A multi-channel three-dimensional ground penetrating radar capable of realizing real-time interpretation of collapse hidden dangers

Abstract

The application discloses the technical field of three-dimensional ground penetrating radar and particularly relates to a multi-channel three-dimensional ground penetrating radar capable of realizing AI real-time interpretation of collapse hidden dangers. First, multi-channel parameter initialization is completed according to a detection scene, and a synchronous and anti-interference signal is generated, and underground radar reflection signals are collected and pretreated; then, based on the pretreated standardized underground radar digital signals, an interpretation model is loaded and initialized, signal features are extracted, collapse hidden dangers are identified, and precision is optimized; based on this, the application realizes the real-time linkage of collected data and AI interpretation by deeply fusing a multi-channel cooperative collection module and an AI real-time interpretation module, the AI interpretation module can call standardized data in a collection process in real time, and the standardized data are cooperatively worked on by units such as model initialization, feature extraction, real-time identification and precision optimization, so that the problem that AI interpretation is disconnected from data collection is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional ground-penetrating radar technology, specifically a multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of potential collapse hazards. Background Technology

[0002] With the rapid advancement of urbanization, urban underground spaces, like an invisible "dark web," harbor various safety hazards such as road subsidence, loose underground soil, and water accumulation, seriously threatening the safety of municipal road traffic, the safe operation of underground pipelines, and the personal and property safety of urban residents. Three-dimensional ground-penetrating radar, as a core device for non-destructive detection of underground hazards, has been widely used in the investigation of subsidence hazards in municipal roads, rail transit, and other fields.

[0003] However, in practical applications, existing 3D ground-penetrating radars may suffer from a disconnect between AI interpretation and data acquisition, making it difficult to achieve real-time linkage between the two. Specifically, the data acquisition system and AI analysis module of most commercial ground-penetrating radars are independent, making real-time linkage difficult. After data acquisition, it needs to be exported to a dedicated terminal for offline analysis, resulting in long interpretation delays. Furthermore, the AI ​​module cannot directly access real-time data from the acquisition process for dynamic adjustments. For example, in the detection of old urban roads, existing equipment needs to spend 2-3 hours for offline interpretation after collecting underground data for a 10km stretch of road. During this period, if a sudden underground collapse hazard is encountered, it cannot be detected and warned in time, which can easily lead to safety accidents. At the same time, the AI ​​interpretation module cannot dynamically optimize the interpretation algorithm based on the real-time acquired antenna frequency parameters and detection depth data, resulting in significant differences in interpretation accuracy for collapse hazards at different depths (such as 0-2m shallow layers and 3-5m deep layers), with a false alarm rate of over 10% for deep hazards. Summary of the Invention

[0004] To address the technical problem of the disconnect between AI interpretation and data acquisition, making it difficult to achieve real-time linkage between data acquisition and AI interpretation, this invention provides the following technical solution:

[0005] A multi-channel 3D ground-penetrating radar capable of real-time AI interpretation of subsidence hazards includes a mobile base. A ground-penetrating radar main unit is fixedly mounted on the top of the mobile base. A bracket is fixedly mounted on the front side of the mobile base. An electric lifting device is fixedly mounted on the top of the bracket. A support plate is fixedly mounted on the top of the electric lifting device. A MIMO array antenna is fixedly mounted on the top of the support plate. A connector for connection to the rear of a vehicle is provided on the front side of the bracket. The system also includes a multi-channel 3D ground-penetrating radar system comprising:

[0006] The multi-channel collaborative acquisition module initializes multi-channel parameters according to the detection scenario, generates synchronization and anti-interference signals, acquires and preprocesses underground radar reflection signals, and outputs preprocessed standardized underground radar digital signals.

[0007] The AI ​​real-time interpretation module includes an interpretation model initialization unit, a radar signal feature extraction unit, a collapse hazard real-time identification unit, and an interpretation accuracy optimization unit. The interpretation model initialization unit loads a preset collapse hazard AI interpretation model and, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module, initializes the core parameters of the interpretation model and outputs the initialized collapse hazard AI interpretation model. The radar signal feature extraction unit, based on the initialized collapse hazard AI interpretation model output by the interpretation model initialization unit, extracts features from the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and outputs the extracted underground medium radar signal feature set. The real-time identification unit, based on the extracted underground medium radar signal feature set output by the radar signal feature extraction unit, performs real-time matching and identification through an AI interpretation model to distinguish between normal underground medium and potential collapse hazards, classifies potential collapse hazards, calculates the specific location, size, and depth parameters of the hazards, and outputs real-time identification results and identification confidence levels. The interpretation accuracy optimization unit, based on the real-time identification results and identification confidence levels output by the real-time identification unit, judges the accuracy of the identification results, and simultaneously calls the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module to perform secondary feature extraction and matching on suspected hazards, optimizes the interpretation model parameters, and outputs optimized collapse hazard identification results and model optimization parameters.

