Personnel real-time positioning early warning method and system based on Beidou

Abstract

The invention belongs to the technical field of Beidou positioning early warning, and discloses a Beidou-based personnel real-time positioning early warning method and system. The method comprises the following steps: monitoring a central server, and constructing a scene digital twinborn map and an artificial intelligence model; the Beidou positioning equipment is used for acquiring real-time Beidou positioning data and real-time personnel identity data of personnel; the monitoring center server is used for correcting the positioning data according to the real-time Beidou positioning data; the monitoring center server performs abnormal behavior identification according to the scene digital twin map, the corrected real-time Beidou positioning data and the real-time personnel identity data; the monitoring center server is used for generating a personnel abnormity early warning strategy according to the real-time abnormal behavior recognition result; and the monitoring center server is used for carrying out visual display by using a scene digital twin map. According to the invention, the problems of insufficient positioning precision, inaccurate abnormal behavior identification and low intelligent degree in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of Beidou positioning and warning, and particularly relates to a method and system for real-time positioning and warning of personnel based on Beidou. Background Art

[0002] With the development of society and the progress of technology, personnel positioning and warning play an increasingly important role in many fields such as security monitoring, emergency response, etc., especially in environments such as high-risk operation areas, mines, etc. Therefore, by online monitoring of personnel to achieve high-precision real-time positioning and abnormal behavior alarm with seamless indoor and outdoor switching is of great significance for ensuring personnel safety and improving emergency response efficiency, and has become the development direction of this field.

[0003] The existing personnel positioning and warning methods mainly have the following problems:

[0004] 1) Insufficient positioning accuracy: Traditional positioning technologies such as GPS, Beidou, etc. are prone to signal interference in complex environments (such as high-rise buildings, underground spaces, etc.), resulting in large positioning errors and often having a certain delay, making it difficult to achieve real-time positioning in the true sense;

[0005] 2) Inaccurate identification of abnormal behaviors: Existing technologies often rely on simple rules or shallow learning algorithms, unable to mine the deep features of positioning data, making it difficult to accurately identify complex and changeable abnormal behaviors, and due to the imperfect identification model, it is easy to misjudge normal behaviors as abnormal behaviors, resulting in a high false alarm rate;

[0006] 3) Low degree of intelligence: Existing personnel abnormal warning often relies on artificial rule-making or threshold triggering, with a low degree of intelligence and unable to meet customized personnel abnormal warning. Summary of the Invention

[0007] In order to solve the problems of insufficient positioning accuracy, inaccurate identification of abnormal behaviors, and low degree of intelligence existing in the prior art, the purpose of the present invention is to provide a method and system for real-time positioning and warning of personnel based on Beidou.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A method for real-time positioning and warning of personnel based on Beidou, comprising the following steps:

[0010] The monitoring center server uses digital twin technology to construct a scene digital twin map, and uses artificial intelligence algorithms to construct a positioning data correction model, an abnormal behavior identification model, and a personnel abnormal warning model;

[0011] The Beidou positioning device collects the real-time Beidou positioning data and real-time personnel identity data of personnel, and sends the real-time Beidou positioning data and real-time personnel identity data to the monitoring center server;

[0012] The monitoring center server uses the positioning data correction model according to the real-time Beidou positioning data to correct the positioning data and obtain the corrected real-time Beidou positioning data;

[0013] The monitoring center server uses the abnormal behavior recognition model according to the scenario digital twin map, the corrected real-time Beidou positioning data and the real-time personnel identity data to perform abnormal behavior recognition and obtain the real-time abnormal behavior recognition result;

[0014] If the monitoring center server determines that the real-time abnormal behavior recognition result indicates the existence of abnormal behavior, it uses the personnel abnormal early warning model according to the real-time abnormal behavior recognition result to generate a personnel abnormal early warning strategy and obtain the real-time personnel abnormal early warning strategy;

[0015] The monitoring center server uses the scenario digital twin map according to the real-time personnel abnormal early warning strategy to visually display the real-time abnormal behavior recognition result, the corrected real-time Beidou positioning data and the real-time personnel identity data.

[0016] Furthermore, the monitoring center server uses digital twin technology to construct a scenario digital twin map and uses artificial intelligence algorithms to construct a positioning data correction model, an abnormal behavior recognition model and a personnel abnormal early warning model, including the following steps:

[0017] The monitoring center server obtains the GIS data of the application scenario and constructs a three-dimensional simulation map of the application scenario according to the GIS data;

[0018] Integrate digital twin data into the three-dimensional simulation map of the scenario to obtain the scenario digital twin map of the application scenario;

[0019] Collect a number of historical Beidou positioning data and corresponding historical personnel identity data, and perform preprocessing to obtain a number of preprocessed historical Beidou positioning data and a number of preprocessed historical personnel identity data;

[0020] Set a corresponding true error value for each preprocessed historical Beidou positioning data to obtain a number of preprocessed historical Beidou positioning data with true error values set;

[0021] According to a number of preprocessed historical Beidou positioning data with true error values set, use the deep learning and decision tree fusion algorithm to construct a positioning data correction model and generate a number of corrected historical Beidou positioning data;

[0022] Based on the scenario digital twin map, a number of corrected historical Beidou positioning data, and the corresponding preprocessed historical personnel identity data, use a multi-modal fusion deep learning algorithm to construct an abnormal behavior recognition model and generate a number of historical abnormal behavior recognition results;

[0023] Based on a number of historical abnormal behavior recognition results, use a reinforcement learning and adversarial training fusion algorithm to construct a personnel abnormal warning model and generate a number of historical warning strategy generation experiences.

[0024] Furthermore, the positioning data correction model is constructed based on the DBN-RF algorithm.

