Deep and large foundation pit deformation monitoring method and system based on close-range photogrammetry

Abstract

The invention relates to the technical field of foundation pit engineering monitoring, and discloses a deep and large foundation pit deformation monitoring method and system based on close-range photogrammetry, and the method comprises the steps: receiving risk region data, and generating a standardized image set; self-adaptive three-dimensional reconstruction is carried out in combination with domain knowledge, and a three-dimensional grid model with semantic annotation is generated; extracting a model deformation mode and performing semantic annotation to form a structured set; based on this, a geometric domain and knowledge domain bidirectional information channel is constructed, and a knowledge-geometry collaborative metric graph is created; a multi-scale space-time memory structure is used for achieving self-evolution learning and resource optimization allocation, a resource allocation strategy is output, and a monitoring knowledge base is updated; according to the deformation mode, semantic interpretation and the updated knowledge base, risk assessment is executed, intervention suggestions are generated, graded early warning information and a multi-level decision support scheme are output, and automatic processing and optimization from data to decision are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of foundation pit engineering monitoring, and more particularly to a method and system for monitoring deformation of a deep and large foundation pit based on close-range photogrammetry. Background Art

[0002] With the continuous advancement of urbanization, large-scale underground projects and ultra-deep foundation pit projects are increasing, and the safety and accuracy of foundation pit deformation monitoring have become particularly important. Traditional foundation pit monitoring methods such as measuring targets, inclinometers and displacement meters can provide monitoring data at discrete points, but it is difficult to obtain global deformation information on the foundation pit surface, and manual operation is intensive and inefficient. In recent years, although new monitoring technologies such as laser scanning and drone photogrammetry have been introduced, these methods often face problems such as insufficient accuracy, poor real-time performance, and high costs in complex environments. Especially in densely built areas or large foundation pits with a depth of more than 30 meters, there is still a lack of monitoring solutions that can simultaneously take into account high precision, high efficiency and low cost.

[0003] Even more serious is the widespread lack of the ability to effectively integrate geometric observation data with engineering knowledge in current foundation pit deformation monitoring systems. This results in a wealth of monitoring data but little intelligence, making it impossible to achieve early identification and early warning of potential risks. Specifically, there is a semantic gap between monitoring data and deformation mechanisms, and the system cannot automatically interpret the engineering reasons behind deformation phenomena; monitoring resource allocation lacks intelligent self-adaptation and cannot dynamically adjust sampling density based on risk levels; and monitoring experience is difficult to accumulate and pass on, and the system's cognitive capabilities cannot improve as project experience grows. This disconnect between geometric data and engineering knowledge, and the monitoring system's lack of self-evolutionary capabilities, has become a key bottleneck restricting the effectiveness of foundation pit safety monitoring. Summary of the Invention

[0004] The present invention provides a method and system for monitoring deformation of deep and large foundation pits based on close-range photogrammetry, which solves the technical problem of the separation of geometric data and engineering knowledge in related technologies.

[0005] The present invention provides a method for monitoring deformation of a deep foundation pit based on close-range photogrammetry, comprising the following steps: Receive risk area data from the knowledge graph, perform adaptive acquisition and preprocessing operations, and generate a standardized image set; The standardized image set is combined with domain knowledge constraints to perform adaptive 3D reconstruction guided by knowledge geometry interaction entropy, generating a 3D mesh model with semantic annotations. Perform deformation pattern extraction and semantic annotation on 3D mesh models based on knowledge graphs to generate a structured deformation pattern set and corresponding semantic interpretations; Based on the 3D mesh model and deformation pattern set, a bidirectional information transmission channel is constructed between the geometry domain and the knowledge domain to generate a knowledge geometry collaborative metric graph. Based on the multi-scale spatiotemporal memory structure, it realizes self-evolutionary learning and resource optimization allocation, outputs adaptive resource allocation strategy and updated monitoring knowledge base; Based on the deformation pattern set, semantic interpretation and updated monitoring knowledge base, risk assessment is performed and intervention recommendations are generated, outputting risk-graded warning information and multi-level decision support solutions.

[0006] In a preferred embodiment, the steps of performing the adaptive acquisition and preprocessing operations include: Constructing a sampling density distribution matrix based on the risk area data, wherein the sampling density distribution matrix defines a sampling density value for each spatial location; Performing adaptive density image acquisition according to the sampling density allocation matrix to form an original image set including multi-angle and multi-position views; The illumination and quality of the original image set are optimized to generate a standardized image set.

[0007] In a preferred embodiment, the sampling density distribution matrix is calculated by the following formula: ; in, represents the sampling density distribution matrix, is the basic sampling density constant, For location The risk level function value at is the geometric uncertainty function value of the position, and are the first and second weight parameters respectively.

[0008] In a preferred embodiment, the adaptive 3D reconstruction steps guided by knowledge geometric interaction entropy include: Extract knowledge-geometry joint features and generate two types of feature representations: geometric features and semantic features; Calculate the feature importance weight matrix to quantify the contribution of each feature point to the reconstruction process; Perform weighted feature point matching and camera parameter estimation to generate a set of camera projection matrices; Generate multi-view depth maps and fused point clouds, including the 3D geometric structure of the foundation pit surface; Construct a 3D mesh model with semantic annotations, including geometric surface and semantic annotation information.

[0009] In a preferred embodiment, the steps of deformation pattern extraction and semantic annotation based on the knowledge graph include: Establish a deformation representation graph to express the geometric and semantic information of the 3D mesh model in the form of a graph structure; Extract knowledge-constrained deformation features and generate deformation pattern feature sets; Identify and classify deformation mode types to obtain a set of discrete deformation mode categories; Generate semantic interpretations of deformation patterns, including deformation type labels, cause explanations, development trend predictions, and risk level assessments.

[0010] In a preferred embodiment, the step of constructing a bidirectional information transmission channel between the geometry domain and the knowledge domain includes: Quantify the uncertainty distribution of geometric reconstruction and generate a three-dimensional spatial field representing the spatial distribution of reconstruction accuracy; Quantify the uncertainty distribution of knowledge reasoning and generate a three-dimensional space field representing the credibility distribution of pattern recognition; Construct a set of geometry-to-knowledge trigger mappings: ; in, represents a set of trigger mappings from geometry to knowledge, represents the geometric uncertainty threshold, Indicates the risk level threshold, Indicates a point The risk level assessment value of the location; Construct a set of trigger mappings from knowledge to geometry to identify areas where geometric reconstruction accuracy needs to be improved first; Generate a knowledge geometry collaborative metric graph that integrates geometric accuracy and semantic importance and its uncertainty.

