249 results about "Algorithm design" patented technology

The invention discloses a large language model enhanced reinforcement learning training method for key automatic driving scene stable decision, and the method comprises the steps: constructing an LLM-based vehicle decision intelligent agent, and providing a high-level guidance strategy through a chain thinking technology and a knowledge experience pool; then, jointly constructing a dynamic intervention mechanism based on the environmental risk sign and the decision boundary parameter, thereby determining an LLM guidance opportunity; finally, through expert guidance algorithm design, the prior knowledge of the LLM is effectively integrated into the deep reinforcement learning strategy network by adopting an adaptive exploration-imitation fusion loss function. The DRL model is better in performance in a vehicle driving decision-making task and shows higher generalization ability in various scenes, and the robustness, the sample efficiency and the overall generalization level of the DRL model can be remarkably improved.

The invention provides a zero-sum game-based distributed elastic consensus optimal control method under denial of service attack of a multi-agent system, and the method specifically comprises the steps: firstly, constructing a multi-agent formation model in which a leader and a plurality of followers cooperatively move through graph theory knowledge and a multi-agent second-order state equation; secondly, in order to reduce the influence of denial of service attack on communication topology, a time-varying weight distributed elastic observer is provided to estimate the state of a leader, and the attacked condition of the leader is considered; then, by constructing an augmentation system, a distributed consistency tracking problem with a leader is converted into a local tracking problem between each follower and a virtual leader thereof; and finally, in order to solve the zero-sum game problem, introducing a Hamiltonian-Jacobian-Ansaxophone equation to realize optimal control input under maximum external disturbance, and realizing algorithm design by using single-evaluation reinforcement learning with experience playback and combining a gradient descent method.

The invention relates to the technical field of digital twinning, particularly provides a digital twinning construction method based on multi-source point cloud data fusion, and solves the problems of low multi-source point cloud data fusion precision, low calculation efficiency and poor adaptability to complex scenes in the prior art. The method comprises the following steps: S1, preprocessing multi-source point cloud data: carrying out data denoising and coordinate unified processing on the point cloud data of each source; s2, designing a point cloud data fusion algorithm: performing feature extraction and description, feature matching and fusion decision and weighted fusion calculation on the preprocessed point cloud data to obtain fused point cloud data; s3, constructing and optimizing a digital twinborn model: performing triangular mesh generation, model optimization processing, texture mapping and attribute addition on the fused point cloud data, and constructing the digital twinborn model; the digital twinning construction method based on multi-source point cloud data fusion can effectively solve the problems that multi-source point cloud data fusion is low in precision, low in calculation efficiency and poor in adaptability to complex scenes.

The invention discloses an intelligent heuristic algorithm generation method for unmanned aerial vehicle routing inspection path planning, belongs to the technical field of unmanned aerial vehicle autonomous navigation and intelligent optimization, and aims to plan a routing inspection path with the total cost as low as possible on a building routing inspection three-dimensional point set for an unmanned aerial vehicle. Aiming at the problems of high path planning complexity and low efficiency of artificial design of a heuristic algorithm in a building coverage inspection task, the method comprises the following steps: firstly, constructing a constraint model of an unmanned aerial vehicle inspection scene, and converting multi-objective optimization requirements such as building coverage rate, total detection time and flight path length into a heuristic algorithm design problem; compared with a traditional manual design method, the method enables the heuristic algorithm to reach or exceed the design level of field experts on indexes such as building coverage integrity, path length and calculation efficiency, and provides an efficient path planning solution for intelligent inspection of the unmanned aerial vehicle.

