Non-coal mine high steep slope point surface monitoring data fusion method based on dynamic weight distribution

Abstract

The invention discloses a non-coal mine high and steep slope point-surface monitoring data fusion method based on dynamic weight distribution, and relates to the technical field of non-coal mine monitoring, and the method comprises the following steps: constructing a time sequence synchronization window based on the instantaneous change rate of point position monitoring data and the overall trend of surface domain monitoring data, performing unified time baseline registration on the sampling rhythms of the point location monitoring data and the area monitoring data to form a continuous and traceable deformation starting band; and establishing a direction symbol identification matrix by utilizing displacement direction inversion information contained in the deformation starting band, and dynamically comparing the displacement direction of the point location monitoring data with the displacement direction of the area monitoring data. According to the invention, accurate alignment of point-surface monitoring data is realized through time sequence synchronization and direction identification, time mismatch is eliminated, and deformation identification precision is improved; and through gradient suppression and time-varying weight smooth regulation and control, a fusion result is kept stable and continuous, real three-dimensional deformation is generated, a false stable state is eliminated, and the reliability and timeliness of slope early warning are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of non-coal mine monitoring, in particular to a non-coal mine high and steep slope point-surface monitoring data fusion method based on dynamic weight distribution. BACKGROUND

[0002] The non-coal mine high and steep slope point-surface monitoring data fusion based on dynamic weight distribution refers to real-time calculation of weights according to time sequence state, spatial consistency and noise level for point monitoring data (such as local fine displacement information obtained by GNSS, crack meter, inclinometer, etc.) and surface monitoring data (such as overall deformation field obtained by laser radar, unmanned aerial vehicle oblique photography, InSAR, etc.), and automatic adjustment of contribution degrees of the two types of data in each round of fusion process. The system will first identify the current deformation mode, disturbance intensity and data stability of the slope, and then assign different dynamic weights to point data and surface data, so that fine but local point monitoring is responsible for describing key cracks, local slip and other details, and surface monitoring with wider coverage but more complex noise is responsible for providing overall deformation trend. Through this weight distribution method which is updated with time and working condition changes, a more stable and complete three-dimensional shape cognition can be generated, thereby improving the reliability and timeliness of slope early warning.

[0003] The prior art has the following disadvantages: In the prior art, the dynamic weight mechanism mainly relies on the instantaneous change rate of point monitoring data and the overall trend of surface monitoring data for comprehensive determination. However, when the slope deformation state suddenly switches from slow tension to rapid compression, the point monitoring will have a direction reversal jump in a short time, and the surface monitoring will still maintain a lagging overall trend due to long acquisition period and large calculation inertia. Such reverse transient scene will cause sign reversal superposition in the dynamic weight calculation process, introducing high-frequency gradient oscillation. If the oscillation amplitude continues to accumulate to the limit edge of the weight determination interval, the fusion model will be forced to converge to an invalid solution, and the output deformation result will collapse to a distorted value instantaneously, and will appear as a sustained false stable state in the time sequence. This false stable state can easily cover up the precursors of accelerated damage, so that the high and steep slope in the critical instability stage is misjudged as having low risk by the prior art, thereby damaging the reliability of the early warning link and causing serious safety hazards.

[0004] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a non-coal mine high and steep slope point-surface monitoring data fusion method based on dynamic weight distribution to solve the problems in the background.

[0006] In order to achieve the above object, the present application provides the following technical scheme: a non-coal mine high and steep slope point and surface monitoring data fusion method based on dynamic weight distribution, comprising the following steps: Step one, based on the instantaneous change rate of point monitoring data and the overall trend of surface monitoring data, a time synchronization window is constructed, the sampling rhythm of point monitoring data and surface monitoring data is unified time baseline registration, so as to realize the time alignment of the two kinds of monitoring data in the reverse transient stage, and form a continuous deformation starting zone which can be tracked; Step two, using the displacement direction reversal information contained in the deformation starting zone, a direction symbol identification matrix is established, the displacement direction of point monitoring data and surface monitoring data is dynamically compared, the direction reversal section is identified and marked as a transition zone, so as to provide boundary input for subsequent gradient constraint calculation; Step three, according to the direction symbol identification result of the transition zone, a gradient suppression function is constructed, the high-frequency gradient oscillation in the dynamic weight model is smoothed and regulated, so that the dynamic weight realizes continuous transition in the direction reversal stage, prevents the fusion result from entering the invalid convergence interval, and maintains the stability of the dynamic weight evolution process; Step four, according to the dynamic weight sequence smoothed by the gradient suppression function, a time-varying weight mapping table is established, and the deformation starting zone is taken as the time anchor point for iterative updating, so that the dynamic weight model adaptively adjusts the contribution proportion of point monitoring data and surface monitoring data in the subsequent fusion period; Step five, based on the dynamic weight updated by the time-varying weight mapping table, the point monitoring data and the surface monitoring data are weighted and fused to generate stable and continuous three-dimensional deformation results, so as to eliminate the false stable state and restore the true deformation trend of the high and steep slope, thereby improving the reliability and timeliness of the slope warning result.

[0007] Preferably, the step of constructing a time synchronization window comprises: The obtained point monitoring data and surface monitoring data are respectively subjected to time indexing processing, the sampling time of point monitoring data is repositioned through time interpolation method with the collection time of surface monitoring data as the main time axis, and a unified time reference framework is established; A movable time synchronization window is established on the unified time baseline, the point monitoring data change rate and deformation direction in each window are compared with the overall deformation trend of surface monitoring data as the reference, the time boundary when the direction is reversed is recorded as the potential reverse transient marker; Based on the reverse transient marker, the time sequences of point monitoring data and surface monitoring data are unified time baseline registration, the time mapping relationship is established through continuous interpolation, and the time offset curve is continuously smoothed; According to the continuity checking result of the time mapping curve, a time section which is continuous in time and obviously changes in direction is extracted to determine a deformation starting zone and to perform time smoothing expansion to form a continuous and traceable deformation starting zone.

