Visual monitoring network and navigation system synchronous settlement early warning method for soft soil foundation structure

Abstract

The invention discloses a soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method, and relates to the technical field of geological monitoring and investigation. The soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method comprises the steps that S1, image data, time signal data and environment data in the soft soil foundation monitoring process are collected and preprocessed, and a standardized soft soil foundation state data set is constructed; s2, evaluating the spatial offset characteristic of the reference point, and dynamically adjusting the stability identifier and the subsequent replacement trigger condition of the reference point; s3, analyzing the image semantic consistency of the candidate reference points, and screening stable replacement points; s4, the coordinate consistency of the replacement reference points is evaluated, and the spatial positions of the reference points are dynamically corrected; and S5, performing difference comparison on the multi-base-station monitoring data after rollback correction, and repairing curve abrupt change. The problems that the visual reference is prone to drifting and an automatic repairing mechanism is lacked under the condition that the reference point in the soft soil foundation is unstable are solved.

Description

Technical Field

[0001] This invention relates to the field of geological monitoring and exploration technology, specifically to a method for synchronous settlement early warning of a visual monitoring network and navigation system for soft soil foundation structures. Background Technology

[0002] With the continuous improvement of the digitalization and intelligence of infrastructure, long-term stability monitoring of soft soil foundation structures has become a key technical aspect of the operation and maintenance of roads, bridges, rail transit, and underground engineering. Current mainstream settlement monitoring methods mostly rely on manual leveling, GNSS fixed-point observation, or periodic re-measurement with total stations. While these methods are applicable to small-scale, high-precision measurements, they suffer from sparse data sampling, insufficient temporal resolution, and delayed monitoring response in large-span, heterogeneous soft soil environments. Although some monitoring systems have introduced automatic data acquisition and wireless transmission mechanisms, they still primarily rely on single-point coordinates or fixed target reflection signals, lacking the ability to fuse multi-view, multi-node synchronous data, making it difficult to achieve continuous identification and spatial differential analysis of the overall structural settlement process.

[0003] For example, the invention patent with announcement number CN113610902B discloses a method for mapping and registering ground-based true aperture radar with point cloud data. The method includes using ground-based true aperture radar to continuously monitor slopes through a large-area point source scan; obtaining three-dimensional image information #imgabs1# and #imgabs2# from the scattered echo signals #imgabs0# of several quasi-static strong scattering targets within the slope monitoring area; and performing windowing filtering on the original scattered echo signals #imgabs3# in the frequency domain. The aim is to provide a method for accurately acquiring information about main scattering targets within the slope space and performing three-dimensional mapping and registration with topographic data from various applications such as mines, landslides, and dams. This facilitates the interpretation and analysis of deformation and displacement in the monitored area and avoids abnormal situations where the imaging target deviates significantly from the spatial position of the slope.

[0004] For example, the invention patent with publication number CN116503312B discloses a shield tunnel lining disease detection method based on ground-based LiDAR point cloud; based on ground-based LiDAR tunnel lining laser point cloud, the point cloud is pretreated to obtain three-dimensional laser point cloud; the three-dimensional laser point cloud is projected to obtain a two-dimensional laser point cloud image, and the two-dimensional laser point cloud image is grayed to obtain a laser point cloud gray image; based on the local pre-fitting energy driven active contour fast model segmentation of the laser point cloud gray image, the tunnel lining disease is detected and analyzed, which has the advantages of high accuracy, recall rate, small false alarm rate, area error and less calculation running time in disease detection, can quickly and accurately identify the disease of shield tunnel lining, and solves the problem that the traditional subway shield tunnel disease monitoring can only obtain limited discrete monitoring point changes and cannot comprehensively reflect the tunnel lining disease condition.

[0005] In a complex soft soil environment, the deformation of the foundation has significant nonlinear and time-lag characteristics, is affected by multiple factors such as rainfall, groundwater level fluctuation, traffic dynamic load and temperature and humidity change, and non-uniform settlement or sudden displacement anomaly often occurs in local areas. The traditional monitoring method relies on a single sensor or fixed sampling point, and cannot maintain measurement stability under the conditions of reference point drift, target identification loss or environmental light change; in addition, the data between the monitoring nodes often lack a unified time reference, and the image and coordinate information of different sites are difficult to accurately time-align, resulting in error accumulation in settlement trend analysis and abnormal early warning.

[0006] In view of the above problems, a soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method are urgently needed. SUMMARY

[0007] Technical problems solved In view of the deficiencies of the prior art, the soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method is provided, which solves the problems of easy drift of the visual reference under the condition of unstable reference point in the soft soil foundation and the lack of automatic repair mechanism.

[0008] Technical scheme To achieve the above object, the application is implemented by the following technical solutions: a soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method, comprising S1, collecting image data, time signal data and environment data in the soft soil foundation monitoring process, and preprocessing the collected image data, time signal data and environment data to construct a standardized soft soil foundation state data set; S2, based on the standardized soft soil foundation state data set, the stability of the spatial offset characteristics of the reference point is evaluated from the continuous multiple image sequence, and the stability identification of the reference point in the monitoring coordinate system and the subsequent replacement trigger condition are dynamically adjusted based on the stability evaluation result; S3, based on the standardized soft soil foundation state data set, the image semantic consistency of the candidate reference point is analyzed, and the stable replacement reference point is selected based on the semantic analysis result; S4, taking the stability evaluation result and the semantic analysis result as input, the coordinate consistency of the replacement reference point is comprehensively evaluated, and the spatial position of the reference point in the structure coordinate system is dynamically corrected based on the comprehensive evaluation result; S5, the multiple base station monitoring data after rollback correction is time synchronized, settlement trend is smoothed and cross base station difference is compared, and the curve mutation caused by reference point replacement is dynamically repaired.

[0009] Further, the specific steps of collecting image data, time signal data and environment data in the soft soil foundation monitoring process are: collecting image data in the soft soil foundation structure deformation monitoring process, the image data including: horizontal pixel coordinate value, vertical pixel coordinate value of the reference point in each frame of image, spatial offset of the reference point between adjacent frames, gray gradient of each pixel point in the image, gray co-occurrence matrix energy value, gray co-occurrence matrix contrast, and calculating and recording the average offset of the reference point in the sliding window range and the sliding window frame number length; collecting time signal data synchronized by the navigation positioning system, the time signal data including: positioning time interval and positioning time synchronization reference value of each frame; collecting environment data in the structure settlement response process, the environment data including: horizontal projection coordinate component, vertical projection coordinate component, depth ranging value, settlement displacement value and ground illumination intensity of each reference point in the structure coordinate system.

[0010] Further, the preprocessing of the collected image data, time signal data and environmental data to construct the standardized soft soil foundation state data set comprises the following steps: inter-frame smoothing processing is performed on the image data to eliminate single-frame jitter interference, and maximum and minimum normalization is performed on the gray gradient value, gray level co-occurrence matrix energy value and gray level co-occurrence matrix contrast; time reference unification is performed on the time signal data of the navigation positioning system, the positioning time interval is standardized according to the synchronization period, and the positioning time synchronization reference value is centralized to make the time sequence present zero mean value distribution in the unified reference frame; statistical calibration is performed on the environmental data, the depth ranging value, settlement displacement value and ground surface light intensity of each reference point are extracted, abnormal ranging points are checked and repaired by interpolation, and the average depth value, settlement displacement mean value, light intensity standard deviation and light intensity mean value of each reference point are calculated and recorded; the normalized image data, time signal data and environmental data are normalized to construct the standardized soft soil foundation state data set.