[0011] In a preferred embodiment, the steps of implementing self-evolutionary learning and resource optimization allocation based on a multi-scale spatiotemporal memory structure include: Construct a multi-scale spatiotemporal memory storage structure containing short-term, medium-term, and long-term memory levels; Perform experience accumulation and extraction operations to generate a structured set of experience knowledge; Generate and evaluate candidate resource allocation plans based on resource constraints and knowledge geometry collaboration metrics, and select the optimal plan that maximizes system benefits; Perform knowledge base updates and evolutionary iterations to achieve continuous improvement of the system's cognitive capabilities.

[0012] In a preferred embodiment, the uncertainty distribution of the geometric reconstruction is calculated by the following formula: ; in, Indicates a point The geometric uncertainty value at Indicates a point The standard deviation of the geometric reconstruction at Indicates a point The average value of the geometric reconstruction at .

[0013] In a preferred embodiment, the uncertainty distribution of knowledge reasoning is calculated by the following formula: ; in, Indicates a point The knowledge uncertainty value at represents the set of deformation mode categories, Represents a given point Features In the case of , the point belongs to the deformation category The probability value of Indicates a point The feature vector extracted at .

[0014] In a preferred embodiment, a deep foundation pit deformation monitoring system based on close-range photogrammetry includes: Adaptive data acquisition and preprocessing module, which is used to receive risk area data in the knowledge graph, perform adaptive acquisition and preprocessing operations, and generate a standardized image set; The knowledge geometry interactive entropy reconstruction module is used to perform adaptive 3D reconstruction by combining standardized image sets with domain knowledge constraints to generate a 3D mesh model with semantic annotations. The deformation pattern extraction and annotation module is used to extract and semantically annotate deformation patterns of 3D mesh models based on knowledge graphs, generating a structured deformation pattern set and corresponding semantic interpretations. The uncertainty transfer mechanism module is used to build a bidirectional information transfer channel between the geometry domain and the knowledge domain based on the 3D mesh model and the deformation pattern set, and generate a knowledge geometry collaborative metric graph; Self-evolutionary learning and resource optimization module, used to achieve continuous evolution of system capabilities and optimal resource allocation based on multi-scale spatiotemporal memory structure; The risk assessment and decision support module is used to perform risk assessment and generate intervention recommendations based on deformation pattern sets, semantic interpretations, and monitoring knowledge bases, and output risk classification warning information and multi-level decision support solutions.

[0015] The beneficial effects of the present invention are: This invention achieves a dual improvement in the accuracy and efficiency of foundation pit deformation monitoring by constructing a knowledge-geometry dual-state symbiotic system. This system organically integrates geometric observation data with engineering domain knowledge, enabling the system to automatically identify, classify, and interpret various deformation patterns. This not only improves the accuracy of foundation pit deformation monitoring from centimeters to millimeters, but also tracks the development trends of small deformations, issuing early warnings before deformation reaches dangerous thresholds, significantly enhancing the safety assurance capabilities of foundation pit projects.

[0016] The system's self-evolving learning mechanism, based on a multi-scale spatiotemporal memory structure, enables it to continuously accumulate experience from past projects and transform it into structured knowledge for application in new projects. Over time, the system improves the efficiency of monitoring resource allocation and the accuracy of identifying key areas. While maintaining high monitoring quality, it also reduces overall operating costs by approximately 35%. This makes it suitable for long-term safety monitoring of various deep and large foundation pit projects, demonstrating its significant engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flow chart of a method for monitoring deformation of a deep and large foundation pit based on close-range photogrammetry. DETAILED DESCRIPTION

[0018] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. Furthermore, features described in some examples may be combined in other examples.

[0019] At least one embodiment of the present invention discloses a method for monitoring deformation of a deep foundation pit based on close-range photogrammetry, such as Figure 1 As shown, the following steps are included: Step 100, close-range photogrammetry data acquisition and preprocessing based on a two-state symbiotic framework; Receive risk area data from the knowledge graph, perform adaptive acquisition and preprocessing operations, and output a standardized image set. The details are as follows: Step 101: Construct a sampling density distribution matrix based on risk area data: Take the risk area data in the existing knowledge graph as input and generate a sampling density distribution matrix through an adaptive sampling density calculation algorithm. This matrix defines each position in the deep foundation pit space The sampling density value of : in, Indicates location The sampling density value at ; represents the basic sampling density constant, which defines the standard sampling density in areas without special risks; Indicates location The risk level function value at the location quantifies the potential risk level of the location; The geometric uncertainty function value of the position reflects the uncertainty of the geometric reconstruction of the position; and are the third and fourth weight coefficients, respectively, which control the influence of risk level and geometric uncertainty on sampling density.

[0020] Step 102, perform adaptive density image acquisition: allocate matrix according to sampling density , control the multi-view image acquisition device to obtain the original image of the deep foundation pit, and form an original image set containing multi-angle and multi-position views ,in 、 、 are the 1st, 2nd, and nth original images respectively, is the total number of images. The original image set contains the visual information of the deep foundation pit surface, but may have problems such as uneven lighting and insufficient contrast.

[0021] Step 103, image illumination and quality optimization: the original image set The image is input into the image enhancement module, which uses the illumination compensation algorithm and quality optimization algorithm to eliminate uneven illumination, enhance texture details, correct color distortion, and generate a standardized image set. ,in 、 、 are the 1st, 2nd, and nth standardized images respectively, is the total number of standardized images. The standardized image set has balanced lighting conditions and enhanced detail features, providing high-quality input data for subsequent 3D reconstruction.

[0022] Step 200: Adaptive 3D reconstruction guided by knowledge-geometric interaction entropy; This step inputs the standardized image set , combined with domain knowledge constraints, outputs a 3D mesh model with semantic annotations. The details are as follows: Step 201, extracting knowledge geometry joint features: standardize the image set The input is fed into the joint feature extractor to generate two types of feature representations, geometric features and semantic features, in parallel: in, Representation knowledge geometric joint feature extractor, used to extract geometric and semantic features from images simultaneously; Represents a set of geometric features extracted from an image, including low-level geometric elements such as feature points, edges, and textures; Represents a set of knowledge features extracted from an image, including high-level semantic elements such as material type and structural component type.

[0023] Step 202, calculate the feature importance weight matrix: based on the extracted joint features , through the knowledge geometry interactive entropy algorithm for each feature point Calculate importance weights: in, Representing feature points The importance weight value quantifies the contribution of the point to the reconstruction process; Representing feature points The probability distribution in the geometric domain reflects the geometry of the point; Representing feature points The probability distribution in the knowledge domain reflects the semantic importance of the point; is a balancing factor used to adjust the relative importance of geometric information and semantic information in weight calculation.

