A mountain foundation pit construction monitoring optimization method, system, device and medium

Abstract

The application discloses a mountainous ground pit construction monitoring optimization method, system, device and medium, relates to the field of safe construction monitoring of power transmission and transformation projects, and comprises the following steps: constructing a construction behavior characteristic library containing the characteristic relationship between various modal data; acquiring multi-modal sensing data, identifying effective construction signals according to a modal filtering rule library, and locking a target construction tower leg foundation pit after path correction of the effective construction signals; extracting multi-modal features according to the position of the target construction tower leg foundation pit, performing feature matching and binding based on a mechanical feature mapping relationship, and acquiring the target construction signal; calculating the matching degree of the construction behavior characteristic library and the modal time sequence features in the target construction signal, determining the compliance of the construction behavior, and acquiring the behavior risk pre-value of the construction behavior according to a grid risk model; and responding to a hierarchical early warning mechanism according to the behavior risk pre-value. The application realizes more accurate construction positioning and accurate compliance judgment, and further improves the construction monitoring efficiency.

Description

Technical Field

[0001] This application relates to the field of safety construction monitoring for power transmission and transformation projects, specifically to an optimized method, system, equipment, and medium for monitoring construction of foundation pits in mountainous areas. Background Technology

[0002] In the construction of power transmission and transformation lines, power towers serve as crucial nodes for the transmission of electricity in the power grid, with many tower foundations located in challenging terrains such as mountains. The concealed nature of construction in mountainous areas makes supervision difficult, limiting on-site safety management for workers. Manual inspections are also constrained by labor costs and coverage limitations, making it difficult for inspectors to efficiently traverse all construction areas, resulting in "vacuum zones" in safety management. Video surveillance and vibration / acoustic wave detection are commonly used monitoring technologies on construction sites. However, due to the influence of terrain, video surveillance has blind spots and cannot accurately distinguish the type of work within the foundation pit. Vibration / acoustic wave detection cannot fully cover construction conditions below the detection benchmark or non-vibration methods, and is susceptible to environmental interference. Vibration detection deviations can lead to inaccurate location of sound sources and construction methods, thus affecting overall construction supervision and maintenance efficiency.

[0003] The patent "A Remote Monitoring Method and System for Intelligent Sensing of Situation, Vibration, and Orientation," publication number CN120538580A, discloses a method for identifying and acquiring dynamic construction behavior patterns. This method constructs a spatiotemporal dependent topology map of the dynamic construction behavior patterns using monitoring and tracing patterns and orientation tracking clusters. It infers the state potential energy of the spatiotemporal dependent topology map based on the evolution of vibration sensing signals and horizontal sensing data to determine the global construction sensing situation of the target monitoring area. Furthermore, it evaluates the phase space evolution trajectory of the global construction sensing situation using a spatial evolution evaluation criterion to obtain a stable node pattern. The method introduces a Lyapunov function to analyze the divergence of the stable node pattern and whether buried facilities are involved, thereby controlling the remote monitoring equipment to issue monitoring and early warning signals. However, this solution only compensates for vibration source errors caused by equipment attitude deviations by correcting the equipment attitude. It lacks multimodal sensing and terrain adaptation compensation for complex mountainous construction environments, and vibration signal errors still exist, resulting in poor adaptability of the solution. Summary of the Invention

[0004] The purpose of this application is to address the problem that existing construction monitoring technologies, when dealing with complex mountainous scenarios, suffer from limited perception dimensions and lack terrain adaptation compensation, making it impossible to accurately identify the compliance of construction activities. This application proposes an optimized method, system, equipment, and medium for monitoring construction in mountainous foundation pits. By using multimodal perception data and path correction, this method enables refined identification of the compliance of construction activities and prediction of risk levels, thereby achieving more comprehensive perception, accurate construction positioning, and precise compliance judgment.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, embodiments of this application provide an optimized method for monitoring construction of foundation pits in mountainous areas, the method comprising: A construction behavior feature library containing the characteristic relationships between various modal data is constructed; multimodal perception data is acquired, and valid construction signals are identified according to the modal filtering rule library. After path correction of the valid construction signals, the target construction tower leg foundation pit is located; multimodal features are extracted according to the location of the target construction tower leg foundation pit, and feature matching and binding are performed based on mechanical feature mapping relationship to obtain the target construction signal; the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal is calculated to determine the compliance of construction behavior and obtain the behavioral risk prediction value of construction behavior according to the gridded risk model; a graded early warning mechanism is implemented based on the behavioral risk prediction value.

[0006] This solution utilizes multimodal sensing data to eliminate interference through cross-verification of multiple sources, avoiding misjudgments caused by single sensing. Path correction of valid construction signals corrects the traditional assumption of linear sound wave propagation, compensating for terrain deviations and the impact of terrain on features, further ensuring the positioning accuracy of the target construction tower leg pit and providing an accurate location benchmark for construction compliance identification. Feature matching and binding based on mechanical feature mapping relationships separates multiple signal sources, acquiring the target construction signal and avoiding signal overlap from multiple machines operating simultaneously. This improves the utilization rate of sensing data and enhances the accuracy of construction machinery type identification. This allows for accurate comparison of the current target construction signal based on the correspondence between machinery type and corresponding features using a construction behavior feature library, ultimately accurately determining construction compliance and providing risk warnings based on the target construction signal. Furthermore, the construction and application of the construction behavior feature library, combined with dynamic matching methods, enables continuous tracking of construction behavior, improving the dynamic switching recognition rate and thus enhancing the efficiency and accuracy of construction behavior monitoring.