[0008] The intelligent control module monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. At the same time, it judges the collaborative logic and outputs control commands including parameter adjustment, antenna raising and lowering, channel switching and execution feedback data of each control action.

[0009] The results output module integrates and detects the preprocessed standardized underground radar digital signals output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. At the same time, it interprets relevant information and performs standardized processing to realize the visualization, local storage and export / import of results.

[0010] The hazard tracing and dynamic prediction module, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data, completes the generation of hazard cause tracing, development trend prediction and prevention and control suggestions;

[0011] The data transmission module sequentially classifies, encrypts, encodes, performs internal data interaction, and transmits external data to the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the results output module, and the hidden danger tracing and dynamic prediction module, thereby achieving bidirectional data communication.

[0012] As a preferred embodiment of the multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards described in this invention, the multi-channel collaborative acquisition module includes:

[0013] The multi-channel parameter configuration unit presets core parameters such as antenna main frequency, number of channels, channel spacing, antenna coverage width, effective detection depth, and detection speed according to the detection scenario. At the same time, it completes the initialization configuration of multi-channel acquisition parameters and outputs the initialized multi-channel acquisition parameters.

[0014] The synchronization signal generation unit generates a synchronization control signal based on the initialized multi-channel acquisition parameters output by the multi-channel parameter configuration unit, controlling each channel to start acquisition work simultaneously; it also generates an anti-interference synchronization signal and outputs the multi-channel synchronous acquisition control signal and the anti-interference synchronization signal.

[0015] The multi-channel radar signal acquisition unit, based on the multi-channel synchronous acquisition control signal and anti-interference synchronization signal output by the synchronization signal generation unit, synchronously acquires the radar reflection signal of the underground medium through the MIMO array antenna, and dynamically adjusts the acquisition sensitivity in combination with the real-time status of the electric lifting device; at the same time, it acquires in parallel through 8 / 10 channels and outputs the original underground radar reflection signal acquired synchronously through multiple channels.

[0016] The raw signal preprocessing unit, based on the underground radar reflection raw signal synchronously acquired by the multi-channel radar signal acquisition unit, uses filtering, denoising, and signal enhancement algorithms to remove interference signals from the raw signal, and performs normalization processing on the signal to convert the signal into a standardized digital signal, and outputs the preprocessed standardized underground radar digital signal.

[0017] As a preferred embodiment of the multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards described in this invention, the intelligent control module includes:

[0018] The data receiving and status monitoring unit monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module; at the same time, it monitors the working status of the electric lifting device, determines whether there is any abnormality, and outputs the working status data of the system / device.

[0019] The collaborative logic judgment unit, based on the system / device working status data output by the data receiving and status monitoring unit, judges whether the collaborative working logic of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module is reasonable; if there is a logical deviation or abnormality, it generates adjustment instruction logic and outputs the collaborative working logic judgment result and adjustment instruction logic.

[0020] The parameter and action control unit, based on the collaborative working logic judgment result and adjustment instruction logic output by the collaborative logic judgment unit, first controls the multi-channel collaborative acquisition module to adjust the acquisition parameters; then controls the electric lifting device to lift and lower, adapting to the antenna height requirements of different detection scenarios; then controls the multi-channel switching to ensure that the acquisition coverage accuracy and detection speed are both taken into account, and outputs control instructions including parameter adjustment, antenna lifting and lowering, channel switching and execution feedback data of each control action;

[0021] The emergency control unit, based on the parameters and the execution feedback data of each control action output by the action control unit, will immediately generate an emergency stop command to stop the operation of the relevant modules if a serious abnormality is detected; at the same time, it will generate an abnormal alarm signal to prompt staff to troubleshoot the fault.

[0022] As a preferred embodiment of the multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards described in this invention, the output module includes:

[0023] The results data integration unit, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, integrates the interpretation results with the acquired data, supplements auxiliary information, and outputs a complete detection results dataset.

[0024] The results standardization processing unit, based on the complete detection results dataset output by the results data integration unit, formats the results data according to a preset standardized format to organize it into standardized tables and text reports; at the same time, it performs standardized rendering of the radar signal spectrum and outputs standardized detection results reports, standardized hidden danger information tables, and standardized radar signal spectrums.