[0025] Furthermore, the abnormal behavior recognition model is constructed based on the 3D U2-Net-LSTM-DBN-Attention-MLP-ICPO algorithm.

[0026] Furthermore, the personnel abnormal warning model is constructed based on the MOPPO-cGAN algorithm.

[0027] Furthermore, the monitoring center server, according to the real-time Beidou positioning data, uses the positioning data correction model to correct the positioning data and obtain the corrected real-time Beidou positioning data, including the following steps:

[0028] The monitoring center server preprocesses the real-time Beidou positioning data to obtain the preprocessed real-time Beidou positioning data and inputs the preprocessed real-time Beidou positioning data into the positioning data correction model;

[0029] Use the positioning data correction model to extract the first real-time positioning data feature of the preprocessed real-time Beidou positioning data;

[0030] According to the first real-time positioning data feature, predict the error value and generate the corresponding real-time error value;

[0031] According to the real-time error value, correct the corresponding preprocessed real-time Beidou positioning data to obtain the corrected real-time Beidou positioning data.

[0032] Furthermore, the monitoring center server, according to the scenario digital twin map, the corrected real-time Beidou positioning data, and the real-time personnel identity data, uses the abnormal behavior recognition model to perform abnormal behavior recognition and obtain the real-time abnormal behavior recognition result, including the following steps:

[0033] The monitoring center server extracts the real-time scene map feature of the scenario digital twin map;

[0034] Extract the second real-time positioning data feature of the corrected real-time Beidou positioning data;

[0035] Preprocess the real-time personnel identity data to obtain the preprocessed real-time personnel identity data, and extract the real-time identity data features of the preprocessed real-time personnel identity data;

[0036] According to the preset attention weight values, perform weighted fusion on the real-time scene map features, the second real-time positioning data features, and the real-time identity data features to obtain the real-time weighted fusion features;

[0037] Perform abnormal behavior recognition on the real-time weighted fusion features to obtain the real-time recognition probability distribution;

[0038] Optimize the recognition result of the real-time recognition probability distribution to obtain the real-time abnormal behavior recognition result.

[0039] Furthermore, for the monitoring center server, if it is determined that the real-time abnormal behavior recognition result indicates the existence of abnormal behavior, then according to the real-time abnormal behavior recognition result, use the personnel abnormal warning model to generate a personnel abnormal warning strategy to obtain the real-time personnel abnormal warning strategy, including the following steps:

[0040] For the monitoring center server, if it is determined that the real-time abnormal behavior recognition result indicates the existence of abnormal behavior, then input the real-time abnormal behavior recognition result into the personnel abnormal warning model;

[0041] Randomly extract several historical strategy generation experiences;

[0042] According to several historical strategy generation experiences and the real-time abnormal behavior recognition result, use the personnel abnormal warning model to generate a personnel abnormal warning strategy to obtain the real-time personnel abnormal warning strategy.

[0043] Furthermore, for the monitoring center server, according to the real-time personnel abnormal warning strategy, use the scene digital twin map to visually display the real-time abnormal behavior recognition result, the corrected real-time Beidou positioning data, and the real-time personnel identity data, including the following steps:

[0044] The monitoring center server generates a corresponding real-time personnel abnormal warning signal according to the real-time personnel abnormal warning strategy;

[0045] Locate the real-time abnormal personnel according to the corrected real-time Beidou positioning data, and highlight the real-time abnormal personnel on the scene digital twin map according to the real-time personnel abnormal warning strategy;

[0046] Visually display the real-time personnel abnormal warning signal, the real-time abnormal behavior recognition result, the corrected real-time Beidou positioning data, and the real-time personnel identity data on the scene digital twin map according to the real-time personnel abnormal warning strategy.

[0047] A Beidou-based personnel real-time positioning and early warning system is used to implement a real-time personnel positioning and early warning method. The system includes a monitoring center server and several Beidou positioning devices. The several Beidou positioning devices are all communicated with the monitoring center server. The monitoring center server includes a model building unit, a positioning data correction unit, an abnormal behavior recognition unit, a personnel abnormality early warning unit and a visualization display unit which are connected in sequence.

[0048] The beneficial effects of the present invention are:

[0049] The present invention provides a Beidou-based personnel real-time positioning and early warning method and system. The Beidou positioning device collects the real-time Beidou positioning data and real-time personnel identity data of the personnel in real time, thereby improving the real-time performance. The real-time Beidou positioning data is corrected by a positioning data correction model, thereby reducing signal interference in a complex environment, realizing high-precision real-time positioning with seamless switching between indoor and outdoor environments, effectively reducing positioning errors, improving positioning accuracy, ensuring accurate acquisition of personnel positions, and providing a reliable data basis for subsequent abnormal behavior recognition and early warning. The advanced artificial intelligence algorithm is used to construct an abnormal behavior recognition model, deeply excavates the deep features of the positioning data, and combines a scene digital twin map and real-time personnel identity data to perform a fusion analysis of multimodal and multi-data sources, thereby improving the accuracy of abnormal behavior recognition, realizing customized analysis based on identity recognition, avoiding misjudging normal behavior as abnormal behavior, and reducing the false alarm rate. Based on real-time data and artificial intelligence models, a real-time personnel abnormal early warning strategy is automatically generated, thereby realizing customized and intelligent personnel abnormal early warning, improving the timeliness and accuracy of the early warning, and gaining precious time for emergency response. The scene digital twin map is used to visualize the real-time positioning and early warning of personnel, thereby improving the convenience and practicality of observation.

[0050] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flowchart of the Beidou-based personnel real-time positioning and early warning method in the present invention.