[0007] Preferably, the construction of a construction behavior feature library containing the characteristic relationships between various modal data includes: labeling historical multimodal time-series data according to construction behavior labels, wherein the construction behavior labels include the correspondence between construction type, intensity, and corresponding modal data features; normalizing the labeled historical multimodal time-series data, and using a hybrid architecture of TCN network and local attention to extract modal time-series fusion features and key modal features of construction behavior types to learn, thereby obtaining the mapping probability between modal time-series feature vectors and construction behavior categories; establishing a feature memory based on a vector index library, and determining construction behavior feature templates according to the mapping probability between modal time-series feature vectors and construction behavior categories to complete the initialization of the feature memory; storing dynamically acquired modal time-series feature vectors, corresponding construction behavior labels, and time-series metadata into the feature memory, completing the update of the feature memory, and using the updated feature memory as the construction behavior feature library.

[0008] Preferably, the step of acquiring multimodal sensing data, identifying valid construction signals based on a modal filtering rule base, and locking the target construction tower leg foundation pit after path correction of the valid construction signals includes: the multimodal sensing data includes at least acoustic signals, vibration signals, infrared data, construction machinery tags, and meteorological sensing data, with each modal sensing data item forming a modal sensing group based on a time series; extracting the critical value of the valid construction signal based on the modal filtering rule base and identifying and filtering the modal sensing group of non-monitoring areas, false thermal targets, and environmental interference items using modal interconnection rules, and acquiring valid construction signals with the remaining modal sensing groups; correcting the deviation of the valid construction signals, and determining the target construction tower leg foundation pit based on the corrected valid construction signals.

[0009] Preferably, the deviation correction of the effective construction signal and the determination of the target construction tower leg pit based on the corrected effective construction signal includes: acquiring three-dimensional terrain parameters of the central pile and each tower leg, including slope, elevation difference, and distribution of surface obstacles, to determine the terrain correction factor; determining the wind speed correction factor based on the angle between the wind direction and the direction of sound wave movement, and correcting the sound wave signal by combining the terrain correction factor and the weighted average sound speed of the sound wave, and simulating the propagation path of the sound wave from the source to the sensor using the ray tracing method; determining the standard propagation time of each path based on the propagation path of the sound wave from the source to the sensor, and performing anomaly verification on the time difference of the actual sound wave from the source to each group of microphones, calculating the spatial coordinates of the construction source target based on the effective time difference and the corrected sound wave signal; calculating the straight-line distance between the construction source target and the center of each tower leg pit based on the spatial coordinates of the construction source target, and determining the target construction tower leg pit by combining the coordinates of the construction machinery.

[0010] Preferably, the process involves calculating the straight-line distance between the spatial coordinates of the construction seismic source target and the center of each tower leg pit, and determining the target construction tower leg pit by combining the coordinates of the construction machinery. This includes: calculating the straight-line distance between the spatial coordinates of the construction seismic source target and the center coordinates of four preset tower leg pits, and taking the tower leg corresponding to the minimum distance as the initial judgment construction tower leg pit; extracting the coordinates of the construction machinery from the effective construction signal, calculating the distance between the coordinates of the construction machinery and the coordinates of the initial judgment construction tower leg pit, and if the distance is within a preset pit range, then the initial judgment construction tower leg pit is taken as the target construction tower leg pit.

[0011] Preferably, the step of extracting multimodal features based on the location of the target construction tower leg pit, performing feature matching and binding based on the mechanical feature mapping relationship, and obtaining the target construction signal includes: extracting construction machinery tag IDs and construction machinery coordinates based on the valid construction signals to filter out construction machinery tags within the monitoring area to which the target construction tower leg pit belongs; extracting multimodal sensing features from the valid construction signals in the spatial domain of each construction machinery tag; comparing the multimodal sensing features with the mechanical features in the mechanical feature mapping relationship to calculate a first feature matching degree, wherein the mechanical feature mapping relationship includes construction machinery, construction machinery tags, and mechanical features bound to the construction machinery; and binding the multimodal sensing features with a first feature matching degree greater than a feature matching degree threshold to the corresponding construction machinery tags and construction machinery to obtain the target construction signal.

[0012] Preferably, the step of calculating the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal to determine the compliance of the construction behavior and obtain the behavioral risk prediction value of the construction behavior according to the gridded risk model includes: extracting vibration temporal features, infrared temporal features, and construction machinery coordinate temporal features from the target construction signal; calculating the second feature matching degree between the vibration temporal features, infrared temporal features, and construction machinery coordinate temporal features and each feature template in the construction behavior feature library based on Euclidean distance; determining the corresponding construction behavior category and judging whether the construction behavior violates the construction specifications based on the second feature matching degree; determining the monitoring grid to which the current construction behavior belongs based on the construction machinery coordinates; calculating the grid risk value of the monitoring grid based on the weight of each modal temporal feature; and using it as the behavioral risk prediction value of the construction behavior.

[0013] Secondly, embodiments of this application provide a monitoring and optimization system for mountain foundation pit construction, comprising: a memory module for constructing a construction behavior feature library containing characteristic relationships between various modal data; an acquisition module for acquiring multimodal sensing data, identifying valid construction signals according to a modal filtering rule library, and locking the target construction tower leg foundation pit after path correction of the valid construction signals; a matching module for extracting multimodal features based on the location of the target construction tower leg foundation pit, performing feature matching and binding based on mechanical feature mapping relationships, and acquiring the target construction signal; a judgment module for calculating the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal, determining the compliance of the construction behavior, and obtaining the behavioral risk prediction value of the construction behavior according to a gridded risk model; and an early warning module for responding to a graded early warning mechanism based on the behavioral risk prediction value.