[0008] Preferably, the time sequence synchronization window adopts a fixed time span and a fixed step length; a reverse transient mark is set only when the direction reversal lasts for more than a preset minimum time span within the window; the time mapping relationship is a monotonous and continuous curve, and the deformation starting zone is time-smoothly expanded after the checking.

[0009] Preferably, the step of establishing the direction symbol recognition matrix comprises: On the basis of the deformation starting zone determination, direction information is extracted from the point monitoring data and the surface area monitoring data, and is marked as a direction state of outward expansion or inward contraction, and a direction state sequence is formed by recording the direction change nodes; On the basis of the direction state sequence, the direction states of the point monitoring data and the surface area monitoring data are compared in time, and a direction symbol recognition matrix is established; when the directions are consistent, it is recorded as a consistent state, and when the directions are opposite, it is recorded as a reverse state; According to the reverse state distribution of the direction symbol recognition matrix, a direction reversal section is identified, and a real direction reversal section is determined through the direction consistency verification of continuous time sections; The time range of the identified direction reversal section is defined as a transition zone, and the transition zone is uniformly marked by time smoothing and spatial aggregation to serve as a boundary input for subsequent gradient constraint calculation.

[0010] Preferably, in the uniform marking process of the transition zone, the time boundaries of adjacent direction reversal sections are smoothly connected, and the direction consistency of point monitoring data at different spatial positions is verified; when the direction change trend is continuous and the spatial distribution is consistent, the adjacent sections are merged into a single transition zone to ensure the time continuity and spatial consistency of the transition zone.

[0011] Preferably, the step of constructing the gradient suppression function comprises: After obtaining the time range and direction attribute of the transition zone, the dynamic weight change characteristics in the transition zone are analyzed, and a time section with a rapid weight change rate and frequent direction jumps is extracted as a high-frequency gradient shock zone; According to the constraint condition of the gradient suppression function constructed according to the direction symbol recognition result in the transition zone, the direction reversal nodes and the dynamic weight change curve are matched, and a time buffer zone is set on both sides of the reversal node to continuously constrain the weight change rate; In the high-frequency gradient shock zone, the weight change curve is smoothly regulated according to the constraint condition, so that the weight remains continuous in the reversal stage and gradually returns to the normal change trend; The continuity check and stability verification are performed on the smoothed weight curve, the stable weight curve is associated with the deformation starting zone and the direction symbol recognition result, and a time-continuous weight evolution record is formed.

[0012] Preferably, the time range of the time buffer zone set on both sides of the inversion node is dynamically determined according to the start time and end time of the transition zone, so that the gradient suppression function uniformly acts on the weight change curve before and after the direction inversion, thereby ensuring the smooth transition of the weight in the inversion stage and preventing the weight change from lagging or advancing.

[0013] Preferably, the step of establishing a time-varying weight mapping table and iteratively updating it with the deformation starting zone as a time anchor point comprises: After obtaining the dynamic weight sequence smoothed by the gradient suppression function, the weight data is time-structured, the weight values are reordered in time sequence, and a unified time reference is established with the start time of the deformation starting zone as the reference point; Based on the structured time weight sequence, a time-varying weight mapping table is constructed, the start weight value, end weight value and change trend of each time period are stored as a weight mapping unit, and the weight boundary values of adjacent time periods are smoothly connected; The time-varying weight mapping table is iteratively updated with the deformation starting zone as a time anchor point, the change direction and rate of the weight mapping unit are adjusted by calculating the time offset to match the new deformation stage; The updated time-varying weight mapping table is checked for stability and verified for adaptability, the weight change trend is confirmed to be continuous and consistent with the deformation trend, and the verified mapping table is stored as the weight reference of the current stage.

[0014] Preferably, during the iterative updating process of the time-varying weight mapping table, when the time anchor point of the deformation starting zone is offset, the start weight value and end weight value of the corresponding weight mapping unit are adjusted synchronously according to the calculated time offset, and the updated weight boundary values are smoothly connected to ensure that the weight change trend remains continuous and stable in the time dimension.

[0015] Preferably, the step of weighted fusion based on the dynamic weight updated by the time-varying weight mapping table comprises: After the time-varying weight mapping table is updated, the point monitoring data and the surface monitoring data are matched and time-aligned, the consistency of the two types of monitoring data in time and space is ensured through geographic coordinate matching and time synchronization registration; The contribution proportion of the two types of monitoring data at each time node is determined according to the dynamic weight updated by the time-varying weight mapping table, the weight of the surface monitoring data is increased in the stable stage of the deformation process, and the weight of the point monitoring data is increased in the direction inversion stage, so as to achieve dynamic balance; The point monitoring data and the surface area monitoring data are weighted and fused according to the determined weight proportion, a continuous three-dimensional deformation result is generated in the surface area monitoring space grid, and a local abnormal value area is processed for smooth transition; The generated three-dimensional deformation result is subjected to continuity inspection and authenticity verification, and through time series smoothing evaluation and comparison with actual measurement, it is confirmed that the deformation result is continuous in time, coordinated in space and truly reflects the slope deformation trend.

[0016] In the above technical solution, the technical effects and advantages provided by the present application are as follows: The present application introduces a time sequence synchronization window and a direction symbol recognition matrix in the monitoring data fusion process, so that the point monitoring data and the surface area monitoring data are accurately aligned in the time dimension, and the continuous association between the data is maintained when the deformation direction is reversed, thereby effectively eliminating the time mismatch problem caused by the sampling period difference and the response inertia. In this way, the fusion result can fully reflect the continuous evolution process of the slope from stability to mutation, significantly improving the accuracy and timeliness of deformation identification, and providing a stable time sequence basis for dynamic monitoring of high and steep slopes.