The invention provides a seed screening method and system based on a laser radar technology, and relates to the technical field of agricultural automation, and the method comprises the steps: extracting geometric features and defect features from optimized point cloud data; extracting mildew and insect pest spectral features from the optimized spectral data; a multi-dimensional feature set is obtained; calculating seed plumpness, roundness and surface roughness based on the optimized point cloud data; calculating a sag index and a crack feature based on point cloud edge detection; fusing the geometric features, the defect features and the pest spectral features to generate a comprehensive score; a machine learning model is utilized to execute final grading, and a first probability predictor based on laser radar features and a second probability predictor based on pest spectral features are trained respectively; and inputting the probability outputs of the two predictors into a second-layer grading model of the meta-learning architecture, and outputting a final quality grade. According to the method, efficient and accurate seed screening is realized through automatic process and algorithm design.

The invention relates to the technical field of power system court management, and particularly discloses a court identification and user binding method based on a special transformer acquisition terminal, which comprises the following steps: separating harmonic interference generated by start and stop of an electric welding machine through a kurtosis index and time-frequency energy entropy fusion detection method, and extracting a fundamental voltage and a pure power signal; constructing a dynamic time bending model containing phase consistency and gradient feature constraint based on the fundamental wave voltage, and calculating voltage track similarity to judge electrical coupling; for the users with electrical coupling, dynamic weight power cross-correlation analysis is fused to generate a comprehensive decision value, and a space-time diagram neural network is utilized to update a transformer area topology structure online in combination with a historical topology change event, so that the dynamic determination of the transformer area to which the users belong is realized; through multi-dimensional feature fusion and anti-interference algorithm design, the problems of phase deviation and waveform distortion caused by electric welding machine harmonic waves are reduced, and the transformer area identification accuracy and real-time performance are improved.

The invention discloses an intelligent verbal skill process configuration system and method based on a dynamic rule engine, and the system comprises a configuration analysis module, a process engine module, an execution engine module and a large model intention recognition module. The problems of rule solidification, high maintenance cost and low intention recognition accuracy in existing verbal skill process configuration are solved. The configuration analysis module analyzes a JSON format verbal skill rule configuration file to realize visual definition and dynamic update of the rule; the process engine module completes rule instantiation through variable replacement and drives dynamic routing based on an intention recognition result; the execution engine module instantiates a service class by using a reflection mechanism to realize decoupling of rules and service logic; the large model intention recognition module innovatively adopts a semantic similarity weighting algorithm, and intention recognition accuracy is remarkably improved by calculating semantic vector cosine similarity of user input and branch intention and combining service weight and context correlation degree.

The invention relates to the technical field of computer science and industrial engineering, in particular to a tool workshop scheduling method, which comprises the following steps of: acquiring production characteristics of a current tool workshop, generating an optimal batch scheduling strategy according to the production characteristics of the current tool workshop based on a preset evolutionary strategy, and automatically designing an algorithm by utilizing a preset large model, and calculating the minimum completion time and the maximum completion time of the optimal batch scheduling strategy, and completing tool workshop scheduling according to the minimum completion time and the maximum completion time based on the optimal batch scheduling strategy. Therefore, the problems of high workshop scheduling complexity and the like caused by strong coupling of batch division and working procedure sorting, strict constraint between working procedures and high resource networking degree in workshop scheduling are solved, an automatic algorithm design framework based on a large model is introduced, the expert algorithm design time is shortened, and the workshop scheduling efficiency is improved. Therefore, the efficiency and feasibility of aircraft tooling production scheduling are improved.

The invention relates to the technical field of air temperature risk assessment, in particular to an AdaBoost-based urban air temperature inference and thermal exposure risk assessment method. The method comprises the following steps: acquiring original city form data and city meteorological data of a city; performing data preprocessing on the original city form and the city meteorological data of the city to generate processed original city form data and city meteorological data; integrating the processed original city form data and city meteorological data into a model training set and a model test set; and designing an artificial neural network architecture by adopting an AdaBoost algorithm, and performing model training on the model training set by utilizing the artificial neural network architecture to generate an urban air temperature inference pre-model. According to the method, by fusing multi-source data, applying an advanced machine learning algorithm and enhancing sensitivity analysis and Monte Carlo simulation, the accuracy and reliability of urban air temperature inference and thermal exposure risk assessment are improved.