[0014] Thirdly, embodiments of this application provide a computer device, including: a processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other through the communication bus; the memory is used to store computer programs; and the processor, when executing the program stored in the memory, implements the steps of the method described in the first aspect above.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect above.

[0016] The beneficial effects of this application are: 1. This application achieves signal complementarity and modal verification to eliminate interference and false signals through multimodal data collaboration, providing multidimensional basis for judging the compliance of construction behavior in complex mountain foundation pits, thereby reducing the misjudgment rate; and constructs a construction behavior feature library to clarify the types of construction machinery and related modal perception features, which is used to perform direct feature matching processing on the real-time acquired multimodal perception data, quickly judge whether the construction behavior is compliant, and avoid compliance misjudgment caused by single-dimensional judgment; 2. Due to environmental biases in the data of each modality during multimodal data collaboration, in order to avoid the environmental biases of modal data and the impact of terrain on construction methods, this application obtains a terrain correction factor through land formation parameters, and then obtains the sound velocity adapted to the terrain by combining the correction term of sound waves. Based on the ray tracing method, the path deviation of the seismic source location is corrected, providing an accurate sound velocity basis for seismic source location, thereby improving the terrain adaptability of construction behavior compliance judgment and ensuring the accuracy of risk warning. 3. This application identifies the specific machinery associated with multimodal sensing data by binding construction machinery tags with coordinates and corresponding mechanical features. By comparing with the mechanical features of the construction machinery benchmark, aliased signals are separated, making the signal attribution clear. Even when multiple machines are operating simultaneously, it can accurately locate which equipment is performing which type of construction on which tower leg. It has multi-machine monitoring capabilities and avoids the problem of not being able to locate specific equipment during the compliance identification process of construction behavior due to the inability to distinguish the signal aliasing of multiple machines operating simultaneously. Furthermore, through the mechanical tags, mechanical coordinates, and bound features in the mechanical feature mapping relationship, construction behavior can be traced, that is, illegal construction can be accurately traced back to specific equipment, further improving the efficiency of construction behavior monitoring. Attached Figure Description

[0017] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0018] Figure 1 A flowchart illustrating an optimized method for monitoring construction of foundation pits in mountainous areas, provided as an embodiment of this application.

[0019] Figure 2 This is a flowchart of step S1 in a method for optimizing monitoring of construction pits in mountainous areas, provided in an embodiment of this application.

[0020] Figure 3 The flowchart of step S2 in the optimization method for monitoring construction of mountain foundation pits provided in the embodiments of this application is shown.

[0021] Figure 4 This is a schematic flowchart of a method for step S23 provided in an embodiment of this application.

[0022] Figure 5 This is a flowchart of step S3 in a method for optimizing monitoring of construction pits in mountainous areas, provided in an embodiment of this application.

[0023] Figure 6 The flowchart of step S4 in the optimization method for monitoring construction of mountain foundation pits provided in the embodiments of this application is shown.

[0024] Figure 7 This is a schematic diagram of a monitoring and optimization system module for mountain foundation pit construction, provided as an embodiment of this application.

[0025] Figure 8 A schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely one preferred embodiment of this application and are only used to explain this application. They do not limit the scope of protection of this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] Example 1: As Figure 1 As shown, an optimized method for monitoring construction of foundation pits in mountainous areas includes the following steps: S1. Construct a feature library of construction behavior that includes the characteristic relationships between various modal data.

[0028] In an optional embodiment, such as Figure 2 As shown, step S1 includes: S11. Label the historical multimodal time series data according to the construction behavior labels. The construction behavior labels include the construction type, intensity and the corresponding relationship of the modal data features. S12. After normalizing the historical multimodal time-series data labeled with tags, a hybrid architecture of TCN network and local attention is used to extract modal time-series fusion features and key modal features of construction behavior types to learn and obtain the mapping probability between modal time-series feature vectors and construction behavior categories. S13. Establish a feature memory based on the vector index library, and determine the construction behavior feature template according to the mapping probability between the modal temporal feature vector and the construction behavior category to complete the feature memory initialization; S14. Store the dynamically acquired modal temporal feature vectors, corresponding construction behavior labels, and temporal metadata into the feature memory, complete the feature memory update, and use the updated feature memory as the construction behavior feature library.

[0029] In this embodiment, construction behavior labels are established based on construction category, achievable intensity, and core modal time series features within a certain time period. Construction categories include parent and child categories. When labeling historical multimodal time series data, a three-level labeling system is used for multiple construction behaviors, following the parent category-child category-intensity level. The construction behavior labels are shown in Table 1. Table 1 Construction Behavior Labels ; Standardization is performed on the historical multimodal time-series data that has been annotated to reduce the data processing dimensionality and improve the efficiency of building the construction behavior feature library. Specifically, the disconnected data of construction machinery labels are completed by linear interpolation, and the three-modal frame alignment between the same feature vector corresponding to sound waves, vibrations and label data is achieved by synchronizing the timestamps. Z-score standardization is performed on the three-modal historical data of the frame alignment to eliminate the difference in dimensions.

[0030] Specifically, based on the fusion architecture of Temporal Convolutional Network (TCN) and Local Attention, key temporal features of construction behavior are extracted and fused from the standardized modal historical data to obtain a 16-dimensional temporal feature embedding vector. This vector is then used for model training. Finally, the model outputs the probabilities of various construction behaviors. The 16-dimensional temporal feature embedding vector and the corresponding construction machinery label are used as the core unit of the feature memory. In the subsequent modal temporal feature matching process of target construction signals, this vector is directly matched.