[0017] The present application constructs a gradient inhibition function and a time-varying weight mapping table, so that the dynamic weight realizes smooth transition in the direction reversal stage, prevents the weight evolution from appearing shock and invalid convergence phenomenon, and ensures that the fusion result remains stable and continuous under complex working conditions. The three-dimensional deformation result after fusion can truly reflect the spatial deformation characteristics and trend change of the slope, eliminate the interference of false stable state on risk identification, and make the monitoring data reveal potential instability signs in time, thereby significantly improving the reliability and response speed of slope early warning. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments or prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0019] Figure 1 The method flowchart of the non-coal mine high and steep slope point-surface monitoring data fusion method based on dynamic weight distribution of the present application. DETAILED DESCRIPTION

[0020] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive aspects to those skilled in the art. Like reference numerals may refer to like elements throughout.

[0021] The present application provides a non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution as shown in Figure 1 The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution as shown in the embodiment of the present application comprises the following steps: Step one, based on the instantaneous change rate of point monitoring data and the overall trend of plane monitoring data, a time sequence synchronization window is constructed, the sampling rhythm of point monitoring data and plane monitoring data is unified time baseline registration, so as to realize the time alignment of the two types of monitoring data in the reverse transient stage, and form a continuous and traceable deformation starting zone. The specific implementation of this step is: After obtaining the original data from the slope monitoring area, the point monitoring data and the plane monitoring data are respectively subjected to time indexing processing. The point monitoring data is usually obtained from high-frequency observation equipment such as global navigation satellite system monitoring points, crack meters and inclinometers arranged at key positions of the slope, and the data time interval is short, usually in minutes or hours. The plane monitoring data is mainly obtained by unmanned aerial vehicle oblique photography, laser radar scanning or synthetic aperture radar interferometric measurement, and the data update period is relatively long, generally between several hours and several days. In order to compare the two types of data under the same time reference, the original time stamp information of the two types of data is first extracted, and a unified time reference framework is established. Taking the collection time of the plane monitoring data as the main time axis, the sampling time of all point monitoring data is repositioned by time interpolation method, so that the data under different monitoring frequencies can be identified in the same time sequence. Through this process, all monitoring data obtain unified time label, eliminating the time offset problem caused by the difference in sampling frequency of different monitoring equipment. In order to further improve the accuracy of time alignment, the time label of each time node is finely corrected at the second or minute level, and the corrected time index table is recorded, so that each data can be correctly referenced under the unified time baseline in the subsequent processing.

[0022] After the time indexing and baseline unification are completed, the timing rhythm matching stage is entered. The stage takes the overall deformation trend of the area domain monitoring data as the rhythm reference, and adjusts the short-term change characteristics of the point monitoring data to match the time rhythm. Specifically, a movable timing synchronization window is established on the time axis of the unified time baseline. The window has a fixed time span, for example, set to several hours or several days, and slides along the time axis with a fixed step, and each sliding forms a data alignment analysis area in a time period. In each sliding window, the displacement change amount, change direction and change rate of the point monitoring data in the corresponding time period are extracted, and the overall deformation trend curve of the area domain monitoring data in the same time period is extracted. By comparing the direction changes of the two types of data in the same time window, it can be identified whether there is consistency in the deformation trend of the two types of data. When the point monitoring data in the window shows a direction reversal from the tension state to the compression state, while the area domain monitoring data still maintains the previous trend, the time boundary of the window is recorded as a potential reverse transient marker. As the window continues to slide along the time axis, these markers gradually form a set of time period set, indicating the time period when the point and area domain monitoring data have differences in deformation direction, providing initial synchronization anchor points for subsequent time registration.

[0023] After obtaining the preliminary synchronization markers, the two types of data are registered on the unified time baseline. The registration process is based on the identified time markers to align the time series of point monitoring data and the time series of area domain monitoring data at these key time nodes. Specifically, at each synchronization marker, the sampling time of the area domain monitoring data is taken as the reference point, and the time series of the point monitoring data is adjusted so that the deformation states of the two types of data in the same time period are consistent. For the data section between adjacent time markers, a time mapping relationship is established by continuous interpolation to shift the time nodes of the area domain monitoring data with large time differences to positions corresponding to the observation intervals of the point monitoring data. In order to ensure the continuity and accuracy of the registration, the change trend of the time offset is calculated in each time period, and the offset curve is continuously smoothed to make the time mapping curve monotonous and continuous globally. The mapping curve reflects the correspondence between the point monitoring data and the area domain monitoring data in the entire time range, thereby realizing the point-by-point correspondence of the two types of data in the time dimension. When the slope deformation occurs in the reverse direction, the time mapping curve can automatically track the corresponding nodes of the two types of data, ensuring that the subsequent deformation analysis and direction identification are carried out on the unified time baseline.

[0024] After the time registration is completed, the registration result is checked for continuity, and the time period forming the deformation starting zone is extracted. The specific implementation is to retrieve all time periods that are continuous in time and have obvious direction changes with reference to the generated time mapping curve. These time periods represent the time zone where the slope changes from a stable stage to a sudden stage, and are key time segments where the deformation trend changes significantly. In this process, first, the time nodes of all continuous time periods are verified for continuity to ensure that there are no time faults or repeated time nodes. Subsequently, the displacement directions of the point monitoring data and the surface monitoring data in the time period are compared and analyzed to identify the starting time and ending time of the displacement direction reversal, and define this time interval as the deformation starting zone. In order to enhance the time sequence stability of the deformation starting zone, the identified time interval is time-smoothed and expanded to cover the entire process from the initial disturbance to the direction reversal of the slope deformation. Finally, the deformation starting zone and its time range are recorded, and a deformation starting zone database is established, which serves as the basis for subsequent direction symbol recognition matrix and dynamic weight smoothing control.

[0025] Through the above steps, the point monitoring data and the surface monitoring data are accurately registered from asynchronous sampling to unified baseline in the time dimension, ensuring that in the reverse transient stage of the transition from slow stretching to rapid compression of slope deformation, both types of data can reflect the time evolution process of the same geological event. The construction of this time sequence synchronization window not only ensures the correspondence of the monitoring data in the time dimension, but also effectively identifies the key time period of slope deformation in the formation of the deformation starting zone.

