Underground cavity tomography method based on existing urban infrastructure excitation source

Abstract

The invention provides an underground cavity tomography method based on a city existing infrastructure excitation source, and relates to the field of underground detection, and the method comprises the steps: collecting an excitation source, and constructing an original observation sequence; extracting the characteristics of the excitation source through a time-frequency analysis method based on the preprocessed data, and constructing an excitation source characteristic sub-library; aligning the original observation sequence according to the excitation source occurrence time, and obtaining a waveform data set by adopting pulse response estimation; carrying out frequency dispersion chromatography inversion based on the waveform data set to obtain a rigidity or wave velocity abnormal body; and performing probability fusion on the anomalous body, constructing a voxel-level probability distribution model, and combining a preset probability threshold and a geometric constraint condition to generate a targeted review list and a routing inspection path. The technical scheme of the invention is suitable for underground cavity rapid general survey, fine review and risk assessment scenes of urban roads, pipe gallery corridors and pipe network dense areas.

Description

Technical Field

[0001] This application relates to the field of underground exploration, and more particularly to a method for underground cavity tomography based on excitation sources from existing urban infrastructure. Background Technology

[0002] Underground cavities are a typical hidden danger in the development and operation of urban underground space. They mainly stem from factors such as insufficient roadbed compaction, pipeline leakage and erosion, and unfinished underground engineering projects. These cavities can easily lead to major engineering disasters such as road collapses and pipeline ruptures, posing a direct threat to urban traffic safety and infrastructure stability. Currently, underground cavity detection technology has formed a technical system based on "active stimulation detection as the main method, passive monitoring as a supplement, and drilling verification as the final means." However, in practical applications in densely populated urban areas (such as commercial districts, utility tunnels, and elevated ramps), existing technologies still face many challenges, specifically in the following aspects: 1. Significant construction disruption and limited coverage: Active source detection requires road closures, traffic interruptions, or surface excavation, making large-scale operations difficult in densely trafficked areas such as commercial districts and elevated ramps; single-line / single-point detection is inefficient, with a single operation typically covering less than 0.1 square kilometers. This cannot meet the needs of a rapid city-level census.

[0003] 2. High excitation costs and poor environmental adaptability: Traditional technologies rely on manual excitation equipment (such as hammers and vibratory source vehicles) or active launch devices, which have high equipment purchase and maintenance costs. When operating at night, manual excitation is prone to noise pollution and is difficult to match the complex underground media conditions in cities (such as densely populated pipeline areas and areas with alternating soft and hard soil layers).

[0004] 3. Insufficient multimodal data fusion: Existing detection methods rely heavily on single data sources (such as GPR, which can only identify shallow interfaces, and drilling, which only reflects single-point information), and lack systematic fusion of multi-source data such as shear wave velocity (VS), ground-penetrating radar reflection, and drilling soil layer classification; data interpretation relies on human experience, which can easily lead to missed or false detections of cavities due to "misjudgment of single indicators", especially in areas with interference around pipelines where accuracy drops significantly.

[0005] 4. Lack of adaptive adjustment capability: During the detection process, the parameters cannot be dynamically optimized according to the intensity of the on-site excitation (such as reduced traffic flow or weak water hammer signal) or the verification results. When the stability of the excitation source is insufficient, it is easy to reduce the inversion accuracy, requiring repeated work and further increasing time and economic costs.

[0006] Furthermore, the "natural excitation sources" widely present in cities, such as transient water hammer / pressure pulsations in municipal pipe networks and road vehicle loads, have the advantages of wide distribution, repeatability, and no additional disturbance. However, current technologies have not been able to effectively utilize these signal sources for reliable detection of underground cavities. Therefore, there is an urgent need for an underground cavity tomographic imaging technology that can rely on existing urban infrastructure for excitation, with low disturbance, wide coverage, and high precision, to address the shortcomings of existing detection methods. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of existing underground cavity detection technologies, such as large construction disturbance, high cost, narrow coverage, insufficient multimodal data fusion, and lack of adaptive adjustment capabilities, which rely on active excitation sources. This invention provides an underground cavity tomography method based on excitation sources from existing urban infrastructure.

