, the distance decay weight, and the time decay weight, and is used to update the unit risk score of the candidate diffusion slope unit number. By introducing the above diffusion control parameters, the risk aggregation result is simultaneously constrained by spatial adjacency and temporal consistency.

[0118] During the hot zone generation phase, a threshold screening is performed on the updated unit risk table. The screening condition is that the unit risk score is not less than the hot zone risk threshold parameter. For the set of slope unit numbers that meet the screening condition, connectivity clustering is performed according to the slope adjacency list. The clustering outputs a set of hot zone units, and a hot zone number is assigned to each set. A hot zone risk score is calculated for each hot zone number. The hot zone risk score is taken as the maximum unit risk score within the corresponding hot zone unit set. The union of the associated event sets of each slope unit number within the hot zone unit set is used to form a set of hot zone associated deep anomaly event indexes. A size constraint is applied to the hot zone unit set. If the number of units in the hot zone unit set is less than the minimum number of hot zone units, the hot zone number is not output. Hot zone numbers that meet the size constraint are retained, forming a risk hot zone set, which is then output to the instability mode discrimination module for use.

[0119] It should be noted that the hot zone risk score and the hot zone-associated deep anomaly event index set are regional-level result parameters calculated during the hot zone generation stage of this invention. The hot zone risk score is used to characterize the risk level of the region corresponding to the hot zone number within the current fusion time window; the hot zone-associated deep anomaly event index set is used to characterize the source of deep anomaly events that form the hot zone risk score. The hot zone risk score and the hot zone-associated deep anomaly event index set are jointly stored under the same hot zone number, enabling the subsequent instability mode identification process to simultaneously use regional risk level and event composition information to perform mode recognition.

[0120] For example, to facilitate full disclosure, the preset intensity upper limit parameter is set to two, the preset low confidence weight parameter is set to 0.5, the diffusion radius parameter is set to three, the distance attenuation parameter is set to one, the time attenuation parameter is set to one, and the hot zone risk threshold parameter and the minimum number of hot zone units parameter are configured in the parameter storage area as required by the project and take effect after being loaded by the fusion positioning module.

[0121] The instability mode discrimination module sums the mapping weight parameters in the anomaly mapping table by slope unit number to obtain unit weight parameters, performs weighted summation to obtain the hot zone representative normal vector, and calculates the hot zone normal consistency parameter; it counts the triggering ratio parameter by triggering channel name and takes the intensity indication parameter with the maximum value of the ratio of triggering feature quantity to the corresponding threshold value; it generates the dominant parameter and corrects the tilting dominant parameter, determines the instability mode label and mode confidence score, writes them into the hot zone mode result set, and outputs them.

[0122] like Figure 4 As shown, the method for obtaining the hot zone model result set includes:

[0123] The instability mode discrimination module reads the risk hotspot set, slope model, anomaly mapping table, and anomaly event table to establish a mode discrimination parameter set. This parameter set includes a mode candidate set, feature summarization rule parameters, a dominant fusion parameter set, a mode determination threshold parameter set, normal consistency threshold parameters, normal correction parameter set, and mode confidence score parameters. The mode candidate set includes tensile cracking mode, shear slip mode, toppling deformation mode, and pore pressure driven mode. The feature summarization rule parameters include weighted summation rules and maximum value rules. The dominant fusion parameter set includes intensity fusion weight parameters and proportion fusion weight parameters, used to summarize intensity indication parameters and channel triggering proportion parameters into dominant mode parameters. The mode determination threshold parameter set is used to determine the execution order of displacement dominant parameters, shear dominant parameters, toppling dominant parameters, and pore pressure dominant parameters.

[0124] It should be noted that the feature aggregation rule parameters are used to aggregate multi-unit, multi-event data of the hot zone unit set and the hot zone associated deep anomaly event index set. The weighted summation rule is used to aggregate unit attributes within the hot zone unit set. The weights of the weighted summation rule are fixed to the unit weight parameters, which are determined by the anomaly mapping table. The calculation method is to sum the mapping weight parameters of all anomaly mapping relationships associated with the same slope unit number. A correspondence is established between the unit weight parameters and the slope unit numbers and written into the unit weight table, which records the slope unit number and the unit weight parameters.

[0125] The weighted summation rule is executed as follows: For each slope element number within the thermal zone set, the summation value corresponding to that slope element number is read from the slope model. This summation value can be one of the following: element aspect, element slope, or element normal vector. Simultaneously, the element weight parameter corresponding to that slope element number is read from the element weight table as the weight value. The weight values ​​of all slope element numbers within the thermal zone set are summed to obtain the total weight. The weight value of each slope element number is divided by the total weight to obtain the normalized weight. Then, the summation value of each slope element number is multiplied by the normalized weight and summed to obtain the weighted summation value. If the total weight is zero, an equal-weight rule is used, which involves directly averaging the summation values ​​within the thermal zone set to obtain the weighted summation value.