[0026] Step two, using the displacement direction reversal information contained in the deformation starting zone, a direction symbol recognition matrix is established to dynamically compare the displacement directions of the point monitoring data and the surface monitoring data, identify the direction reversal section and mark it as a transition zone, to provide boundary input for subsequent gradient constraint calculation; The specific implementation of this step is: On the basis of the deformation starting zone determination, the direction information of the point monitoring data and the surface monitoring data is extracted. The point monitoring data usually reflects the displacement change of the local key position, and the numerical change in the deformation starting zone often shows the direction mutation characteristics; while the surface monitoring data represents the deformation trend of the overall slope, and its direction change is relatively flat but covers a wider range. In order to make full use of the differences and complementarities of the two types of data, first, from the starting time to the ending time of the deformation starting zone, the displacement increase and decrease along the main sliding direction in the point monitoring data is extracted hour by hour, and is respectively identified as the direction state of outward expansion or inward contraction. At the same time, the overall surface deformation direction is extracted from the surface monitoring data in the same time range, and by calculating the displacement difference of adjacent time periods, it is identified as the direction state of overall stretching or overall compression. In order to ensure the reliability of the direction extraction result, the direction state of each time period is checked for continuity, and when the direction state of a time period and the direction state of the previous time period appear to be mutated, the mutation time is recorded as a direction change node, and is marked in the time index. Through this way of hour-by-hour extraction and direction identification, a direction state sequence covering the entire deformation starting zone is formed, providing basic data for subsequent comparison and identification.

[0027] After the direction state sequence is formed, the direction states of the point monitoring data and the surface monitoring data are dynamically compared to establish a direction symbol identification matrix. The construction of the matrix takes the time index of the deformation starting zone as the core, and compares the point monitoring direction state and the surface monitoring direction state at the same time node one by one. The specific implementation way is: scanning point by point along the time axis, when the point monitoring direction and the surface monitoring direction are consistent, recording the direction symbol of the time node as consistent state; when the directions are opposite, recording as reverse state, and marking the difference symbol in the corresponding position of the matrix. Through this process, a direction symbol identification matrix is formed, with time as the horizontal axis and the direction state comparison result as the vertical axis. The matrix not only records the direction consistency of the two types of monitoring data in each time period, but also reflects the relative delay relationship of the direction changes of the two types of data. In order to ensure the accuracy of the identification result, the short-term reverse phenomenon between the time nodes is continuously judged during the comparison process, and only when the direction reversal lasts more than the set minimum time span, it is confirmed as the real direction reversal state, so as to avoid the direction misjudgment caused by instantaneous noise or sampling anomaly.

[0028] After the direction symbol recognition matrix is constructed, the direction reversal section is recognized by using the reverse state distribution in the matrix. The recognition process takes time continuity as the criterion, and the time region in which the direction symbol continuously reverses in adjacent time sections is regarded as a potential direction reversal section. Specifically, starting from the time starting point of the deformation starting zone, the time axis is searched backward step by step, and when it is detected that the direction state of a plurality of continuous time nodes is in the reverse marking state, the time interval is determined as a candidate zone of a direction reversal section. Subsequently, the stability of the direction state at both ends of the candidate section is analyzed, and when the direction reversal at the starting end of the section changes from single reversal to continuous reversal and the direction state at the end of the section is consistent with the direction of the surface area monitoring data again, it is confirmed that the interval is a real direction reversal section. In order to enhance the spatial consistency of the section recognition, after the recognition is completed, the direction consistency of the point monitoring data at different spatial positions in the same time section is verified to ensure that the recognized direction reversal section is not only continuous in time but also shows the same direction change trend in spatial distribution. Through this process, the direction reversal section is accurately extracted from the complex deformation time sequence, laying a foundation for the subsequent transition analysis of the slope deformation.

[0029] After the direction reversal section is recognized, the transition zone is marked to provide boundary input for the subsequent gradient constraint calculation. Specifically, the time range of each recognized direction reversal section is defined as a transition zone, and the starting time and ending time are marked on the time axis. For the case that there is a slight direction difference at different spatial positions in the same time section, adjacent direction reversal sections are merged into a unified transition zone through time smoothing and spatial aggregation to ensure time continuity and spatial consistency. In the marking process, the transition zone is associated with the deformation starting zone, so that each transition zone corresponds to a specific deformation starting stage, thereby forming an ordered time partition structure. Each transition zone records its direction change characteristics, duration and direction difference information with the surrounding time section. After the marking is completed, the time range and direction attribute of the transition zone are included in the input data of the subsequent gradient constraint calculation, which is used to determine the boundary regulation range of the dynamic weight model in the direction reversal stage.

[0030] Through the above steps, the direction reversal information contained in the deformation starting zone is fully mined and converted into a structured direction symbol recognition matrix, and then the direction reversal section is recognized and marked as a transition zone, so that the two types of monitoring data have clear boundary characteristics in the direction change stage. The embodiment not only ensures the time sequence continuity and spatial consistency of direction recognition, but also provides clear boundary input for dynamic weight smoothing regulation, so that the subsequent weight allocation process can realize smooth transition in the direction mutation stage, thereby effectively avoiding weight shock and result distortion in the monitoring data fusion process.