The invention provides a vehicle control method and device, electronic equipment and a medium, and relates to the technical field of automatic driving, and the method comprises the steps: constructing a three-part loss function comprising a decision variable, a state variable and a control variable, and adding the outputs of the three parts in each time step to obtain a first sum value, the first sum values of the multiple time steps are accumulated to form a unified target sum value, a driving variable sequence for minimizing the target sum value is jointly solved under the constraint of a dynamics mathematical model of the first vehicle and a first constraint condition, and finally the vehicle is directly controlled through the optimal target driving variable sequence. Therefore, a coupling channel of discrete decision making (namely lane decision making) and continuous trajectory planning is effectively opened, and the lane decision making and trajectory generation are synchronized and coordinated under the same optimization target, so that the safety, the driving efficiency and the driving comfort of the automatic driving vehicle in a complex traffic scene are improved; the overall driving performance better than that of a traditional split algorithm design is achieved.

The invention relates to the field of deep learning, and discloses a multi-agent collaborative task state embedding method based on a multi-scale hypergraph and multi-dimensional aggregation. In order to solve the problems that an existing graph neural network is difficult to capture a high-order interaction relation, poor in dynamic environment adaptability and limited in communication, the method comprises the steps of constructing a state observation graph; calculating an adjacent matrix according to explicit states such as position and speed; generating a latent layer feature adjacency matrix through nonlinear conversion and similarity calculation of a graph convolutional network, and fusing the latent layer feature adjacency matrix with the Hadamard product of the interactive graph; constructing a multi-scale hypergraph (containing S scales) according to the latent layer matrix, and searching a high-density sub-matrix to form hyperedges; building a two-stage information aggregation model: integrating multi-dimensional features and calculating association degree and interaction types in a hyperedge aggregation stage, and updating node features by using a graph attention network GAT in a node aggregation stage; a multi-agent soft behavior-commentator algorithm MASAC is fused, a behavior and reward function is designed, and an MHGNN-MASAC model is formed; the method is applied to cooperative control task decision.

The invention provides a radio frequency device hyper-parameter automatic optimization algorithm library. The radio frequency device hyper-parameter automatic optimization algorithm library comprises an interface module, a directional optimization algorithm module and a multi-modal optimization engine module, according to the optimization algorithm library, the automation level and the performance optimization efficiency of radio frequency device design are remarkably improved by constructing a domain knowledge graph covering six core devices including a radio frequency filter, a power amplifier, a numerical control attenuator and a low-noise amplifier and combining a customized optimization algorithm and a cross-level cooperation strategy. On the algorithm design level, aiming at the specific characteristics of the device, different hybrid optimization strategies are adopted, the global search capability of a genetic algorithm and a particle swarm optimization algorithm and the local optimization advantage of a gradient descent method are combined, and dynamic weight adjustment is matched, so that the targeted radio frequency device optimization process design is realized. According to the optimization algorithm library, through an independent calling interface for packaging the automatic design algorithm, the use difficulty of the automatic design algorithm is simplified, and the optimization process is normalized.

The invention discloses a port field knowledge graph automatic construction method based on cooperation of multiple intelligent Agents, and the method comprises the steps: processing structured data, semi-structured data and non-structured data of a port field through intelligent Agents with an autonomous decision-making capability, and achieving the entity recognition, relation extraction and knowledge fusion; a plurality of intelligent Agents with the cooperation function are used for executing knowledge discovery, new knowledge verification, conflict detection, knowledge fusion and graph updating tasks respectively, cooperative communication among the intelligent Agents is achieved through a message queue, and dynamic evolution of a knowledge graph is completed; processing the knowledge graph by adopting a customized knowledge graph embedding method and a semantic fusion algorithm designed for professional terms and knowledge structures in the port field; and storing the processed knowledge graph data based on a distributed architecture, wherein the distributed architecture supports high concurrent processing and real-time response. According to the method, the port domain knowledge graph can be efficiently and automatically constructed.