[0126] When the values ​​to be summarized are unit normal vectors, the weighted summation rule is executed as follows: the unit normal vectors of each slope unit number within the hot zone unit set are weighted and accumulated according to normalized weights to obtain a composite normal vector. The composite normal vector is then normalized to obtain the representative normal vector of the hot zone, ensuring that the representative normal vector of the hot zone is a unit vector. If there are missing unit normal vectors within the hot zone unit set, the missing terms are ignored, and the remaining weights are re-normalized before the weighted accumulation is performed.

[0127] The maximum value rule is executed as follows: For each deep anomaly event index within the hotspot-associated deep anomaly event index set, the trigger feature quantity and corresponding threshold value are read from the anomaly event table, and a feature ratio parameter is calculated. The feature ratio parameter is the trigger feature quantity divided by the threshold value. When the corresponding threshold value is less than or equal to zero, the threshold value corresponding to the trigger channel name is read from the threshold parameter set for calculation. If a valid threshold value still cannot be obtained, the deep anomaly event index is marked as an invalid ratio event and does not participate in the maximum value calculation. The maximum feature ratio parameter is obtained by taking the maximum value of the feature ratio parameters of all valid ratio events, and the maximum feature ratio parameter is used as the extreme indicator value of the hotspot. If multiple deep anomaly event indices have the same feature ratio parameter and are all maximum values, the deep anomaly event index with the newer trigger timestamp is selected first. If the trigger timestamps are still the same, the mapping weight parameter is retrieved from the anomaly mapping table according to the deep anomaly event index, and the deep anomaly event index with the larger mapping weight parameter is selected to ensure the determinism of the maximum value rule.

[0128] The process iterates through each hotspot number in the risk hotspot set, retrieving the corresponding hotspot unit set, hotspot risk score, and hotspot-related deep anomaly event index set. Based on the hotspot unit set, it extracts the unit normal vector and unit center coordinates corresponding to each slope unit number from the slope model, forming a hotspot geometric data package. Based on the hotspot-related deep anomaly event index set, it retrieves the trigger channel name, trigger feature quantity, threshold value, and trigger timestamp corresponding to each deep anomaly event index from the anomaly event table, forming a hotspot event data package. Furthermore, it summarizes the unit weight parameters based on the anomaly mapping table. The unit weight parameters are calculated by summing the mapping weight parameters of all anomaly mapping relationships associated with the same slope unit number, establishing a correspondence between slope unit numbers and unit weight parameters for use as the weight source for subsequent weighted summation rules.

[0129] Furthermore, the mode discrimination parameter set is the parameter configuration data of this invention, derived from the parameter configuration file and field calibration records during the system deployment phase, and written into the parameter storage area for the instability mode discrimination module to call. The mode discrimination parameter set includes a mode candidate set, feature summary rule parameters, a mode judgment threshold parameter set, a normal consistency threshold parameter, a normal correction parameter set, and a mode confidence score parameter. The mode candidate set is a fixed enumeration set, with elements including tensile cracking mode, shear slip mode, toppling deformation mode, and pore pressure driven mode. The mode candidate set remains unchanged during operation after being written into the parameter storage area. The intensity fusion weight parameter and the proportion fusion weight parameter in the dominant fusion parameter set are determined by statistical results of historical monitoring data and field debugging results, and satisfy the condition that the sum of the two is one.

[0130] The feature summarization rule parameters are used to fix the summarization scope. The feature summarization rule parameters include weighted summation rules and maximum value rules. The weight of the weighted summation rule is fixed to the unit weight parameter, which is obtained by summing the mapping weight parameters under the same slope unit number in the anomaly mapping table. The maximum value rule is used to select the event with the largest feature ratio parameter from the set of deep anomaly event indexes associated with hot zones as the channel intensity indicator.

[0131] The set of mode determination threshold parameters, normal consistency threshold parameters, normal penalty coefficient parameters, intensity fusion weight parameters, proportion fusion weight parameters, difference weight parameters, risk weight parameters, difference normalization upper limit parameters, risk normalization upper limit parameters, and confidence score truncation parameters are all configuration parameters during the deployment period. The configuration is based on the geological conditions of the field monitoring object, the noise level of deep sensor measurements, the point cloud registration quality, and the statistical results of historical stable periods. After configuration, the parameters are written to the parameter storage area with the parameter version number and effective timestamp to ensure that the discrimination criteria are consistent and reproducible within the same time period.

[0132] For ease of understanding, a set of example parameter values ​​are provided. The example values ​​for the mode determination threshold parameter set are: pore pressure dominant threshold parameter 1.3, tilting dominant threshold parameter 1.2, shear dominant threshold parameter 1.1, and displacement dominant threshold parameter 1.1. The example value for the normal consistency threshold parameter is 8 degrees, and the example value for the normal penalty coefficient parameter is 0.7. The example value for the dominant fusion parameter set is: intensity fusion weight parameter 0.6, and proportion fusion weight parameter 0.4. The example value for the mode confidence score parameters is: difference weight parameter 0.6, risk weight parameter 0.4, difference normalization upper limit parameter 1, risk normalization upper limit parameter 10, confidence score lower limit parameter 0, and confidence score upper limit parameter 1. These values ​​are examples used to illustrate the source and usage of the parameters. The parameter storage area can be updated according to project needs, and different configurations are distinguished by parameter version numbers.