[0024] Step 203: Perform weighted feature point matching and camera parameter estimation: The feature point set and its weight matrix Input into the multi-view matching algorithm, perform weighted feature matching, and generate a camera projection matrix set: in, Indicates the The projection matrix of each perspective defines the mapping relationship from three-dimensional space to two-dimensional image; represents the total number of projection matrices, Represents the camera internal parameter matrix, including internal parameters such as focal length and principal point coordinates; Represents the rotation matrix, describing the orientation of the camera; Represents the translation vector, describing the position of the camera.

[0025] Step 204: Generate multi-view depth map and fused point cloud: Use camera projection matrix set and standardized image sets , apply the variational multi-view stereo matching algorithm to generate a set of depth maps for each view: ; in, 、 、 Respectively represent the 1st, 2nd, Depth map of each viewpoint; Indicates the total number of depth maps; The depth map is then converted and fused into a unified 3D point cloud : in, Represents the inverse projection function, which converts the two-dimensional pixel coordinates and depth value Through the projection matrix Convert to three-dimensional space coordinate points ; The fused 3D point cloud contains the complete geometric structure of the foundation pit surface.

[0026] Step 205: Construct a 3D mesh model with semantic annotations: and semantic features Input into the knowledge-constrained Poisson surface reconstruction algorithm to generate a 3D mesh model with semantic annotations : in, Represents a three-dimensional mesh model, containing a set of vertices , edge set , noodle set and semantic annotation sets ; Represents the data item, which measures the degree of fit between the model and the point cloud; Represents the smoothing term to ensure the continuity of the surface; Represents knowledge constraints to ensure that the reconstruction results conform to known physical laws and engineering experience; and are weight coefficients, which control the strength of smoothness constraint and knowledge constraint respectively.

[0027] Step 300: Extracting deformation patterns and semantic annotation based on knowledge graph; This step converts the semantically annotated 3D mesh model into a structured set of deformation patterns and corresponding semantic interpretations. The details are as follows: Step 301: Create a deformation representation map: transform the semantically annotated 3D mesh model Convert to deformation representation , the mathematical representation of the graph is: in, The deformation representation graph is a structured graph representation; Represents a set of nodes in the graph, each node Corresponding to a key point on the 3D model; Represents the set of edges in the graph, each edge Represents the spatial connection relationship between key points; Represents the attribute matrix, which contains the characteristic attributes of nodes and edges, such as geometric deformation, material type, structure type, etc.

[0028] Step 302: Extract the deformation features of knowledge constraints: transform the deformation representation graph and domain prior knowledge Input into the knowledge-guided graph convolutional network to generate a deformation pattern feature set : in, It represents the deformation mode feature set and captures the essential characteristics of foundation pit deformation; Represents a knowledge-guided graph convolutional network function used to extract features from the graph structure; Represents a set of prior knowledge including physical constraints, engineering experience, etc., which is used to guide the feature extraction process.

[0029] Step 303: Identify and classify deformation pattern types: Input into the unsupervised clustering algorithm based on physical constraints to obtain a set of discrete deformation mode categories : in, Represents a set of deformation mode categories, including a variety of typical deformation modes; represents the clustering function, which is used to group similar features; Represents a set of clustering constraints, including physical rationality constraints, spatiotemporal continuity constraints, etc. Indicates the deformation pattern categories, including spatial distribution maps, quantitative values of deformation degree, deformation rate time series and other multi-dimensional information.

[0030] Step 304: Generate semantic interpretation of deformation pattern: Set deformation pattern category Input into the knowledge reasoning engine to generate the corresponding deformation description set : in, Represents a set of deformation descriptions, which contains professional explanations of various deformation modes; Represents the semantic interpretation function, which converts numerical features into understandable text descriptions; Represents a knowledge base that includes professional knowledge such as the causal relationship of deformation types and deformation feature risk level mapping; Indicates the The semantic description of each deformation pattern includes deformation type label, possible cause explanation, development trend prediction and risk level assessment.

[0031] Step 400, establishing a bidirectional information flow of the uncertainty transfer mechanism; This step builds a bidirectional information transmission channel between the geometry domain and the knowledge domain based on the 3D model and deformation pattern data, and outputs a knowledge geometry collaborative measurement graph. The details are as follows: Step 401, quantify the uncertainty distribution of geometric reconstruction: transform the 3D mesh model and original point cloud data As input, we calculate the value of each point in space by statistical analysis method. The geometric reconstruction uncertainty value of : in, Indicates a point The geometric uncertainty value at Indicates a point The standard deviation of the geometric reconstruction at Indicates a point The geometric uncertainty distribution generated is the average value of the geometric reconstruction at . It is a three-dimensional spatial field that represents the spatial distribution of reconstruction accuracy in the entire foundation pit area.

[0032] Step 402, quantify the uncertainty distribution of knowledge reasoning: transform the deformation pattern feature set and deformation mode category sets As input, calculate each point in space Uncertainty value of knowledge reasoning : in, Indicates a point The knowledge uncertainty value at Represents a given point Features In the case of , the point belongs to the deformation category The probability value of Denotes the deformation category, represents the set of deformation mode categories, Indicates a point The generated knowledge uncertainty distribution It is also a three-dimensional spatial field, which represents the credibility distribution of pattern recognition and semantic understanding in the entire foundation pit area.

[0033] Step 403: Constructing a geometry-to-knowledge trigger mapping set: combining geometric uncertainty distribution and risk level distribution , determine the set of spatial points where geometric information should trigger knowledge updates : in, represents a set of trigger mappings from geometry to knowledge, represents the geometric uncertainty threshold, Indicates the risk level threshold, Indicates a point The risk level value at Represents a point in space. This mapping set identifies which areas in the foundation pit need to be prioritized for knowledge update and semantic interpretation due to high geometric uncertainty and high risk level.

[0034] Step 404: Construct a set of trigger mappings from knowledge to geometry: Combined with knowledge uncertainty distribution and knowledge uncertainty gradient distribution , determine the set of spatial points where knowledge information should trigger geometric accuracy improvement : in, represents a set of trigger mappings from knowledge to geometry, represents the knowledge uncertainty threshold, represents the knowledge uncertainty gradient threshold, Indicates a point The spatial gradient value of knowledge uncertainty at Represents a point in space. This mapping set identifies which areas in the foundation pit need to be prioritized to improve the geometric reconstruction accuracy due to high uncertainty and rapid changes in knowledge.