The invention provides a CF / PEEK milling surface roughness modeling method based on an L1-LSSVR algorithm. The method comprises the steps that S1, static milling process parameters serve as experimental variables, an experimental scheme is designed, dynamic signals in the milling process are collected in the experimental process, and after the experiment is finished, the CF / PEEK three-dimensional surface roughness is measured; s2, time domain features of the dynamic signals are extracted, and the time domain features are used for L1-LSSVR model training after being combined with static milling process parameters; s3, analyzing the correlation between the features through a Pearson's correlation coefficient, and eliminating redundant features with similar influence rules on the roughness of the three-dimensional surface; and performing feature selection by using L1 norm regularization, and screening out features having high correlation with the three-dimensional surface roughness. S4, establishing a CF / PEEK milling surface roughness model by taking the features screened in the step S3 as the input of an LSSVR algorithm and the CF / PEEK milling three-dimensional surface roughness as the output; according to the algorithm designed by the invention, modeling of the milling three-dimensional surface roughness of the CF / PEEK material can be effectively realized.

The invention discloses a reinforcement learning-driven public building user energy consumption behavior simulation method, which comprises the following steps of: 1, acquiring basic data, and building a simulation environment model for simulating a mapping relationship between a user energy consumption behavior and building energy consumption as well as an environment comfort degree; 2, defining a state space and an action space of the intelligent agent, selecting a depth deterministic strategy gradient algorithm as a core learning algorithm of the intelligent agent, designing a multi-target comprehensive reward function of the core learning algorithm, and embedding a public building energy consumption control target; step 3, using an intelligent agent to interact with a simulation environment, simulating an energy consumption process of a user, and identifying different scenes according to a preset rule; 4, the intelligent agent iteratively updates network parameters in the core learning algorithm based on the number of samples collected in the interaction process, and after training convergence, a user energy consumption behavior optimization scheme is output.

The invention belongs to the technical field of advertisement media, and particularly relates to a reinforcement learning optimization channel weight double-attribution advertisement effect analysis system and method, and the method comprises the steps: collecting multi-source heterogeneous data, and carrying out the preprocessing, so as to obtain standardized data; a reinforcement learning algorithm is designed, weights are distributed according to dynamic time sequence attention, a double-weight coordination action space is constructed, and a multi-modal coding state and action requirements are obtained; multi-modal data is coded, text, time sequence and structured data are coded respectively, and feature vectors matched with the model are output through modal attention fusion; designing a reward function, and generating a reward signal for real-time optimization; and a real-time optimization mechanism is constructed, parameters are updated in a layered manner, gray scale verification is performed, multi-source heterogeneous data preprocessing and enhancement algorithm design are promoted, and a closed loop is formed. According to the method, the effectiveness and accuracy of advertisement effect analysis are effectively improved, and the adaptability to multi-source heterogeneous data and dynamic scenes is enhanced.

The invention discloses a large and small model collaborative robot autonomous assembly method and system, and belongs to the field of advanced manufacturing technology intellectualization. The problems that traditional robot assembly depends on manual programming, adaptability is poor, and an algorithm cannot be optimized autonomously are solved. The method comprises the following steps: receiving an assembly project request submitted by a user in a natural language; the request is decoupled into a structured action primitive through an arrangement agent, and tasks are allocated; an algorithm design agent retrieves and adapts an algorithm from an open source library based on the action primitive, and completes training in a sandbox environment; the verification agent is responsible for automatic data acquisition and labeling, and carrying out overall performance test and iterative optimization on an integrated algorithm in a simulation environment; and finally generating an executable field control script. According to the method, the problems of algorithm autonomous design, simulation verification and closed-loop optimization when the robot faces a complex assembly task are solved, the development threshold and the deployment risk are reduced, and the intelligent level and the flexible adaptive capacity of an assembly system are remarkably improved.