[0133] The thermal zone normal consistency parameter is calculated on the thermal zone geometry data package. First, the representative normal vector of the thermal zone is calculated according to the weighted summation rule in the feature summary rule parameters. The calculation method is as follows: for each slope unit number in the thermal zone unit set, the unit normal vector and the corresponding unit weight parameter are read as weight values. The weight values ​​are summed to obtain the weight sum and normalized. Then, the normal vectors of each unit are weighted and accumulated according to the normalized weight to obtain the composite normal vector. The composite normal vector is normalized to obtain the representative normal vector of the thermal zone. When the weight sum is equal to zero, the unit normal vectors are averaged according to the equal weight rule and then normalized to obtain the representative normal vector of the thermal zone. Subsequently, for each slope unit number in the thermal zone unit set, the angle between the unit normal vector and the representative normal vector of the thermal zone is calculated to obtain the normal deviation angle set. When the weight sum is greater than zero, the thermal zone normal consistency parameter is the weighted average of the normal deviation angle set calculated according to the unit weight parameter; when the weight sum is equal to zero, the thermal zone normal consistency parameter is the arithmetic mean of the normal deviation angle set.

[0134] The set of channel triggering percentage parameters and the set of intensity indication parameters are calculated on the hot zone event data packet. The set of channel triggering percentage parameters is statistically analyzed by triggering channel name, yielding triggering percentage parameters for displacement change rate, strain change rate, tilt angle change rate, and pore pressure change rate. Each percentage parameter is calculated by dividing the number of events in the hot zone-related deep anomaly event index set where the triggering channel name equals the corresponding change rate channel by the total number of events. The set of intensity indication parameters is calculated according to the maximum value rule in the feature summary rules. Specifically, for each deep anomaly event index in the hot zone-related deep anomaly event index set, the triggering feature quantity and threshold value are read, and the feature ratio parameter is calculated by dividing the triggering feature quantity by the threshold value. When the threshold value is less than or equal to zero, the event is marked as an invalid ratio event and does not participate in the maximum value calculation. The parameter for indicating a sudden increase in pore pressure in the hot zone is the maximum value of the characteristic ratio parameter of the effective ratio event of the pore pressure change rate channel, which is equal to the trigger channel name. The parameter for indicating a tilt in the hot zone is the maximum value of the characteristic ratio parameter of the effective ratio event of the tilt angle change rate channel, which is equal to the trigger channel name. The parameter for indicating a displacement strength in the hot zone is the maximum value of the characteristic ratio parameter of the effective ratio event of the displacement change rate channel, which is equal to the trigger channel name. The parameter for indicating a strain strength in the hot zone is the maximum value of the characteristic ratio parameter of the effective ratio event of the strain change rate channel, which is equal to the trigger channel name. When there is no effective ratio event for a certain channel, the corresponding strength indicator parameter is set to one.

[0135] It should be noted that the threshold value is the threshold value field of the record corresponding to the deep anomaly event index in the anomaly event table. The threshold value field is generated when the anomaly detection process writes to the anomaly event table. The generation method is to read the value of the corresponding threshold parameter from the threshold parameter set according to the trigger channel name and solidify it into the threshold value field. Specifically, when the trigger channel name is the displacement change rate channel, the threshold value is the displacement change rate threshold value; when the trigger channel name is the strain change rate channel, the threshold value is the strain change rate threshold value; when the trigger channel name is the tilt angle change rate channel, the threshold value is the tilt angle change rate threshold value; and when the trigger channel name is the pore pressure change rate channel, the threshold value is the pore pressure change rate threshold value.

[0136] For example, if the displacement change rate threshold is 0.05 mm / s, and the trigger channel name of a certain deep anomaly event index is the displacement change rate channel, and the trigger feature quantity is 0.08 mm / s, then the corresponding threshold value is 0.05 mm / s, and the ratio of the trigger feature quantity to the corresponding threshold value is 1.6.

[0137] A feature set for mode discrimination is constructed. The instability mode discrimination module reads the normal correction parameter set and the dominant fusion parameter set. The dominant fusion parameter set includes intensity fusion weight parameters and proportion fusion weight parameters, and the sum of intensity fusion weight parameters and proportion fusion weight parameters is equal to one. The displacement dominant parameter is calculated by multiplying the hot zone displacement intensity indicator parameter by the intensity fusion weight parameter and summing it with the displacement change rate trigger proportion parameter multiplied by the proportion fusion weight parameter. The shear dominant parameter is calculated by multiplying the hot zone strain intensity indicator parameter by the intensity fusion weight parameter and summing it with the strain change rate trigger proportion parameter multiplied by the proportion fusion weight parameter. The pore pressure dominant parameter is calculated by multiplying the hot zone pore pressure surge indicator parameter by the intensity fusion weight parameter and summing it with the pore pressure change rate trigger proportion parameter multiplied by the proportion fusion weight parameter. The initial value of the tilting dominant parameter is calculated by multiplying the hot zone tilting indicator parameter by the intensity fusion weight parameter and summing it with the tilt angle change rate trigger proportion parameter multiplied by the proportion fusion weight parameter.