[0035] Step 405: Generate a knowledge geometry collaboration metric graph: Based on the geometric accuracy index , knowledge relevance index and its corresponding uncertainty distribution and , calculate the knowledge geometry collaboration metric in the entire three-dimensional space : in, Indicates a point The knowledge geometry collaborative metric value at Indicates a point The geometric accuracy index value at Indicates a point The knowledge relevance index value at and is the weight coefficient, represents a point in space, Indicates a point The geometric uncertainty value at Indicates a point The knowledge uncertainty value at . The generated knowledge geometry collaborative measurement graph It is a three-dimensional scalar field that integrates geometric accuracy, semantic importance and their respective uncertainties, providing a unified reference for system resource allocation.

[0036] Step 500: self-evolutionary learning and resource optimization allocation based on memory enhancement; This step inputs the knowledge geometry collaborative measurement graph, historical deformation data, and system resource constraints to optimize system resource allocation and enhance cognitive capabilities, and outputs an adaptive resource allocation strategy and an updated monitoring knowledge base. The details are as follows: Step 501: Construct a multi-scale spatiotemporal memory storage structure: transform the historical deformation data and current knowledge geometry collaborative metric graph Input into the memory storage module to build a hierarchical spatiotemporal memory structure : in, Represents a collection of multi-scale spatiotemporal memory structures, including short-term memory (Store high-precision raw data for the last 7 days), medium-term memory (Stores the main deformation patterns and statistical features of the last 790 days) and long-term memory (Stores more than 90 days of abstract deformation patterns and empirical knowledge.) This memory structure organizes data at different time scales and levels of abstraction, enabling the system to efficiently store and access historical information.

[0037] Step 502: Execute experience accumulation and extraction operations: convert the multi-scale spatiotemporal memory structure Input into the experience extraction module and generate experience knowledge set through pattern mining algorithm : in, Represents an empirical knowledge record, including a pattern description (Formal representation of deformation patterns), environmental conditions (Description of environmental conditions that trigger this mode), time characteristics (time evolution characteristics of the pattern) and frequency statistics (the frequency and probability distribution of the pattern), Represents the index of the experience knowledge record. The generated experience knowledge set Extract the regular knowledge implicit in historical data in a structured form to provide an empirical basis for subsequent optimization decisions.

[0038] Step 503: Generate a set of candidate allocation solutions based on resource constraints: , knowledge geometry collaborative measurement graph and experiential knowledge set Input into the resource allocation optimization module to generate a set of candidate resource allocation solutions: in, 、 、 Respectively represent the 1st, 2nd, candidate resource allocation schemes; Indicates the total number of candidate resource allocation solutions; in, Indicates the candidate resource allocation schemes, Indicates position in space Amount of allocated resources , represents the total resource constraint, represents the total number of locations after spatial discretization, An index representing a spatial position, Indicates the index of the candidate solution. Each candidate solution It is a feasible resource allocation strategy that satisfies the condition that the total resources do not exceed the constraint.

[0039] Step 504: Simulation evaluation and optimal solution selection: The candidate resource allocation solutions are combined Input to the performance simulation evaluation module, combined with the knowledge geometry collaborative measurement graph and experiential knowledge set , calculate the expected benefit of each plan : in, Representation scheme The expected return value of represents the resource benefit conversion function, Represents the position based on empirical knowledge The importance evaluation function, Indicates location The knowledge geometry collaborative metric value at represents the total number of locations after spatial discretization, An index representing a spatial position, Represents the index of the candidate solution. By comparing the expected benefits of each solution, the optimal resource allocation solution is selected. : Optimal resource allocation plan It explicitly specifies how the system should allocate limited computing, storage, and network resources to maximize the overall monitoring benefits.

[0040] Step 505: perform knowledge base update and evolution iteration: set the current deformation mode , deformation description set and experiential knowledge set Input into the knowledge base update module to generate an updated monitoring knowledge base : in, represents the knowledge base update function, represents the original monitoring knowledge base, represents the updated knowledge base, Represents the current deformation mode set, represents a set of deformation descriptions, Represents the set of empirical knowledge. The updated knowledge base It not only includes updated deformation pattern causal relationships and risk assessment rules, but also includes newly discovered deformation laws and the system's adaptability experience to different situations, reflecting the continuous improvement of the system's cognitive ability.

[0041] Step 600, risk assessment and proactive intervention decision support; This step inputs the updated monitoring knowledge base, the current deformation pattern set, and the resource allocation strategy, performs risk assessment and generates intervention recommendations, and outputs risk classification warning information and multi-level decision support solutions. The details are as follows: Step 601, perform dynamic risk threshold calculation: update the monitoring knowledge base and the current deformation mode set Input into the risk threshold calculation module to generate the adaptive risk threshold matrix : in, Representation type Deformation in the area The risk threshold, represents the basic threshold constant, represents the impact factor calculation function, taking into account the empirical rules in the knowledge base, the characteristics of the current deformation mode, and the deformation type and regional attributes. An index representing the deformation type, The index of the region. represents the updated monitoring knowledge base, Represents the current deformation mode set. The generated risk threshold matrix It provides a criterion for judging environmental adaptability for subsequent risk assessment.

[0042] Step 602: Generate a multi-dimensional risk score map: Set the current deformation mode , deformation description set and risk threshold matrix Input into the risk scoring module to calculate the multi-dimensional risk scoring map : in, Indicates the Dimensional risk score, represents the weight coefficient, represents the risk calculation function, Indicates the deformation mode, Indicates the A deformation description, Representation type Deformation in the area The risk threshold, An index representing the deformation type, The index of the region. Indicates the index of the risk dimension. Multi-dimensional risk score map The scoring includes multiple dimensions such as deformation rate risk, cumulative deformation risk, trend risk, and abnormal risk, which comprehensively characterizes the risk status of foundation pit deformation.

[0043] Step 603: Construct a causal risk propagation network: transform the risk score map and monitoring knowledge base Input into the risk propagation model to construct a causal risk propagation network : in, represents the causal risk propagation network, Represents a set of nodes (corresponding to different risk events), represents the edge set (corresponding to the causal relationship between risk events), Indicates risk from events Propagate to events The conditional probability of represents the risk propagation probability calculation function, represents the updated monitoring knowledge base, represents the risk score map, and Represents a risk event node. Causal risk propagation network The formalization represents the correlation between different risk events, enabling the system to predict the possible propagation path and scope of risks.