[0025] The results display and export unit, based on the standardized detection results report, standardized hidden danger information table, and standardized radar signal spectrum output by the results standardization processing unit, enables the visualization of the results; it also supports local storage and export of the results data.

[0026] As a preferred embodiment of the multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards described in this invention, the hazard tracing and dynamic prediction module includes:

[0027] The hazard tracing unit, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data, analyzes the core causes of collapse hazards through tracing algorithms, and at the same time associates the location, depth, and size parameters of the hazard to clarify the correspondence between the causes and the characteristics of the hazard, and outputs a collapse hazard tracing report and tracing credibility parameters;

[0028] The dynamic prediction unit for hidden dangers, based on the tracing report and credibility parameters of the cause of the collapse hidden dangers output by the cause tracing unit, combined with meteorological data and past regional hidden danger evolution data, dynamically predicts the development trend of the collapse hidden dangers through a time-series prediction algorithm, while setting the prediction period and warning threshold, and outputting the dynamic prediction results, prediction warning thresholds and periodic prediction reports for the hidden dangers.

[0029] The prevention and control suggestion generation unit generates targeted and implementable prevention and control suggestions for collapse hazards based on the dynamic prediction results of the hidden dangers, the prediction and early warning thresholds, and the periodic prediction reports output by the dynamic prediction unit of the hidden dangers. At the same time, it transmits the prevention and control suggestion instructions to the intelligent control module and transmits the standardized prevention and control suggestion documents to the results output module.

[0030] As a preferred embodiment of the multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards described in this invention, the data transmission module includes:

[0031] The data encoding unit, based on the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the result output module, and the hidden danger tracing and dynamic prediction module, classifies and encodes different types of data, and uses an encryption encoding algorithm to ensure that the data is not leaked or lost during transmission; at the same time, it converts the data into a format that adapts to the transmission protocol, and outputs the encrypted and encoded classified data and data transmission format identifier.

[0032] The internal data interaction unit, based on the encrypted and encoded classification data and data transmission format identifier output by the data encoding unit, adopts a high-speed transmission protocol to realize real-time data interaction between the multi-channel collaborative acquisition module, AI real-time interpretation module, intelligent control module, results output module, and hidden danger tracing and dynamic prediction module; at the same time, it monitors the data interaction status and outputs encrypted data of real-time interaction and data interaction status feedback.

[0033] The external data transmission unit transmits the encrypted and encoded classified data and data transmission format identifier output by the data encoding unit to an external terminal or the city's smart management and control platform via wired / wireless transmission; at the same time, it receives control commands from the external terminal or the management and control platform and transmits them to the intelligent control module.

[0034] Compared with existing technologies:

[0035] 1. By setting up a multi-channel collaborative acquisition module, it can realize synchronous acquisition and anti-interference processing of multiple channels, effectively solve the problem of poor multi-channel coordination, and at the same time, it can achieve both detection efficiency and coverage accuracy, avoid channel data gaps and hidden danger omissions, and improve the coordination consistency and data acquisition quality of the multi-channel acquisition system.

[0036] 2. By deeply integrating the multi-channel collaborative acquisition module with the AI ​​real-time interpretation module, it can achieve real-time linkage between acquired data and AI interpretation. The AI ​​interpretation module can call up standardized data in the acquisition process in real time. Through the collaborative work of units such as model initialization, feature extraction, real-time recognition and accuracy optimization, it effectively solves the problem of the disconnect between AI interpretation and data acquisition. At the same time, it can also realize real-time interpretation of subsidence hazards, dynamic optimization of interpretation accuracy, and timely issuance of warnings of suspicious hazards, avoiding safety accidents caused by interpretation delays. It breaks down the barriers between acquisition and interpretation, and significantly improves the timeliness and accuracy of interpretation. Attached Figure Description

[0037] Figure 1 This is a front view schematic diagram of the overall structure of the present invention;

[0038] Figure 2 This is a top view of the overall structure of the present invention;

[0039] Figure 3 This is a schematic diagram of the multi-channel three-dimensional ground-penetrating radar system framework of the present invention;

[0040] Figure 4 This is a schematic diagram of the multi-channel collaborative acquisition module framework of the present invention;

[0041] Figure 5 This is a schematic diagram of the AI ​​real-time interpretation module framework of the present invention;

[0042] Figure 6This is a schematic diagram of the intelligent control module framework of the present invention;

[0043] Figure 7 This is a schematic diagram of the output module framework of the present invention;

[0044] Figure 8 This is a schematic diagram of the hazard tracing and dynamic prediction module framework of the present invention;

[0045] Figure 9 This is a schematic diagram of the data transmission module framework of the present invention.