[0011] Further, the stability evaluation of the spatial offset characteristics of the reference point from the continuous multiple frame image sequence based on the standardized soft soil foundation state data set comprises the following steps: the sum of the square value of the horizontal pixel coordinate value difference and the square value of the vertical pixel coordinate value difference of the reference point in adjacent two frames of images is calculated, and the square root is taken to obtain a single-frame offset value; the single-frame offset values in all adjacent frame numbers before the current frame are added, and then multiplied by the reciprocal of the length of the sliding window frame number minus one to obtain a spatial offset stability value; the difference between the settlement displacement value of the reference point in the current frame image and the settlement displacement mean value is squared to obtain a single-frame settlement fluctuation value; the sum of the single-frame settlement fluctuation values of the reference point in the current frame number and all frame numbers before the current frame number is calculated, and then multiplied by the reciprocal of the length of the sliding window frame number to obtain a spatial smooth fluctuation value; the spatial offset stability value and the spatial smooth fluctuation value are added to obtain a reference point stability evaluation value.

[0012] Further, the dynamic adjustment of the stability identification and the subsequent replacement trigger condition of the reference point in the monitoring coordinate system based on the stability evaluation result comprises the following steps: the reference point stability evaluation value of the current reference point is compared with the stability evaluation threshold value in real time; when the reference point stability evaluation value is less than or equal to the stability evaluation threshold value, the effective state of the reference point in the monitoring coordinate is maintained, the navigation positioning system timestamp of the current reference point is written into the multi-base station data buffer area synchronously, and is recorded as a stable reference point, and the image sampling alignment instruction of the adjacent monitoring base station is triggered; when the reference point stability evaluation value is greater than the stability evaluation threshold value, an abnormal marking instruction is sent to the visual recognition unit, the data writing of the reference point is paused, the same area image of the adjacent frame in the redundant image buffer is called to perform semantic consistency replacement matching, and the coordinates of the replacement reference point are mapped to the navigation positioning system for re-registration and update.

[0013] Further, the semantic analysis of the image semantic consistency of the candidate reference point based on the standardized soft soil base state data set comprises the following specific steps: collecting the total number of pixels in the image in the neighborhood window of the substitute reference point, calculating the sum of the absolute values of the gray scale gradients of all pixels, dividing the sum by the total number of pixels, multiplying the ratio of the energy value of the gray scale co-occurrence matrix and the contrast of the gray scale co-occurrence matrix, and obtaining the local edge definition; adding one to the ratio of the standard deviation of the illumination intensity of the current substitute reference point to the average value of the illumination intensity, and obtaining the illumination balance correction value; calculating the structural similarity index of the current substitute reference point and the stable reference point by the structural similarity algorithm, dividing the structural similarity index by the standard deviation of the illumination intensity to obtain the structural similarity evaluation value; taking the natural logarithm of the absolute value of the difference between the depth ranging value of the current frame of the substitute reference point and the average depth value divided by the average depth value plus one, to obtain the depth direction continuity value; adding the product of the structural similarity evaluation value and the depth direction continuity value to the ratio of the local edge definition to the illumination balance correction value, to obtain the semantic consistency discrimination value of the current substitute reference point.

[0014] Further, the filtering of the stable substitute reference point based on the semantic analysis result comprises the following specific steps: real-time calculation of the semantic consistency discrimination value of the current substitute reference point and writing into the buffer, observation and recording of the semantic consistency discrimination value change trend of the substitute reference point: when the semantic consistency discrimination value of the current substitute reference point keeps rising trend in the adjacent observation period, the image sampling frequency is immediately increased, and the light flow residual correction algorithm is simultaneously called to perform local registration update; when the semantic consistency discrimination value of the current substitute reference point keeps unchanged in the adjacent observation period, the existing substitute reference point is maintained as the main reference input, and the navigation positioning system timestamp, light flow field vector and depth matching result are simultaneously written into the data cache; when the semantic consistency discrimination value of the current substitute reference point appears downward trend in the adjacent observation period, the current substitute reference point is suspended, the substitute reference point is reacquired, the reference point position is transferred and continuous observation is registered.

[0015] Further, the step of comprehensively evaluating the coordinate consistency of the substitute reference point based on the stability evaluation result and the semantic analysis result is: obtaining the horizontal pixel coordinate value and the vertical pixel coordinate value of the substitute reference point input as the main reference in the current image frame, calculating and recording the spatial displacement distance difference between the substitute reference point input as the main reference and the last valid reference point; obtaining a stable decay correction value by multiplying the spatial displacement distance difference by a reference point stable state evaluation value; obtaining a time synchronization correction value by multiplying the semantic consistency discrimination value by the positioning and timing time interval divided by the positioning time synchronization reference value; obtaining a horizontal projection correction value by calculating the difference between the horizontal projection coordinate component of the substitute reference point after updating in the same structure coordinate system and the horizontal projection coordinate component of the last valid reference point in the structure coordinate system; obtaining a vertical projection correction value by calculating the difference between the vertical projection coordinate component of the substitute reference point after updating in the same structure coordinate system and the vertical projection coordinate component of the last valid reference point in the structure coordinate system; obtaining a geometric projection correction value by taking the sum of the horizontal projection correction value and the vertical projection correction value as the numerator, and taking the square root of the sum of the square of the horizontal projection correction value and the square of the vertical projection correction value as the denominator; and obtaining a spatial rollback correction value by adding the stable decay correction value, the time synchronization correction value, and the geometric projection correction value.

[0016] Further, the step of dynamically correcting the spatial position of the reference point in the structure coordinate system based on the comprehensive evaluation result is: real-time comparison of the current spatial rollback correction value and the rollback correction threshold value, the rollback correction threshold value including a first correction threshold value and a second correction threshold value; when the spatial rollback correction value is less than or equal to the second correction threshold value, maintaining the existing coordinate state of the substitute reference point, keeping the original reference point identifier unchanged, and simultaneously recording the current navigation positioning system timestamp and the optical flow field vector as a continuous observation sample in the stable state interval; when the spatial rollback correction value is greater than the second correction threshold value and less than or equal to the first correction threshold value, calling the auxiliary projection in the adjacent base station cross field to perform linear offset correction, updating the projection coefficient matrix of the substitute reference point, and writing the intermediate correction identifier in the data buffer; when the spatial rollback correction value is greater than the first correction threshold value, freezing the current data writing channel, reconstructing the main projection parameter set of the current substitute reference point in the structure coordinate system, and redefining the time synchronization index of the current substitute reference point through the navigation positioning system.

[0017] Further, the time synchronization, settlement trend smoothing and cross-base station difference comparison of the back-off corrected multi-base station monitoring data, the specific steps for dynamically repairing the curve mutation caused by the reference point replacement are: reading the reference point coordinates after the spatial back-off correction, and calling the navigation positioning system to extract the time label of each base station, performing continuity analysis on the settlement displacement of the reference point of each base station, analyzing the settlement trend of the reference point in the current frame based on time series smoothing algorithm, and identifying the mutation point and the discontinuous section; after obtaining the settlement trend, a secondary correction process is performed, the steady-state evaluation value and the semantic consistency discrimination value of the reference point are superimposed on the time series curve, the local smoothing repair is performed on the instantaneous displacement sudden increase caused by the replacement of the reference point, and the difference comparison is performed on the image data between the base stations, the settlement trend graph and the abnormal identification list of the current monitoring period are generated, and the corrected settlement trend, the reference point identification number and the time synchronization index are written into the monitoring database; after completing the writing of the monitoring database, the settlement early warning data output process is entered, and the real-time settlement distribution graph, the early warning state code and the corresponding base station number are output.