[0044] Step 604: Generate graded warning information and spatial distribution map: Generate multi-dimensional risk score map and causal risk propagation networks Input into the warning generation module to generate graded warning information and early warning spatial distribution map : in, , Indicates the risk level is The set of spatial points ( corresponding to four levels: safety, attention, warning and danger). represents the risk level determination function, represents the risk score map, represents the causal risk propagation network, Indicates spatial coordinate points. Warning spatial distribution map It is a three-dimensional color-coded map that intuitively displays the risk distribution of the entire foundation pit area, making the monitoring results visual and easy to understand.

[0045] Step 605: Generate a multi-level decision support plan: , warning spatial distribution map and monitoring knowledge base Input into the decision support module to generate a multi-level decision support plan : in, Indicates immediate intervention options (including safety measures that need to be taken immediately), Indicates a short-term intervention plan (including reinforcement measures to be completed within 17 days). Indicates the long-term intervention plan (including monitoring strategy adjustments and protection system optimization suggestions), 、 and represent the immediate, short-term and long-term solution generation functions respectively, Indicates graded warning information. represents the early warning spatial distribution map, Represents the updated monitoring knowledge base. Multi-level decision support scheme Specific operational recommendations are provided for engineering decisions at different time scales to ensure engineering safety.

[0046] In one embodiment of the present invention, an application example of the aforementioned method for monitoring deformation of a deep foundation pit based on close-range photogrammetry is provided: This example focuses on a large-scale underground complex foundation pit project with the following characteristics: Foundation pit area: approximately 68,500 square meters; Foundation pit depth: The deepest reaches 38.6 meters; Surrounding environment: The north side is adjacent to the subway line 13 tunnel (about 23 meters away), the west side is a historical protected building complex (about 35 meters away), and the east side is a high-rise office building complex (about 40 meters away); Geological conditions: The upper part is a fill layer (35 meters) and a sandy silt layer (812 meters), the middle part is a silty clay layer (1015 meters), and the lower part is moderately weathered granite; Groundwater conditions: The shallow groundwater level is about 57 meters deep, and the pressure water level is about 2530 meters deep; Support form: underground continuous wall + three steel pipe supports + prestressed anchor cable combined support structure; The main technical challenges faced by this project include: sensitive surrounding environment and strict requirements for deformation control; complex geological conditions, with unevenly weathered rock layers and fault zones; a long construction period (estimated to be 24 months), requiring long-term stable monitoring; and changeable meteorological conditions with obvious seasonal influences.

[0047] The project adopted this implementation method to build a self-evolving foundation pit cognitive system with knowledge-geometry dual-state symbiosis. The deployment configuration of the monitoring system is shown in Tables 1 and 2: Table 1: Monitoring system hardware configuration table Table 2: Distribution of foundation pit risk areas During the initial project implementation, the system automatically generated a sampling density allocation matrix based on the risk area distribution and levels in Table 2. For example, the RA01 area on the north side of the foundation pit, near the subway tunnel, had the highest risk level (9 points), and the system assigned it a higher sampling density than other areas. Table 3 shows a comparison of sampling densities for different risk areas: Table 3: Example of adaptive sampling density allocation in risk areas The sampling density allocation matrix guides the dynamic scheduling of cameras. For example, in the RA01 area, the system performs 24 high-density image acquisitions per day, while in general areas with lower risk levels, only four acquisitions are performed. This differentiated acquisition strategy based on risk level ensures efficient use of limited monitoring resources.

[0048] During the 3D reconstruction process, the system uses the knowledge-based geometry interaction entropy algorithm to weight the importance of feature points. For example, during a reconstruction on July 15, 2023, the system detected a significant increase in geometric uncertainty in area RA05 (the intersection of the fault zone in the middle of the foundation pit). The feature point weight matrix calculated using the knowledge-based geometry interaction entropy algorithm improved the reconstruction accuracy of this area. Table 4 shows the distribution of feature point weights for different areas on that day: Table 4: Example of feature point weight distribution based on knowledge geometric interaction entropy (July 15, 2023) This example clearly demonstrates the dynamic control capability of knowledge geometric interaction entropy on the reconstruction process. and knowledge domain probability When the weights are lower (0.45 and 0.25, respectively), indicating a high uncertainty in that area, the system automatically increases the weight of that area (3.65), resulting in a 285% improvement in reconstruction accuracy. In contrast, due to the low uncertainty in general areas, the weight is only 0.18, and the reconstruction accuracy improves by only 12%.

[0049] Through adaptive weight allocation, the system implements a refined monitoring strategy for key areas with limited resources, creating a positive correlation between reconstruction accuracy and risk level.

[0050] During the long-term monitoring process, the system gradually built a rich library of deformation patterns. Taking the monitoring of area RA04 (the corner of the south foundation pit) from August 10 to August 30, 2023 as an example, the system successfully identified and classified a potential risk deformation pattern through deformation representation graph structure and knowledge-constrained deformation feature extraction, as shown in Table 5: Table 5: Examples of RA04 region deformation pattern extraction and semantic annotation This example demonstrates how the system, through graph structure representation and knowledge-constrained feature extraction, successfully identifies localized corner instability risk patterns without pre-defined corner instability patterns, providing accurate semantic interpretation and risk assessment. The system not only identifies abnormal growth trends in deformation but also correlates multi-dimensional information such as support axial force, anchor cable prestress, and surrounding ground settlement, providing a comprehensive assessment of potential risks.

[0051] One of the core innovations of this system is the establishment of a bidirectional information transfer mechanism between the geometric domain and the knowledge domain. For example, on September 5, 2023, the system monitored area RA08 (a weak interlayer at the bottom of the foundation pit). The system detected abnormal changes in geometric uncertainty and knowledge uncertainty in this area, triggering the bidirectional information transfer mechanism, as shown in Tables 6 and 7. Table 6: Examples of RA08 regional uncertainty transfer mechanisms (September 5, 2023) Table 7: Comparison of the collaborative improvement effect between reconstruction accuracy and risk perception Data analysis shows that this implementation improves reconstruction accuracy in high-risk areas (RA01, RA05, and RA04) by an average of 259%, while in standard areas it only improves by 5%. Furthermore, risk identification accuracy improves by an average of 18.9 percentage points in high-risk areas, while it only improves by 4.9 percentage points in standard areas. This fully demonstrates that this implementation can dynamically adjust system resource allocation based on risk level, achieving a coordinated improvement in accuracy and risk.