[0046] In the figure: mobile base 10, ground penetrating radar main unit 20, bracket 30, electric lifting device 40, bearing plate 50, MIMO array antenna 60, connector 70. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0048] This invention provides a multi-channel 3D ground-penetrating radar capable of real-time AI interpretation of potential collapse hazards. Please refer to [link / reference]. Figures 1-2 The system includes a mobile base 10, on the top of which a ground-penetrating radar host 20 is fixedly mounted. A bracket 30 is fixedly mounted on the front side of the mobile base 10. An electric lifting device 40 is fixedly mounted on the top of the bracket 30. A support plate 50 is fixedly mounted on the top of the electric lifting device 40. A MIMO array antenna 60 is fixedly mounted on the top of the support plate 50. A connector 70 for connecting to the rear of the vehicle is provided on the front side of the bracket 30.

[0049] Also includes: a multi-channel three-dimensional ground-penetrating radar system, please refer to... Figure 3 The multi-channel three-dimensional ground-penetrating radar system includes:

[0050] The multi-channel collaborative acquisition module initializes multi-channel parameters according to the detection scenario, generates synchronization and anti-interference signals, acquires and preprocesses underground radar reflection signals, and outputs preprocessed standardized underground radar digital signals.

[0051] The AI ​​real-time interpretation module loads and initializes the interpretation model based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module. At the same time, it extracts signal features, identifies potential collapse hazards, optimizes accuracy, and outputs the optimized collapse hazard identification results and model optimization parameters.

[0052] The intelligent control module monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. At the same time, it judges the collaborative logic and outputs control commands including parameter adjustment, antenna raising and lowering, channel switching and execution feedback data of each control action.

[0053] The results output module integrates and detects the preprocessed standardized underground radar digital signals output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. At the same time, it interprets relevant information and performs standardized processing to realize the visualization, local storage and export / import of results.

[0054] The hazard tracing and dynamic prediction module, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data, completes the generation of hazard cause tracing, development trend prediction and prevention and control suggestions;

[0055] The data transmission module sequentially classifies, encrypts, encodes, performs internal data interaction, and transmits external data to the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the results output module, and the hidden danger tracing and dynamic prediction module, thereby achieving bidirectional data communication.

[0056] Please see Figure 4 The multi-channel collaborative acquisition module includes:

[0057] The multi-channel parameter configuration unit presets core parameters such as antenna main frequency (200MHz / 400MHz switchable), number of channels (8 / 10 channels selectable), channel spacing (fixed 14cm), antenna coverage width (2.0m standard width, supports adaptive adjustment), effective detection depth (0-5m), and detection speed (maximum 70km / h) according to the detection scenario (such as municipal roads and community side roads). At the same time, it completes the initial configuration of multi-channel acquisition parameters and outputs the initialized multi-channel acquisition parameters.

[0058] The synchronization signal generation unit generates a synchronization control signal based on the initialized multi-channel acquisition parameters output by the multi-channel parameter configuration unit. This control signal enables each channel to start acquisition simultaneously, preventing data gaps caused by asynchronous channel data acquisition. Simultaneously, it generates an anti-interference synchronization signal to suppress interference from vehicle vibration and external electromagnetic signals on the acquisition process, and outputs both the multi-channel synchronization acquisition control signal and the anti-interference synchronization signal.

[0059] The multi-channel radar signal acquisition unit, based on the multi-channel synchronous acquisition control signal and anti-interference synchronization signal output by the synchronization signal generation unit, synchronously acquires radar reflection signals of the underground medium through the MIMO array antenna 60. It also dynamically adjusts the acquisition sensitivity by combining the real-time status (such as lifting height) of the electric lifting device 40 (set as an electric push rod) to ensure signal acquisition integrity at different detection depths (0-5m). Simultaneously, it achieves high coverage acquisition with a 2.0m antenna coverage width through 8 / 10-channel parallel acquisition, minimizing detection blind spots, adapting to the rapid detection requirements of 70km / h, and outputting the original underground radar reflection signals acquired synchronously through multiple channels.

[0060] The raw signal preprocessing unit, based on the underground radar reflection raw signal synchronously acquired by the multi-channel radar signal acquisition unit, uses filtering, denoising, and signal enhancement algorithms to remove interference signals (such as clutter generated by vehicle vibration and external electromagnetic interference) from the raw signal, and performs normalization processing on the signal to convert the signal into a standardized digital signal, and outputs the preprocessed standardized underground radar digital signal.