[0052] Figure 2 It is a structural block diagram of the Beidou-based personnel real-time positioning and early warning system in the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0054] Embodiment 1:

[0055] like Figure 1 As shown, this embodiment provides a Beidou-based personnel real-time positioning warning method, comprising the following steps:

[0056] S1: The monitoring center server uses digital twin technology to construct a scene digital twin map and uses artificial intelligence algorithms to construct a positioning data correction model, an abnormal behavior recognition model, and a personnel abnormal warning model, including the following steps:

[0057] S1-1: The monitoring center server obtains the GIS data of the application scenario and constructs a three-dimensional simulation map of the application scenario according to the Geographic Information System (GIS) data;

[0058] S1-2: Integrate digital twin data for the three-dimensional simulation map of the scene to obtain a scene digital twin map of the application scenario;

[0059] S1-3: Collect a number of historical Beidou positioning data and corresponding historical personnel identity data, and perform preprocessing to obtain a number of preprocessed historical Beidou positioning data and a number of preprocessed historical personnel identity data;

[0060] S1-4: Set a corresponding true error value for each preprocessed historical Beidou positioning data to obtain a number of preprocessed historical Beidou positioning data with true error values set;

[0061] S1-5: According to a number of preprocessed historical Beidou positioning data with true error values set, use a deep learning and decision tree fusion algorithm to construct a positioning data correction model and generate a number of corrected historical Beidou positioning data;

[0062] The positioning data correction model is constructed based on the Deep Belief Network (DBN)-Random Forest (RF) algorithm, and the positioning data correction model includes a first positioning data feature extraction module constructed based on the DBN algorithm and a positioning data correction module constructed based on the RF algorithm, which are connected in sequence;

[0063] The first positioning data feature extraction module performs deep-level feature extraction on the positioning data. Using the multi-layer structure of the deep belief network (DBN), it learns the complex patterns and internal relationships in the positioning data, and the extracted features are more representative and abstract, which helps the subsequent correction module to more accurately understand and process the positioning data. The positioning data correction module uses the Random Forest (RF) algorithm to correct the extracted positioning data features. Through the ensemble learning of multiple decision trees, it classifies or regresses and corrects the positioning data to improve the accuracy and reliability of the data. Each tree randomly selects a feature subset during training, enhancing the generalization ability and anti-overfitting ability of the model;

[0064] According to a number of preprocessed historical Beidou positioning data with true error values, using a deep learning and decision tree fusion algorithm, construct a positioning data correction model, and generate a number of corrected historical Beidou positioning data, including the following steps:

[0065] S1-5-1: Use the DBN-RF algorithm to construct an initial positioning data correction model; the initial positioning data correction model includes an initial first positioning data feature extraction module and an initial positioning data correction module;

[0066] S1-5-2: According to a number of preprocessed historical Beidou positioning data with true error values, train and optimize the initial first positioning data feature extraction module to obtain an optimized first positioning data feature extraction module, and generate the first historical positioning data features of each preprocessed historical Beidou positioning data;

[0067] S1-5-3: According to a number of first historical positioning data features, train and optimize the initial positioning data correction module to obtain an optimized positioning data correction module, and generate the importance scores of each feature component of the first historical positioning data features;

[0068] S1-5-4: According to the importance scores of the feature components, select the first several feature components as the key feature components of the first historical positioning data features;

[0069] S1-5-5: Based on a number of key feature components, use the optimized positioning data correction module to predict the error value and generate the historical error values of each historical positioning data feature;

[0070] S1-5-6: Statistically compare the historical error values with the true error values of the corresponding preprocessed historical Beidou positioning data to obtain the prediction accuracy rate. If the prediction accuracy rate is greater than the accuracy threshold, output the final positioning data correction model;

[0071] S1-5-7: According to the historical error values, correct the corresponding preprocessed historical Beidou positioning data to obtain a number of corrected historical Beidou positioning data;

[0072] S1-6: According to the scenario digital twin map, a number of corrected historical Beidou positioning data, and the corresponding preprocessed historical personnel identity data, use a multi-modal fusion deep learning algorithm to construct an abnormal behavior recognition model, and generate a number of historical abnormal behavior recognition results;

[0073] The abnormal behavior recognition model is constructed based on the 3D U-shaped network (3D Unet with U-shaped architecture, 3D U2-Net) - Long Short-Term Memory (LSTM) algorithm - Deep Belief Network (DBN) - Attention - Multilayer Perceptron (MLP) - Improved Crested Porcupine Optimizer (ICPO) algorithm. The abnormal behavior recognition model includes a scene map feature extraction module constructed based on the 3D U2-Net algorithm, an identity data feature extraction module constructed based on the LSTM algorithm, a second positioning data feature extraction module constructed based on the DBN algorithm, an attention weight module constructed based on the Attention mechanism, an abnormal behavior recognition module constructed based on the MLP algorithm, and a recognition result optimization module constructed based on the ICPO algorithm. The scene map feature extraction module, the identity data feature extraction module, and the second positioning data feature extraction module are all connected to the attention weight module, and the attention weight module, the abnormal behavior recognition module, and the recognition result optimization module are connected in sequence;