The invention relates to the field of intelligent scheduling allocation, in particular to an edge cloud computing resource allocation optimization method based on deep learning, which adopts a space-time prediction algorithm based on multi-head attention and gating mechanism optimization to design time coding and space coding. The spatial relationship and interaction between time sequence characteristics of the computing power load and edge server nodes are captured, and meanwhile, a multi-head attention mechanism and expansion causal convolution are combined, so that instantaneous computing power load fluctuation can be captured, and the long-term trend of the computing power load can be mined; therefore, a reliable basis is provided for subsequent computing power scheduling by predicting an accurate computing power load. The invention designs an alternating direction multiplier method based on genetic algorithm optimization, which is not only suitable for a nonlinear and multi-constraint optimization problem, but also can be expanded to a larger-scale distributed edge node cloud computing system, and meanwhile, a global optimal solution is quickly approached through the genetic algorithm, so that the quality of an initial solution is improved, and model convergence is accelerated; and the distributed collaborative allocation scheduling efficiency is improved.

The invention relates to the field of airplane airborne pod servo control algorithm design, and provides a method and a system for calculating external force disturbance resistance according to current feedback of a servo motor. The method comprises the steps that three-phase currents Ia, Ib and Ic and an electrical angle theta are collected, two-phase currents Id and Iq are obtained through Clarke conversion and Park conversion, and the Iq is defined as an external force disturbance resistance estimator IQ; establishing a set value r (t) and a deviation e (t) = r (t)-IQ, calculating an output u (t) of a PID controller, and writing the u (t) as a q-axis control reference into a driver: mapping to Vq and Vd in a voltage control mode for inverse transformation and PWM modulation, mapping to q-axis current reference Iqref in a current control mode, converting the q-axis current reference Iqref into a corresponding voltage instruction by a driver current loop, and driving a servo motor to generate an electromagnetic torque for counteracting disturbance and reactance. A complex motor model and a high-end inertial sensor are not needed, the calculation amount is small, disturbance can be rapidly and accurately estimated, and the stable control performance of the aircraft pod in a complex flight environment is guaranteed.

The invention discloses a beam focusing method and system based on a liquid neural network and meta learning, and belongs to the technical field of wireless communication, and the method comprises the following steps: determining a system model; determining an optimization target; and joint optimization. The LNNM-BF algorithm with high robustness and unsupervised learning is provided, the LNNM-BF algorithm designs a phase shift matrix theta through an index matrix S and a phase compensation matrix phi, then an analog precoding matrix FPS and a digital precoding matrix FBB are designed at a BS end, and the near-field dual-beam splitting effect can be well relieved without introducing an extra true time delayer. Besides, the LNNM-BF algorithm enhances the adaptability of beam focusing to imperfect CSI through a meta-learning framework and improves the robustness of the algorithm, and the embedded LNN dynamically adjusts the topological structure and the connection weight between neurons through the currently input and previously implied iteration information to realize the maximization of SE.

The invention discloses a heterogeneous client autonomous collaboration method based on a twin channel model in an air-ground clustering federated learning scene, and belongs to the technical field of machine training. Dividing a client cluster by adopting an AP algorithm and preferably selecting a part of unmanned aerial vehicles from the unmanned aerial vehicles as cluster center nodes; an HT-UCB algorithm is adopted in the cluster, a multi-dimensional reward function is designed, and clients participating in global training are dynamically selected; the cluster center nodes and the air-ground heterogeneous clients selected in the cluster execute federated learning, and finally, the center node aggregates all cluster center node parameters to update a global model and perform loop iteration until the model converges; according to the invention, the identification precision and convergence rate of the overall model and the overall utilization rate of resources in the cooperative inspection task of the unmanned aerial vehicle and the unmanned vehicle are effectively improved.