[0061] Please see Figure 5 The AI ​​real-time interpretation module includes:

[0062] The interpretation model initialization unit loads a preset AI interpretation model for subsidence hazards (trained based on massive amounts of underground subsidence hazard data and radar signal feature data, covering subsidence hazard features of different depths and sizes from 0 to 5m), and initializes the core parameters of the interpretation model (such as feature extraction threshold and recognition accuracy parameters) based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module, and outputs the initialized AI interpretation model for subsidence hazards.

[0063] The radar signal feature extraction unit, based on the initialized AI interpretation model of the subsidence hazard output by the interpretation model initialization unit, performs feature extraction on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module. It extracts core features such as the amplitude, frequency, and phase of the reflected wave of the underground medium to identify key feature parameters related to subsidence hazards (such as underground cavities and loose soil), and outputs the extracted underground medium radar signal feature set.

[0064] The real-time subsidence hazard identification unit, based on the extracted underground medium radar signal feature set output by the radar signal feature extraction unit, performs real-time matching and identification through an AI interpretation model to distinguish between normal underground medium and subsidence hazards (such as underground cavities, loose soil, water-rich areas, etc.), and classifies the subsidence hazards (by size into small, medium, and large, and by depth into shallow, medium, and deep layers). At the same time, it calculates the specific location (based on detection speed and channel location conversion), size, depth, and other parameters of the hazard, and outputs the real-time identification results and identification confidence level.

[0065] The interpretation accuracy optimization unit, based on the real-time identification results and identification confidence level output by the real-time collapse hazard identification unit, judges the accuracy of the identification results (results with a confidence level below 90% are considered suspicious hazards). At the same time, it calls the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module to perform secondary feature extraction and matching on suspicious hazards and optimize the interpretation model parameters. If the identification confidence level is still below 85%, an early warning signal is generated to prompt staff to conduct key verification, and the optimized collapse hazard identification results (confidence level ≥ 85%) and model optimization parameters are output.

[0066] Please see Figure 6 The intelligent control module includes:

[0067] The data receiving and status monitoring unit monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module (e.g., whether the acquisition unit is acquiring data normally and whether the AI ​​interpretation module is operating normally) based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module; at the same time, it monitors the working status of the electric lifting device 40, determines whether there are any abnormalities (e.g., antenna lifting jamming), and outputs the working status data of the system / device.

[0068] The collaborative logic judgment unit, based on the system / device operating status data output by the data receiving and status monitoring unit, judges whether the collaborative working logic of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module is reasonable (e.g., whether the acquisition parameters are adapted to the AI ​​interpretation requirements, and whether the antenna status affects the acquisition accuracy); if there is a logical deviation or abnormality, it generates adjustment instruction logic and outputs the collaborative working logic judgment result and adjustment instruction logic (if any).

[0069] The parameter and action control unit, based on the collaborative working logic judgment result and adjustment instruction logic output by the collaborative logic judgment unit, first controls the multi-channel collaborative acquisition module to adjust the acquisition parameters (such as optimizing parameters according to the AI ​​interpretation model, adjusting the antenna main frequency and acquisition sensitivity); then controls the electric lifting device 40 to lift and lower, adapting to the antenna height requirements of different detection scenarios; then controls the multi-channel switching (8 / 10 channels) to ensure that the acquisition coverage accuracy and detection speed are both taken into account, and outputs control instructions including parameter adjustment, antenna lifting and lowering, channel switching and execution feedback data of each control action;

[0070] The emergency control unit, based on the parameters and the execution feedback data of each control action output by the action control unit, will immediately generate an emergency stop command to stop the operation of the relevant modules if a serious anomaly is detected (such as antenna lifting failure, data acquisition interruption, or AI interpretation module crash). At the same time, it will generate an anomaly alarm signal to prompt staff to troubleshoot the fault. If the anomaly is minor (such as slight interference or low confidence level), it will generate an emergency adjustment command to control the relevant units to perform self-repair.

[0071] Please see Figure 7 The output module includes:

[0072] The results data integration unit integrates the preprocessed standardized underground radar digital signals output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. It also supplements auxiliary information such as detection time, detection location (latitude and longitude, which can be connected to the Beidou / GPS positioning system), and detection parameters, and outputs a complete dataset of detection results.

[0073] The results standardization processing unit, based on the complete detection results dataset output by the results data integration unit, formats the results data according to a preset standardized format (adapting to the import requirements of the intelligent management and control platform) to organize the collapse hazard information (location, size, depth, type), acquisition parameters, interpretation confidence level, etc. into standardized tables and text reports; at the same time, it performs standardized rendering of the radar signal spectrum for easy viewing by staff, and outputs standardized detection results reports, standardized hazard information tables, and standardized radar signal spectrums.