[0018] Advantages The present application has the following advantages: (1) The soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method, by constructing a multi-base station based visual acquisition network, realizes multi-angle and continuous monitoring of the settlement process of the soft soil foundation structure, and significantly improves the time and space coverage accuracy of the monitoring data.

[0019] (2) The soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method, by introducing the Beidou timing synchronization mechanism, realizes data acquisition and settlement information alignment of each monitoring base station under the unified time reference, and ensures the time consistency and trend comparability of the cross-base station data.

[0020] (3) The soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method, by the steady-state evaluation and semantic consistency discrimination of the reference point, can automatically identify and replace the substitute point under the condition of reference point instability or shielding, and ensure the continuity and reliability of the monitoring process.

[0021] (4) The soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method, by combining the continuity detection logic of the multi-base station synchronous settlement curve, realizes the secondary correction and false alarm suppression of the sudden settlement abnormality, and enhances the stability and reliability of the early warning result.

[0022] Of course, implementing any product of the present application does not necessarily require achieving all the advantages described above at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The flow chart of the soft soil foundation structure visual monitoring network and navigation system synchronous settlement early warning method of the present application; Figure 2 A space back-off correction value column chart related to the present application; Figure 3 A multi-base station synchronous lifting and falling trend diagram related to the present application; Figure 4 A real-time settlement distribution map related to the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0025] Please refer to Figures 1-4 The embodiment of the present application provides a technical solution: a soft soil foundation structure visual monitoring network and a navigation system synchronous settlement early warning method, comprising S1, collecting image data, time signal data and environment data in the soft soil foundation monitoring process, and preprocessing the collected image data, time signal data and environment data to construct a standardized soft soil foundation state data set; S2, based on the standardized soft soil foundation state data set, the stability of the spatial offset characteristics of the reference point is evaluated from the continuous multiple image sequences, and the stability of the reference point in the monitoring coordinate system is dynamically adjusted based on the stability evaluation result and the subsequent replacement trigger condition; S3, based on the standardized soft soil foundation state data set, the image semantic consistency of the candidate reference point is analyzed, and the stable replacement reference point is selected based on the semantic analysis result; S4, taking the stability evaluation result and the semantic analysis result as input, the coordinate consistency of the replacement reference point is comprehensively evaluated, and the spatial position of the reference point in the structure coordinate system is dynamically corrected based on the comprehensive evaluation result; S5, the multi-base station monitoring data after back-off correction is time-synchronized, settlement trend is smoothed, and cross-base station difference comparison is performed, and the curve mutation caused by the replacement of the reference point is dynamically repaired Specifically, the specific steps of collecting image data, time signal data and environmental data in the soft soil foundation monitoring process are as follows: collecting image data in the soft soil foundation structure deformation monitoring process, the image data including: the lateral pixel coordinate value and the longitudinal pixel coordinate value of the reference point in each frame of image, which are used to construct the spatial position sequence of the reference point in the image coordinate system; the spatial offset of the reference point between adjacent frames, which is used to reflect the subtle displacement change in the continuous observation period; the gray gradient of each pixel point in the image, which is used to describe the local structure edge change; the gray level co-occurrence matrix energy value and the gray level co-occurrence matrix contrast, which are used to quantify the texture uniformity and gray difference of the image area. Through continuous frame sampling, the average offset of the reference point in the sliding window range and the sliding window frame number length are calculated and recorded, so as to ensure that the image time sequence data has stable spatial reference relationship and time consistency in the continuous observation period.

[0026] The time signal data collected by the Beidou timing synchronization includes: the positioning timing time interval of each frame and the positioning time synchronization reference value. Through the synchronous collection of time signal, each observation station captures images and records data under the unified timing reference, ensures the accurate consistency of the time sequence between each frame of data in the subsequent reference point steady-state evaluation and spatial rollback correction process, and provides a standardized reference for the time alignment of global monitoring.

[0027] The environmental data collected in the structure settlement response process includes: the horizontal projection coordinate component and the vertical projection coordinate component of each reference point in the structure coordinate system, which are used to describe the geometric position of the monitoring point in the actual structure space; the depth ranging value and the settlement displacement value, which are used to reflect the compression of soft soil layer and the deformation degree of structure; the ground illumination intensity, which is used to represent the influence of external light environment change on the quality of visual monitoring. The synchronous acquisition of such data can realize the dynamic correlation of environmental factors and spatial response, and provide multi-dimensional support data for subsequent semantic consistency discrimination and settlement trend correction.

[0028] In the embodiment, the data foundation framework of the visual monitoring network of the soft soil foundation structure is constructed, the fusion support of the three types of information of space, time and environment is realized by synchronously collecting image data, time signal data and environment data. The acquisition of the image data provides pixel-level feature basis for the steady-state evaluation and semantic consistency discrimination of the reference point, and can identify the tiny displacement change and texture difference; the collection of the time signal data establishes a unified time reference through the time synchronization mechanism of the navigation positioning system, ensures that the observation data of the multiple base stations are strictly aligned in time sequence, and provides time consistency guarantee for the subsequent spatial rollback correction and cross-base station trend comparison; the acquisition of the environment data can quantify the settlement response and external light interference under the actual spatial coordinates of the structure, and provides auxiliary variables on the physical level for the settlement trend analysis. The three types of data cooperatively constitute the core input channel of the settlement monitoring, so that the subsequent steady-state evaluation, substitute point discrimination and dynamic correction process have quantifiable, comparable and traceable basic data support.

[0029] Specifically, the collected image data, time signal data and environment data are preprocessed, and the specific steps of constructing the standardized soft soil foundation state data set are as follows: the image data is subjected to inter-frame smoothing processing, the high-frequency jitter components between continuous frames are filtered out through a sliding window, and the influence of environmental light changes, camera unit vibration and short-time pixel flicker on the reference point recognition accuracy is weakened. Subsequently, the maximum and minimum normalization is performed on the gray gradient value, the energy value of the gray level co-occurrence matrix and the contrast of the gray level co-occurrence matrix, so that different image regions have a unified numerical scale in the feature space, and it is ensured that in the subsequent steady-state evaluation and semantic consistency discrimination links, each texture feature can participate in the structure change recognition in an equal weight manner. The time reference unification processing is performed on the Beidou time signal data, the positioning and timing time interval is standardized according to the observation synchronization period, the time drift deviation caused by satellite timing delay and multipath error is eliminated; at the same time, the centralization operation is performed on the positioning time synchronization reference value, so that the time sequence is distributed with zero mean value in the unified reference frame, thereby realizing the accurate alignment and unified time scale mapping of the observation data between multiple base stations on the time axis, and providing a stable time reference for spatial rollback correction. The statistical calibration is performed on the environment data, the depth ranging value, settlement displacement value and ground surface light intensity of each reference point are extracted from the original sampling record, the abnormal ranging point is detected and repaired, and the continuity and reliability of the data are maintained; after the calibration is completed, the average depth value, settlement displacement mean value, light intensity standard deviation and light intensity mean value of each reference point are calculated and recorded, which provides statistical quantity support for the subsequent settlement trend correction and light adaptation processing. The standardized image data, Beidou time signal data and environment data are subjected to unified normalization processing, and the standardized soft soil foundation state data set is constructed.