An algorithm design method based on a large-scale language model comprises the steps that algorithm design requirements are received, and the algorithm design requirements comprise problem description; generating at least one candidate algorithm based on the algorithm design requirements; the to-be-evolved algorithm is iteratively evolved through the large-scale language model to obtain a target algorithm conforming to algorithm design requirements, the to-be-evolved algorithm comprises an evolved algorithm and / or a candidate algorithm, in each iteration process, a prompt is generated based on the to-be-evolved algorithm, and the prompt is used for guiding the large-scale language model to evolve. Therefore, through an explicit guidance mode, in the process of generating the algorithm by using the large-scale language model, clear guidance can be provided for the large-scale language model, so that the large-scale language model automatically generates the algorithm with relatively high accuracy, and the finally designed algorithm can be comparable with or even surpass the algorithm of manual customization design; and the difficulty of algorithm design is reduced.

This invention discloses a chip layout image generation method based on a boundary tracking scan line algorithm. By designing a sub-image-polygon data structure and establishing a maximum matrix of adjacent polygons in the sub-image, this method improves the boundary scan line algorithm to quickly segment the complete chip layout into sub-images and generate corresponding binary images. The algorithm designs a sub-image bounding box-polygon bounding box data structure to quickly locate which polygons are most likely to intersect with the sub-image. Saving this data constructs the maximum bounding box of all polygons contained in the current sub-image, eliminating the need for the algorithm to calculate the intersection points of polygons with the sub-image, thus reducing the time consumption for intersection point calculation. Furthermore, a boundary tracking method is designed to improve the scan line algorithm. By saving each point in the boundary, when scanning a row, it is not necessary to calculate the intersection points of the scan line and the polygon boundary, further reducing the algorithm complexity.

This application provides a vehicle control method, device, electronic device, and medium, relating to the field of autonomous driving technology. It constructs a three-part loss function comprising decision variables, state variables, and control variables, and sums the outputs of these three components at each time step to obtain a first sum. The first sums from multiple time steps are then accumulated to form a unified target sum. Under the constraints of a first vehicle's dynamic mathematical model and first constraints, a sequence of driving variables that minimizes this target sum is jointly solved. Finally, the vehicle is directly controlled using this optimal target driving variable sequence. This effectively bridges the coupling channel between discrete decision-making (i.e., lane decision) and continuous trajectory planning, enabling lane decision-making and trajectory generation to work synchronously and collaboratively under the same optimization objective. This improves the safety, driving efficiency, and ride comfort of autonomous vehicles in complex traffic scenarios, achieving superior overall driving performance compared to traditional fragmented algorithm designs.

The present invention discloses a method for constructing a heterogeneous parallel Helmholtz operator for a DCU cluster, belonging to the technical field of computational fluid dynamics. The present invention proposes a method for constructing a heterogeneous parallel Helmholtz operator for a DCU cluster, which mainly includes two parts: block matrix multiplication memory access optimization and task decomposition algorithm design. Compared with the existing design, the present invention no longer performs fine-grained access to the main memory, but instead fully utilizes the on-chip shared memory and registers on the DCU to perform block-parallel prefetching of matrices, significantly reducing memory access overhead. In addition, the present invention enables reasonable allocation of tasks on the DCU, fully utilizing the computing power of the DCU, and the acceleration effect increases accordingly with the increase of the interpolation order. Finally, the load balancing problem between the program CPU and DCU, and between multiple DCUs is solved, thereby improving the scalability of the program. The use of multiple DCUs can further improve the acceleration ratio.

The present application relates to the technical field of power system stability control simulation analysis, and provides a transient simulation analysis and decision method, device and equipment for power grid harmonic oscillation, which comprises the following steps: combining the model and model parameters of electromagnetic transient offline simulation with the topology of the power grid to be analyzed to obtain an electromagnetic transient example, and then describing the calculation process of electromagnetic transient with a hierarchical directed acyclic graph, and taking each calculation model with redistributed resources as the electromagnetic transient algorithm; performing risk analysis on the harmonic oscillation condition according to the simulation result, and making an optimized decision on the harmonic oscillation suppression measures; through example design and electromagnetic transient algorithm design, an electromagnetic online simulation model is constructed to meet the simulation requirements of large power grids and realize the analysis and suppression decision of harmonic oscillation.