[0008] The above-mentioned objective of this application is achieved through the following technical solution: S1: Collect the excitation sources generated by urban infrastructure; simultaneously deploy a surface sensor array and a distributed fiber optic acoustic sensing system to collect data and obtain the original observation sequence; S2: Preprocess the original observation sequence; and based on the preprocessed data, extract the features of the excitation source using time-frequency analysis to construct an excitation source feature sub-library; S3: Align the original observation sequence according to the occurrence time of the excitation source, and use the impulse response estimation and cross-correlation method to solve the medium propagation function to obtain the waveform dataset; perform dispersion tomography inversion based on the waveform dataset to obtain the shallow shear wave velocity and equivalent stiffness distribution, and obtain the stiffness or wave velocity anomaly. S4: Probabilistically fuse anomalies with ground-penetrating radar profiles, drilling data, and the spatial layout of underground pipelines to construct a voxel-level probability distribution model; S5: Based on the preset probability threshold and geometric constraints, combined with the voxel-level probability distribution model, output the three-dimensional spatial range, depth estimate and confidence interval of the suspected cavity, and generate a targeted review list and inspection path.

[0009] Optionally, step S1 includes: Within the target area, identify and collect repeatable or semi-repeatable excitation sources generated by urban infrastructure, including transient water hammer or pressure pulsation of municipal pipelines, as well as dynamic loads and environmental vibrations generated by road vehicle traffic. The surface sensor array uses MEMS accelerometers with an element spacing of 2–15 meters and a sampling frequency of 200–4000 Hz; the distributed fiber optic acoustic sensing system has a measurement spacing of 5–20 meters and an equivalent frequency bandwidth of 10–2000 Hz.

[0010] Optionally, step S2 includes: Preprocessing includes: detrending and denoising; The amplitude, frequency band, duration, and repetitiveness characteristics of the excitation source are extracted using time-frequency analysis methods, and a feature sub-library of the excitation source is constructed. The time-frequency analysis method includes short-time Fourier transform and continuous wavelet transform, and uses a repeatability score R to screen for excitation sources with high stability. Incentives should be given priority for inclusion in the incentive source sub-pool.

[0011] Optionally, step S3 includes: The original observation sequence was aligned according to the excitation source occurrence time. The medium propagation function was solved by impulse response estimation and cross-correlation methods to obtain the waveform dataset related to the subsurface medium response. The surface wave dispersion curve was extracted based on the waveform dataset, and two-dimensional or three-dimensional dispersion tomography inversion was performed to obtain the shallow shear wave velocity and equivalent stiffness distribution. A multi-input multi-output impulse response estimation method is adopted, and a time-varying filter is introduced into the vehicle excitation to compensate for the phase drift caused by changes in vehicle speed; The tomographic inversion uses a grid size of 1 to 10 meters and a surface wave frequency range of 0.5 to 50 Hz. Two-dimensional Rayleigh wave phase velocity tomography is performed first, followed by three-dimensional joint inversion to constrain the spatial distribution of shear wave velocity.

[0012] Optionally, step S4 includes: By probabilistically fusing anomalies with ground-penetrating radar profiles, drilling data, and the spatial layout of underground pipelines, a voxel-level probability distribution model of underground cavities or loose areas is constructed. The probabilistic fusion employs a Bayesian update framework, where the negative anomaly amplitude relative to the baseline is measured by the shear wave velocity. Velocity gradient norm Using the dispersion fitting residual as the likelihood term, and with the ground-penetrating radar interface reflection characteristics, drilling soil layer classification results, and pipeline space buffer as prior information, the probability of void existence for each voxel is calculated. Thus, a voxel-level probability distribution model was obtained.

[0013] Optionally, step S5 includes: The thresholding process employs the Otsu method or quantile thresholding. And set geometric constraints, including minimum connected volume. The volume is 0.01 to 0.5 cubic meters, with a minimum equivalent radius. The distance is 0.1 to 0.5 meters to eliminate discrete noise points; Based on the on-site verification results and newly collected data, the excitation source feature sub-library, inversion parameter weights, and probability thresholds are dynamically updated.

[0014] An electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform a method for underground cavity tomography based on excitation sources of existing urban infrastructure.