[0074] The results display and export unit, based on the standardized detection results report, standardized hidden danger information table, and standardized radar signal spectrum output by the results standardization processing unit, enables the visualization of results (such as real-time display of the hidden danger distribution and radar spectrum of the current detection section); it also supports local storage and export of results data (such as exporting to Excel or PDF format), and supports direct import of standardized results into the city's smart management and control platform to achieve "one map" management of above-ground and underground areas.

[0075] Please see Figure 8 The hazard tracing and dynamic prediction module includes:

[0076] The hazard tracing unit, based on the preprocessed standardized underground radar digital signals output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data (such as regional soil layer distribution and groundwater level data), analyzes the core causes of collapse hazards (such as loose underground soil, water-rich erosion, pipeline leakage, etc.) through tracing algorithms. At the same time, it associates parameters such as hazard location, depth, and size to clarify the correspondence between causes and hazard characteristics, and outputs a collapse hazard tracing report and tracing credibility parameters.

[0077] The dynamic hazard prediction unit, based on the collapse hazard cause tracing report and tracing credibility parameters output by the hazard cause tracing unit, combined with meteorological data (such as rainfall and temperature) and regional past hazard evolution data, dynamically predicts the development trend of collapse hazards (such as the speed of hazard expansion, possible secondary hazards, and changes in risk level) through a time-series prediction algorithm. At the same time, it sets the prediction period and early warning threshold, and outputs the dynamic hazard prediction results, prediction early warning thresholds, and periodic prediction reports.

[0078] The prevention and control suggestion generation unit generates targeted and implementable prevention and control suggestions for collapse hazards (such as local reinforcement, drainage treatment, and retesting frequency in key areas) based on the dynamic prediction results of the hidden dangers, prediction and early warning thresholds, and periodic prediction reports output by the dynamic prediction unit of the hidden dangers. At the same time, it transmits the prevention and control suggestion instructions to the intelligent control module and transmits the standardized prevention and control suggestion documents to the results output module.

[0079] Please see Figure 9 The data transmission module includes:

[0080] The data encoding unit, based on the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the result output module, and the hidden danger tracing and dynamic prediction module, classifies and encodes different types of data, and uses an encryption encoding algorithm to ensure that the data is not leaked or lost during transmission; at the same time, it converts the data into a format that adapts to the transmission protocol, and outputs the encrypted and encoded classified data and data transmission format identifier.

[0081] The internal data interaction unit, based on the encrypted and encoded classification data and data transmission format identifier output by the data encoding unit, adopts a high-speed transmission protocol to realize real-time data interaction between the multi-channel collaborative acquisition module, AI real-time interpretation module, intelligent control module, results output module, and hidden danger tracing and dynamic prediction module; at the same time, it monitors the data interaction status, and if data loss or delay occurs, it generates a retransmission instruction and outputs encrypted data of real-time interaction and data interaction status feedback.

[0082] The external data transmission unit transmits the encrypted and encoded classified data and data transmission format identifier output by the data encoding unit to an external terminal (such as a staff handheld terminal) or a smart city management and control platform (such as a smart pipeline platform) via wired / wireless transmission. At the same time, it receives control commands (such as parameter adjustment commands and detection task commands) from the external terminal or management and control platform and transmits them to the intelligent control module.

[0083] In practical use, the specific steps are as follows:

[0084] S1, through the multi-channel collaborative acquisition module, initializes multi-channel parameters according to the detection scenario, generates synchronization and anti-interference signals, and simultaneously acquires and preprocesses underground radar reflection signals, outputting preprocessed standardized underground radar digital signals.

[0085] S2, through the AI ​​real-time interpretation module, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module, loads and initializes the interpretation model, and at the same time extracts signal features, identifies potential collapse hazards and optimizes accuracy, and outputs the optimized collapse hazard identification results and model optimization parameters;

[0086] S3, through the intelligent control module, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification result and model optimization parameters output by the AI ​​real-time interpretation module, monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module, and at the same time judges the collaborative logic, outputs control commands including parameter adjustment, antenna raising and lowering, channel switching and execution feedback data of each control action;

[0087] S4, through the results output module, integrates and detects the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, while interpreting relevant information and performing standardized processing to realize the results visualization, local storage and export / import.

[0088] S5, through the hidden danger tracing and dynamic prediction module, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hidden danger identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data, the hidden danger cause tracing, development trend prediction and prevention and control suggestions are generated.

[0089] S6, through the data transmission module, the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the result output module and the hidden danger tracing and dynamic prediction module are sequentially classified, encrypted and encoded, internally interacted and externally transmitted, to achieve bidirectional data communication.