[0030] In this implementation plan, a highly consistent input foundation for soft soil foundation settlement monitoring is constructed through the standardization and unification of multi-source data. Inter-frame smoothing and feature normalization of image data effectively eliminate local instability factors caused by illumination fluctuations, camera shake, and texture differences, ensuring the comparability and continuity of visual features in subsequent steady-state analysis. The benchmark unification and centralized processing of BeiDou time signal data ensures the synchronization and zero-bias distribution of multi-base station observation data along the time axis, providing precise time-scale support for spatial backtracking correction and cross-base station settlement trend alignment. Statistical calibration and anomaly repair of environmental data improve the reliability of depth ranging, settlement displacement, and illumination information, ensuring the matching of environmental impact factors and structural response data in terms of dimensions and accuracy. Through the joint normalization of these three types of data, this step achieves coordinated consistency of soft soil foundation state information across spatial, temporal, and environmental dimensions, providing a stable, balanced, and directly computable basic data structure for subsequent steady-state assessment, semantic discrimination, and dynamic correction.

[0031] Specifically, based on a standardized soft soil foundation state dataset, the stability assessment of the spatial offset characteristics of a reference point from a series of consecutive image frames involves the following steps: Calculate the sum of the squares of the differences in the horizontal and vertical pixel coordinates of the reference point in two adjacent frames, and then take the square root to obtain the single-frame offset value; sum the single-frame offset values ​​of the reference point in all adjacent frames before the current frame, and multiply this sum by the reciprocal of the sliding window frame length minus one to obtain the stable spatial offset value; calculate the square of the difference between the settlement displacement value and the mean settlement displacement of the reference point in the current frame to obtain the single-frame settlement fluctuation value; calculate the sum of the single-frame settlement fluctuation values ​​of the reference point in the current frame and all frames before the current frame, and multiply this sum by the reciprocal of the sliding window frame length to obtain the stable spatial fluctuation value; finally, add the stable spatial offset value and the stable spatial fluctuation value to obtain the steady-state assessment value of the reference point.

[0032] The formula for calculating the benchmark steady-state assessment value is: ; In the formula, It represents the horizontal pixel coordinate value of the reference point in the k-th frame image. It is used to quantify the displacement change of the reference point in the horizontal direction in the image coordinate system. It is the basic parameter for calculating the horizontal drift between adjacent frames and comes from the image pixel matrix analysis result output by the visual sensing device in the multi-reference redundant acquisition module. It represents the vertical pixel coordinate value of the reference point in the k-th frame image. It is used to quantify the displacement change of the reference point in the vertical direction in the image coordinate system. It is a key parameter reflecting the vertical settlement trend and the sensitivity to height change. It comes from the image frame positioning data synchronized and calibrated by Beidou time synchronization in the multi-reference redundant acquisition module. The settlement displacement value of the reference point in the kth frame, used to reflect the small settlement change amount of the reference point in the vertical direction, derived from the depth ranging module and the stereovision ranging result; The settlement displacement mean value in the sliding window range, used to calculate the time stationarity term, derived from the settlement displacement difference value average of the continuous frames; The sliding window frame length, used to limit the number of continuous frames participating in the calculation in the steady state evaluation process, is a time scale parameter balancing response sensitivity and stability, derived from the continuous sampling period configuration.

[0033] In this embodiment, the stability level of the reference point in the continuous observation period is quantified, which comprehensively reflects its dynamic consistency in space and time dimensions, and is used to judge the stability degree of the reference point in soft soil deformation monitoring.

[0034] Specifically, the specific steps of dynamically adjusting the stability identification of the reference point in the monitoring coordinate system and the subsequent replacement trigger condition based on the stability evaluation result are: real-time comparison of the reference point steady state evaluation value of the current reference point with the steady state evaluation threshold, and judgment of the stability level of the reference point in the soft soil settlement process through continuous monitoring.

[0035] When the reference point steady state evaluation value is less than or equal to the steady state evaluation threshold, the effective state of the reference point in the monitoring coordinate is maintained, and the Beidou timestamp of the current reference point is written into the multi-base station data buffer area synchronously, marked as a stable reference point, and the image sampling alignment instruction of the adjacent monitoring base station is triggered to ensure the synchronization sampling accuracy of the spatial distribution data under the same time reference, providing a stable reference for the subsequent replacement point candidate area identification. When the reference point steady state evaluation value is greater than the steady state evaluation threshold, an abnormal marking instruction is immediately sent to the visual recognition unit, the data writing of the reference point is suspended, and the same area image of the adjacent frame in the redundant image buffer is called to perform semantic consistency replacement matching to search for potential replacement points; at the same time, the coordinates of the replacement reference point are remapped to the Beidou timing reference framework for registration and update, ensuring the synchronization correlation of the replacement point in the time and space dimensions, and providing a unified space-time coordinate basis for the subsequent spatial rollback correction process.

[0036] In this embodiment, by real-time comparison of the benchmark point steady-state evaluation value and the steady-state threshold value, a dynamic judgment and automatic replacement mechanism of the benchmark point state is established, thereby guaranteeing the continuity and reliability of the space-time benchmark in the process of soft soil foundation settlement monitoring. When the benchmark point is within the steady-state threshold value range, through Beidou timestamp synchronization and multi-base station image sampling alignment, stable observation data are realized in time reference and space distribution, thereby providing a stable reference benchmark for subsequent trend analysis. When the steady-state evaluation value exceeds the threshold value, an abnormal response is triggered immediately, the data writing of the unstable point is suspended, and the semantic consistency replacement matching is realized to quickly replace and re-register the coordinates of the benchmark point, thereby ensuring that the monitoring network can still maintain the overall continuous observation ability when local drift occurs. This step realizes real-time state screening, abnormal self-repair and automatic connection of the replacement point of the benchmark point, so that the soft soil settlement monitoring has adaptive space-time benchmark updating capability, and provides high-reliability data support for subsequent spatial rollback correction and settlement warning output.

[0037] Specifically, based on the standardized soft soil foundation state data set, the semantic analysis of the image semantic consistency of the candidate benchmark point comprises the following steps: collecting the total number of pixels in the image in the neighborhood window of the replacement benchmark point, calculating the sum of the absolute values of the gray gradient of all pixel points and dividing the total number of pixels, then multiplying the ratio of the energy value of the gray co-occurrence matrix and the contrast of the gray co-occurrence matrix to obtain the local edge sharpness; the ratio of the standard deviation of the current replacement benchmark point to the average value of the illumination intensity is added to one to obtain the illumination balance correction value; the structural similarity index of the current replacement benchmark point and the stable benchmark point is calculated by the structural similarity algorithm, and the structural similarity index is divided by the standard deviation of the illumination intensity to obtain the structural similarity evaluation value; the absolute value of the difference between the current frame depth ranging value of the replacement benchmark point and the average depth value is divided by the average depth value, and then one is added to take the natural logarithm to obtain the depth direction continuity value; the product of the structural similarity evaluation value and the depth direction continuity value is added to the ratio of the local edge sharpness to the illumination balance correction value to obtain the semantic consistency discrimination value of the current replacement benchmark point.