[0015] A computer-readable storage medium storing instructions that, when executed, perform a method for underground cavity tomography based on excitation sources from existing urban infrastructure.

[0016] The beneficial effects of the technical solution provided in this application are: This invention utilizes transient water hammer / pressure pulsations in municipal pipelines and road vehicle loads as available excitations, combining surface MEMS arrays and distributed fiber optic DAS for data acquisition and observation. The medium response is obtained through time-frequency analysis and source deconvolution, surface wave dispersion is extracted, and 2D / 3D tomographic inversion is performed to reconstruct the shallow shear wave velocity and equivalent stiffness distribution. Bayesian fusion is then performed with GPR, drilling, and pipeline spatial priors to construct a voxel-level cavity probability volume, outputting the 3D range, depth, and confidence interval of suspected cavities, and generating a targeted verification list and scheduling strategy. Compared to traditional active source detection, this invention reduces construction disturbance and cost, has a wider coverage area, and is suitable for rapid surveying and detailed verification of urban roads, utility tunnels, and densely networked areas. Attached Figure Description

[0017] The present application will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a step diagram of an embodiment of this application; Figure 2 This is the overall technical roadmap in the embodiments of this application; Figure 3 This is a data perception and incentive filtering diagram in the embodiments of this application; Figure 4 This is a diagram illustrating the signal processing and medium response extraction in an embodiment of this application; Figure 5 This is a speed modeling and probability fusion graph in an embodiment of this application; Figure 6 This is the target recognition and adaptive optimization graph in the embodiments of this application; Figure 7 This is a schematic diagram of the electronic device structure in the embodiments of this application. Detailed Implementation

[0018] To provide a clearer understanding of the technical features, objectives, and effects of this application, the specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0019] The embodiments of this application provide a method for underground cavity tomography based on excitation sources from existing urban infrastructure.

[0020] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the steps of a method for underground cavity tomography based on excitation sources from existing urban infrastructure, as described in this application, including: S1: Collect the excitation sources generated by urban infrastructure; simultaneously deploy a surface sensor array and a distributed fiber optic acoustic sensing system to collect data and obtain the original observation sequence; S2: Preprocess the original observation sequence; and based on the preprocessed data, extract the features of the excitation source using time-frequency analysis to construct an excitation source feature sub-library; S3: Align the original observation sequence according to the occurrence time of the excitation source, and use the impulse response estimation and cross-correlation method to solve the medium propagation function to obtain the waveform dataset; perform dispersion tomography inversion based on the waveform dataset to obtain the shallow shear wave velocity and equivalent stiffness distribution, and obtain the stiffness or wave velocity anomaly. S4: Probabilistically fuse anomalies with ground-penetrating radar profiles, drilling data, and the spatial layout of underground pipelines to construct a voxel-level probability distribution model; S5: Based on the preset probability threshold and geometric constraints, combined with the voxel-level probability distribution model, output the three-dimensional spatial range, depth estimate and confidence interval of the suspected cavity, and generate a targeted review list and inspection path.

[0021] In one specific implementation of this application, Figure 2 This is the overall technical roadmap in the embodiments of this application; Figure 3 This is a data perception and incentive filtering diagram in the embodiments of this application; Figure 4 This is a diagram illustrating the signal processing and medium response extraction in an embodiment of this application; Figure 5 This is a speed modeling and probability fusion graph in an embodiment of this application; Figure 6 This is the target recognition and adaptive optimization diagram in the embodiments of this application.

[0022] Step S1 includes: Within the target area, identify and collect repeatable or semi-repeatable excitation sources generated by urban infrastructure, including transient water hammer or pressure pulsation of municipal pipelines, as well as dynamic loads and environmental vibrations generated by road vehicle traffic. The surface sensor array uses MEMS accelerometers with an element spacing of 2–15 meters and a sampling frequency of 200–4000 Hz; the distributed fiber optic acoustic sensing system has a measurement spacing of 5–20 meters and an equivalent frequency bandwidth of 10–2000 Hz.