[0138] It should be noted that the hot zone normal consistency parameter, the channel triggering proportion parameter set, the channel intensity indication parameter, and the dominant parameter of each mode are discrimination feature parameters calculated by this invention during the instability mode discrimination process. The hot zone normal consistency parameter is calculated from the deviation between the normal vector of the unit within the hot zone unit set and the normal vector of the hot zone representative, and is used to characterize the consistency of the hot zone geometry, and serves as the correction input for the dominant parameter. The channel triggering proportion parameter set is used to characterize the triggering composition ratio of each rate of change channel in the hot zone associated deep anomaly event index set. The channel intensity indication parameter is used to characterize the upper bound of the superthreshold intensity of the corresponding channel within the hot zone. The dominant parameter of each mode is generated by weighting the channel triggering proportion parameter and the channel intensity indication parameter according to the dominant fusion parameter set, and is used to convert the multi-channel event composition into a comparable mode discrimination input. Through the hierarchical calculation of the above discrimination feature parameters, the determination of the instability mode label is simultaneously constrained by geometry, event composition, and triggering intensity.

[0139] Normal consistency correction is performed on the dominant dumping parameter to ensure that the thermal zone normal consistency parameter actually participates in the discrimination chain. The instability mode discrimination module reads the normal consistency threshold parameter and the normal penalty coefficient parameter, where the normal penalty coefficient parameter is a real number greater than zero and less than one. When the thermal zone normal consistency parameter is not greater than the normal consistency threshold parameter, the dominant dumping parameter remains unchanged. When the thermal zone normal consistency parameter is greater than the normal consistency threshold parameter, the dominant dumping parameter is multiplied by the normal penalty coefficient parameter to obtain the corrected dominant dumping parameter, and this corrected dominant dumping parameter is used to replace the original dominant dumping parameter in the mode discrimination feature set. This ensures that the discrimination conditions for the dumping mode automatically tighten when the thermal zone geometric consistency is insufficient.

[0140] The instability mode label is determined by comparing the mode discrimination feature set with a preset set of mode judgment threshold parameters. The preset set of mode judgment threshold parameters includes displacement-dominant threshold parameters, shear-dominant threshold parameters, toppling-dominant threshold parameters, and pore pressure-dominant threshold parameters. The discrimination rule is as follows: when the pore pressure-dominant parameter is not less than the pore pressure-dominant threshold parameter, the pore pressure-driven mode is selected as the instability mode label from the mode candidate set; otherwise, when the toppling-dominant parameter is not less than the toppling-dominant threshold parameter, the toppling deformation mode is selected as the instability mode label from the mode candidate set; otherwise, when the shear-dominant parameter is not less than the shear-dominant threshold parameter, the shear slip mode is selected as the instability mode label from the mode candidate set; otherwise, when the displacement-dominant parameter is not less than the displacement-dominant threshold parameter, the tension cracking mode is selected as the instability mode label from the mode candidate set; if none of the above conditions are met, the instability mode label is determined as a mode to be confirmed, and the discrimination is repeated after updating the risk hot zone set in the subsequent fusion time window.

[0141] It should be noted that the discrimination rules are executed sequentially according to a preset priority order. First, it is determined whether the pore pressure dominant parameter meets the pore pressure dominant threshold parameter. If it does, the instability mode label is directly determined as the pore pressure driven mode and the current discrimination ends. If it does not meet the threshold parameter, it is further determined whether the tilting dominant parameter meets the tilting dominant threshold parameter. If it does, the instability mode label is directly determined as the tilting deformation mode and the current discrimination ends. If it does not meet the threshold parameter, it is further determined whether the shear dominant parameter meets the shear dominant threshold parameter. If it does, the instability mode label is directly determined as the shear slip mode and the current discrimination ends. If it does not meet the threshold parameter, it is further determined whether the displacement dominant parameter meets the displacement dominant threshold parameter. If it does, the instability mode label is directly determined as the tension cracking mode and the current discrimination ends. If none of the above conditions are met, the instability mode label is determined as a mode to be confirmed.

[0142] In the instability mode identification stage, a discrimination rule is adopted that is executed sequentially according to priority. This is mainly used to solve the problems of mode conflict and label uniqueness when multiple channel anomalies occur simultaneously within the same hot zone. The risk hot zone set consists of a set of deep anomaly event indexes associated with the hot zone. Within a fusion time window, the set of deep anomaly event indexes associated with the hot zone may simultaneously contain trigger events for displacement change rate, strain change rate, tilt angle change rate, and pore pressure change rate. If parallel or disordered discrimination is used, the same hot zone number may be assigned multiple instability mode labels within the same time window, making it impossible for subsequent handling processes to be uniformly invoked based on a single label. Therefore, this invention sets a deterministic sequential discrimination to output a unique label.