[0031] Specifically, a feature memory is constructed using a vector retrieval library. This feature memory stores 16-dimensional temporal feature embedding vectors, construction machinery tags, and temporal metadata. The temporal metadata includes modal data acquisition time, foundation pit geological parameters, and meteorological data. Specifically, 16-dimensional temporal feature embedding vectors with a prediction probability exceeding 99% are selected as feature templates based on the construction behavior probabilities output by the model, thus initializing the feature memory. Depending on different construction scenarios, the predicted behavior probabilities of new modal temporal data are obtained, and feature vectors meeting the probability criteria are stored in the feature memory, enabling dynamic updates to the feature memory. The updated feature memory serves as the construction behavior feature library.

[0032] In this embodiment, the dynamic updating of the vector retrieval library and feature memory library allows for adaptation to new construction machinery or geological conditions without retraining modal data. The established construction behavior feature library clarifies the standards for judging the compliance of construction behaviors, thereby providing accurate feature matching templates for compliance judgments and ensuring the accuracy and reliability of the construction compliance judgment results.

[0033] S2. Acquire multimodal sensing data, identify valid construction signals according to the modal filtering rule base, and lock the target construction tower leg foundation pit after path correction of the valid construction signals.

[0034] In an optional embodiment, such as Figure 3 As shown, step S2 includes: S21. The multimodal sensing data includes at least acoustic signals, vibration signals, infrared data, construction machinery tags, and meteorological sensing data. Each piece of modal sensing data forms a modal sensing group based on a time series. S22. Based on the modal filtering rule base, extract the critical value of the effective construction signal and identify and filter the modal perception group of non-monitoring area, false thermal target and environmental interference item by modal interconnection rule, and obtain the effective construction signal with the remaining modal perception group; S23. Correct the deviation of the effective construction signal, and determine the target construction tower leg foundation pit based on the corrected effective construction signal.

[0035] In this embodiment, the construction machinery tags are collected by the UWB ultra-wideband positioning module. The UWB ultra-wideband positioning module includes three anchor points, which are distributed in a triangle at the edge of the monitoring area and are equidistant from the center pile of the foundation pit. Each construction machinery is identified by a unique tag. When the construction machinery enters the monitoring area, the UWB ultra-wideband positioning module collects the UWB tag. The UWB tag includes the construction machinery tag ID and the machinery type.

[0036] In some embodiments, the modal filtering rule base is constructed based on modal sensing data constraints generated by non-construction machinery. The rule base includes threshold values ​​for meteorological and acoustic data collected in the construction scenario. For example, when the wind speed is ≥5 and there is no construction machinery tag signal or infrared thermal target, it is determined to be natural interference. When the boundary acoustic sensor detects that the sound wave propagation direction comes from outside the monitoring area and the construction machinery tag signal is within the monitoring area, and the coordinates of the infrared thermal target and the construction machinery do not match, it is determined to be an external environmental interference item. When the infrared thermal target moves at a speed ≥5m / s (such as a vehicle) and there is no corresponding vibration signal, it is determined to be a false thermal target. When the sound wave is ≥60dB, the vibration is ≥0.1g, the infrared thermal target is ≥0.2m² and the duration is ≥5s, and the construction machinery tag is within the monitoring area, the modal data is marked as a valid construction signal.

[0037] Understandably, in complex scenarios, natural disturbances (such as heavy rain and falling rocks), external disturbances (such as distant road construction), and false thermal targets (such as passing vehicles) are easily misinterpreted as construction signals. Relying solely on single-modal data to filter out some of these disturbances has limited effectiveness. Therefore, by constructing a modal filtering rule base, the above three types of core disturbances are proactively excluded, thereby reducing the misjudgment rate. Simultaneously, the detected multimodal sensing data can serve as a preliminary judgment on the existence of illegal construction activities. If modal sensing data excluding the aforementioned disturbances is detected without a construction plan, it can be identified as illegal construction, triggering a high-risk warning.

[0038] In an optional embodiment, such as Figure 4 As shown, step S23 includes: S231. Obtain the three-dimensional terrain parameters of the central pile and each tower leg, including slope, elevation difference and distribution of surface obstacles, in order to determine the terrain correction factor; S232. Determine the wind speed correction factor based on the angle between the wind direction and the direction of sound wave movement, and correct the sound wave signal by combining the terrain correction factor and the weighted average sound speed of the sound wave, and use the ray tracing method to simulate the propagation path of the sound wave from the source to the sensor. S233. Determine the standard propagation time of each path based on the propagation path of the sound wave from the source to the sensor, and perform anomaly verification on the time difference of the actual sound wave from the source to each group of microphones. Calculate the spatial coordinates of the construction source target based on the effective time difference and the corrected sound wave signal. S234. Calculate the straight-line distance between the target construction seismic source and the center of each tower leg pit based on the spatial coordinates of the target construction source, and determine the target construction tower leg pit by combining the coordinates of the construction machinery.

[0039] In this embodiment, the acoustic signal is collected based on four sets of MEMS microphones along the four legs of the tower base. The three-dimensional terrain parameters of the four sets of MEMS microphones and the potential seismic source area (20m around each tower leg pit) are extracted. These parameters include the average slope from the i-th microphone to the j-th tower leg pit, the vertical height difference between the microphone and the tower leg pit (i.e., the elevation difference), and obstacles (such as piles of rocks and bushes) on the acoustic propagation path, with corresponding attenuation coefficients. For example, the rock coefficient is 1.2, the vegetation coefficient is 1.05, and the coefficient is 1.0 when there are no obstacles.