[0023] Step S2 includes: Preprocessing includes: detrending and denoising; The amplitude, frequency band, duration, and repetitiveness characteristics of the excitation source are extracted using time-frequency analysis methods, and a feature sub-library of the excitation source is constructed. The time-frequency analysis method includes short-time Fourier transform and continuous wavelet transform, and uses a repeatability score R to screen for excitation sources with high stability. Incentives should be given priority for inclusion in the incentive source sub-pool.

[0024] Step S3 includes: The original observation sequence was aligned according to the excitation source occurrence time. The medium propagation function was solved by impulse response estimation and cross-correlation methods to obtain the waveform dataset related to the subsurface medium response. The surface wave dispersion curve was extracted based on the waveform dataset, and two-dimensional or three-dimensional dispersion tomography inversion was performed to obtain the shallow shear wave velocity and equivalent stiffness distribution. A multi-input multi-output impulse response estimation method is adopted, and a time-varying filter is introduced into the vehicle excitation to compensate for the phase drift caused by changes in vehicle speed; The tomographic inversion uses a grid size of 1 to 10 meters and a surface wave frequency range of 0.5 to 50 Hz. Two-dimensional Rayleigh wave phase velocity tomography is performed first, followed by three-dimensional joint inversion to constrain the spatial distribution of shear wave velocity.

[0025] Step S4 includes: By probabilistically fusing anomalies with ground-penetrating radar profiles, drilling data, and the spatial layout of underground pipelines, a voxel-level probability distribution model of underground cavities or loose areas is constructed. The probabilistic fusion employs a Bayesian update framework, where the negative anomaly amplitude relative to the baseline is measured by the shear wave velocity. Velocity gradient norm Using the dispersion fitting residual as the likelihood term, and with the ground-penetrating radar interface reflection characteristics, drilling soil layer classification results, and pipeline space buffer as prior information, the probability of void existence for each voxel is calculated. Thus, a voxel-level probability distribution model was obtained.

[0026] Step S5 includes: The thresholding process employs the Otsu method or quantile thresholding. And set geometric constraints, including minimum connected volume. The volume is 0.01 to 0.5 cubic meters, with a minimum equivalent radius. The distance is 0.1 to 0.5 meters to eliminate discrete noise points; Based on the on-site verification results and newly collected data, the excitation source feature sub-library, inversion parameter weights, and probability thresholds are dynamically updated. Adaptive scheduling includes: increasing the weight of shear wave velocity anomalies in the likelihood function when the positive rate of the verification is lower than the preset target; and triggering zoned pipeline pressure pulsation tests to enhance the excitation signal when the reduced traffic flow at night leads to insufficient excitation intensity.

[0027] In one specific embodiment of this application, a subsurface cavity tomographic imaging system is designed, comprising: an excitation identification unit for identifying water hammer and vehicle traffic excitations from municipal pipelines and constructing an excitation source feature sub-library; an acquisition and alignment unit for simultaneously acquiring surface array and distributed fiber optic data and performing spatiotemporal alignment and source deconvolution processing; a dispersion and tomographic inversion unit for extracting surface wave dispersion curves and reconstructing the three-dimensional distribution of shear wave velocity or stiffness; a multimodal fusion unit for probabilistically fusing the inversion results with ground-penetrating radar, drilling, and pipeline priors; and a target output and scheduling unit for outputting the cavity probability volume, generating a checklist, and adaptively updating system parameters.

[0028] The acquisition and alignment unit includes a time synchronization module, an inter-track cross-correlation module, a multi-input multi-output impulse response estimation module, and a vehicle speed adaptive phase compensation module.

[0029] The dispersion and tomography inversion unit includes an automatic dispersion curve picking module, a two-dimensional tomography imaging module, a three-dimensional inversion module, and a result quality control module.

[0030] The multimodal fusion unit includes a prior construction module, a likelihood assessment module, and a Bayesian update module. The prior construction module integrates ground-penetrating radar reflection intensity, borehole soil layer labels, and pipeline space buffer information. The likelihood assessment module performs calculations based on wave velocity anomalies, gradients, and dispersion residuals.

[0031] In one embodiment, this application has the following technical effects: 1. Low disturbance and wide coverage: No manual active excitation is required; existing urban infrastructure (pipeline network, traffic flow) is directly used as the excitation source, avoiding road occupation and traffic interruption. The coverage area of ​​a single operation can reach 2-3 [units / areas]. It is suitable for rapid surveys in densely populated areas such as commercial districts and utility tunnels.