[0038] The semantic consistency discrimination value calculation formula is: ; In the formula, represents the pixel of the neighborhood window of the replacement point, which is used to determine the statistical range of the average gray gradient and is a normalization parameter for calculating the edge sharpness. The value is directly determined by the total number of pixels in the neighborhood window of the current replacement point and is derived from the original image resolution collected by the visual monitoring camera and the local analysis window size; represents the gray gradient of the i-th pixel point, which is used to quantify the gray change intensity in the candidate area and is the basic data reflecting the edge sharpness, which is derived from the image gray matrix output by the camera; Gray level co-occurrence matrix energy value, used to measure the aggregation degree of the texture of the candidate region, the greater the value, the stronger the texture directionality, derived from the local gray scale statistics of the image collected by the visual acquisition device; Gray level co-occurrence matrix contrast, used to represent the change amplitude of the texture details, is an important factor for judging the continuity of the texture, derived from the difference result of the image pixel intensity; Light intensity standard deviation, used to measure the uniformity of the light distribution, the greater the value, the stronger the light disturbance, derived from the brightness sampling sequence of the camera exposure channel; Light intensity mean value, used to normalize and correct the light disturbance, derived from the global brightness mean value calculation of the current image frame; Alternative reference point Structural similarity index of the stable reference point , used to measure the geometric feature consistency, derived from the real-time comparison of the template matching unit on the image blocks of the two reference points; Depth ranging value of the candidate point in the current frame, used to reflect the spatial distance of the point from the monitored structure surface, derived from the stereo vision ranging and structured light module; Average depth value, used to quantify the consistency of the candidate point in the spatial direction, derived from the depth average value of the last stable period.

[0039] In this embodiment, the semantic consistency degree of the alternative reference point candidate region in the image space is quantified, which is used to judge the similarity level of the alternative point and the original reference point in terms of texture features, brightness distribution and structure matching, so as to select the optimal alternative target. The formula quantifies the local texture, light distribution and structure similarity comprehensively, ensures the semantic reliability and spatial correspondence accuracy of the alternative reference point, and provides a stable image feature basis for subsequent spatial rollback correction and settlement trend updating.

[0040] Specifically, based on the semantic analysis estimation result, the specific steps of selecting the stable alternative reference point are as follows: the semantic consistency discrimination value of the current alternative reference point is calculated in real time and written into the buffer, and the semantic change trend of the alternative reference point in different observation periods is continuously monitored to realize dynamic tracking of the state of the alternative point.

[0041] When the semantic consistency discrimination value of the current alternative reference point in the adjacent observation period keeps rising, it means that the texture stability and structure matching degree of the alternative point are continuously enhanced, the image sampling frequency is immediately increased to improve the short-time change resolution, and the optical flow residual correction algorithm is called synchronously to perform local registration update, so that the structure mapping of the alternative reference point in the image space is more accurate.

[0042] The semantic consistency discrimination value of the current substitute reference point in the adjacent observation period remains unchanged, indicating that the substitute point maintains relatively stable texture and brightness structure in continuous observation, maintains the existing substitute reference point as the main reference input, and simultaneously writes the Beidou timestamp, optical flow field vector and depth matching result into the data buffer area, thereby providing a stable space-time reference for subsequent spatial rollback correction and settlement trend analysis.

[0043] When the semantic consistency discrimination value of the current substitute reference point in the adjacent observation period shows a downward trend, it indicates that the texture features of the substitute point degrade and the illumination disturbance enhances, and the current substitute reference point data writing is immediately suspended, the candidate area is re-extracted and a new substitute reference point is selected, the position transfer and continuous observation are connected and registered, and the monitoring sequence is ensured to have stable continuity and structural matching reliability when the visual consistency decreases.

[0044] In the embodiment, dynamic monitoring and adaptive updating of the semantic consistency of the substitute reference point are implemented to maintain the stability and continuity of the visual reference point in the soft soil foundation settlement monitoring. By calculating the semantic consistency discrimination value in real time and analyzing its trend, the stability state of the substitute point is identified in time when the image texture, illumination condition and geometric structure fluctuate. Through dynamic perception of the semantic change trend and multi-layer adaptive adjustment, the stable transmission and visual consistency of the substitute reference point in a complex environment are ensured, and a reliable reference is provided for subsequent spatial rollback correction and settlement trend calculation.

[0045] Specifically, the coordinate consistency of the substitute reference point is comprehensively evaluated by taking the stability evaluation result and the semantic analysis result as inputs. The specific steps are as follows: the horizontal pixel coordinate value and the vertical pixel coordinate value of the substitute reference point serving as the main reference input in the current image frame are obtained, and the spatial displacement distance difference value between the substitute reference point serving as the main reference input and the last valid reference point is calculated and recorded; the product of the difference between the reference point stability evaluation value and the spatial displacement distance difference value is obtained as the stability attenuation correction value; the product of the time interval of positioning and timing after being divided by the positioning time synchronization reference value and the semantic consistency discrimination value is obtained as the time synchronization correction value; the difference between the horizontal projection coordinate component of the substitute reference point after being updated in the same structure coordinate system and the horizontal projection coordinate component of the last valid reference point in the structure coordinate system is calculated as the horizontal projection correction value; the difference between the vertical projection coordinate component of the substitute reference point after being updated in the same structure coordinate system and the vertical projection coordinate component of the last valid reference point in the structure coordinate system is calculated as the vertical projection correction value; the sum of the horizontal projection correction value and the vertical projection correction value is taken as the numerator, and the square root of the sum of the square values of the horizontal projection correction value and the vertical projection correction value is taken as the denominator, thereby obtaining the geometric projection correction value; the sum of the stability attenuation correction value, the time synchronization correction value and the geometric projection correction value is obtained as the spatial rollback correction value, wherein the navigation positioning system is the Beidou system.

[0046] The space rollback correction value calculation formula is:

[0047] In the formula, represents the reference point steady-state evaluation value, is used to reflect the stability level of the replacement point in the continuous time sequence, and is the main control factor for correction strength calculation; represents the spatial displacement distance difference value of the replacement point and the last effective reference point, is used to represent the position change amplitude on the structure, is the basic space measurement parameter for rollback correction, and is derived from the Beidou timing synchronous coordinate difference calculation; represents the semantic consistency discrimination value, is used to quantify the matching consistency of the replacement reference point in the image structure, texture and semantic level, and is the adjustment basis for time continuity correction; represents the Beidou timing time interval between the current frame and the last effective frame, is used to quantify the time offset in the sampling period, is the key parameter for time synchronization correction, and is derived from the timestamp data of the Beidou timing module; represents the Beidou time synchronization reference value, is used to normalize the time interval, so that the time sequence offset of different sampling periods has comparability, and is derived from the fixed synchronization parameter configuration of the Beidou timing system; represents the horizontal projection coordinate component of the replacement reference point after updating in the same coordinate system, is used to re-establish the spatial direction correlation, and is derived from the replacement point re-projection positioning module; represents the horizontal projection coordinate component of the last effective reference point in the structure coordinate system, is used to calculate the relative offset direction of the replacement point in the plane, and is derived from the multi-base station joint perspective projection data; represents the vertical projection coordinate component of the replacement reference point after updating in the same coordinate system, is used to correct the vertical spatial offset error, and is derived from the replacement point re-projection positioning module; represents the vertical projection coordinate component of the last effective reference point in the structure coordinate system, is used to reflect the relative position relationship of the point in the vertical direction, and is derived from the multi-base station perspective geometric mapping calculation.