[0143] The priority of sequential discrimination is arranged in the following order: pore pressure driving mode, toppling deformation mode, shear slip mode, and tensile cracking mode. This is based on the physical driving correlation strength and temporal characteristics differences of each dominant parameter. The pore pressure change rate channel reflects the rapid change of pore pressure over a short period. This change can directly alter the effective stress state and have a dominant impact on slope structural stability. Furthermore, pore pressure anomalies can appear as driving clues before displacement and strain have accumulated significantly. Therefore, when the dominant pore pressure parameter reaches a threshold, the pore pressure driving mode is prioritized to avoid displacement or strain indicators overshadowing the interpretation of pore pressure driving in subsequent discrimination. The tilt angle change rate channel reflects the dominant motion pattern of block rotation or toppling. This invention further introduces a thermal zone normal consistency parameter to correct the toppling dominant parameter, requiring toppling discrimination to simultaneously satisfy deep tilt angle triggering evidence and slope geometric consistency constraints. When these conditions are met, the toppling mode has a stronger directional indication of thermal zone motion patterns, and therefore has a higher priority than shear and tensile modes. The strain rate of change channel reflects the cumulative trend of shear deformation on structural surfaces or potential slip zones. Its directionality is stronger than that of simple displacement rate of change. However, when multiple modes coexist, strain anomalies may occur alongside pore pressure-driven or toppling deformation. Therefore, it is placed after toppling to avoid overshadowing the toppling mode. The displacement rate of change channel is used to characterize overall or local displacement changes. Its triggering can occur in various instability mechanisms, representing a relatively universal anomaly. To reduce the overshadowing of specific mechanism labels by general indicators, the dominant displacement parameter is determined last.

[0144] Based on the above sequential discrimination, the instability mode discrimination module judges each item from top to bottom according to the preset priority in each discrimination. Once a certain dominant parameter meets the corresponding threshold, the instability mode label is determined and the discrimination ends. This ensures that the same hot zone number outputs a unique and definite instability mode label within the same fusion time window, and maintains consistency with the multi-channel triggering characteristics of the hot zone associated deep abnormal event index set.

[0145] The pattern confidence score is calculated based on the pattern confidence score parameters. The instability mode discrimination module selects the largest dominant parameter from the pattern discrimination feature set as the first dominant value and the second largest dominant parameter as the second dominant value. The difference score parameter is the first dominant value minus the second dominant value. The pattern confidence score parameters include the difference weight parameter, the risk weight parameter, the difference normalization upper limit parameter, the risk normalization upper limit parameter, and the confidence score truncation parameter. The sum of the difference weight parameter and the risk weight parameter is equal to one. The difference score parameter is divided by the difference normalization upper limit parameter to obtain the normalized difference score parameter and truncated to zero or one. The hot zone risk score is divided by the risk normalization upper limit parameter to obtain the normalized risk score parameter and truncated to zero or one. The normalized difference score parameter is multiplied by the difference weight parameter and summed with the normalized risk score parameter multiplied by the risk weight parameter to obtain the pattern confidence score. Then, it is truncated to a preset range according to the confidence score truncation parameter. The instability mode discrimination module writes the hot zone number, instability mode label, and pattern confidence score into the hot zone pattern result set and outputs it.

[0146] It should be noted that the instability mode discrimination module writes the hot zone number, instability mode label, and mode confidence score into the hot zone mode result set and outputs it. The hot zone mode result set is the final discrimination output result of this invention. The hot zone number is used to maintain a one-to-one correspondence with the hot zone numbers in the risk hot zone set, so that each record in the hot zone mode result set can be used to look up the corresponding hot zone unit set, hot zone risk score, and hot zone associated deep anomaly event index set through the hot zone number. Furthermore, it can look up the trigger channel name, trigger feature quantity, and threshold value in the anomaly event table through the hot zone associated deep anomaly event index set, and look up the point cloud frame index, anomaly spatial coordinates, and slope unit number through the anomaly mapping table, thereby forming a traceable association link from the final discrimination result to the input evidence data.

[0147] The instability mode label is used to characterize the instability type judgment conclusion of the risk hot zone corresponding to the hot zone number within the current fusion time window, and the mode confidence score is used to characterize the degree of certainty of the instability mode label within the current fusion time window. When outputting the hot zone mode result set, the hot zone numbers can be sorted by mode confidence score, and the corresponding slope unit number range, instability mode label, and mode confidence score can be output using the hot zone number as an index. This allows the present invention to complete the unified output of spatial positioning and instability mode judgment results for abnormal areas of rock slopes without adding other modules.

[0148] The hot zone pattern result set can be continuously written to the result storage area according to the update cycle of the fusion time window and retain historical records. The unstable mode label and mode confidence score corresponding to the same hot zone number under different unified timestamps form a hot zone discrimination sequence. The hot zone discrimination sequence is used to characterize the discrimination change process of the hot zone number in the time dimension, and is used to compare and update with the latest discrimination result when the subsequent time window arrives, so as to ensure that the output results of this invention can be continuously recorded, reproduced and updated in the time dimension.