[0074] The scene map feature extraction module extracts the scene map features of the scene digital twin map. Through the encoder-decoder structure, it captures the spatial information and context relationship of the scene, effectively extracts the three-dimensional spatial features of the scene map, provides rich information for subsequent abnormal behavior recognition, and effectively extracts the three-dimensional spatial features of the scene map, provides rich information for subsequent abnormal behavior recognition; The second positioning data feature extraction module uses the deep belief network structure of DBN to extract features from the second positioning data, deeply mines the internal features of the second positioning data, improves the expression ability of the features, and enhances the model's understanding and analysis ability of the positioning data; The identity data feature extraction module uses the long short-term memory ability of LSTM to extract temporal features from the identity data, accurately extracts the temporal features of the identity data, provides an important basis for abnormal behavior recognition, and realizes the effect of customized analysis of the positioning data of different permission identities; The attention weight module uses the Attention mechanism to allocate weights to the scene map features, identity data features, and second positioning data features, effectively focuses on the key features, improves the accuracy of abnormal behavior recognition, and enhances the model's adaptability and generalization ability to complex scenes; The abnormal behavior recognition module uses the multi-layer perceptron structure of MLP to recognize abnormal behaviors from the fused features. Through the non-linear activation function and multi-layer structure, it learns complex behavior patterns, has strong non-linear modeling ability, and adapts to various abnormal behavior patterns; The improved crown porcupine optimizer (ICPO) algorithm is used to optimize the recognition results, effectively improves the accuracy and stability of the recognition results, enhances the model's optimization ability, and makes the recognition results more reliable;

[0075] According to the scene digital twin map, several corrected historical Beidou positioning data, and the corresponding preprocessed historical personnel identity data, use the multi-modal fusion deep learning algorithm to construct an abnormal behavior recognition model and generate several historical abnormal behavior recognition results, including the following steps:

[0076] S1-6-1: Use the 3D U2-Net-LSTM-DBN-Attention-MLP-ICPO algorithm to construct an initial abnormal behavior recognition model; The initial abnormal behavior recognition model includes an initial scene map feature extraction module, an initial identity data feature extraction module, an initial second positioning data feature extraction module, an initial attention weight module, an initial abnormal behavior recognition module, and an initial recognition result optimization module;

[0077] S1-6-2: Combine the first loss function of the initial scene map feature extraction module, the second loss function of the initial identity data feature extraction module, the third loss function of the initial second positioning data feature extraction module, and the fourth loss function of the initial abnormal behavior recognition module to obtain a comprehensive loss function;

[0078] S1-6-3: Iteratively optimize and train the initial abnormal behavior recognition model based on the scenario digital twin map, a number of corrected historical Beidou positioning data, and the corresponding preprocessed historical personnel identity data, and use the comprehensive loss function to obtain the historical comprehensive loss value for each iteration of the optimization training;

[0079] S1-6-4: If the number of iterations of the optimization training is greater than the iteration number threshold, or the historical comprehensive loss value is less than the loss value threshold, then output the final abnormal behavior recognition model; otherwise, continue the iterative optimization training of the abnormal behavior recognition model;

[0080] S1-7: Based on a number of historical abnormal behavior recognition results, use the reinforcement learning and adversarial training fusion algorithm to construct a personnel abnormal warning model and generate a number of historical warning strategy generation experiences;

[0081] The personnel abnormal warning model is constructed based on the Multi-Objective Proximal Policy Optimization (MOPPO)-Conditional Generative Adversarial Network (cGAN) algorithm, and the personnel abnormal warning model includes a personnel abnormal warning strategy generation module constructed based on the MOPPO algorithm and an adversarial training module constructed based on the cGAN algorithm that are connected in sequence. The personnel abnormal warning strategy generation module is provided with a set of objective functions, an experience replay pool, an Actor network, a Critic network, and an agent, and the adversarial training module is provided with a generator and a discriminator;

[0082] The Actor network is responsible for outputting the probability distribution of the actions that should be taken in a given state. The goal is to learn an optimal strategy, that is, to maximize the long-term cumulative reward. In the continuous action space, the Actor network usually outputs a mean value and optional variance parameters to describe the probability distribution of the actions. The Critic network is responsible for evaluating the value of a given state, that is, predicting the expected return that can be obtained starting from this state and following the current strategy, and usually outputs a scalar value representing the value of the state or the state-action value. The experience replay pool is used to store historical experiences for reuse during the training process. The set of objective functions includes functions that define multiple personnel abnormal warning objectives, including maximizing the sensitivity of personnel abnormal warning, minimizing the false alarm rate of personnel abnormal warning, maximizing the real-time performance of personnel abnormal warning, etc.; The generator of the adversarial training module is used to generate strategies, and the discriminator is used to distinguish the generated strategies from the optimal strategies. Through this adversarial training process, the personnel abnormal warning strategy generation module can learn more effective strategies, and at the same time, the adversarial training module helps to ensure the diversity and quality of the strategies, so as to find a better optimal solution in the multi-objective optimization problem;

[0083] According to the recognition results of a number of historical abnormal behaviors, using a fusion algorithm of reinforcement learning and adversarial training, construct a personnel abnormal early warning model, and generate a number of historical early warning strategy generation experiences, including the following steps:

[0084] S1-7-1: Use the MOPPO-GAN algorithm to construct an initial personnel abnormal early warning model; the initial personnel abnormal early warning model includes an initial personnel abnormal early warning strategy generation module and an initial adversarial training module;

[0085] S1-7-2: Set a set of objective functions, an experience replay pool, an Actor network, a Critic network, and an agent for the initial personnel abnormal early warning strategy generation module;

[0086] S1-7-3: Take the personnel abnormal early warning strategy generation problem as the simulation environment of the initial personnel abnormal early warning strategy generation module, and set an action space and a state space for the agent;

[0087] S1-7-4: Based on any objective function in the set of objective functions, according to the recognition results of a number of historical abnormal behaviors, pre-train the initial personnel abnormal early warning strategy generation module to obtain a pre-trained personnel abnormal early warning strategy generation module, and generate a number of historical personnel abnormal early warning strategies and corresponding historical strategy generation experiences;