[0031] Step three, according to the direction symbol recognition result of the transition zone, a gradient inhibition function is constructed to smooth and regulate the high-frequency gradient oscillation in the dynamic weight model, so that the dynamic weight realizes continuous transition in the direction reversal stage, prevents the fusion result from entering the invalid convergence interval, and maintains the stability of the dynamic weight evolution process; The specific implementation of this step is: After obtaining the time range and direction attribute of the transition zone, the difference analysis of the dynamic weight change characteristics in the transition zone is performed to determine the time interval that needs to be subjected to gradient inhibition. Specifically, the point monitoring data and the time series weight change of the surface monitoring data in the transition zone are compared, and the time period with rapid weight change rate and frequent direction jump in the direction reversal stage is extracted as the high-risk oscillation zone. In this process, the starting time and the ending time of the transition zone are taken as boundaries to divide the time evolution of the dynamic weight into three continuous sections, i.e., the pre-reversal section, the reversal section, and the post-reversal section. By comparing the weight change amplitudes in these three sections, the section with abnormal fluctuation in the weight change curve can be clearly identified. For the part with weight increase or decrease amplitude exceeding the set threshold or the change direction appearing multiple times in a short time, it is determined as the high-frequency gradient oscillation zone. This determination result is used to determine the action range of the gradient inhibition function, ensuring that the subsequent smoothing and regulation process can specifically inhibit the high-frequency fluctuation without affecting the normal weight evolution trend.

[0032] After determining the high-frequency gradient oscillation zone, the constraint conditions of the gradient inhibition function are constructed according to the direction symbol recognition result in the transition zone. This process takes the time interval of direction reversal as the core, extracts the reversal markers in the corresponding time period of the direction symbol recognition matrix, corresponds each reversal node to the dynamic weight change curve, and forms the time matching relationship between direction change and weight change. In order to ensure the synchronization of the gradient inhibition process and the direction change, the time range on both sides of each direction reversal node is taken as a buffer zone, so that the gradient inhibition effect can be evenly distributed before and after the direction change, avoiding the situation of inhibition lag or advance. Subsequently, in each buffer zone, the slope of the weight change curve is continuously constrained to ensure that the weight change rate gradually slows down before the direction reversal, remains stable during the reversal, and gradually recovers and increases after the reversal. In this way, the constraint conditions of the gradient inhibition function are not only based on the weight change characteristics, but also combined with the dynamic evolution of the direction symbol, so that the weight change process is consistent with the deformation direction turning, thereby realizing the natural and smooth weight transition.

[0033] After the establishment of the gradient inhibition constraint condition is completed, the weight change curve in the high-frequency gradient oscillation zone is smoothed and regulated. The regulation process is based on a determined time buffer to continuously modify the weight change curve, so that it maintains a smooth time evolution trend in the transition zone. Specifically, starting from the pre-reversal section, the weight change rate is gradually reduced, so that the weight curve does not suddenly increase or decrease near the direction reversal node; after entering the reversal middle section, the weight change amplitude is controlled within a stable range, so that the weight curve presents a smooth transition state in this interval; when entering the post-reversal section, the normal trend of the weight change is gradually restored, so that it can smoothly transition to the new equilibrium interval. In order to avoid the weight change lag caused by continuous smoothing, the direction change of the transition zone boundary is continuously tracked during the entire regulation process, and when the direction reversal ends and the weight change tends to be stable, the gradient inhibition effect is gradually weakened, so that the weight evolution returns to the natural change state. Through this dynamic smoothing and regulation, the dynamic weight can maintain continuity and adapt to the change of the deformation direction during the direction reversal phase, realizing the stable weight transition in the complex deformation phase.

[0034] After the smoothing and regulation is completed, the continuity of the weight change trend at each time node in the transition zone is detected to ensure that the weight curve does not have new mutation points or unreasonable fluctuations. Then, the overall change trend of the weight curve before and after the inhibition is compared and analyzed to evaluate the influence of the smoothing and regulation on the weight evolution. When it is found that the fluctuation amplitude of the weight curve after the inhibition is significantly reduced, the change direction is consistent with the deformation trend, and there is no convergence to abnormal value during the reversal phase, it is confirmed that the gradient inhibition process is effective. In order to further improve the adaptive ability of the dynamic weight, the stable weight curve is associated with the deformation starting zone and the direction symbol recognition result of the previous period to establish a time-continuous weight evolution record. In this way, the smoothed weight change trend can be dynamically updated in the subsequent monitoring period, so as to continuously maintain the stability of the weight evolution and prevent the high-frequency gradient oscillation caused by new deformation disturbance.

[0035] Through the above implementation steps, the present application realizes accurate control and continuous smoothing of the dynamic weight change process during the direction reversal phase. The construction of the gradient inhibition function is not only based on the direction symbol recognition result and the determination of the transition zone time interval, but also combines the time sequence characteristics of the weight change, so that the weight evolution process can naturally connect different deformation stages, thereby effectively avoiding the sharp fluctuation or convergence of the weight curve during the direction reversal period. The implementation of the present application inhibits the high-frequency gradient oscillation, so that the dynamic weight can maintain stable transition in the complex deformation environment, significantly improving the continuity and reliability of the fusion result of the point monitoring data and the surface monitoring data.

[0036] Step four, according to the dynamic weight sequence smoothed by the gradient inhibition function, a time-varying weight mapping table is established, and the deformation starting band is taken as the time anchor point for iterative updating, so that the dynamic weight model can adaptively adjust the contribution proportion of point monitoring data and area monitoring data in the subsequent fusion period; The specific implementation of this step is: After obtaining the dynamic weight sequence smoothed by the gradient inhibition function, the weight data is time-structured to form a time-continuous sequence that can be used for subsequent mapping. Specifically, the smoothed dynamic weight values are reordered in time sequence, and the weight values of each time node are recalibrated in combination with the time range of the deformation starting band formed in the previous period. Since the weight sequence shows a more smooth and continuous evolution trend after gradient inhibition, by matching the weight values of each time node with the time label of the deformation starting band, the continuous change relationship of dynamic weight in the deformation starting stage, transition stage and stable stage can be clearly depicted. In this process, the starting time of the deformation starting band is taken as the reference point of the weight sequence, and the weight change of all subsequent time nodes is recorded relative to this reference point, so that the dynamic weight sequence forms a relatively unified coordinate baseline in the time dimension. Through this time-anchored weight structuring method, an ordered time reference is provided for establishing the time-varying weight mapping table.