[0149] Example 2

[0150] Please see Figure 2 As shown, this embodiment provides a method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors, including:

[0151] Acquire multi-temporal point cloud data output by a 3D laser scanner and deep monitoring sequences output by a deep sensor, construct the binding relationship between point cloud frames and deep sequence windows, and output a set of point cloud frame indexes and a set of deep anomaly event indexes when an anomaly is triggered in the deep sequence window;

[0152] Retrieve the point cloud frame data volume corresponding to the point cloud frame index and construct the slope model; based on the depth of the abnormal measurement point corresponding to the deep abnormal event index and combined with the borehole opening coordinates and borehole trajectory parameters, calculate the abnormal spatial coordinates, project the abnormal spatial coordinates onto the slope model to obtain the slope unit number, generate an abnormal mapping table and output the abnormal mapping relationship set.

[0153] Based on the set of anomaly mapping relationships and the slope model, the deep anomaly event index is mapped and aggregated according to the slope unit number. The unit risk score is generated according to the mapping confidence parameter and the triggering feature quantity. The unit risk score is diffused and aggregated according to the slope adjacency relationship to obtain a risk hot zone set including hot zone number, hot zone unit set, hot zone risk score and hot zone associated deep anomaly event index set.

[0154] The mapping weight parameters in the anomaly mapping table are summed by slope unit number to obtain unit weight parameters. The hot zone representative normal vector is obtained by weighted summation, and the hot zone normal consistency parameter is calculated. The triggering proportion parameter is calculated by triggering channel name, and the intensity indication parameter is obtained by taking the maximum value of the ratio of the triggering feature quantity to the corresponding threshold value. The dominant parameter is generated and the tilting dominant parameter is corrected. The instability mode label and mode confidence score are determined, written into the hot zone mode result set, and output.

[0155] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0156] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors, characterized in that, include: Acquire multi-temporal point cloud data output by a 3D laser scanner and deep monitoring sequences output by a deep sensor, construct the binding relationship between point cloud frames and deep sequence windows, and output a set of point cloud frame indexes and a set of deep anomaly event indexes when an anomaly is triggered in the deep sequence window; Retrieve the point cloud frame data volume corresponding to the point cloud frame index and construct the slope model; based on the depth of the abnormal measurement point corresponding to the deep abnormal event index and combined with the borehole opening coordinates and borehole trajectory parameters, calculate the abnormal spatial coordinates, project the abnormal spatial coordinates onto the slope model to obtain the slope unit number, generate an abnormal mapping table and output the abnormal mapping relationship set. Based on the set of anomaly mapping relationships and the slope model, the deep anomaly event index is mapped and aggregated according to the slope unit number. The unit risk score is generated according to the mapping confidence parameter and the triggering feature quantity. The unit risk score is diffused and aggregated according to the slope adjacency relationship to obtain a risk hot zone set including hot zone number, hot zone unit set, hot zone risk score and hot zone associated deep anomaly event index set. The mapping weight parameters in the anomaly mapping table are summed by slope unit number to obtain unit weight parameters. The hot zone representative normal vector is obtained by weighted summation, and the hot zone normal consistency parameter is calculated. The triggering proportion parameter is calculated by triggering channel name, and the intensity indication parameter is obtained by taking the maximum value of the ratio of the triggering feature quantity to the corresponding threshold value. The dominant parameter is generated and the tilting dominant parameter is corrected. The instability mode label and mode confidence score are determined, written into the hot zone mode result set, and output.

2. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 1, characterized in that, The method for obtaining the hot zone pattern result set includes: Traverse the risk hot zone set by hot zone number to construct hot zone geometric data packets and hot zone event data packets; calculate the hot zone normal consistency parameter, channel triggering ratio parameter set, and intensity indication parameter set according to the feature aggregation rule parameters based on the anomaly mapping table summary unit weight parameters; generate displacement dominant parameters, shear dominant parameters, toppling dominant parameters, and pore pressure dominant parameters based on the dominant fusion parameter set, and correct the toppling dominant parameters based on the normal consistency threshold parameter and normal correction parameter set; determine the instability mode label based on the mode determination threshold parameter set, calculate the mode confidence score based on the mode confidence score parameter, and write it into the hot zone mode result set.

3. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 2, characterized in that, The construction of the hot zone geometry data packet and hot zone event data packet, as well as parameter calculation, includes: The hot zone geometric data package is formed by reading the element normal vectors and element center coordinates corresponding to the slope element numbers from the slope model according to the hot zone element set; the hot zone event data package is formed by reading the trigger channel name, trigger characteristic quantity, threshold value and trigger timestamp from the anomaly event table according to the hot zone associated deep anomaly event index set; the element weight parameters are obtained by summing the mapping weight parameters in the anomaly mapping table according to the slope element number; the hot zone representative normal vector is calculated according to the weighted summation rule, and the hot zone normal consistency parameter is calculated based on the angle between each element normal vector and the hot zone representative normal vector; the proportion parameters of displacement change rate triggering, strain change rate triggering, tilt angle change rate triggering, and pore pressure change rate triggering are statistically analyzed according to the trigger channel name, and the hot zone displacement intensity indication parameters, hot zone strain intensity indication parameters, hot zone tilting indication parameters and hot zone pore pressure surge indication parameters are calculated according to the maximum value rule.

4. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 3, characterized in that, The determination of the instability mode label and mode confidence score includes: Based on the intensity fusion weight parameter and the proportion fusion weight parameter in the dominant fusion parameter set, the displacement dominant parameter is obtained by weighted summation of the hot zone displacement intensity indicator parameter and the displacement change rate trigger proportion parameter; the shear dominant parameter is obtained by weighted summation of the hot zone strain intensity indicator parameter and the strain change rate trigger proportion parameter; the toppling dominant parameter is obtained by weighted summation of the hot zone toppling indicator parameter and the tilt angle change rate trigger proportion parameter; and the pore pressure dominant parameter is obtained by weighted summation of the hot zone pore pressure surge indicator parameter and the pore pressure change rate trigger proportion parameter. When the hot zone normal consistency parameter is greater than the normal consistency threshold parameter, the toppling dominant parameter is corrected according to the normal penalty coefficient parameter. The instability mode label is determined by comparing the pore pressure dominant parameter, toppling dominant parameter, shear dominant parameter, and displacement dominant parameter with the corresponding threshold in that order. The mode confidence score is calculated based on the first dominant value, the second dominant value, and the hot zone risk score according to the mode confidence score parameter.

5. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 1, characterized in that, The methods for obtaining the risk hot zone set include: The abnormal mapping relationship set is read from the abnormal mapping table by point cloud frame index; the trigger timestamp, trigger feature quantity, and threshold value are read from the abnormal event table by deep abnormal event index; and the unified timestamp is read from the point cloud frame index table by point cloud frame index. The maximum value of the unified timestamp in the abnormal mapping relationship set is used to determine the fusion reference timestamp. Abnormal mapping relationships whose time difference between the unified timestamp and the fusion reference timestamp is not greater than the fusion time window length parameter are selected. For the selected abnormal mapping relationships, the abnormal intensity parameter is calculated as the ratio of the trigger feature quantity to the threshold value and normalized according to the preset intensity upper limit parameter to obtain the normalized abnormal intensity parameter. The confidence weight parameter is obtained according to the mapping confidence parameter, with a value of 1 for high confidence mapping and a preset low confidence weight parameter for low confidence mapping. The normalized abnormal intensity parameter is multiplied by the confidence weight parameter to obtain the mapping weight parameter, and the mapping weight parameter is accumulated according to the slope unit number to obtain the unit risk score. At the same time, the deep abnormal event index set corresponding to the slope unit number is summarized.

6. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 5, characterized in that, The method for obtaining the risk hot zone set also includes: A slope adjacency table is established based on the unit center coordinates corresponding to the slope unit number in the slope model and the unit adjacency distance threshold parameter. The slope unit number matched in the anomaly mapping relationship set is used as the diffusion starting point, and diffusion is carried out layer by layer along the slope adjacency table to the number of layers limited by the diffusion radius parameter. The path length parameter is taken as the number of adjacent layers, and the normalized distance is taken as the ratio of the path length parameter to the diffusion radius parameter, and the distance attenuation weight is calculated accordingly. The trigger timestamp is read according to the deep anomaly event index and the absolute difference is calculated with the unified timestamp corresponding to the diffusion starting point point cloud frame index to obtain the time difference parameter. The normalized time is taken as the ratio of the time difference parameter to the fusion time window length parameter, and the time attenuation weight is calculated accordingly. The diffusion contribution value is obtained by multiplying the mapping weight parameter, distance decay weight, and time decay weight, and then added to the unit risk score of the candidate slope unit number. At the same time, the deep anomaly event index set corresponding to the candidate slope unit number is updated. The unit risk score is filtered according to the hot zone risk threshold parameter and clustered according to the slope adjacency list to obtain the hot zone unit set and hot zone number. The hot zone risk score is the maximum unit risk score in the hot zone unit set. The hot zone associated deep anomaly event index set is the union of the deep anomaly event index sets corresponding to the hot zone unit set, and the risk hot zone set is obtained by filtering according to the minimum hot zone unit number parameter.

7. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 1, characterized in that, The methods for obtaining the set of abnormal mapping relationships include: Traverse the deep anomaly event index set and read the trigger timestamp, trigger channel name, trigger feature quantity and threshold comparison result from the anomaly event table. Determine the point cloud frame index associated with the deep anomaly event index based on the binding relationship table and obtain the deep sequence window data packet. The data packets of the deep sequence window are grouped by the measurement point number field, and the channel value field corresponding to the trigger channel name is compared with the threshold in the threshold comparison result. The measurement point over-threshold count value is counted according to the preset continuous over-threshold count parameter, the abnormal measurement point number is determined, and the abnormal measurement point depth parameter is read from the measurement point depth field and written into the abnormal event table. Anomaly spatial coordinates are calculated based on anomaly measurement point depth parameters, borehole opening coordinates, and borehole trajectory parameters; a slope grid cell set is obtained from the slope model based on the associated point cloud frame index, and a projection matching process is performed according to the projection method parameters to determine the matching slope cell number; mapping distance parameters, trajectory deviation upper limit parameters, and point cloud resolution parameters are calculated, and mapping confidence parameters are marked according to confidence threshold parameters; the confidence threshold parameters include distance threshold parameters and deviation threshold parameters; the distance threshold parameters are set according to multiples of the grid resolution parameters, and the deviation threshold parameters are set according to multiples of the grid resolution parameters; Write the point cloud frame index, deep anomaly event index, anomaly spatial coordinates, matching slope unit number, and mapping confidence parameters into the anomaly mapping table and output it as a set of anomaly mapping relationships.