[0090] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of potential collapse hazards, comprising a mobile base (10), a ground-penetrating radar host (20) fixedly mounted on the top of the mobile base (10), a bracket (30) fixedly mounted on the front side of the mobile base (10), an electric lifting device (40) fixedly mounted on the top of the bracket (30), a support plate (50) fixedly mounted on the top of the electric lifting device (40), a MIMO array antenna (60) fixedly mounted on the top of the support plate (50), and a connector (70) for connecting to the rear of a vehicle provided on the front side of the bracket (30), characterized in that, It also includes: a multi-channel three-dimensional ground-penetrating radar system, the multi-channel three-dimensional ground-penetrating radar system comprising: The multi-channel collaborative acquisition module initializes multi-channel parameters according to the detection scenario, generates synchronization and anti-interference signals, acquires and preprocesses underground radar reflection signals, and outputs preprocessed standardized underground radar digital signals. The AI ​​real-time interpretation module includes an interpretation model initialization unit, a radar signal feature extraction unit, a collapse hazard real-time identification unit, and an interpretation accuracy optimization unit. The interpretation model initialization unit loads a preset collapse hazard AI interpretation model and, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module, initializes the core parameters of the interpretation model and outputs the initialized collapse hazard AI interpretation model. The radar signal feature extraction unit, based on the initialized collapse hazard AI interpretation model output by the interpretation model initialization unit, extracts features from the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and outputs the extracted underground medium radar signal feature set. The real-time identification unit, based on the extracted underground medium radar signal feature set output by the radar signal feature extraction unit, performs real-time matching and identification through an AI interpretation model to distinguish between normal underground medium and potential collapse hazards, classifies potential collapse hazards, calculates the specific location, size, and depth parameters of the hazards, and outputs real-time identification results and identification confidence levels. The interpretation accuracy optimization unit, based on the real-time identification results and identification confidence levels output by the real-time identification unit, judges the accuracy of the identification results, and simultaneously calls the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module to perform secondary feature extraction and matching on suspected hazards, optimizes the interpretation model parameters, and outputs optimized collapse hazard identification results and model optimization parameters. The intelligent control module monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. At the same time, it judges the collaborative logic and outputs control commands including parameter adjustment, antenna raising and lowering, channel switching and execution feedback data of each control action. The results output module integrates and detects the preprocessed standardized underground radar digital signals output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module. At the same time, it interprets relevant information and performs standardized processing to realize the visualization, local storage and export / import of results. The hazard tracing and dynamic prediction module, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data, completes the generation of hazard cause tracing, development trend prediction and prevention and control suggestions; The data transmission module sequentially classifies, encrypts, encodes, performs internal data interaction, and transmits external data to the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the results output module, and the hidden danger tracing and dynamic prediction module, thereby achieving bidirectional data communication.

2. The multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards according to claim 1, characterized in that, The multi-channel collaborative acquisition module includes: The multi-channel parameter configuration unit presets core parameters such as antenna main frequency, number of channels, channel spacing, antenna coverage width, effective detection depth, and detection speed according to the detection scenario. At the same time, it completes the initialization configuration of multi-channel acquisition parameters and outputs the initialized multi-channel acquisition parameters. The synchronization signal generation unit generates a synchronization control signal based on the initialized multi-channel acquisition parameters output by the multi-channel parameter configuration unit, controlling each channel to start acquisition work simultaneously; it also generates an anti-interference synchronization signal and outputs the multi-channel synchronous acquisition control signal and the anti-interference synchronization signal. The multi-channel radar signal acquisition unit, based on the multi-channel synchronous acquisition control signal and anti-interference synchronization signal output by the synchronization signal generation unit, synchronously acquires the radar reflection signal of the underground medium through the MIMO array antenna (60), and dynamically adjusts the acquisition sensitivity in combination with the real-time status of the electric lifting device (40); at the same time, it acquires in parallel through 8 / 10 channels and outputs the original underground radar reflection signal acquired by the multi-channel synchronous acquisition. The raw signal preprocessing unit, based on the underground radar reflection raw signal synchronously acquired by the multi-channel radar signal acquisition unit, uses filtering, denoising, and signal enhancement algorithms to remove interference signals from the raw signal, and performs normalization processing on the signal to convert the signal into a standardized digital signal, and outputs the preprocessed standardized underground radar digital signal.