The invention provides a power system black-start partition decision method based on lightweight reinforcement learning, and the method comprises the steps: constructing a black-start model of a power system containing multiple physical constraints, and formalizing the black-start model into a Markov decision process, the Markov decision process comprises a state space, an action space, a reward function and a transfer process; decomposing the power system into a plurality of sub-regions by adopting a dynamic partition recovery strategy; based on the Markov decision process, a depth deterministic strategy gradient algorithm is adopted to design and train a lightweight strategy network of deep reinforcement learning; enabling the intelligent agent to interact with the environment in each sub-region through the lightweight strategy network of deep reinforcement learning, and learning an optimal decision strategy; and according to the optimal decision strategy, performing partition power supply capability recovery on the power system. According to the invention, black-start rapid optimization decision in a complex power grid environment is realized.

The application provides a robot system distributed formation control method and device based on U-K theory, which comprises the following steps: establishing a dynamic model of a networked robot system by adopting a Newton-Euler mechanical method according to system parameters in the networked robot system moving in a two-dimensional space, establishing a communication relationship between robots according to a preset algebraic graph theory strategy, setting a communication topological graph of a directed weighted graph with a directed spanning tree, converting an expected formation control target into a constraint equation, combining the dynamic model with a U-K equation to calculate a servo control force required by a non-complete constraint robot system to complete the expected formation control target, and designing a distributed formation controller in a leaderless case based on the servo control force by adopting a preset distributed formation control algorithm. The technical scheme solves the technical problem of lack of universality in the related art, expands the application range, is more general and universal, and guarantees the formation control stability on the basis of simplifying the calculation of the control force.

The present invention discloses a method and system for blur removal and jitter correction in visual monitoring of super high-rise buildings, comprising the following steps: a visual sensor acquires a video stream at a fixed frame rate, and locates an initial region of interest (ROI) by mapping prior information; the spatial gradient field is calculated for the ROI image of the current frame, and a Gaussian-weighted two-dimensional gradient structure tensor matrix is ​​constructed; for the clear image sequence that passes blur detection, variational mode decomposition is then used in the time domain to decouple the displacement time series into eigenmode functions of different frequencies, and the low-frequency components are reconstructed to preserve the true deformation of the building; the reconstructed low-frequency displacement signal is output as the final monitoring result. The present invention employs a lightweight algorithm design throughout, significantly reducing the matrix operation dimension of subsequent processing through a dynamic ROI clipping mechanism; the structural tensor eigenvalues ​​are solved using a direct algebraic analytical method, avoiding complex matrix iterative decomposition.

The invention provides a method and system for evaluating the integrated photovoltaic potential of buildings in a high-density urban area, and the method comprises the steps: obtaining the color point cloud data of a city through the photographing and surveying of an unmanned plane, introducing the physical priori knowledge through algorithm design, and assisting a neural network in extracting a feature vector for representing the characteristics of a building group from the data; comprising the steps of point cloud sampling, normal estimation, window-wall ratio estimation and geographic orientation angle calculation, so that the model can learn implicit relation between the urban form and a photovoltaic potential prediction result; aiming at a scene of comprehensively realizing photovoltaic integration popularization of medium and large-scale buildings, the power generation efficiency of building facades and photovoltaic modules mounted at different positions of windows and walls is considered, feature vectors are input into a regression prediction framework formed by a one-dimensional point cloud neural network, evaluation errors are reduced, systematic deviation of resolution is avoided, and the evaluation accuracy is improved. Data acquisition and prediction are facilitated, and the function of predicting the maximum photovoltaic power generation potential after the building is comprehensively paved with the BIPV is realized.