8. The method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 1, characterized in that, The methods for obtaining the point cloud frame index set and the deep anomaly event index set include: Establish a device time offset table. The device time offset table records the time offset of the 3D laser scanner and the time offset of the deep sensor. The initial value of the device time offset table is written by the device registration process. The device registration process records the device identifier, communication delay measurement results and initial clock difference measurement results. Each time a 3D laser scanner generates a frame of point cloud data, it triggers the point cloud frame storage process. The point cloud frame storage process receives the original acquisition time of the point cloud frame and generates a unified timestamp by combining it with the device time offset table. Based on the unified timestamp, a point cloud frame index is generated and written into the point cloud frame index table. The deep sensor outputs a deep monitoring sequence according to a preset sampling period. The deep monitoring sequence includes displacement data, strain data, tilt angle data and pore pressure data inside the borehole. During the deep sequence storage process, the original deep sampling time is received and combined with the equipment time offset table to generate a unified timestamp and write it into the deep sequence index table. Using the point cloud frame index table as the main index, a deep sequence window is generated for each point cloud frame index. The deep sequence window is limited by a preset time window length parameter. The start and end points of the window are determined based on the unified timestamp corresponding to the point cloud frame index. The deep monitoring sequence within the range from the start to the end of the window is retrieved from the deep sequence index table to form a deep sequence window data packet, which is then written to the binding relationship table. When the binding relationship table is generated or updated, an anomaly detection process is performed, and the point cloud frame index set and the deep anomaly event index set are output.

9. A method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and deep sensors according to claim 8, characterized in that, Methods for performing anomaly detection procedures when binding relationship tables are generated or updated include: Using deep sequence window data packets as input, the set of characteristic quantities within the window is calculated based on the abnormal triggering condition parameters. The set of characteristic quantities within the window includes the displacement change rate, strain change rate, tilt angle change rate, and pore pressure change rate within the window. The anomaly detection process compares the feature set within the window with the threshold parameter set, and determines whether the deep sequence window meets the anomaly triggering conditions based on the number of consecutive threshold exceedances. When the anomaly triggering conditions are met, a deep anomaly event index is assigned to the anomaly event, and the deep anomaly event index, trigger timestamp, trigger channel name, and comparison results of trigger features and thresholds are written into the anomaly event table. The point cloud frame index that meets the anomaly triggering conditions is added to the point cloud frame index set, and the corresponding deep anomaly event index is added to the deep anomaly event index set.

10. A system for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and a deep sensor, used to implement the method for real-time monitoring of rock slope stability based on a three-dimensional laser scanner and a deep sensor as described in any one of claims 1-9, characterized in that, include: The multi-source acquisition and timing module is used to acquire multi-temporal point cloud data output by the 3D laser scanner and deep monitoring sequences output by the deep sensor, construct the binding relationship between point cloud frames and deep sequence windows, and output the point cloud frame index set and the deep anomaly event index set when an anomaly is triggered in the deep sequence window. The spatial mapping module retrieves the point cloud frame data volume corresponding to the point cloud frame index and constructs a slope model; based on the depth of the abnormal measurement point corresponding to the deep anomaly event index and combined with the borehole opening coordinates and borehole trajectory parameters, it calculates the abnormal spatial coordinates, projects the abnormal spatial coordinates onto the slope model to obtain the slope unit number, generates an anomaly mapping table and outputs the set of anomaly mapping relationships. The fusion positioning module is used to map and aggregate the deep anomaly event index based on the anomaly mapping relationship set and the slope model, according to the slope unit number, generate unit risk scores based on the mapping confidence parameters and trigger feature quantities, and diffuse and aggregate the unit risk scores according to the slope adjacency relationship to obtain a risk hot zone set including hot zone number, hot zone unit set, hot zone risk score and hot zone associated deep anomaly event index set; The instability mode discrimination module sums the mapping weight parameters in the anomaly mapping table by slope unit number to obtain unit weight parameters, performs weighted summation to obtain the hot zone representative normal vector, and calculates the hot zone normal consistency parameter; it counts the triggering ratio parameter by triggering channel name and takes the intensity indication parameter with the maximum value of the ratio of triggering feature quantity to the corresponding threshold value; it generates the dominant parameter and corrects the tilting dominant parameter, determines the instability mode label and mode confidence score, writes them into the hot zone mode result set, and outputs them.

Images