[0032] 2. Low cost and high adaptability: Reuse the "natural incentive" of existing urban pipeline network and road traffic flow to significantly reduce equipment purchase and operation and maintenance costs; Support MEMS and DAS multi-sensor collaboration, and can be adapted to different detection scenarios (such as traffic flow dominance during the day and water hammer in pipeline network at night).

[0033] 3. High precision and high reliability: By using multimodal Bayesian fusion (VS inversion + GPR + drilling + pipeline prior), the problem of misjudgment based on single data is solved, and the detection rate of suspected voids is improved by more than 30% compared with traditional methods; the introduction of repeatability scoring to screen the excitation source reduces the inversion residual by 15% to 20%.

[0034] To verify the high accuracy and reliability of the proposed method, a comparative experiment was conducted between the proposed multimodal Bayesian fusion method and traditional single-data methods. The comparison objects included: Method A (traditional method): Anomaly identification is based solely on shear wave velocity (VS) tomography inversion results, without integrating GPR, drilling and pipeline prior information, and without using repeatability scoring to screen excitation sources; Method B (Comparative Method): A simple superposition interpretation of VS inversion + GPR + drilling + pipeline prior is used, but the Bayesian probabilistic fusion framework is not used and the repeatability score R is not introduced to screen the stimulus source. Method C (the method of this application): Multimodal Bayesian fusion (VS inversion + GPR + drilling + pipeline prior) is adopted, and excitation records with repeatability score R≥0.6 are introduced to construct an excitation source feature sub-library, and excitations with low repeatability and poor signal-to-noise ratio are eliminated.

[0035] Within the test area, 19 actual cavities / loose bodies were confirmed based on on-site verification (GPR mobile survey + drilling). Using this as the "true value" set, the detection of suspected cavities by different methods and the inversion residuals were statistically analyzed, and the results are shown in Table 1.

[0036] Table 1 Comparison of the effectiveness of different methods in identifying underground cavities

[0037] As can be seen from Table 1: Compared with the traditional method A, the method C of this application, under the same review conditions, increases the detection rate of suspected cavities from about 63% to about 84%, which is a relative increase of about 33%, consistent with the field experience statistics that "the detection rate of suspected cavities is more than 30% higher than that of the traditional method"; Meanwhile, the false alarm rate of method C was reduced from about 29% of method A to about 24%, which improved detection while avoiding a large number of false alarms; Regarding the dispersion fitting residuals, the Rrms of method C is reduced by about 18% compared to method A (from 1.00 to 0.82), falling within the range of "reduction of inversion residuals by 15% to 20%", indicating that after introducing the repeatability score R to screen the excitation source, the surface wave dispersion fitting quality and VS inversion stability are significantly improved.

[0038] 4. Adaptive and easy to implement: Parameters (such as weights and incentive supplements) can be dynamically adjusted according to the intensity of on-site excitation and the results of verification, reducing repetitive work; the output target verification checklist can directly guide on-site inspections, reducing the verification mileage by more than 25% and significantly improving detection efficiency.

[0039] In one embodiment, a rapid survey of a commercial district road-pipeline complex scenario is conducted: Area and sensor configuration: Approximately 2.5 kilometers of old commercial districts in a certain city were selected. The area is characterized by dense road networks and a complex network of DN300–DN800 water supply / sewage pipelines. MEMS surface arrays (sampling at 1kHz) are deployed every 5m along the roads, and DAS (measurement spacing 10m, bandwidth 10–1500Hz) are laid along the main pipelines.

[0040] Excitation and data collection: During the day, traffic flow is used as a continuous excitation, and at night, a small water hammer (pressure pulsation <0.1MPa, duration <1s) is introduced through the operation of zoned valves, triggering twice per night.

[0041] Time-frequency analysis: Perform STFT and CWT on the array and DAS data, and statistically analyze the pulsation energy envelope, repeatability score R, and records with R≥0.6 are added to the excitation sub-library.