[0052] In addition, by comparing the system resource allocation in different periods, it is found that this implementation automatically allocates most computing and storage resources to high-risk areas: Computing resource allocation: High-risk areas (21% of the total area) received 82.7% of the computing resources; Storage resource allocation: Data storage accuracy in high-risk areas is 3.5 times higher than that in normal areas; Detection frequency allocation: The monitoring frequency in high-risk areas is 6 times higher than that in ordinary areas.

[0053] To verify the system's self-evolution capabilities, the project team conducted a systematic evaluation of the system's performance at different time periods (initial deployment, 3 months later, 6 months later, and 12 months later). The specific evaluation data is shown in Table 8: Table 8: Verification data of system self-evolution capability over time Compared with traditional systems deployed during the same period (using fixed parameters and algorithms), the performance indicators of traditional systems remained basically unchanged or slightly decreased, as shown in Table 9: Table 9: 12-month performance comparison between the traditional system and this implementation The above data fully demonstrates that the self-evolution mechanism of this implementation method continuously improves system capabilities over time, with key performance indicators increasing by an average of over 150% within one year. In particular, in terms of deformation pattern recognition, the system autonomously discovered and verified 26 new deformation patterns, which played a significant role in subsequent monitoring. The early identification of five key deformation patterns helped the project avoid potential engineering risks and accidents.

[0054] In one embodiment of the present invention, in order to adapt to the application scenarios of large-scale foundation pit projects with a wide monitoring area, numerous points, and complex and changing environments, and to improve the coverage and collaborative efficiency of the monitoring system, the deformation monitoring method for deep and large foundation pits based on close-range photogrammetry also includes: Step 10: Build a multi-level asynchronous decision-making architecture; This step first performs the initialization step of the deep and large foundation pit deformation monitoring method based on close-range photogrammetry, and then constructs a three-level computing architecture.

[0055] Specifically, this step receives system topology data, computing node resource information, and collaborative task requirement description as input data, and generates a multi-level asynchronous decision architecture through a hierarchical time scale mapping algorithm, including: Cloud layer node configuration: a list of central server nodes and their computing power parameters, responsible for executing the global planning task with a 1030-second update cycle; Fog layer node configuration: a list of regional edge nodes and their computing power parameters, responsible for performing regional coordination tasks with a 13-second update cycle; Edge layer node configuration: a list of robot body nodes and their computing power parameters, responsible for executing local response tasks with a 50-200 millisecond update cycle; Inter-layer communication link table: defines the communication topology and bandwidth parameters between nodes at each layer; Decision-making authority allocation table: specifies the initial decision-making scope and authority limits of nodes at each level; The output multi-level asynchronous decision-making architecture configuration includes the functional positioning, communication relationship and initial resource allocation of each node in the system, providing infrastructure support for the execution of subsequent asynchronous decisions.

[0056] Step 20, perform time-sensitive task decomposition and allocation; This step uses the time-sensitive task decomposition algorithm to decompose the overall collaborative task and assign it to the appropriate decision-making level.

[0057] The input data includes: Collaborative task description list: contains all collaborative tasks to be executed and their attributes Task time parameter set: contains time constraint information such as the deadline and expected execution time of each task Computing resource status table: contains the available computing resources and load status of each computing node Multi-level asynchronous decision architecture configuration output in step 31; The time-sensitive task decomposition algorithm calculates the time sensitivity index of each task, and its formula is: in, Indicates a task The time sensitivity index of , the larger the value, the more sensitive it is to time constraints; Indicates a task The deadline, in seconds; It represents the gradient of the impact of response time on task success rate, and represents the change in task success rate caused by a one-unit change in response time. It represents the inter-task dependency strength index, which measures the degree to which the task is dependent on other tasks; 、 、 is the weight coefficient, which is used to adjust the relative importance of each factor.

[0058] Based on the calculated time sensitivity index, the algorithm generates a hierarchical task allocation table, which includes: Task level mapping relationship: specifies at which decision level each task is executed Task update frequency configuration: specifies the decision update cycle of each task Task resource requirement list: specifies the amount of computing and communication resources required for each task; The output hierarchical task allocation table provides a detailed execution plan for collaborative tasks in the three-level computing architecture, providing task-level guidance for the implementation of asynchronous decision-making.

[0059] Step 30: Generate inter-layer information transmission strategy and decision coordination plan; This step integrates the information entropy-based information transfer algorithm and the multi-time-scale decision coordination algorithm to ensure system consistency in an asynchronous decision-making environment.

[0060] The input data includes: Multi-level node status dataset: contains the current computing and communication status of each node at each level Communication resource allocation matrix: records the available communication resources in the system and their distribution Historical decision record database: stores past decisions and their effects Multi-level asynchronous decision architecture configuration output from step 31 Hierarchical task allocation table output from step 32; The information transfer algorithm based on information entropy calculates the value index of inter-layer information transfer: in, Indicates at a point in time The information Slave nodes Pass to node Value indicators; Display information The information entropy quantifies the degree of reduction in information uncertainty; represents the expected increment of the decision utility of the receiving node after information transmission; Indicates at a point in time Delivering information communication costs; Indicates the current time point.

[0061] The multi-timescale decision coordination algorithm calculates the degree of coordination between different decisions: in, Representing a decision and decision-making The degree of coordination between them, the higher the value, the better the coordination; Indicates the A similarity measurement function to evaluate the similarity of decisions in different aspects; Indicates the A conflict measurement function to evaluate the degree of conflict in different aspects of decision making; Indicates the The weight of the similarity measure; Indicates the overall weight coefficient of conflict factors; and Represents the number of types of similarity measures and conflict measures respectively.

[0062] Combining the calculation results of the above two algorithms, this step finally outputs: Inter-layer information transmission strategy table: contains the following contents: information transmission trigger condition rule set: defines when information needs to be transmitted between layers; information compression coding scheme: defines how to compress the transmitted information to reduce communication costs; information transmission priority list: defines the transmission priority of different types of information; Decision coordination plan: includes the following contents: decision execution sequence table: specifies the execution order of decisions at different levels and different nodes; conflict resolution rule set: defines the handling method when decision conflicts occur; collaborative behavior pattern library: defines the standard behavior pattern of collaborative decision-making between nodes; the output inter-layer information transmission strategy table and decision coordination plan together constitute the key tools for maintaining system consistency in an asynchronous decision-making environment, ensuring the coordination and consistency of distributed nodes under limited information exchange conditions.

[0063] Step 40, calculate the dynamic resource allocation table and conflict resolution strategy; This step applies load-aware task deployment algorithms and resource conflict detection algorithms to achieve dynamic optimization of system resources and prevent potential conflicts.