[0088] S1-7-5: According to the recognition results of a number of historical abnormal behaviors and the corresponding historical personnel abnormal early warning strategies, optimize and train the initial generator of the initial adversarial training module to obtain an optimized generator, and generate a number of generated personnel abnormal early warning strategies;

[0089] S1-7-6: According to a number of historical personnel abnormal early warning strategies and the corresponding generated personnel abnormal early warning strategies, optimize and train the initial discriminator of the initial adversarial training module to obtain an optimized discriminator, and generate a number of historical discrimination results;

[0090] S1-7-7: Use the Critic network of the pre-trained personnel abnormal early warning strategy generation module to obtain a number of rewards for the generated personnel abnormal early warning strategies, and optimize the Actor network of the pre-trained personnel abnormal early warning strategy generation module according to the number of rewards to obtain an optimized Actor network;

[0091] S1-7-8: Optimize the Critic network of the pre-trained personnel abnormal early warning strategy generation module according to the number of rewards for the generated personnel abnormal early warning strategies and the corresponding historical discrimination results to obtain an optimized Critic network;

[0092] S1-7-9: Traverse all objective functions in the objective function set, repeat the above adversarial training steps, and obtain an optimized personnel anomaly warning strategy generation module with an optimized Actor network and an optimized Critic network, and an optimized adversarial training module with an optimized discriminator and an optimized discriminator;

[0093] S1-7-10: Integrate the optimized personnel anomaly warning strategy generation module and the optimized adversarial training module to obtain the final personnel anomaly warning model, and store several historical strategy generation experiences in the experience replay pool;

[0094] S2: The Beidou positioning device collects the real-time Beidou positioning data and real-time personnel identity data of the personnel, and sends the real-time Beidou positioning data and real-time personnel identity data to the monitoring center server;

[0095] The Beidou positioning device is generally set in the safety helmet of the personnel to facilitate the real-time collection of the Beidou positioning data and personnel identity data of the personnel;

[0096] S3: The monitoring center server uses the positioning data correction model according to the real-time Beidou positioning data to perform positioning data correction and obtain the corrected real-time Beidou positioning data, including the following steps:

[0097] S3-1: The monitoring center server preprocesses the real-time Beidou positioning data to obtain the preprocessed real-time Beidou positioning data, and inputs the preprocessed real-time Beidou positioning data into the positioning data correction model;

[0098] S3-2: Use the first positioning data feature extraction module of the positioning data correction model to extract the first real-time positioning data feature of the preprocessed real-time Beidou positioning data;

[0099] S3-3: Based on several key feature components, according to the first real-time positioning data feature, use the positioning data correction module of the positioning data correction model to predict the error value and generate the corresponding real-time error value;

[0100] S3-4: According to the real-time error value, correct the corresponding preprocessed real-time Beidou positioning data to obtain the corrected real-time Beidou positioning data;

[0101] S4: The monitoring center server uses the abnormal behavior recognition model according to the scene digital twin map, the corrected real-time Beidou positioning data, and the real-time personnel identity data to perform abnormal behavior recognition and obtain the real-time abnormal behavior recognition result, including the following steps:

[0102] S4-1: The monitoring center server uses the scene map feature extraction module of the abnormal behavior recognition model to extract the real-time scene map feature of the scene digital twin map;

[0103] S4-2: The second positioning data feature extraction module using the abnormal behavior recognition model extracts the second real-time positioning data features of the corrected real-time Beidou positioning data;

[0104] S4-3: Preprocess the real-time personnel identity data to obtain the preprocessed real-time personnel identity data, and use the identity data feature extraction module of the abnormal behavior recognition model to extract the real-time identity data features of the preprocessed real-time personnel identity data;

[0105] S4-4: According to the preset attention weight values, use the attention weight module of the abnormal behavior recognition model to perform weighted fusion on the real-time scene map features, the second real-time positioning data features, and the real-time identity data features to obtain the real-time weighted fusion features;

[0106] S4-5: Use the abnormal behavior recognition module of the abnormal behavior recognition model to perform abnormal behavior recognition on the real-time weighted fusion features to obtain the real-time recognition probability distribution;

[0107] S4-6: Use the recognition result optimization module of the abnormal behavior recognition model to optimize the real-time recognition probability distribution to obtain the real-time abnormal behavior recognition result, including the following steps:

[0108] S4-6-1: According to the real-time recognition probability distribution, set the real-time adjustment parameter of the real-time recognition probability distribution as the solution vector of the classification result optimization module, and perform initialization according to the solution vector to obtain a number of initial solutions;

[0109] Specifically, use the Circle chaotic mapping sequence for initialization to generate a number of initial solutions of the classification result optimization module, and obtain an initial ICPO population composed of a number of initial ICPO individuals (initial solutions);

[0110] The formula is:

[0111]

[0112] In the formula, is the initial ICPO individual of the Circle chaotic mapping; is the randomly generated initial ICPO individual; i' is the ICPO individual indicator; mod(*) is the remainder function;

[0113] S4-6-2: Taking the minimization of the classification error as the optimization goal, use the optimization goal as the fitness function, and set the ICPO population parameters and the maximum number of iterations of the ICPO optimization algorithm;

[0114] The formula is:

[0115] f(x i) = minMSE

[0116] Wherein, f(x i ) is the fitness function of the ICPO individual x i ; MSE is the classification error value; x i is the ICPO individual variable; i' is the ICPO individual indicator;

[0117] S4 - 6 - 3: Introduce a cyclic population reduction mechanism to limit the number of individuals in the ICPO population parameters to obtain the updated ICPO population parameters for the next iteration;

[0118] The formula is:

[0119]

[0120] Wherein, S t+1 is the number of individuals in the ICPO population parameters at the (t + 1)-th iteration; S t is the number of individuals in the ICPO population parameters at the t-th iteration; S min is the minimum value of the number of individuals in the ICPO population parameters; a' is the function evaluation parameter; V is the function evaluation loop parameter; V max is the maximum function evaluation loop parameter; t is the iteration number indicator;

[0121] S4 - 6 - 4: Calculate the initial fitness values of the initial ICPO individuals in the initial ICPO population according to the fitness function;

[0122] S4 - 6 - 5: Update the initial ICPO population using the first defense strategy, the second defense strategy, the third defense strategy, and the fourth defense strategy according to the initial fitness values and the updated ICPO population parameters to obtain the updated ICPO population;

[0123] The formula of the first defense strategy is:

[0124]

[0125] Wherein, is the updated ICPO individual within the first defense range; is the initial ICPO individual within the first defense range; τ 1 is a random number based on the normal distribution; τ 2 is a random value in the interval [0, 1]; is the optimal solution within the first defense range; is the vector generated between the true optimal solution and the randomly selected optimal solution from the ICPO population within the first defense range; i' is the ICPO individual indicator; t is the iteration indicator;

[0126] The formula for the second defense strategy is:

[0127]

[0128] In the formula, is the updated ICPO individual within the second defense range; is the initial ICPO individual within the second defense range; is the search upper limit vector of the second defense range; τ 3 is a random value in the interval [0, 1]; are the r1-th and r2-th initial ICPO individuals respectively; r1 and r2 are both two random integers between [1, S]; is the vector generated between the true optimal solution and the randomly selected optimal solution from the ICPO population within the second defense range;

[0129] The formula for the third defense strategy is:

[0130]

[0131] In the formula, is the updated ICPO individual within the third defense range; is the initial ICPO individual within the third defense range; is the search upper limit vector of the third defense range; are the r2-th and r3-th initial ICPO individuals respectively; r3 is a random integer between [1, S]; is the odor diffusion factor defined by the fitness function; λ t is the defense factor; is the search direction control parameter;

[0132] The formula for the fourth defense strategy is:

[0133]

[0134] In the formula, is the updated ICPO individual within the fourth defense range; is the initial ICPO individual within the fourth defense range; is the optimal solution within the fourth defense range; τ 4 and τ 5 are both random values in the interval [0, 1]; λ t is the defense factor; is the search direction control parameter; is the average force affecting the search direction; a' is the convergence speed factor;

[0135] S4-6-6: Use the dynamic reverse learning algorithm to perform dynamic reverse learning on the updated ICPO population to generate a dynamically reversed ICPO population;

[0136] The formula is:

[0137]

[0138] In the formula, is an individual of the dynamically reversed ICPO; γ' is a decreasing inertia coefficient; L max and L min are the maximum and minimum values of the vector space respectively; is an updated ICPO individual;

[0139] S4-6-7: According to the fitness function, calculate the fitness values of all ICPO individuals in the updated ICPO population and the dynamically reversed ICPO population, take the ICPO individual with the minimum fitness value as the optimal individual, and retain the optimal individual;

[0140] S4-6-8: Decode the solution vector of the optimal individual to obtain the optimal real-time adjustment parameter, and adjust the real-time recognition probability distribution according to the optimal real-time adjustment parameter to obtain the adjusted real-time recognition probability distribution;

[0141] S4-6-9: Take the real-time abnormal behavior recognition prediction label with the highest probability in the adjusted real-time recognition probability distribution as the real-time abnormal behavior recognition result;

[0142] S5: For the monitoring center server, if it is determined that the real-time abnormal behavior recognition result indicates the existence of abnormal behavior, then according to the real-time abnormal behavior recognition result, use the personnel abnormal warning model to generate a personnel abnormal warning strategy to obtain a real-time personnel abnormal warning strategy, including the following steps:

[0143] S5-1: For the monitoring center server, if it is determined that the real-time abnormal behavior recognition result indicates the existence of abnormal behavior, then input the real-time abnormal behavior recognition result into the personnel abnormal warning model;

[0144] S5-2: Randomly extract several historical strategy-generated experiences from the experience replay pool and parse the historical strategy-generated experiences to obtain several real-time personnel abnormal warning actions;

[0145] S5-3: According to several historical strategy-generated experiences and the real-time abnormal behavior recognition result, use the personnel abnormal warning model to generate a personnel abnormal warning strategy to obtain a real-time personnel abnormal warning strategy, including the following steps:

[0146] S5-3-1: The personnel anomaly warning strategy generation module of the user anomaly warning model selects the most appropriate objective function from the set of objective functions according to the real-time anomaly behavior recognition result;

[0147] S5-3-2: Analyze the real-time anomaly behavior recognition result to obtain several real-time anomaly behavior states, and update the state space of the intelligent agent according to the several real-time anomaly behavior states to obtain an updated state space;

[0148] S5-3-3: Update the action space of the intelligent agent according to several real-time personnel anomaly warning actions to obtain an updated action space;

[0149] S5-3-4: Based on the most appropriate objective function, use the intelligent agent to control the Actor network to generate the probability distribution of all possible real-time personnel anomaly warning actions in the updated action space for each real-time anomaly behavior state in the updated state space;

[0150] S5-3-5: Take the real-time personnel anomaly warning action with the highest probability distribution as the execution real-time personnel anomaly warning action corresponding to the real-time anomaly behavior state, and integrate all the execution real-time personnel anomaly warning actions to obtain a real-time personnel anomaly warning strategy;

[0151] S6: The monitoring center server uses the scenario digital twin map to visually display the real-time anomaly behavior recognition result, the corrected real-time Beidou positioning data, and the real-time personnel identity data according to the real-time personnel anomaly warning strategy, including the following steps:

[0152] S6-1: The monitoring center server generates a corresponding real-time personnel anomaly warning signal according to the real-time personnel anomaly warning strategy;

[0153] S6-2: Locate the real-time abnormal personnel according to the corrected real-time Beidou positioning data, and highlight the real-time abnormal personnel on the scenario digital twin map according to the real-time personnel anomaly warning strategy;

[0154] S6-3: Visually display the real-time personnel anomaly warning signal, the real-time anomaly behavior recognition result, the corrected real-time Beidou positioning data, and the real-time personnel identity data on the scenario digital twin map according to the real-time personnel anomaly warning strategy.