[0048] In this embodiment, S of Example 1 is set to 0.82, is set to 0.35, and Q is set to 0.60, is set to 1.20, is set to 2.50, is set to 0.48, is set to 0.40, is set to 0.52, is set to 0.46; S of Example 2 is set to 0.75, is set to 0.30, and Q is set to 0.55, Set to 1.00, Set to 2.00, Set to 0.46, Set to 0.38, Set to 0.50, Set to 0.42; S of Example 3 set to 0.68, Set to 0.28, Q set to 0.62, Set to 1.40, Set to 2.30, Set to 0.50, Set to 0.43, Set to 0.53, Set to 0.44; S of Example 4 set to 0.80, Set to 0.32, Q set to 0.58, Set to 1.10, Set to 2.10, Set to 0.47, Set to 0.41, Set to 0.49, Set to 0.43; S of Example 5 set to 0.73, Set to 0.33, Q set to 0.63, Set to 1.30, Set to 2.40, Set to 0.52, Set to 0.45, Set to 0.54, Set to 0.47; S of Example 6 set to 0.78, Set to 0.31, Q set to 0.56, Set to 1.00, Set to 2.20, Set to 0.49, Set to 0.42, Set to 0.51, Set to 0.45; S of Example 7 set to 0.70, Set to 0.29, Q set to 0.59, Set to 1.50, Set to 2.60, Set to 0.53, Set to 0.46, Set to 0.55. The value is set to 0.48. The spatial rollback correction value for each embodiment is calculated, as shown in Table 1, Spatial Rollback Correction Value Data Table.

[0049] Table 1. Spatial Backtracking Correction Values ​​Data Table

[0050] like Figure 3 The table shown is a bar chart of spatial rollback correction values ​​provided in this application example. (See Table 1 and...) Figure 3 As can be seen, Instance 6 has the lowest spatial backtracking correction value, indicating that it has the smallest offset and the strongest steady-state retention during time synchronization and spatial coordinate correction. The replacement reference point can maintain the stability of the monitoring coordinates without additional adjustment. This instance is marked as a steady-state reference point to support continuous observation and threshold calibration. Instance 3 has the highest spatial backtracking correction value, indicating that the instance has significant deviations in the time alignment and spatial reprojection stages. It requires strong coordinate correction and global replacement operations. Its correction frequency is high and its spatial coupling complexity is large. It will be included in the priority intervention sequence to avoid anomaly accumulation. From the overall distribution trend, the lower the spatial backtracking correction value, the higher the matching stability between the replacement point and the structural coordinate system, and the more inclined it is to maintain a static observation state. The higher the spatial backtracking correction value, the more it reflects the potential drift risk. According to the threshold level, the corresponding fine-tuning, reconstruction and replacement mechanisms are dynamically triggered to ensure the stability and continuity of global synchronous monitoring.

[0051] Specifically, the steps for dynamically correcting the spatial position of the reference point in the structural coordinate system based on the comprehensive evaluation results are as follows: The current spatial backtracking correction value is compared with the backtracking correction threshold in real time, and a dynamic hierarchical threshold mechanism is used to achieve fine-grained control over the spatial stability of the substitute reference point. The backtracking correction threshold includes a first correction threshold and a second correction threshold, which are used to distinguish between the steady-state maintenance region, the fine-tuning correction region, and the global reconstruction region, respectively.

[0052] When the spatial backtracking correction value is less than or equal to the second correction threshold, it indicates that the fluctuation of the substitute reference point in time and space coordinates is within a controllable range. The existing coordinate state of the substitute reference point is maintained, and no geometric adjustment operation is triggered. At the same time, the original reference point identifier remains unchanged, and the current BeiDou timestamp and optical flow field vector are synchronously written into the buffer as continuous observation samples in the steady state interval to support subsequent trend analysis and time series consistency verification.

[0053] When the space rollback correction value is greater than the second correction threshold and less than or equal to the first correction threshold, it indicates that the alternative reference point has a slight spatial deviation, and the adjacent base station cross-field auxiliary projection data is immediately called to perform linear deviation correction to eliminate local parallax error and update the projection coefficient matrix of the alternative reference point in real time, so that the multi-base station spatial relationship is re-aligned, and an intermediate correction identifier is written in the data buffer to provide a reference for subsequent incremental correction and trend tracking.

[0054] When the space rollback correction value is greater than the first correction threshold, it indicates that the spatial deviation of the alternative reference point has exceeded the safe interval, and the current data writing channel is immediately frozen to prevent abnormal data from spreading, and the global reconstruction process of the alternative reference point is started to recalculate the main projection parameter set of the point in the structure coordinate system, and the time synchronization index is redefined through the Beidou timing mechanism to ensure the overall consistency of the spatial geometric relationship and the global traceability of the monitoring data.

[0055] In the embodiment, dynamic stability control and multi-level correction decision of the alternative reference point in the spatial dimension are realized to ensure the continuity and consistency of the settlement monitoring data in time and space. By comparing the space rollback correction value with the hierarchical threshold in real time, the maintenance, fine-tuning and reconstruction operations can be automatically selected according to the different deviation amplitudes of the reference point, so as to ensure the monitoring accuracy while avoiding excessive adjustment. This mechanism realizes multi-level coordination from local to global, so that the alternative reference point can still maintain the stability of the spatial geometric constraint relationship and the integrity of the data link under complex working conditions, and provides a high-confidence spatial reference for subsequent settlement trend calculation and early warning output.

[0056] Specifically, the multi-base station monitoring data after rollback correction is time-synchronized, settlement trend is smoothed, and cross-base station difference is compared, and the specific steps of dynamically repairing the curve mutation caused by the replacement of the reference point are as follows: reading the coordinates of the reference point after the space rollback correction, and calling the Beidou timing module to extract the time labels of each monitoring base station to realize the synchronization and alignment of the observation data of each base station under the unified time reference. Through the continuity analysis of the settlement displacement sequence of the multi-base station monitoring reference point, the time series smoothing algorithm is used to extract the settlement change trend of the current frame, and the mutation points and discontinuous sections in the time series are identified and marked to exclude the non-real fluctuations caused by short-time disturbance, light change and observation overlap error, and ensure the time series smoothing and physical rationality of the settlement trend.

[0057] After obtaining the preliminary settlement trend, a secondary correction process is performed to superimpose the benchmark point steady-state evaluation value and semantic consistency discrimination value onto the time series curve, locally smooth the sudden displacement surge caused by benchmark replacement, and make the settlement trajectory of the replaced benchmark seamlessly connect with the previous data on the time axis. At the same time, the image features and settlement results between each base station are compared for differences, the spatial gradient changes across base stations are extracted, the multi-base station synchronous settlement trend graph and abnormal identification list for the current monitoring period are generated, and the corrected settlement trend value, benchmark identification number and Beidou time synchronization index are written into the monitoring database, providing complete data chain for subsequent statistical analysis and visual output.

[0058] After completing the writing of the monitoring database, the settlement warning data output process is entered, and the settlement distribution graph, warning state code and corresponding base station number are generated and output in real time, so that the monitoring center can intuitively master the overall deformation trend and abnormal area distribution of the soft soil foundation in the current period, and provide high-precision, traceable real-time monitoring results for structure safety evaluation and dynamic warning decision-making.