[0037] After obtaining the structured time weight sequence, a time-varying weight mapping table is constructed. The mapping table takes time as the main axis and weight change trend as the vertical parameter, and is used to describe the contribution proportion relationship of point monitoring data and area monitoring data in different time periods. In the specific implementation process, the structured weight sequence is segmented according to time, and each time period represents a stable stage or transition stage of dynamic weight. For each time period, the starting weight value, ending weight value and change trend of the intermediate time period are extracted, and these data are stored as a weight mapping unit. Each weight mapping unit not only records the weight change in the time interval, but also includes the corresponding deformation state identifier, direction change characteristics and related monitoring data fluctuation degree. When all weight mapping units are connected in turn, a time-varying weight mapping table covering the entire monitoring period is formed. This mapping table can intuitively reflect the change law of dynamic weight in different deformation stages, so that the subsequent weight update and data fusion can be automatically adjusted according to the mapping relationship. At the same time, in order to ensure the continuity of the mapping table, the weight boundary values of adjacent time periods are smoothly connected, so that the entire table forms a continuous transition in the time dimension, avoiding the problem of weight jump caused by sudden change of time interval.

[0038] After the time-varying weight mapping table is constructed, the weight mapping table is iteratively updated with the deformation starting zone as the time anchor point. The core of this step is to enable the dynamic weight model to automatically adjust the weight mapping relationship according to the new time anchor point and deformation characteristics after each new monitoring data input, thereby realizing adaptive evolution. Specifically, at the beginning of each new monitoring period, first detect whether the deformation starting zone has changed, including the advance, delay or extension of its starting time. When the time range of the deformation starting zone is detected to be adjusted, compare the new time anchor point with the anchor point of the last period, calculate the time offset, and apply the offset to the corresponding time axis section in the current time-varying weight mapping table to realize anchor point synchronization. Subsequently, update the weight mapping unit after the time offset to adjust the weight trend direction and change rate to match the new deformation stage. During the iterative update process, the latest monitoring data trend also needs to be fused with the original mapping table. When the new point monitoring data and the surface monitoring data show obvious directional changes, update the weight distribution proportion of the corresponding time period. Through this iterative updating mechanism with the deformation starting zone as the time anchor point, the weight mapping table can reflect the dynamic changes of the slope deformation process in real time, ensuring that the dynamic weight model maintains continuous adaptive ability.

[0039] After the iterative update is completed, the updated time-varying weight mapping table is checked for stability and adaptability. First, check the global continuity of the weight change trend in the entire mapping table to confirm that there is no sudden jump or reverse fault between adjacent time periods. Subsequently, compare the weight evolution trajectories before and after updating to evaluate whether the weight adjustment can accurately reflect the actual contribution proportion of point monitoring data and surface monitoring data in different stages. When it is detected that the weight changes too quickly or the change direction does not match the deformation trend in a certain time period, rebalance the weight parameters in that time period to return to a stable state. To further verify the adaptability of the time-varying weight mapping table, match the updated mapping table with the actual deformation results in the latest monitoring period, and judge whether the weight update is effective by comparing the consistency of the deformation trend. If the weight change trend and the deformation trend are consistent and the monitoring fusion result is stable, it means that the iterative update of the time-varying weight mapping table achieves the expected effect and can be used to guide the data fusion process in the next period. Finally, store the verified time-varying weight mapping table as the weight reference of the current stage for subsequent fusion stages to call, thereby realizing the continuous optimization and evolution of the dynamic weight model in different monitoring periods.

[0040] Through the above steps, the dynamic weight sequence smoothed by the gradient inhibition function is further structured and converted into a time-varying weight mapping table, so that the dynamic weight distribution has a clear evolution track and updating mechanism in the time dimension. Through the iterative updating mode with the deformation starting zone as the time anchor point, the dynamic weight model can continuously and adaptively adjust the contribution proportion of the point monitoring data and the area monitoring data, which not only maintains the stability of the fusion result, but also improves the response sensitivity of the model to the change of the deformation state.

[0041] Step five, based on the dynamic weight updated by the time-varying weight mapping table, the point monitoring data and the area monitoring data are weighted and fused to generate stable and continuous three-dimensional deformation results, so as to eliminate the false stable state and restore the true deformation trend of the high and steep slope, thereby improving the reliability and timeliness of the slope warning result. The specific implementation of this step is: After updating the time-varying weight mapping table, the point monitoring data and the area monitoring data are matched and time-aligned to ensure that the two types of monitoring data participating in the weighted fusion are consistent in the time dimension and the spatial dimension. Specifically, according to the deformation starting zone and the time-varying weight mapping table established in the last stage, the observation values of the two types of monitoring data at the same time node are extracted. The point monitoring data is usually obtained from high-frequency displacement monitoring points at key positions of the slope, such as crack monitoring points, GPS monitoring points or inclinometer points; the area monitoring data includes the overall ground deformation field obtained by unmanned aerial oblique photography, laser radar scanning or interferometric radar imaging. On this basis, first, the spatial correspondence of the two types of monitoring data is established to ensure that each point monitoring data can find the corresponding spatial position or area in the area monitoring data. For this purpose, the geographic coordinates of the point monitoring position are matched with the area monitoring grid, and when there is a coordinate offset, the position is interpolated according to the local surface deformation gradient of the area data, so as to realize the spatial alignment of the point data. At the same time, the time dimension is adjusted synchronously, so that the time sampling difference produced by different monitoring methods is re-registered on the unified time axis, ensuring that at each time node, the two types of data can represent the deformation state of the slope in the same time period.

[0042] After completing data matching and time alignment, the contribution proportion of point monitoring data and surface monitoring data at each time node is determined according to the dynamic weight updated by the time-varying weight mapping table. Specifically, the weight value corresponding to the time period is extracted from the time-varying weight mapping table, and is respectively assigned to the point monitoring data and the surface monitoring data. Since the time-varying weight mapping table is constructed according to the deformation starting zone as the time anchor point and is updated by multiple rounds of iteration, the weight value therein not only reflects the deformation state characteristics in the current time period, but also considers the dynamic change trend of the previous and subsequent time periods. In the weight distribution process, when the monitoring time period is in the stable stage of the deformation direction, the surface monitoring data weight is relatively high to enhance the continuity of the overall deformation trend; when the monitoring time period is in the deformation direction reversal or local disturbance stage, the point monitoring data weight is relatively high to enhance the ability to capture local deformation characteristics. Through this weight distribution method based on the joint constraint of time and state, the fusion process can dynamically balance the contribution of the two types of data, thereby achieving a reasonable coordination between the overall trend and local details.