[0155] Example 2:

[0156] As Figure 2As shown in the figure, this embodiment provides a Beidou-based real-time personnel positioning and warning system for implementing a real-time personnel positioning and warning method. The system includes a monitoring center server and a number of Beidou positioning devices. The number of Beidou positioning devices are all communicatively connected to the monitoring center server. The monitoring center server includes a model construction unit, a positioning data correction unit, an abnormal behavior recognition unit, a personnel abnormal warning unit, and a visualization display unit that are connected in sequence;

[0157] The Beidou positioning device is used to collect the real-time Beidou positioning data and real-time personnel identity data of the personnel, and send the real-time Beidou positioning data and real-time personnel identity data to the monitoring center server;

[0158] The model construction unit is used to use digital twin technology to construct a scene digital twin map, and use artificial intelligence algorithms to construct a positioning data correction model, an abnormal behavior recognition model, and a personnel abnormal warning model;

[0159] The positioning data correction unit is used to correct the positioning data according to the real-time Beidou positioning data using the positioning data correction model to obtain the corrected real-time Beidou positioning data;

[0160] The abnormal behavior recognition unit is used to recognize abnormal behaviors according to the scene digital twin map, the corrected real-time Beidou positioning data, and the real-time personnel identity data using the abnormal behavior recognition model to obtain the real-time abnormal behavior recognition result;

[0161] The personnel abnormal warning unit is used to generate a personnel abnormal warning strategy according to the real-time abnormal behavior recognition result using the personnel abnormal warning model to obtain the real-time personnel abnormal warning strategy when the real-time abnormal behavior recognition result indicates the existence of abnormal behavior;

[0162] The visualization display unit is used to visually display the real-time abnormal behavior recognition result, the corrected real-time Beidou positioning data, and the real-time personnel identity data according to the real-time personnel abnormal warning strategy using the scene digital twin map.

[0163] The present invention provides a Beidou-based personnel real-time positioning and early warning method and system. The Beidou positioning device collects the real-time Beidou positioning data and real-time personnel identity data of the personnel in real time, thereby improving the real-time performance. The real-time Beidou positioning data is corrected by a positioning data correction model, thereby reducing signal interference in a complex environment, realizing high-precision real-time positioning with seamless switching between indoor and outdoor environments, effectively reducing positioning errors, improving positioning accuracy, ensuring accurate acquisition of personnel positions, and providing a reliable data basis for subsequent abnormal behavior recognition and early warning. The advanced artificial intelligence algorithm is used to construct an abnormal behavior recognition model, deeply excavates the deep features of the positioning data, and combines a scene digital twin map and real-time personnel identity data to perform a fusion analysis of multimodal and multi-data sources, thereby improving the accuracy of abnormal behavior recognition, realizing customized analysis based on identity recognition, avoiding misjudging normal behavior as abnormal behavior, and reducing the false alarm rate. Based on real-time data and artificial intelligence models, a real-time personnel abnormal early warning strategy is automatically generated, thereby realizing customized and intelligent personnel abnormal early warning, improving the timeliness and accuracy of the early warning, and gaining precious time for emergency response. The scene digital twin map is used to visualize the real-time positioning and early warning of personnel, thereby improving the convenience and practicality of observation.

[0164] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.

Claims

1. A Beidou-based personnel real-time positioning and early warning method, characterized in that: The steps include: The monitoring center server uses digital twin technology to build a scene digital twin map, and uses artificial intelligence algorithms to build a positioning data correction model, an abnormal behavior recognition model, and a personnel abnormality warning model; Beidou positioning equipment collects real-time Beidou positioning data and real-time personnel identity data of personnel, and sends the real-time Beidou positioning data and real-time personnel identity data to the monitoring center server; The monitoring center server uses the positioning data correction model to correct the positioning data based on the real-time Beidou positioning data to obtain the corrected real-time Beidou positioning data; The monitoring center server uses the abnormal behavior recognition model to identify abnormal behaviors based on the scene digital twin map, the corrected real-time Beidou positioning data, and the real-time personnel identity data, and obtains real-time abnormal behavior recognition results; If the monitoring center server determines that the real-time abnormal behavior recognition result is abnormal behavior, it uses the personnel abnormality warning model to generate a personnel abnormality warning strategy based on the real-time abnormal behavior recognition result to obtain a real-time personnel abnormality warning strategy; The monitoring center server uses the scene digital twin map to visualize the real-time abnormal behavior recognition results, corrected real-time Beidou positioning data, and real-time personnel identity data based on the real-time personnel abnormal warning strategy.