[0042] Source deconvolution and alignment: Using the water hammer triggering time as a reference, cross-correlation alignment of each channel is performed, and MIMO impulse response estimation is used to compensate for the phase delay difference between adjacent test sections.

[0043] Dispersion and tomography: 0.8–20Hz surface wave phase velocity curves were extracted with a grid resolution of 5m. 2D tomography was performed first to obtain the phase velocity field, and then 3D VS inversion was performed with the joint constraints of surface array + DAS.

[0044] Multimodal fusion: , The likelihood of "loose / void tendency" is obtained by increasing the residual of the dispersion fitting to the threshold, taking the reflection intensity of the historical GPR profile, the existing borehole soil layer classification and pipeline buffer zone (±2m) as priors, and using Bayesian update to obtain the void probability volume.

[0045] Target output: Threshold , =0.02 , =0.2m. Output 3 suspected cavity hotspots, depth 0.8–1.6m, confidence interval ±0.2m, and generate a checklist (priority inspection routes, recommended GPR frequencies and measurement line lengths).

[0046] Verification and Scheduling: The following day, GPR mobile survey and verification of two shallow drilling sites were conducted. Two sites were positive and one was suspected. The system corrected the verification results for this batch, improving efficiency. The weight is 10%, and a low-amplitude water hammer is added at night to improve the signal-to-noise ratio.

[0047] Results: Compared with the baseline of "no excitation sub-library / no fusion", the detection rate of suspected voids increased by about 30%, the review mileage decreased by about 25%, and the on-site operation disturbance was significantly reduced.

[0048] In another embodiment, the elevated ramp-traffic main excitation scenario: Scenario: Vehicle speed fluctuates significantly and traffic volume is high at the curve of the elevated ramp.

[0049] Key points of the method: Only traffic flow excitation is used, vehicle speed estimation is provided by camera triggering or radar speed measurement, time-varying phase compensation is introduced in source deconvolution, and the dispersion band is extended to 2–35Hz to cover thinner road surfaces and overlying fill layers.

[0050] Results: Two longitudinally extending negative VS anomaly zones were identified. After integrating the priors of ramp expansion joints and rainwater branch pipes, sheet-like bodies with a probability of "loose zone" ≥ 0.8 were output, and it is recommended to verify with GPR.

[0051] In another embodiment, the utility tunnel corridor—a DAS-led silent census: Scenario: Municipal integrated utility tunnel corridor, where it is inconvenient to deploy surface arrays during the day.

[0052] Key points of the method: Based on DAS, the ultra-low amplitude water hammer test is carried out at night by the pipeline network dispatcher, triggered 1 to 2 times, with the surface wave frequency band set at 0.5–10Hz, the inversion resolution set at 10m, and fused with the existing GPR vertical short profile and roadway inspection records.

[0053] Result: A small cavity (~0.03m³) was output, prompting "secondary priority review". Two weeks later, the review confirmed that it was a localized loose filling material, and low-pressure grouting was implemented.

[0054] This application also discloses an electronic device. (See reference...) Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. The electronic device 500 may include: at least one processor 501, at least one network interface 504, a user interface 503, a memory 505, and at least one communication bus 502.

[0055] The communication bus 502 is used to enable communication between these components.

[0056] The user interface 503 may include a display screen, and optionally, the user interface 503 may also include a standard wired interface or a wireless interface.

[0057] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0058] This application also discloses a computer-readable storage medium storing multiple instructions adapted for loading by a processor to execute the above-described method for underground cavity tomography based on excitation sources of existing urban infrastructure.

[0059] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure.

[0060] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for tomographic imaging of underground cavities based on excitation sources from existing urban infrastructure, characterized in that, The method includes the following steps: S1: Collect incentive sources generated by urban infrastructure; Simultaneously deploy a surface sensor array and a distributed fiber optic acoustic sensing system to collect data and obtain the original observation sequence; S2: Preprocess the original observation sequence; and based on the preprocessed data, extract the features of the excitation source using time-frequency analysis to construct an excitation source feature sub-library; S3: Align the original observation sequence according to the occurrence time of the excitation source, and use the impulse response estimation and cross-correlation method to solve the medium propagation function to obtain the waveform dataset; Based on the waveform dataset, dispersion tomography inversion is performed to obtain the shallow shear wave velocity and equivalent stiffness distribution, and stiffness or wave velocity anomaly is obtained. S4: Probabilistically fuse anomalies with ground-penetrating radar profiles, drilling data, and the spatial layout of underground pipelines to construct a voxel-level probability distribution model; S5: Based on the preset probability threshold and geometric constraints, combined with the voxel-level probability distribution model, output the three-dimensional spatial range, depth estimate and confidence interval of the suspected cavity, and generate a targeted review list and inspection path.