3. A multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards according to claim 1, characterized in that, The intelligent control module includes: The data receiving and status monitoring unit monitors the working status of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module; at the same time, it monitors the working status of the electric lifting device (40), determines whether there is any abnormality, and outputs the working status data of the system / device. The collaborative logic judgment unit, based on the system / device working status data output by the data receiving and status monitoring unit, judges whether the collaborative working logic of the multi-channel collaborative acquisition module and the AI ​​real-time interpretation module is reasonable; if there is a logical deviation or abnormality, it generates adjustment instruction logic and outputs the collaborative working logic judgment result and adjustment instruction logic. The parameter and action control unit, based on the collaborative work logic judgment result and adjustment instruction logic output by the collaborative logic judgment unit, first controls the multi-channel collaborative acquisition module to adjust the acquisition parameters; then controls the electric lifting device (40) to lift and lower, adapting to the antenna height requirements of different detection scenarios; then controls the multi-channel switching to ensure that the acquisition coverage accuracy and detection speed are both taken into account, and outputs control instructions including parameter adjustment, antenna lifting and lowering, channel switching and execution feedback data of each control action; The emergency control unit, based on the parameters and the execution feedback data of each control action output by the action control unit, will immediately generate an emergency stop command to stop the operation of the relevant modules if a serious abnormality is detected; at the same time, it will generate an abnormal alarm signal to prompt staff to troubleshoot the fault.

4. A multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards according to claim 1, characterized in that, The output module includes: The results data integration unit, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, integrates the interpretation results with the acquired data, supplements auxiliary information, and outputs a complete detection results dataset. The results standardization processing unit, based on the complete detection results dataset output by the results data integration unit, formats the results data according to a preset standardized format to organize it into standardized tables and text reports; at the same time, it performs standardized rendering of the radar signal spectrum and outputs standardized detection results reports, standardized hidden danger information tables, and standardized radar signal spectrums. The results display and export unit, based on the standardized detection results report, standardized hidden danger information table, and standardized radar signal spectrum output by the results standardization processing unit, enables the visualization of the results; it also supports local storage and export of the results data.

5. A multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards according to claim 1, characterized in that, The hazard tracing and dynamic prediction module includes: The hazard tracing unit, based on the preprocessed standardized underground radar digital signal output by the multi-channel collaborative acquisition module and the optimized collapse hazard identification results and model optimization parameters output by the AI ​​real-time interpretation module, combined with geological background data, analyzes the core causes of collapse hazards through tracing algorithms, and at the same time associates the location, depth, and size parameters of the hazard to clarify the correspondence between the causes and the characteristics of the hazard, and outputs a collapse hazard tracing report and tracing credibility parameters; The dynamic prediction unit for hidden dangers, based on the tracing report and credibility parameters of the cause of the collapse hidden dangers output by the cause tracing unit, combined with meteorological data and past regional hidden danger evolution data, dynamically predicts the development trend of the collapse hidden dangers through a time-series prediction algorithm, while setting the prediction period and warning threshold, and outputting the dynamic prediction results, prediction warning thresholds and periodic prediction reports for the hidden dangers. The prevention and control suggestion generation unit generates targeted and implementable prevention and control suggestions for collapse hazards based on the dynamic prediction results of the hidden dangers, the prediction and early warning thresholds, and the periodic prediction reports output by the dynamic prediction unit of the hidden dangers. At the same time, it transmits the prevention and control suggestion instructions to the intelligent control module and transmits the standardized prevention and control suggestion documents to the results output module.

6. A multi-channel three-dimensional ground-penetrating radar capable of real-time AI interpretation of subsidence hazards according to claim 1, characterized in that, The data transmission module includes: The data encoding unit, based on the data output by the multi-channel collaborative acquisition module, the AI ​​real-time interpretation module, the intelligent control module, the result output module, and the hidden danger tracing and dynamic prediction module, classifies and encodes different types of data, and uses an encryption encoding algorithm to ensure that the data is not leaked or lost during transmission; at the same time, it converts the data into a format that adapts to the transmission protocol, and outputs the encrypted and encoded classified data and data transmission format identifier. The internal data interaction unit, based on the encrypted and encoded classification data and data transmission format identifier output by the data encoding unit, adopts a high-speed transmission protocol to realize real-time data interaction between the multi-channel collaborative acquisition module, AI real-time interpretation module, intelligent control module, results output module, and hidden danger tracing and dynamic prediction module; at the same time, it monitors the data interaction status and outputs encrypted data of real-time interaction and data interaction status feedback. The external data transmission unit transmits the encrypted and encoded classified data and data transmission format identifier output by the data encoding unit to an external terminal or the city's smart management and control platform via wired / wireless transmission; at the same time, it receives control commands from the external terminal or the management and control platform and transmits them to the intelligent control module.

Images