[0040] Furthermore, based on slope, elevation difference, and the distribution of surface obstacles, a terrain correction factor is determined, and the calculation formula is expressed as follows: (1); in, Indicates the terrain correction factor. This represents the slope parameter; the steeper the slope, the greater the slope. The smaller, This represents the average slope from the i-th microphone to the j-th tower leg pit. This indicates the distribution of obstacles on the ground; the denser the obstacles, the better. The larger.

[0041] The wind speed correction factor is expressed as: (2); in, The angle between the wind direction and the direction of sound wave propagation is 0° when the wind is downwind and 180° when the wind is upwind. The corrected speed of sound is expressed as: (3); Ray tracing was used to simulate the actual path of sound waves from the potential seismic source to the four microphones. Specifically, starting from the center of each tower leg pit and ending at the microphone, a polygonal propagation path L was generated based on the aforementioned terrain parameters, avoiding obstacles and conforming to the terrain undulations. The actual path length was calculated, and the propagation time of each path was recorded. It is used for time difference verification.

[0042] Furthermore, the time difference is calculated based on the time it takes for the sound to reach the four sets of microphones, and the rationality of the actual time difference is verified by combining the propagation time of each path. If the absolute value of the difference between the two is greater than the preset time difference threshold, it is determined to be a data anomaly, and multimodal sensing data is re-collected.

[0043] In some examples, the time difference is calculated based on the time it takes for the sound to reach the four microphones respectively. For example, using microphone 1 (S1, corresponding to the direction of tower leg A) as a reference, the time difference between the sound waves arriving at each microphone is calculated based on the actual acquisition time of the sound waves at S1 and S2, S1 and S3, and S1 and S4 respectively. , , .

[0044] In an optional embodiment, step S234 includes: Calculate the straight-line distance between the spatial coordinates of the construction seismic source target and the center coordinates of the four pre-set tower leg foundation pits, and take the tower leg corresponding to the minimum distance as the initial judgment of the construction tower leg foundation pit; Extract the coordinates of the construction machinery from the valid construction signal, calculate the distance between the coordinates of the construction machinery and the coordinates of the initially judged construction tower leg pit, and if the distance is within the preset pit range, then the initially judged construction tower leg pit is taken as the target construction tower leg pit.

[0045] S3. Extract multimodal features based on the location of the target construction tower leg pit, perform feature matching and binding based on mechanical feature mapping relationship, and obtain the target construction signal.

[0046] In an optional embodiment, such as Figure 5 The aforementioned step S3 includes: S31. Extract the construction machinery tag ID and construction machinery coordinates based on the valid construction signal to filter out the construction machinery tags in the monitoring area to which the target construction tower leg foundation pit belongs. S32. Extract the multimodal sensing features from the valid construction signals within the spatial domain of each construction machinery tag; S33. Compare the multimodal perception features with the mechanical features in the mechanical feature mapping relationship, and calculate the first feature matching degree, wherein the mechanical feature mapping relationship includes construction machinery, construction machinery tags, and mechanical features bound to construction machinery; S34. Bind the multimodal sensing features with the first feature matching degree greater than the feature matching degree threshold to the corresponding construction machinery label and construction machinery to obtain the target construction signal.

[0047] In some embodiments, the source coordinates are set as P(x,y,z). Based on the principle that the sound wave propagation distance difference = sound speed × time difference, a system of nonlinear equations is established: (4); (5); (6); in, , , and Let each of the four tower legs represent a coordinate; by solving the above system of equations, the precise coordinates of the earthquake source can be obtained. .

[0048] If there are construction machinery tag signals near the epicenter, then calculate the precise coordinates of the epicenter. If the distance from the label coordinates is less than or equal to the mechanical operating radius, the seismic source coordinates are considered valid. Further, by associating the seismic source coordinates with the locations of each tower leg foundation pit, the specific tower leg where construction occurred is located. This specifically includes: calculating the center coordinates of the four pre-set tower leg foundation pits based on the system's settings. The straight-line distance to the center of the four tower leg foundation pits is used to obtain the tower leg corresponding to the minimum distance, which is used as the initial judgment tower leg foundation pit. If there are mechanical tags around the tower leg foundation pit, the construction machinery type, the bound mechanical features and the mechanical coordinates are extracted based on the mechanical feature mapping relationship to further confirm the judgment result. When the tag coordinates are within the effective distance range of the initial judgment tower leg foundation pit, the initial judgment tower leg foundation pit is used as the target tower leg foundation pit.

[0049] In some embodiments, multimodal sensing features are extracted from the valid construction signals in the spatial domain of each construction machinery tag. These features refer to the valid construction signals within a certain range around the current coordinates of the machinery that uniquely identifies each construction machinery tag. Three features are extracted: vibration frequency or vibration peak value, infrared thermal target area or temperature, and sound wave peak value or sound wave frequency. The extracted features are then compared with the mechanical features in the mechanical feature mapping relationship.

[0050] Specifically, the mechanical feature mapping relationship refers to a relationship lookup table that includes the construction machinery tag ID, machinery type and mechanical feature binding relationship. The relationship lookup table stores the binding data of each machinery type and its corresponding tag ID and machinery operation features. The data storage format is: tag ID: machinery type: machinery feature, for example, Tag001 (tag ID): rotary drilling rig: vibration 10-20Hz + infrared 0.5-1㎡.

[0051] It should be noted that the extracted multimodal sensing features are compared with the mechanical features in the mechanical feature mapping relationship to calculate the first feature matching degree. This is used to separate multiple signal sources, thereby solving the problem of inaccurate construction behavior recognition caused by the aliasing of multiple signal sources. For example, if the multimodal sensing features of construction machinery tag Tag001 are found to have a matching degree of 92% with the mechanical features bound to "rotary drilling rig", and the feature matching degree threshold is 80%, then the multimodal sensing features are bound to construction machinery tag Tag001 and rotary drilling rig, and then the rotary drilling rig signal, pneumatic pick signal, etc., in the valid construction signals are separated.