[0043] After completing the weight distribution, the point monitoring data and the surface monitoring data are weighted and fused according to the weight proportion, and a continuous three-dimensional deformation result is generated in the spatial range. In the specific implementation process, based on the spatial distribution of the surface monitoring data, the weighted point monitoring data is projected into the corresponding surface grid element, and through the mapping of time continuity, each grid element contains both local high-precision displacement information and global deformation trend information. At the same time node, the displacement values of all grid elements are fused to form a continuous two-dimensional deformation field; then, the two-dimensional deformation fields of each time node are stacked in time sequence to build a complete three-dimensional deformation time sequence result. In this process, the deformation starting zone and transition zone information established in the previous stage are used to limit the boundary of deformation evolution to ensure that the spatial distribution of the deformation result is continuous and has no abrupt change in the direction reversal stage. At the same time, in order to avoid the interference of local monitoring point outliers on the fusion result, the area with large local deformation value change amplitude is smoothed to make it consistent with the deformation trend of the surrounding grid. Through this weighted fusion method, the generated three-dimensional deformation result not only retains the high-precision characteristics of point monitoring, but also inherits the overall coverage advantage of surface monitoring, thereby forming a uniform spatial and continuous temporal shape recognition result.

[0044] After generating the three-dimensional deformation result, continuity test and authenticity verification are performed on the fusion result to ensure that the obtained deformation result can accurately reflect the true deformation trend of the slope and effectively eliminate the false stable state. The continuity test mainly evaluates the smoothness of the deformation difference value of the continuous time nodes through time series analysis to confirm that there is no mutation or abnormal fault in the time curve of the deformation change in the entire monitoring period. The authenticity verification takes the actual monitoring records or independent observation means as a reference to judge the reliability of the fusion result by comparing the deformation trend of the key positions in the three-dimensional deformation result with the measured data. When detecting unreasonable deformation stagnation or reverse change in the fusion result in a certain time period, it is checked whether there is an abnormality in the dynamic weight distribution in the time period, and the time-varying weight mapping table is adjusted and corrected until the fusion result is continuous in time and coordinated in space. After the inspection and correction process, the final three-dimensional deformation result can truly reflect the whole process of the slope from stability to disturbance to accelerated deformation, and successfully eliminates the false stable state caused by the difference in data sampling period or weight shock.

[0045] Through the above steps, the dynamic weight updated through the time-varying weight mapping table is effectively applied to the weighted fusion of point monitoring data and surface monitoring data, so that the fused three-dimensional deformation result is continuous in time, complete in space and stable in trend. The embodiment not only realizes the deep cooperation and dynamic balance of the two types of monitoring data, but also maintains the stability and authenticity of the fusion result in the direction reversal and disturbance stage, thereby significantly improving the timeliness and reliability of slope deformation identification and early warning.

[0046] The present application introduces time sequence synchronization window and direction symbol recognition matrix in the monitoring data fusion process, so that the point monitoring data and surface monitoring data are accurately aligned in time dimension, and the continuous association between the data is maintained when the deformation direction is reversed, thereby effectively eliminating the time mismatch problem caused by the difference in sampling period and response inertia. In this way, the fusion result can fully reflect the continuous evolution process of the slope from stability to mutation, significantly improving the accuracy and timeliness of deformation identification and providing a stable time sequence basis for dynamic monitoring of high and steep slopes.

[0047] The present application constructs gradient inhibition function and time-varying weight mapping table to make the dynamic weight smoothly transition in the direction reversal stage, prevent the weight evolution from appearing shock and invalid convergence phenomenon, and ensure that the fusion result remains stable and continuous under complex working conditions. The fused three-dimensional deformation result can truly reflect the spatial deformation characteristics and trend change of the slope, eliminate the interference of false stable state on risk identification, and make the monitoring data reveal potential instability signs in time, thereby significantly improving the reliability and response speed of slope early warning.

[0048] The foregoing merely illustrates some exemplary embodiments of the application, and it will be appreciated that those skilled in the art will be able to devise various modifications without departing from the spirit and scope of the application. The appended drawings and description are illustrative only, and are not intended to be limiting.

Claims

1. A non-coal mine high and steep slope point-surface monitoring data fusion method based on dynamic weight distribution, characterized in that, The method comprises the following steps: Step one, based on the instantaneous change rate of point monitoring data and the overall trend of surface monitoring data, a time synchronization window is constructed, the sampling rhythm of point monitoring data and surface monitoring data is unified time baseline registration, and a continuous traceable deformation starting zone is formed; Step two, using the displacement direction reversal information contained in the deformation starting zone, a direction symbol identification matrix is established, the displacement direction of point monitoring data and surface monitoring data is dynamically compared, the direction reversal section is identified and marked as a transition zone; Step three, according to the direction symbol identification result of the transition zone, a gradient suppression function is constructed, the high-frequency gradient oscillation in the dynamic weight model is smoothed and regulated, and the dynamic weight realizes continuous transition in the direction reversal stage; Step four, according to the dynamic weight sequence smoothed by the gradient suppression function, a time-varying weight mapping table is established, and the deformation starting zone is iteratively updated as a time anchor point, so that the dynamic weight model adaptively adjusts the contribution proportion of point monitoring data and surface monitoring data in the subsequent fusion period; Step five, based on the dynamic weight updated by the time-varying weight mapping table, the point monitoring data and the surface monitoring data are weighted and fused to generate stable and continuous three-dimensional deformation results, and the false stable state is eliminated and the true deformation trend of the high and steep slope is restored.

2. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 1, characterized in that, The step of constructing a time synchronization window comprises: The obtained point monitoring data and surface monitoring data are respectively subjected to time indexing treatment, the sampling time of the point monitoring data is repositioned by time interpolation method with the acquisition time of the surface monitoring data as the main time axis, and a unified time reference framework is established; A movable time synchronization window is established on the unified time baseline, the point monitoring data change rate and deformation direction in each window are compared with the overall deformation trend of the surface monitoring data as a reference, the time boundary when the direction is reversed is recorded as a potential reverse transient marker; Based on the reverse transient marker, the time sequences of the point monitoring data and the surface monitoring data are unified time baseline registration, the time mapping relationship is established by continuous interpolation, and the time offset curve is continuously smoothed; According to the continuity checking result of the time mapping curve, the time section which is continuous in time and changes in direction is extracted, and is determined as the deformation starting zone and is subjected to time smoothing expansion, so as to form a continuous traceable deformation starting zone.

3. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 2, characterized in that, The time synchronization window adopts fixed time span and fixed step length; only when the direction reversal lasts more than a preset minimum time span in the window, the reverse transient marker is set.

4. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 3, characterized in that, The step of establishing a direction symbol identification matrix comprises: Based on the determination of the deformation starting zone, the direction information of the point monitoring data and the surface monitoring data is extracted, and the direction state is identified as outward expansion or inward contraction, and the direction change node is recorded to form a direction state sequence; Based on the direction state sequence, the direction states of the point monitoring data and the surface monitoring data are compared in time, a direction symbol identification matrix is established, and when the directions are consistent, it is recorded as a consistent state, and when the directions are opposite, it is recorded as a reverse state; According to the inverse state distribution of the direction symbol recognition matrix, a direction reversal section is identified, and a real direction reversal section is determined through continuous time period direction consistency verification; The identified direction reversal section time range is defined as a transition zone, and the transition zone is uniformly marked through time smoothing and spatial aggregation, serving as boundary input for subsequent gradient constraint calculation.

5. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 4, characterized in that, In the uniform marking process of the transition zone, the time boundaries of adjacent direction reversal sections are smoothly connected, and the direction consistency of point position monitoring data at different spatial positions is verified. When the direction change trend is continuous and the spatial distribution is consistent, the adjacent sections are merged into a single transition zone.

6. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 4, characterized in that, The steps of constructing the gradient inhibition function include: After obtaining the time range and direction attribute of the transition zone, the dynamic weight change characteristics in the transition zone are analyzed, and the time period with fast weight change rate and frequent direction jump is extracted as a high-frequency gradient shock zone; According to the direction symbol recognition result in the transition zone, the constraint condition of the gradient inhibition function is constructed, the direction reversal node is matched with the dynamic weight change curve, and a time buffer zone is set on both sides of the reversal node to constrain the continuity of the weight change rate; In the high-frequency gradient shock zone, the weight change curve is smoothed and regulated according to the constraint condition, so that the weight remains continuous and gradually recovers to the normal change trend in the reversal stage; The continuously smoothed and regulated weight curve is verified for continuity and stability, the stable weight curve is associated with the deformation starting zone and the direction symbol recognition result, and a time-continuous weight evolution record is formed.

7. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 6, characterized in that, The time range of the time buffer zone set on both sides of the reversal node is dynamically determined according to the start time and end time of the transition zone, so that the gradient inhibition function uniformly acts on the weight change curve before and after the direction reversal, ensuring the smooth transition of the weight in the reversal stage.

8. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 6, characterized in that, The steps of establishing a time-varying weight mapping table and iteratively updating it with the deformation starting zone as the time anchor point include: After obtaining the dynamic weight sequence smoothed by the gradient inhibition function, the weight data is time-structured, the weight values are reordered in time sequence, and a unified time reference is established with the start time of the deformation starting zone as the reference point; Based on the structured time weight sequence, a time-varying weight mapping table is constructed, the start weight value, end weight value and change trend of each time period are stored as a weight mapping unit, and the weight boundary values of adjacent time periods are smoothly connected; The time-varying weight mapping table is iteratively updated with the deformation starting zone as the time anchor point, the change direction and rate of the weight mapping unit are adjusted by calculating the time offset, so as to match the new deformation stage; The updated time-varying weight mapping table is checked for stability and verified for adaptability, the weight change trend is confirmed to be continuous and consistent with the deformation trend, and the verified mapping table is stored as the weight reference of the current stage.

9. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 8, characterized in that, In the iterative updating process of the time-varying weight mapping table, when the time anchor point of the deformation starting zone is offset, the start weight value and end weight value of the corresponding weight mapping unit are adjusted synchronously according to the calculated time offset, and the updated weight boundary values are smoothly connected.

10. The non-coal mine high and steep slope point and plane monitoring data fusion method based on dynamic weight distribution according to claim 8, characterized in that, The step of weighted fusion based on the dynamic weight updated by the time-varying weight mapping table comprises: After the time-varying weight mapping table is updated, the point monitoring data and the surface area monitoring data are matched and time-aligned, and through the geographic coordinate matching and time synchronization registration, the consistency of the two types of monitoring data in time and space is ensured; According to the dynamic weight updated by the time-varying weight mapping table, the contribution proportion of the two types of monitoring data at each time node is determined, so that the surface area monitoring is allocated with high weight in the stable stage, and the weight of the surface area monitoring data is increased in the stable stage of the deformation process; the weight of the point monitoring data is increased in the direction reversal stage; According to the determined weight proportion, the point monitoring data and the surface area monitoring data are weighted and fused, the continuous three-dimensional deformation result is generated in the surface area monitoring space grid, and the local abnormal value area is smoothly transitioned; The generated three-dimensional deformation result is continuously tested and verified, and through the time series smoothing evaluation and the comparison with the actual measurement, it is confirmed that the deformation result is continuous in time, coordinated in space and truly reflects the trend of the slope deformation.

Images