[0064] Input data includes: Real-time status report of computing nodes: including real-time operating indicators such as CPU utilization, memory usage, and energy consumption of each computing node; Decision sequence data stream: including a list of decisions currently being executed and planned to be executed by nodes at each level; Current resource allocation status: describing the current allocation of various resources in the system; Environmental complexity assessment results: representing the complexity and changing trends of the current environmental conditions; Inter-layer information transmission strategy table and decision coordination plan output in step 33; The load-aware task deployment algorithm calculates the task deployment decision metric: in, Indicates at a point in time The task Deploy to nodes The comprehensive score of , where a higher value indicates a more reasonable deployment; Indicates that the node Execute tasks on The calculation energy consumption is in joules; Indicates that the task Deploy to nodes The communication energy consumption generated, in joules; Indicates that the node Execute tasks on The delay time, in milliseconds; Representation node At the time point Resource utilization, the value is between 0 and 1; 、 、 、 is the weight coefficient that is dynamically adjusted according to the system status.

[0065] The resource conflict detection algorithm identifies possible decision conflicts: in, Representing a decision and decision-making Whether there is a resource conflict between them, 1 means there is a conflict, 0 means there is no conflict; Representing a decision Resources The usage set includes the amount of resources used and the time interval of use; Representing a decision Resources The usage set includes the amount of resources used and the time interval of use; Representing a decision The spatial impact area, that is, the physical space affected by the decision; Representing a decision The spatial impact area, that is, the physical space affected by the decision; Represents the collection of all resources in the system.

[0066] Based on the calculation results of the above algorithm, this step generates the following output: Dynamic resource allocation table: contains the following contents: Task migration instruction sequence: specifies which tasks need to be migrated to which nodes; Resource reallocation plan: details the optimal allocation method for system resources; Decision authority temporary transfer list: specifies the temporary transfer rules of decision authority under specific conditions; Conflict resolution strategy document: contains the following content: Decision priority adjustment list: specifies the priority order between conflicting decisions; Resource competition mediation plan: details how to resolve resource competition issues; Spatial impact area coordination rules: specifies the operation coordination method for overlapping impact areas; The output dynamic resource allocation table and conflict resolution strategy document jointly ensure the efficient utilization of system resources and minimize conflicts in decision execution, making the asynchronous decision framework efficient and stable in actual operation.

[0067] Step 50, generating an adaptive adjustment parameter set and a performance optimization solution; This step combines the adaptive time step control algorithm and the decision performance evaluation algorithm to achieve continuous optimization and adjustment of system parameters.

[0068] Input data includes: Execution efficiency monitoring data set: Contains efficiency indicators of decision execution at all levels, such as decision accuracy, execution delay, etc. Environment change rate record: Records the change rate and trend of various parameters in the environment Historical decision effect database: Stores the execution effect and performance data of historical decisions Resource utilization time series: Records the time changes in system resource utilization Dynamic resource allocation table and conflict resolution strategy document output by Step 40; The adaptive time step control algorithm calculates the optimal update period for each level of decision making: in, Indicates the The decision update period of the layer at the next time point, in milliseconds; Indicates the The decision update period of the layer at the current time point, in milliseconds; Indicates the The current execution efficiency of the layer, the normalized value is between 0 and 1; Indicates the target execution efficiency, usually the preset optimal efficiency value; Indicates the rate of change of environmental complexity and characterizes the speed of environmental change; and It is an adjustment coefficient used to balance the impact of execution efficiency and environmental changes.

[0069] The decision performance evaluation algorithm calculates the overall performance score of the system: in, Indicates the system at a point in time The overall performance score of , where higher values indicate better performance; Indicates the number of completed tasks; Indicates the total number of tasks; Indicates the total amount of available resources; Indicates the total amount of resources used; Represents the reference delay time, usually the ideal delay for system design; Indicates the actual measured average delay time; Indicates the number of conflicting decisions; represents the total number of decisions; 、 、 、 is the weight coefficient of each indicator.

[0070] Based on the performance score, the system optimization goal can be expressed as: in, Represents the set of adjustable parameters of the system, including the weight coefficients and thresholds of each algorithm; Indicates the end time point Historical data; Indicates that the parameter and historical data Condition, future time point Expected performance score; Indicates the prediction time interval, that is, the length of time predicted from the current time point to the future.

[0071] Based on the calculation results of the above algorithm, this step generates the following output: Adaptive time step parameter set: Contains the following: Decision update cycle table for each level: specifies the optimal decision update frequency for each level; Time step adjustment trigger condition: specifies when the decision update cycle needs to be adjusted; Computing resource dynamic allocation instructions: specifies how to adjust computing resources according to the decision frequency; System performance optimization plan: includes the following: Algorithm parameter adjustment list: specifies the parameter values that need to be adjusted in each algorithm; Decision boundary fine-tuning rules: specifies how to adjust the boundaries between decision levels; Resource allocation optimization strategy: details the optimization direction of resource allocation; The output adaptive time step parameter set and system performance optimization scheme jointly realize the continuous self-optimization of the system, enabling the asynchronous decision-making framework to adapt to environmental changes and dynamic adjustments of task requirements and maintain long-term efficient operation.

[0072] This implementation achieves the following specific technical effects through a multi-level asynchronous decision-making framework and a three-level edge-fog-cloud computing architecture: Computational load distribution and communication overhead reduction; Computing load dispersion effect: Compared with the traditional centralized architecture, the computing load on the central server is reduced, and computing tasks are reasonably distributed across nodes at different levels; Communication bandwidth reduction: The overall communication bandwidth requirement of the system is reduced, and the information entropy-based transmission algorithm reduces the amount of unnecessary information transmitted; Improved communication delay: The average communication delay between nodes is reduced; the real-time response capability is improved; Shortened response time: The average system response time to environmental changes is shortened from seconds to milliseconds; Decision-making and execution efficiency: The ability to handle emergency tasks is improved, achieving millisecond-level local response capabilities; Processing throughput: The number of decisions the system can process per second increases; Optimize the balance between coordination, consistency and autonomy; Reduced coordination overhead: The coordination communication overhead between layers is reduced while maintaining the overall consistency of the system; Decision consistency index: The collaborative consistency of key decision points is improved, ensuring the global consistency of system behavior; Improved local autonomy: The autonomous decision-making capabilities of edge nodes are improved, reducing the decision-making pressure of central nodes; Enhanced system scalability and robustness; Scalability: The system can seamlessly expand the number of robots it supports from dozens to hundreds. Fault tolerance: When 50% of edge nodes fail randomly, the system can still maintain 85% of its functionality, improving system robustness; Abnormal recovery speed: The system recovery time from local failures is reduced from minutes to seconds; Comprehensive optimization of resource utilization efficiency; Computing resource utilization: The average utilization of system computing resources is improved, reducing idle resources; Energy consumption reduction: When completing the same task, the overall energy consumption of the system is reduced; Task completion quality: The successful completion rate of tasks is improved.