[0052] In this embodiment, the specific machinery associated with multimodal sensing data is identified by binding construction machinery tags with coordinates and corresponding mechanical features. By comparing with the mechanical features of the construction machinery baseline, aliased signals are separated, making the signal attribution clear. Even when multiple machines are operating simultaneously, it is possible to accurately locate which equipment is performing which type of construction on which tower leg, providing multi-machine monitoring capabilities and solving the problem of multimodal data fragmentation. Furthermore, through the machinery tags, machinery coordinates, and bound features in the machinery feature mapping relationship, construction behavior can be traced, meaning that illegal construction can be accurately traced back to the specific equipment, further improving the efficiency of construction behavior monitoring.

[0053] S4. Calculate the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal, determine the compliance of the construction behavior, and obtain the behavioral risk prediction value of the construction behavior based on the gridded risk model.

[0054] In an optional embodiment, such as Figure 6 As shown, step S4 includes: S41. Extract vibration timing features, infrared timing features, and construction machinery coordinate timing features based on the target construction signal; S42. Calculate the second feature matching degree between the vibration time series features, infrared time series features, and construction machinery coordinate time series features and each feature template in the construction behavior feature library based on Euclidean distance; S43. Determine the corresponding construction behavior category and judge whether the construction behavior violates the construction specifications based on the second feature matching degree; S44. Determine the monitoring grid to which the current construction behavior belongs based on the coordinates of the construction machinery, calculate the grid risk value of the monitoring grid based on the weight of the temporal characteristics of each modality, and use it as the behavioral risk prediction value of the construction behavior.

[0055] In some embodiments, vibration time-series features, infrared time-series features, and construction machinery coordinate time-series features are extracted from each target construction signal according to a time window. Then, Euclidean distance is used to calculate the second feature matching degree between the current time-series feature and each feature template in the construction behavior feature library. The formula for calculating the second feature matching degree is as follows: (7); in, The second feature matching degree is represented by V, which represents the vibration time difference, IR, which represents the infrared time difference, and D, which represents the construction machinery coordinate time difference. max This represents the maximum temporal difference, which is the difference between the current feature and the template feature. The maximum temporal difference is a preset threshold that can be dynamically set based on scene conditions.

[0056] It should be noted that construction activities change dynamically over time. If only instantaneous signals from various modal sensing data are relied upon to identify construction activities, it is easy to miss construction switching behaviors due to signal interruptions or sudden changes, especially low-intensity illegal construction activities at night. Therefore, this embodiment constructs and applies a construction activity feature library to enable continuous tracking of construction activities by combining it with a dynamic matching degree method calculated using second feature matching. By cross-verifying the signals from various modalities, the recognition rate of dynamic construction switching is improved, thereby enhancing the efficiency and accuracy of construction activity monitoring.

[0057] In some examples, if the second feature matching degree is greater than the preset second feature matching threshold, it is determined to be the corresponding construction behavior; if different feature templates are matched for at least two consecutive time windows, it is determined to be a construction behavior switch; if it is not a planned time period, it is matched with the feature template of manual excavation type, and it is initially determined to be illegal construction.

[0058] S5. A graded early warning mechanism based on the predicted behavioral risk value.

[0059] In some optional implementations, the corresponding risk prediction response tiered early warning mechanism includes: Based on the predicted behavioral risk value, the corresponding construction risk level is obtained, and then an early warning response is initiated according to the early warning rules corresponding to the construction risk level.

[0060] In this embodiment, to assess the global risk superposition of multiple regional construction projects on buried facilities (e.g., low-risk construction on tower legs 1 and 2, which, after superposition, leads to excessive stress in the intermediate cable, lacking a global perspective), a 5m×5m gridded risk model is constructed by introducing buried facility topology data. Based on the baseline weights of vibration intensity (30%), the straight-line distance from the real-time coordinates of construction machinery to the buried facility (25%), infrared intensity (20%), and construction type (25%), the risk value of each grid is calculated (out of 100 points). Adjacent high-risk grids covering the same facility weak point are judged as risk superposition, realizing the quantitative assessment and superposition identification of global risks, thereby improving the accuracy of risk warning for weak points of buried facilities and avoiding global safety hazards caused by single-point assessment.

[0061] Furthermore, based on the coordinates of the construction machinery, the current construction activity is assigned to a 5m×5m grid (e.g., Tag001 coordinates (x1, y1) belong to grid G01). The predicted risk value for the activity is calculated according to the aforementioned weights, and compared with the risk threshold to determine the corresponding risk level. Risk levels include low risk, medium risk, high risk, and overlapping risk. If adjacent grids are all high-risk and cover the same weak point in buried infrastructure (e.g., cable joints), it is considered overlapping risk. Different risk levels trigger different colored indicator lights and sound alerts, and generate early warning information.

[0062] In some examples, a behavioral risk prediction score below 50 is considered low risk, with a local green indicator light remaining constantly on, data recorded in the background, and no personnel notified; a behavioral risk prediction score between 50 and 79 is considered medium risk, with a local yellow indicator light flashing and an audible alert issued, generating a warning message containing specific grid coordinates, risk level, and construction type and sending it to maintenance personnel; a behavioral risk prediction score above 80 is considered high risk, with a local red flashing light and a high-pitched alarm issued, generating a warning message containing grid coordinates, UWB coordinate trajectory, and infrared thermal image and sending it to the construction management platform.