2. The underground cavity tomography method based on excitation sources from existing urban infrastructure as described in claim 1, characterized in that, Step S1 includes: Within the target area, identify and collect repeatable or semi-repeatable excitation sources generated by urban infrastructure, including transient water hammer or pressure pulsation of municipal pipelines, as well as dynamic loads and environmental vibrations generated by road vehicle traffic. The surface sensor array uses MEMS accelerometers with an element spacing of 2–15 meters and a sampling frequency of 200–4000 Hz; the distributed fiber optic acoustic sensing system has a measurement spacing of 5–20 meters and an equivalent frequency bandwidth of 10–2000 Hz.

3. The underground cavity tomography method based on excitation sources from existing urban infrastructure as described in claim 1, characterized in that, Step S2 includes: Preprocessing includes: detrending and denoising; The amplitude, frequency band, duration, and repetitiveness characteristics of the excitation source are extracted using time-frequency analysis methods, and a feature sub-library of the excitation source is constructed. The time-frequency analysis method includes short-time Fourier transform and continuous wavelet transform, and uses a repeatability score R to screen for excitation sources with high stability. Incentives should be given priority for inclusion in the incentive source sub-pool.

4. The underground cavity tomography method based on excitation sources from existing urban infrastructure as described in claim 1, characterized in that, Step S3 includes: The original observation sequence was aligned according to the excitation source occurrence time. The medium propagation function was solved by impulse response estimation and cross-correlation methods to obtain the waveform dataset related to the subsurface medium response. The surface wave dispersion curve was extracted based on the waveform dataset, and two-dimensional or three-dimensional dispersion tomography inversion was performed to obtain the shallow shear wave velocity and equivalent stiffness distribution. A multi-input multi-output impulse response estimation method is adopted, and a time-varying filter is introduced into the vehicle excitation to compensate for the phase drift caused by changes in vehicle speed; The tomographic inversion uses a grid size of 1 to 10 meters and a surface wave frequency range of 0.5 to 50 Hz. Two-dimensional Rayleigh wave phase velocity tomography is performed first, followed by three-dimensional joint inversion to constrain the spatial distribution of shear wave velocity.

5. The underground cavity tomography method based on excitation sources from existing urban infrastructure as described in claim 1, characterized in that, Step S4 includes: By probabilistically fusing anomalies with ground-penetrating radar profiles, drilling data, and the spatial layout of underground pipelines, a voxel-level probability distribution model of underground cavities or loose areas is constructed. The probabilistic fusion employs a Bayesian update framework, where the negative anomaly amplitude relative to the baseline is measured by the shear wave velocity. Velocity gradient norm Using the dispersion fitting residual as the likelihood term, and with the ground-penetrating radar interface reflection characteristics, drilling soil layer classification results, and pipeline space buffer as prior information, the probability of void existence for each voxel is calculated. Thus, a voxel-level probability distribution model was obtained.

6. The underground cavity tomography method based on excitation sources from existing urban infrastructure as described in claim 1, characterized in that, Step S5 includes: The thresholding process employs the Otsu method or quantile thresholding. And set geometric constraints, including minimum connected volume. The volume is 0.01 to 0.5 cubic meters, with a minimum equivalent radius. The distance is 0.1 to 0.5 meters to eliminate discrete noise points; Based on the on-site verification results and newly collected data, the excitation source feature sub-library, inversion parameter weights, and probability thresholds are dynamically updated.

7. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform the underground cavity tomography method based on the excitation source of existing urban infrastructure as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a computer, perform the underground cavity tomography method based on excitation sources of existing urban infrastructure as described in any one of claims 1-6.

Images