2. A Beidou-based personnel real-time positioning and early warning method according to claim 1, characterized in that: The monitoring center server uses digital twin technology to build a scene digital twin map, and uses artificial intelligence algorithms to build a positioning data correction model, an abnormal behavior recognition model, and a personnel abnormality warning model, including the following steps: The monitoring center server obtains the GIS data of the application scenario and builds a three-dimensional simulation map of the application scenario based on the GIS data; Integrate digital twin data of the scene three-dimensional simulation map to obtain the scene digital twin map of the application scenario; Collecting a number of historical Beidou positioning data and corresponding historical personnel identity data, and preprocessing them to obtain a number of preprocessed historical Beidou positioning data and a number of preprocessed historical personnel identity data; Setting a corresponding true error value for each preprocessed historical Beidou positioning data, and obtaining a plurality of preprocessed historical Beidou positioning data with true error values ​​set therein; According to a number of pre-processed historical Beidou positioning data with real error values, a positioning data correction model is constructed using a deep learning and decision tree fusion algorithm, and a number of corrected historical Beidou positioning data are generated; Based on the scene digital twin map, several corrected historical Beidou positioning data and the corresponding pre-processed historical personnel identity data, a multimodal fusion deep learning algorithm is used to build an abnormal behavior recognition model and generate several historical abnormal behavior recognition results; Based on several historical abnormal behavior recognition results, a personnel abnormality warning model is constructed using the reinforcement learning and adversarial training fusion algorithm, and several historical warning strategy generation experiences are generated.

3. A Beidou-based personnel real-time positioning and early warning method according to claim 2, characterized in that: The positioning data correction model is constructed based on the DBN-RF algorithm.

4. The Beidou-based personnel real-time positioning and early warning method according to claim 3 is characterized in that: The abnormal behavior recognition model is built based on the 3D U2-Net-LSTM-DBN-Attention-MLP-ICPO algorithm.

5. The Beidou-based personnel real-time positioning and early warning method according to claim 4 is characterized in that: The personnel abnormality warning model is constructed based on the MOPPO-cGAN algorithm.

6. The Beidou-based personnel real-time positioning and early warning method according to claim 5 is characterized in that: The monitoring center server uses the positioning data correction model to correct the positioning data based on the real-time Beidou positioning data to obtain the corrected real-time Beidou positioning data, including the following steps: The monitoring center server preprocesses the real-time Beidou positioning data to obtain the preprocessed real-time Beidou positioning data, and inputs the preprocessed real-time Beidou positioning data into the positioning data correction model; Using the positioning data correction model, extracting the first real-time positioning data feature of the pre-processed real-time Beidou positioning data; According to the first real-time positioning data feature, an error value is predicted to generate a corresponding real-time error value; According to the real-time error value, the corresponding pre-processed real-time Beidou positioning data is corrected to obtain the corrected real-time Beidou positioning data.

7. The Beidou-based personnel real-time positioning and early warning method according to claim 6 is characterized in that: The monitoring center server uses the abnormal behavior recognition model to identify abnormal behaviors based on the scene digital twin map, the corrected real-time Beidou positioning data, and the real-time personnel identity data, and obtains real-time abnormal behavior recognition results, including the following steps: The monitoring center server extracts the real-time scene map features of the scene digital twin map; Extracting the second real-time positioning data feature of the corrected real-time Beidou positioning data; Preprocessing the real-time personnel identity data to obtain preprocessed real-time personnel identity data, and extracting real-time identity data features of the preprocessed real-time personnel identity data; According to a preset attention weight value, the real-time scene map feature, the second real-time positioning data feature and the real-time identity data feature are weighted and fused to obtain a real-time weighted fusion feature; Perform abnormal behavior recognition on real-time weighted fusion features to obtain real-time recognition probability distribution; The recognition results are optimized for the real-time recognition probability distribution to obtain the real-time abnormal behavior recognition results.

8. The Beidou-based personnel real-time positioning and early warning method according to claim 7 is characterized in that: If the monitoring center server determines that the real-time abnormal behavior recognition result is abnormal behavior, it uses the personnel abnormality warning model to generate a personnel abnormality warning strategy based on the real-time abnormal behavior recognition result, and obtains the real-time personnel abnormality warning strategy, including the following steps: The monitoring center server, if it is determined that the real-time abnormal behavior recognition result is abnormal behavior, inputs the real-time abnormal behavior recognition result into the personnel abnormal warning model; Randomly extract several historical strategies to generate experience; According to several historical strategy generation experiences and real-time abnormal behavior recognition results, the personnel abnormality warning model is used to generate personnel abnormality warning strategies to obtain real-time personnel abnormality warning strategies.

9. A Beidou-based personnel real-time positioning and early warning method according to claim 8, characterized in that: The monitoring center server uses the scene digital twin map to visualize the real-time abnormal behavior recognition results, the corrected real-time Beidou positioning data, and the real-time personnel identity data according to the real-time personnel abnormal warning strategy, including the following steps: The monitoring center server generates corresponding real-time personnel abnormal warning signals according to the real-time personnel abnormal warning strategy; According to the corrected real-time Beidou positioning data, the real-time abnormal personnel are located, and according to the real-time personnel abnormal warning strategy, the real-time abnormal personnel are highlighted on the scene digital twin map; According to the real-time personnel abnormal warning strategy, the real-time personnel abnormal warning signal, real-time abnormal behavior recognition results, corrected real-time Beidou positioning data and real-time personnel identity data are visualized on the scene digital twin map.

10. A Beidou-based personnel real-time positioning and warning system, used to implement the personnel real-time positioning and warning method as claimed in any one of claims 1 to 9, characterized in that: The system includes a monitoring center server and several Beidou positioning devices, and the several Beidou positioning devices are all communicatively connected to the monitoring center server. The monitoring center server includes a model building unit, a positioning data correction unit, an abnormal behavior recognition unit, a personnel abnormality warning unit and a visualization display unit which are connected in sequence.

Images