[0063] Example 2, as Figure 7 As shown in the figure, this application provides a monitoring and optimization system for mountain foundation pit construction, including: The memory module is used to build a feature library of construction behavior that contains the characteristic relationships between data of various modalities; The acquisition module is used to acquire multimodal sensing data, identify valid construction signals according to the modal filtering rule base, and lock the target construction tower leg foundation pit after path correction of the valid construction signals. The matching module is used to extract multimodal features based on the location of the target construction tower leg pit, perform feature matching and binding based on mechanical feature mapping relationship, and obtain the target construction signal; The judgment module is used to calculate the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal, determine the compliance of the construction behavior, and obtain the behavioral risk prediction value of the construction behavior based on the gridded risk model. The early warning module is used to respond to a graded early warning mechanism based on the predicted risk value of the behavior.

[0064] In this embodiment, acquiring multimodal sensing data allows for the elimination of interfering data through cross-verification of multiple sources, avoiding misjudgments caused by single sensing. Path correction of valid construction signals corrects the traditional assumption of linear propagation of sound waves, thereby compensating for terrain deviations and the influence of terrain on features, further ensuring the positioning accuracy of the target construction tower leg pit and providing an accurate location benchmark for construction compliance identification. Feature matching and binding based on mechanical feature mapping relationships allows for the separation of multiple signal sources and the acquisition of target construction signals, avoiding signal overlap from multiple machines operating simultaneously, improving the utilization rate of sensing data, and enhancing the accuracy of construction machinery type identification. This facilitates accurate comparison of the current target construction signal based on the correspondence between machinery type and corresponding features, ultimately accurately determining construction compliance and providing risk warnings based on the target construction signals. Furthermore, constructing and applying a construction behavior feature library allows for continuous tracking of construction behavior using dynamic matching methods, improving the dynamic switching recognition rate and ultimately enhancing the efficiency and accuracy of construction behavior monitoring.

[0065] This application also provides a computer device, such as... Figure 8 As shown, it includes a processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements an optimized method for monitoring construction in mountainous foundation pits.

[0066] The communication bus mentioned in the above computer equipment can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0067] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0068] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0069] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements an optimization method for monitoring construction of mountain foundation pits.

[0070] The above-described embodiments are preferred embodiments of this application and are not intended to limit the specific scope of this application. The scope of this application includes but is not limited to the specific embodiments described above. All equivalent changes made in accordance with the shape, structure, and method of this application are within the protection scope of this application.

Claims

1. An optimized method for monitoring construction of foundation pits in mountainous areas, characterized in that: Includes the following steps: Construct a feature library of construction behavior that includes the characteristic relationships between various modal data; The process involves acquiring multimodal sensing data, identifying valid construction signals based on a modal filtering rule base, and then correcting the paths of the valid construction signals to locate the target construction tower leg foundation pit. The steps for correcting the paths of valid construction signals are as follows: acquiring the three-dimensional terrain parameters of the central pile and each tower leg, including slope, elevation difference, and distribution of surface obstacles, to determine the terrain correction factor; determining the wind speed correction factor based on the angle between the wind direction and the direction of sound wave movement; and correcting the sound wave signal by combining the terrain correction factor and the weighted average sound speed of the sound wave. Multimodal features are extracted based on the location of the target construction tower leg pit, and feature matching and binding are performed based on the mechanical feature mapping relationship to obtain the target construction signal. The steps of feature matching and binding based on the mechanical feature mapping relationship are as follows: extract the construction machinery tag ID and construction machinery coordinates based on the valid construction signal to filter out the construction machinery tags within the monitoring area to which the target construction tower leg pit belongs; extract the multimodal sensing features in the valid construction signal within the spatial domain of each construction machinery tag; compare the multimodal sensing features with the mechanical features in the mechanical feature mapping relationship to calculate the first feature matching degree, wherein the mechanical feature mapping relationship includes the construction machinery, the construction machinery tag, and the mechanical features bound to the construction machinery; bind the multimodal sensing features with the first feature matching degree greater than the feature matching degree threshold to the corresponding construction machinery tag and construction machinery. Calculate the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal, determine the compliance of the construction behavior, and obtain the behavioral risk prediction value of the construction behavior based on the gridded risk model; The behavioral risk prediction response graded early warning mechanism is based on this.

2. The optimized method for monitoring construction of foundation pits in mountainous areas according to claim 1, characterized in that: The construction of a construction behavior feature library containing the characteristic relationships between various modal data includes: Historical multimodal time series data are labeled based on construction behavior tags, which include the correspondence between construction type, intensity, and corresponding modal data features. After normalizing the historical multimodal time-series data labeled with tags, a hybrid architecture of TCN network and local attention is used to extract modal temporal fusion features and key modal features of construction behavior types to learn and obtain the mapping probability between modal temporal feature vectors and construction behavior categories. A feature memory is established based on a vector index library. The feature templates for construction behavior are determined according to the mapping probabilities between modal temporal feature vectors and construction behavior categories, so as to complete the initialization of the feature memory. The dynamically acquired modal temporal feature vectors, corresponding construction behavior labels, and temporal metadata are stored in the feature memory, the feature memory is updated, and the updated feature memory is used as the construction behavior feature library.