[0073] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make more forms of equivalent embodiments based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A method for monitoring deformation of deep foundation pits based on close-range photogrammetry, characterized in that: The following steps are involved: Receive risk area data from the knowledge graph, perform adaptive acquisition and preprocessing operations, and generate a standardized image set; The standardized image set is combined with domain knowledge constraints to perform adaptive 3D reconstruction guided by knowledge geometry interaction entropy, generating a 3D mesh model with semantic annotations. Perform deformation pattern extraction and semantic annotation on 3D mesh models based on knowledge graphs to generate a structured deformation pattern set and corresponding semantic interpretations; Based on the 3D mesh model and deformation pattern set, a bidirectional information transmission channel is constructed between the geometry domain and the knowledge domain to generate a knowledge geometry collaborative metric graph. Based on the multi-scale spatiotemporal memory structure, it realizes self-evolutionary learning and resource optimization allocation, outputs adaptive resource allocation strategy and updated monitoring knowledge base; Based on the deformation pattern set, semantic interpretation and updated monitoring knowledge base, risk assessment is performed and intervention recommendations are generated, outputting risk-graded warning information and multi-level decision support solutions.

2. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 1, characterized in that: The steps to perform adaptive acquisition and preprocessing operations include: Constructing a sampling density distribution matrix based on the risk area data, wherein the sampling density distribution matrix defines a sampling density value for each spatial location; Performing adaptive density image acquisition according to the sampling density allocation matrix to form an original image set including multi-angle and multi-position views; The illumination and quality of the original image set are optimized to generate a standardized image set.

3. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 2, characterized in that: The sampling density distribution matrix is calculated by the following formula: ； in, represents the sampling density distribution matrix, is the basic sampling density constant, For location The risk level function value at is the geometric uncertainty function value of the position, and are the first and second weight parameters respectively.

4. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 1, characterized in that: The steps of adaptive 3D reconstruction guided by knowledge geometry interaction entropy include: Extract knowledge-geometry joint features and generate two types of feature representations: geometric features and semantic features; Calculate the feature importance weight matrix to quantify the contribution of each feature point to the reconstruction process; Perform weighted feature point matching and camera parameter estimation to generate a set of camera projection matrices; Generate multi-view depth maps and fused point clouds, including the 3D geometric structure of the foundation pit surface; Construct a 3D mesh model with semantic annotations, including geometric surface and semantic annotation information.

5. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 1, characterized in that: The steps of deformation pattern extraction and semantic annotation based on knowledge graph include: Establish a deformation representation graph to express the geometric and semantic information of the 3D mesh model in the form of a graph structure; Extract knowledge-constrained deformation features and generate deformation pattern feature sets; Identify and classify deformation mode types to obtain a set of discrete deformation mode categories; Generate semantic interpretations of deformation patterns, including deformation type labels, cause explanations, development trend predictions, and risk level assessments.

6. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 1, characterized in that: The steps to build a bidirectional information transmission channel between the geometry domain and the knowledge domain include: Quantify the uncertainty distribution of geometric reconstruction and generate a three-dimensional spatial field representing the spatial distribution of reconstruction accuracy; Quantify the uncertainty distribution of knowledge reasoning and generate a three-dimensional space field representing the credibility distribution of pattern recognition; Construct a set of geometry-to-knowledge trigger mappings: ； in, represents a set of trigger mappings from geometry to knowledge, represents the geometric uncertainty threshold, Indicates the risk level threshold, Indicates a point The risk level assessment value of the location; Construct a set of trigger mappings from knowledge to geometry to identify areas where geometric reconstruction accuracy needs to be improved first; Generate a knowledge geometry collaborative metric graph that integrates geometric accuracy and semantic importance and its uncertainty.

7. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 1, characterized in that: The steps to achieve self-evolutionary learning and optimal resource allocation based on a multi-scale spatiotemporal memory structure include: Construct a multi-scale spatiotemporal memory storage structure containing short-term, medium-term, and long-term memory levels; Perform experience accumulation and extraction operations to generate a structured set of experience knowledge; Generate and evaluate candidate resource allocation plans based on resource constraints and knowledge geometry collaboration metrics, and select the optimal plan that maximizes system benefits; Perform knowledge base updates and evolutionary iterations to achieve continuous improvement of the system's cognitive capabilities.

8. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 6, characterized in that: The uncertainty distribution of the geometric reconstruction is calculated by the following formula: ； in, Indicates a point The geometric uncertainty value at Indicates a point The standard deviation of the geometric reconstruction at Indicates a point The average value of the geometric reconstruction at .

9. The method for monitoring deformation of deep foundation pits based on close-range photogrammetry according to claim 6, characterized in that: The uncertainty distribution of knowledge reasoning is calculated by the following formula: ； in, Indicates a point The knowledge uncertainty value at represents the set of deformation mode categories, Represents a given point Features In the case of , the point belongs to the deformation category The probability value of Indicates a point The feature vector extracted at .

10. The deep and large foundation pit deformation monitoring system based on close-range photogrammetry is characterized by: The method for performing the deformation monitoring method for deep and large foundation pits based on close-range photogrammetry according to any one of claims 1 to 9 comprises: Adaptive data acquisition and preprocessing module, which is used to receive risk area data in the knowledge graph, perform adaptive acquisition and preprocessing operations, and generate a standardized image set; The knowledge geometry interactive entropy reconstruction module is used to perform adaptive 3D reconstruction by combining standardized image sets with domain knowledge constraints to generate a 3D mesh model with semantic annotations. The deformation pattern extraction and annotation module is used to extract and semantically annotate deformation patterns of 3D mesh models based on knowledge graphs, generating a structured deformation pattern set and corresponding semantic interpretations. The uncertainty transfer mechanism module is used to build a bidirectional information transfer channel between the geometry domain and the knowledge domain based on the 3D mesh model and the deformation pattern set, and generate a knowledge geometry collaborative metric graph; Self-evolutionary learning and resource optimization module, used to achieve continuous evolution of system capabilities and optimal resource allocation based on multi-scale spatiotemporal memory structure; The risk assessment and decision support module is used to perform risk assessment and generate intervention recommendations based on deformation pattern sets, semantic interpretations, and monitoring knowledge bases, and output risk classification warning information and multi-level decision support solutions.

Images