[0059] As Figure 3 The multi-base station synchronous settlement trend diagram provided by the present application is shown in the figure, which shows the settlement change characteristics and synchronous correction effect of each observation base station of the soft soil foundation structure in the continuous monitoring period. The horizontal axis is time, and the vertical axis is the settlement amount Ah, with a unit of meters. The three colored main curves correspond to the settlement trend of base station 1, base station 2 and base station 3 at the same time sequence, and the blue, orange and purple curves are the settlement results after steady-state evaluation and semantic consistency correction. The gray scatter points in the figure represent the original observation values of each base station, reflecting the fluctuation characteristics affected by light, environmental disturbance and benchmark drift in actual monitoring. The red triangular mark is the benchmark replacement node, indicating the automatic replacement correction process after detecting steady-state degradation. The green dashed line is the cross-base station settlement difference distribution curve, which is used to depict the relative settlement gradient between different observation areas. The yellow belt area is the warning threshold range, which is used to indicate the safe settlement interval, and when the settlement curve exceeds this area, an abnormal warning is automatically triggered. Overall, the multi-base station synchronous settlement trend diagram intuitively reflects the continuity and consistency of the settlement trend after multi-base station synchronous monitoring, time alignment and spatial backtracking correction, providing data support for subsequent structure settlement warning.

[0060] As Figure 4The real-time settlement distribution diagram provided by the present application is shown, which is used to show the spatial distribution of settlement along the soft soil foundation monitoring area at the same time. The horizontal axis represents the spatial position of the monitoring point, in meters, and the vertical axis represents the settlement amount Dh at the current time, in meters. The gray dots are the original observation values of each monitoring point, reflecting the measured settlement data without steady-state correction; the blue main curve is the settlement value after multi-base station synchronous correction, which reflects the smooth settlement trend after time synchronization and spatial rollback compensation; the green dotted line is the fitting result of the overall settlement trend, which is used to describe the overall deformation form of the structure at that time; the yellow belt area is the warning threshold range, which is used to distinguish the safe area and the potential risk area; the red triangle is the abnormal settlement point mark, indicating that the monitoring value exceeds the threshold interval and triggers the warning. Overall, this diagram intuitively shows the spatial settlement distribution characteristics and abnormal point positioning effect of the soft soil foundation under real-time monitoring conditions, providing visual verification basis for the synchronous correction results of the multi-base station visual monitoring network.

[0061] In the present embodiment, the fusion analysis and dynamic warning output of the settlement data after spatial rollback correction in time and space dimensions are realized, ensuring the continuity, accuracy and traceability of the monitoring results. By calling the Beidou timing tag to align the observation data of the multi-base station, the timing error caused by timing error can be effectively eliminated, so that the settlement sequences between different base stations have comparability in the same reference frame; through the time series smoothing algorithm and the mutation identification mechanism, local abnormalities caused by environmental disturbance, benchmark replacement and visual jitter can be detected and repaired, maintaining the overall continuity and dynamic stability of the settlement trend curve; the secondary correction process realizes adaptive smoothing compensation of instantaneous error by superimposing the steady-state evaluation value of the reference point and the semantic consistency discrimination value, thereby improving the continuity of the replacement reference point in the time dimension. The closed-loop linkage from data correction to dynamic warning of soft soil foundation settlement monitoring is realized, and a high-time-efficiency and high-precision multi-base station synchronous settlement warning mechanism is constructed.

[0062] It should be noted that, in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0063] The preferred embodiments of the application disclosed above are only to facilitate the elucidation of the application. The preferred embodiments do not describe all the details of the application and limit the application to the specific embodiments described. Obviously, many modifications and variations can be made in light of the teachings above. The description is chosen and described in order to best explain the principles of the application and its practical application to thereby enable others skilled in the art to best utilize the application and get the best results from the application. The application is only limited by the claims and their full scope and equivalents.

Claims

1. A method for synchronous settlement early warning of soft soil foundation structures using a visual monitoring network and navigation system, characterized in that: include: S1, collect image data, time signal data and environmental data during the monitoring process of soft soil foundation, and preprocess the collected image data, time signal data and environmental data to construct a standardized soft soil foundation state dataset; S2, based on a standardized soft soil foundation state dataset, evaluates the stability of the spatial offset characteristics of the benchmark point from a series of consecutive image frames, and dynamically adjusts the stability label of the benchmark point in the monitoring coordinate system and subsequent alternative triggering conditions based on the stability evaluation results. S3, based on a standardized soft soil foundation state dataset, performs semantic analysis on the semantic consistency of candidate benchmark points in images, and selects stable alternative benchmark points based on the semantic analysis results; S4, taking stability assessment results and semantic analysis results as input, comprehensively evaluates the coordinate consistency of the alternative reference point, and dynamically corrects the spatial position of the reference point in the structural coordinate system based on the comprehensive evaluation results; S5 performs time synchronization, settlement trend smoothing, and cross-base station difference comparison on the backtracked multi-base station monitoring data, and dynamically repairs curve abrupt changes caused by benchmark point replacement.

2. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for collecting image data, time signal data, and environmental data during the soft soil foundation monitoring process are as follows: Image data was collected during the deformation monitoring of soft soil foundation structures. The image data included: the horizontal and vertical pixel coordinates of the reference points in each frame, the spatial offset of the reference points between adjacent frames, the gray-level gradient of each pixel in the image, the energy value of the gray-level co-occurrence matrix, and the contrast of the gray-level co-occurrence matrix. At the same time, the average offset of the reference points within the sliding window range and the length of the sliding window frame were calculated and recorded. Collect time signal data for timing synchronization of the navigation and positioning system. The time signal data includes: the positioning and timing time interval of each frame and the positioning time synchronization reference value; Environmental data were collected during the structural settlement response process. The environmental data included: the horizontal and vertical projection coordinate components of each reference point in the structural coordinate system, the depth measurement value, the settlement displacement value, and the surface light intensity.

3. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for preprocessing the collected image data, time signal data, and environmental data to construct a standardized soft soil foundation state dataset are as follows: Inter-frame smoothing is performed on image data to eliminate single-frame jitter interference, and the gray-level gradient value, gray-level co-occurrence matrix energy value, and gray-level co-occurrence matrix contrast are normalized by maximum and minimum. The time reference of the navigation and positioning system is unified, the positioning and timing time interval is standardized according to the synchronization period, and the positioning time synchronization reference value is centered so that the time series has a zero mean distribution in the unified reference frame. Statistical calibration is performed on the environmental data, and the depth measurement value, settlement displacement value and surface light intensity of each benchmark point are extracted. Abnormal measurement points are checked and interpolated for repair. At the same time, the average depth value, mean settlement displacement, standard deviation of light intensity and mean light intensity of each benchmark point are calculated and recorded. The standardized image data, time signal data, and environmental data are normalized to construct a standardized soft soil foundation state dataset.

4. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for evaluating the stability of the spatial offset characteristics of the reference point from a series of consecutive image frames based on the standardized soft soil foundation state dataset are as follows: Calculate the sum of the squares of the differences in the horizontal and vertical pixel coordinates of the reference point in two adjacent frames, and then take the square root to obtain the single-frame offset value. The spatial offset stability value is obtained by adding the single-frame offset values ​​of the reference point in all adjacent frames before the current frame and multiplying them by the reciprocal of the sliding window frame length minus one. The difference between the settlement displacement value of the reference point in the current frame image and the mean settlement displacement is calculated and then squared to obtain the settlement fluctuation value of a single frame. The sum of the single-frame settling fluctuation values ​​of the reference point in the current frame and all frames before the current frame is calculated, and then multiplied by the reciprocal of the sliding window frame length to obtain the spatial stability fluctuation value. The steady-state assessment value of the benchmark point is obtained by adding the stable value of the spatial offset to the stable value of the spatial fluctuation.

5. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for dynamically adjusting the stability indicator and subsequent alternative triggering conditions of the benchmark point in the monitoring coordinate system based on the stability assessment results are as follows: Real-time comparison of the current benchmark steady-state evaluation value with the steady-state evaluation threshold: When the steady-state evaluation value of the reference point is less than or equal to the steady-state evaluation threshold, the reference point is kept in a valid state in the monitoring coordinates, and the navigation and positioning system timestamp of the current reference point is synchronously written into the multi-base station data buffer and recorded as a stable reference point. At the same time, the image sampling alignment command of the adjacent monitoring base station is triggered. When the steady-state evaluation value of the reference point is greater than the steady-state evaluation threshold, an abnormal marking instruction is immediately sent to the visual recognition unit to suspend the writing of reference point data, and the images of the same area of ​​adjacent frames in the redundant image cache are called to perform semantic consistency replacement matching. At the same time, the coordinates of the replacement reference point are mapped to the navigation and positioning system for re-registration and update.

6. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for semantic analysis of the image semantic consistency of candidate reference points based on the standardized soft soil foundation state dataset are as follows: The total number of pixels in the image within the neighborhood window of the alternative reference point is collected. The sum of the absolute values ​​of the gray-level gradients of all pixels is calculated and divided by the total number of pixels. Then, it is multiplied by the ratio of the gray-level co-occurrence matrix energy value and the gray-level co-occurrence matrix contrast value of the image to obtain the local edge sharpness. The illumination equalization correction value is obtained by adding one to the ratio of the standard deviation of the current replacement benchmark point to the mean of the illumination intensity. The structural similarity index between the current alternative benchmark point and the stable benchmark point is calculated using a structural similarity algorithm. The structural similarity index is then divided by the standard deviation of the illumination intensity to obtain the structural similarity assessment value. Divide the absolute value of the difference between the depth ranging value of the current frame and the average depth value of the replacement reference point by the average depth value, add one and take the natural logarithm to obtain the continuous value of the depth direction. The semantic consistency discriminant value of the current alternative benchmark point is obtained by adding the product of the structural similarity evaluation value and the depth direction continuity value, plus the ratio of the local edge sharpness to the illumination equalization correction value.

7. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for selecting stable alternative benchmarks based on semantic analysis estimation results are as follows: Calculate the semantic consistency discriminant value of the current alternative benchmark point in real time and write it to the buffer; observe and record the changing trend of the semantic consistency discriminant value of the alternative benchmark point. When the semantic consistency discrimination value of the current alternative reference point continues to rise in adjacent observation periods, the image sampling frequency is immediately increased, and the optical flow residual correction algorithm is called simultaneously to perform local registration update. When the semantic consistency discrimination value of the current alternative reference point remains unchanged in adjacent observation periods, the existing alternative reference point is maintained as the main reference input, and the navigation and positioning system timestamp, optical flow field vector and depth matching results are synchronously written to the data cache. When the semantic consistency discrimination value of the current alternative benchmark point shows a downward trend in adjacent observation periods, the writing of the current alternative benchmark point is paused, the alternative benchmark point is reacquired, and the benchmark point location is transferred and continuous observation is registered.

8. The method for synchronous settlement early warning of soft soil foundation structure visual monitoring network and navigation system according to claim 1, characterized in that: The specific steps for comprehensively evaluating the coordinate consistency of the alternative benchmark point, using stability assessment results and semantic analysis results as input, are as follows: Obtain the horizontal and vertical pixel coordinates of the alternative reference point used as the main reference input in the current image frame, and calculate and record the spatial displacement distance difference between the alternative reference point used as the main reference input and the previous valid reference point. Multiply the result of subtracting the steady-state evaluation value from the benchmark point by the difference in spatial displacement distance to obtain the steady-state attenuation correction value; The time synchronization correction value is obtained by dividing the positioning time synchronization time interval by the positioning time synchronization reference value and then multiplying it by the semantic consistency discrimination value. The difference between the horizontal projection coordinate components of the previous valid reference point in the structural coordinate system and the horizontal projection coordinate components of the previous valid reference point in the structural coordinate system after the replacement reference point is updated is calculated to obtain the horizontal projection correction value. The vertical projection correction value is obtained by subtracting the vertical projection coordinate components of the previous valid reference point from the vertical projection coordinate components of the previous valid reference point in the same structural coordinate system after the replacement reference point is updated. The sum of the horizontal and vertical projection correction values ​​is used as the numerator, and the square root of the sum of the squares of the horizontal and vertical projection correction values ​​is used as the denominator to obtain the geometric projection correction value. The spatial backoff correction value is obtained by adding the steady-state decay correction value, the time synchronization correction value, and the geometric projection correction value.

9. The method for synchronous settlement early warning of a visual monitoring network and navigation system for soft soil foundation structures according to claim 1, characterized in that: The specific steps for dynamically correcting the spatial position of the reference point in the structural coordinate system based on the comprehensive evaluation results are as follows: Real-time comparison of the current spatial rollback correction value with the rollback correction threshold, which includes a first correction threshold and a second correction threshold: When the spatial backtracking correction value is less than or equal to the second correction threshold, the existing coordinate state of the substitute reference point is maintained, the original reference point identifier remains unchanged, and the current navigation and positioning system timestamp and optical flow field vector are recorded simultaneously as continuous observation samples in the steady-state interval. When the spatial backoff correction value is greater than the second correction threshold and less than or equal to the first correction threshold, the auxiliary projection in the cross field of view of the adjacent base station is called to perform linear offset correction, and the projection coefficient matrix of the replacement reference point is updated. At the same time, the intermediate correction identifier is written into the data cache. When the spatial backoff correction value is greater than the first correction threshold, the current data writing channel is frozen, the main projection parameter set of the current alternative reference point in the structural coordinate system is reconstructed, and the time synchronization index of the current alternative reference point is redefined through the navigation and positioning system.

10. The method for synchronous settlement early warning of a visual monitoring network and navigation system for soft soil foundation structures according to claim 1, characterized in that: The specific steps for performing time synchronization, settlement trend smoothing, and cross-base station difference comparison on the backtracked multi-base station monitoring data, and dynamically repairing curve abrupt changes caused by benchmark point replacement, are as follows: Read the coordinates of the reference points after spatial backtracking correction, and call the navigation and positioning system to extract the time tags of each base station. Perform a continuity analysis on the settlement displacement of the monitoring reference points of each base station, and analyze the settlement trend of the reference points in the current frame based on the time series smoothing algorithm to identify abrupt change points and discontinuities. After obtaining the settlement trend, a secondary correction process is performed. The steady-state evaluation value of the benchmark point and the semantic consistency judgment value are superimposed on the time series curve. Local smoothing repair is performed for the instantaneous displacement surge caused by the benchmark point replacement. At the same time, the image data between each base station is compared to generate a settlement trend map and an anomaly identification list for the current monitoring period. The corrected settlement trend, benchmark point identification number and time synchronization index are written into the monitoring database. After the monitoring database is written, the settlement early warning data output process begins, outputting a real-time settlement distribution map, early warning status code, and corresponding base station number.

Images