3. The optimized method for monitoring construction of foundation pits in mountainous areas according to claim 1, characterized in that: The process of acquiring multimodal sensing data, identifying valid construction signals based on a modal filtering rule base, and then fixing the target construction tower leg foundation pit after path correction of the valid construction signals includes: The multimodal sensing data includes at least acoustic signals, vibration signals, infrared data, construction machinery tags, and meteorological sensing data. Each piece of modal sensing data forms a modal sensing group based on a time series. Based on the modal filtering rule base, the critical value for extracting effective construction signals is used, and modal interconnection rules are used to identify and filter modal sensing groups that are not monitored, false thermal targets, and environmental interference items. The remaining modal sensing groups are then used to obtain effective construction signals. The effective construction signal is corrected for deviation, and the target construction tower leg foundation pit is determined based on the corrected effective construction signal.

4. The optimized method for monitoring construction of foundation pits in mountainous areas according to claim 3, characterized in that: The step of correcting the deviation of the effective construction signal and determining the target construction tower leg foundation pit based on the corrected effective construction signal includes: Obtain the three-dimensional terrain parameters of the central pile and each tower leg, including slope, elevation difference, and distribution of surface obstacles, in order to determine the terrain correction factor; The wind speed correction factor is determined based on the angle between the wind direction and the direction of sound wave movement. The sound wave signal is corrected by combining the terrain correction factor and the weighted average sound speed of the sound wave. The ray tracing method is used to simulate the propagation path of the sound wave from the source to the sensor. The standard propagation time of each path is determined based on the propagation path of sound waves from the seismic source to the sensor, and the time difference of the actual sound waves from the seismic source to each group of microphones is checked for anomalies. The spatial coordinates of the construction seismic source target are calculated based on the effective time difference and the corrected sound wave signal. Calculate the straight-line distance between the target construction seismic source and the center of each tower leg pit based on its spatial coordinates, and determine the target construction tower leg pit by combining the coordinates of the construction machinery.

5. The optimized method for monitoring construction of foundation pits in mountainous areas according to claim 4, characterized in that: Calculate the straight-line distance between the target seismic source and the center of each tower leg pit based on its spatial coordinates, and determine the target tower leg pit by combining the coordinates of the construction machinery, including: Calculate the straight-line distance between the spatial coordinates of the construction seismic source target and the center coordinates of the four pre-set tower leg foundation pits, and take the tower leg corresponding to the minimum distance as the initial judgment of the construction tower leg foundation pit; Extract the coordinates of the construction machinery from the valid construction signal, calculate the distance between the coordinates of the construction machinery and the coordinates of the initially judged construction tower leg pit, and if the distance is within the preset pit range, then the initially judged construction tower leg pit is taken as the target construction tower leg pit.

6. The optimized method for monitoring construction of foundation pits in mountainous areas according to claim 2, characterized in that: The process of calculating the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal to determine the compliance of construction behavior and obtain the behavioral risk prediction value of construction behavior based on the gridded risk model includes: Based on the target construction signal, extract vibration timing features, infrared timing features, and construction machinery coordinate timing features; The second feature matching degree between the vibration time series features, infrared time series features, and construction machinery coordinate time series features and each feature template in the construction behavior feature library is calculated based on Euclidean distance. The corresponding construction behavior category is determined based on the second feature matching degree, and it is judged whether the construction behavior violates the construction specifications. The monitoring grid to which the current construction activity belongs is determined based on the coordinates of the construction machinery. The grid risk value of the monitoring grid is calculated based on the weights of the temporal characteristics of each modality, and this value is used as the behavioral risk prediction value of the construction activity.

7. A monitoring and optimization system for construction of foundation pits in mountainous areas, characterized in that, An optimized method for monitoring construction of foundation pits in mountainous areas, as described in any one of claims 1-6, includes: The memory module is used to build a feature library of construction behavior that contains the characteristic relationships between data of various modalities; The acquisition module is used to acquire multimodal sensing data, identify valid construction signals according to the modal filtering rule base, and locate the target construction tower leg foundation pit after path correction of the valid construction signals. The steps for path correction of valid construction signals are as follows: acquiring the three-dimensional terrain parameters of the central pile and each tower leg, including slope, elevation difference and distribution of surface obstacles, to determine the terrain correction factor; determining the wind speed correction factor based on the angle between the wind direction and the direction of sound wave movement; and correcting the sound wave signal by combining the terrain correction factor and the weighted average sound speed of the sound wave. The matching module is used to extract multimodal features based on the location of the target construction tower leg pit, perform feature matching and binding based on the mechanical feature mapping relationship, and obtain the target construction signal. The steps of feature matching and binding based on the mechanical feature mapping relationship are as follows: extract the construction machinery tag ID and construction machinery coordinates based on the valid construction signal to filter out the construction machinery tags in the monitoring area to which the target construction tower leg pit belongs; extract the multimodal sensing features in the valid construction signal in the spatial domain of each construction machinery tag; compare the multimodal sensing features with the mechanical features in the mechanical feature mapping relationship to calculate the first feature matching degree, wherein the mechanical feature mapping relationship includes construction machinery, construction machinery tags, and mechanical features bound to the construction machinery; bind the multimodal sensing features with the first feature matching degree greater than the feature matching degree threshold to the corresponding construction machinery tags and construction machinery. The judgment module is used to calculate the matching degree between the construction behavior feature library and the modal temporal features in the target construction signal, determine the compliance of the construction behavior, and obtain the behavioral risk prediction value of the construction behavior based on the gridded risk model. The early warning module is used to respond to a graded early warning mechanism based on the predicted risk value of the behavior.

8. A computer device, characterized in that: include: The system includes a processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; the memory is used to store computer programs; and the processor is used to execute the programs stored in the memory to implement the steps of the mountain foundation pit construction monitoring optimization method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, performs the steps of the optimization method for monitoring construction of a mountain foundation pit as described in any one of claims 1-6.

Images