Real-time simulation method for detecting photoelectric tracking equipment

Abstract

The invention relates to the technical field of simulation, and particularly discloses a real-time simulation method for detecting photoelectric tracking equipment, which comprises the following steps of: performing feature decoupling processing on photoelectric signals through a dual-channel adaptive neural network architecture, and constructing an equipment mathematical model by adopting a dynamic gating fusion mechanism based on a processing result; the method comprises the following steps: establishing a multi-physics field coupling simulation environment based on a heterogeneous computing architecture, introducing an equipment mathematical model, performing three-field co-evolution through light transmission modeling, electromagnetic field distribution calculation and target motion prediction, and generating a dynamic test scene containing space-time relevance; according to the method, a mode of combining mixed feature analysis and fuzzy reasoning is adopted, multi-dimensional performance indexes are extracted from simulation data, and multi-target dynamic optimization is carried out through a strategy of combining a quantum evolution algorithm and swarm intelligence optimization; by means of a digital twin platform and a hardware-in-loop interface, real-time verification and closed-loop optimization are achieved, and the response speed and tracking precision of photoelectric tracking equipment in a complex environment are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the field of simulation technology, and particularly relates to a real-time simulation method for detecting optoelectronic tracking devices. Background Art

[0002] Optoelectronic tracking devices are widely used in military, aviation, aerospace and other fields, and their main function is to accurately detect and track targets. However, with the increase in environmental complexity, traditional optoelectronic tracking technologies face many challenges. Especially in a dynamically changing complex environment, their performance is easily affected by various interference factors.

[0003] Most of the existing simulation methods for optoelectronic tracking devices are based on single physical field modeling, ignoring the interaction between different physical fields, which results in limited accuracy and practicality of the simulation results. Common simulation methods mostly focus on signal processing, but mostly use simplified linear models, which are difficult to handle complex scenarios with various interferences and noises. Especially when the signal is affected by electromagnetic interference, atmospheric turbulence and target motion uncertainty, the robustness of traditional methods is poor. In addition, optoelectronic tracking devices usually rely on static optimization schemes, which means that it is difficult for the devices to adapt and adjust their control parameters in real time when facing environmental changes, resulting in the inability to maintain high tracking accuracy and response speed in a dynamic environment. In a complex environment, traditional methods often fail to fully consider the coupling effect of multiple physical fields, resulting in insufficient processing ability of optoelectronic signals, further affecting the anti-interference ability and real-time response ability of the devices.

[0004] Therefore, it is necessary to propose a real-time simulation method for detecting optoelectronic tracking devices to solve the problem of how to achieve real-time and efficient simulation and dynamic optimization of optoelectronic tracking devices in the prior art.

[0005] The above information disclosed in this background art is only used to increase the understanding of the background art of the present invention. Therefore, it may include prior art that is not known to those of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a real-time simulation method for detecting optoelectronic tracking devices to solve the problems raised in the above background art.

[0007] To achieve the above purpose, the present invention provides the following technical solutions: A real-time simulation method for detecting optoelectronic tracking devices, comprising: Performing feature decoupling processing on optoelectronic signals through a dual-channel adaptive neural network architecture, and constructing a device mathematical model based on the processing results by using a dynamic gating fusion mechanism; Build a multi - physical - field coupling simulation environment based on a heterogeneous computing architecture, introduce the device mathematical model, and perform three - field collaborative evolution through ray - transmission modeling, electromagnetic - field distribution calculation, and target - motion prediction to generate a dynamic test scenario with spatio - temporal correlation; Obtain simulation data of multi - dimensional physical fields based on the dynamic test scenario, and use a method combining hybrid feature analysis and fuzzy reasoning to extract multi - dimensional performance indicators from the simulation data and establish a parameter - optimization mapping relationship; Based on the parameter - optimization mapping relationship, adopt a hybrid strategy that combines the quantum - evolution algorithm and swarm - intelligence optimization, and perform multi - objective dynamic optimization of control parameters through an environment - perception weighted mechanism to generate multi - objective dynamic optimization results; Build a virtual - real data interaction channel through a digital - twin platform, input the multi - objective dynamic optimization results into a photon - level simulation engine, and complete the closed - loop iterative optimization of the control algorithm through real - time verification via a hardware - in - the - loop interface.

[0008] Preferably, the dual - channel adaptive neural - network architecture includes a spatial - noise processing channel and a time - varying - interference processing channel; Use the spatial - noise processing channel to perform non - local - mean denoising on the optoelectronic signal, retain the high - frequency edge features, and obtain a denoised optoelectronic signal; Input the denoised optoelectronic signal into a fourth - order residual network to extract shallow - layer texture features, and use the cross - attention module of SwinTransformer to enhance the global context of the shallow - layer texture features to generate a feature matrix; Process the feature matrix through multi - scale spatial convolution and adaptive - entropy weighted fusion to generate a noise distribution map; Process the noise distribution map through a pre - trained conditional generative adversarial network to generate a pixel - level noise - compensation signal; Use the time - varying - interference processing to perform a short - time Fourier transform on the optoelectronic signal to generate a time - frequency spectrum map; Based on the time - frequency spectrum map, extract temporal features through the causal dilated - convolution layer of a probabilistic time - convolutional network; Perform probabilistic modeling on the phase jitter of the temporal features through a variational auto - encoder to generate interference - feature encoding; Use the Kalman smoothing algorithm to post - process the interference - feature encoding to generate a time - varying - interference suppression - compensation signal.

[0009] Preferably, the steps of constructing the device mathematical model by using a dynamic gating fusion mechanism based on the processing results include: Respectively extract features from the pixel - level noise - compensation signal and the time - varying - interference suppression - compensation signal to generate a spatial - feature vector and a time - varying - feature vector; Based on the spatial feature vector and the time-varying feature vector, a two-channel feature similarity matrix is constructed through a multi-layer perceptron and a cosine similarity function; Based on the two-channel feature similarity matrix, a policy gradient reinforcement learning network is used to generate dynamic fusion weights with the tracking error entropy and control efficiency weighted as the reward function; Design a gating function Perform probabilistic channel selection, maximize the cumulative reward, and output the device mathematical model; ; In the formula, is the action value function, is the channel selection action, is the temperature coefficient.

[0010] Preferably, the multi-physical field coupling simulation environment built based on the heterogeneous computing architecture includes: Based on the configuration information of the heterogeneous computing architecture, the heterogeneous computing architecture is dynamically identified and initialized through an adaptive resource scheduling algorithm; Based on the initialized heterogeneous computing architecture, various computing resources are predicted and allocated according to the multi-physical field coupling simulation requirements by using a load balancing algorithm; Introduce the mathematical models and interaction rule sets of each physical field, analyze the spatio-temporal correlation parameters through an intelligent fusion algorithm, and establish a multi-physical field coupling topology architecture; Based on the simulation target constraint conditions and the historical data set, trigger the intelligent optimization algorithm to retrieve the strategy library and generate the parameter configuration tables of each physical field; Input the configuration parameters of each physical field into the established multi-physical field coupling topology architecture, define the data exchange protocol, and establish a time step alignment mechanism for each physical field to form a multi-physical field coupling simulation environment.

[0011] Preferably, the light transmission modeling includes: Initialize the atmospheric environment parameters to construct a dynamic meteorological field, and use an improved Mie scattering model and a GPU-accelerated Monte Carlo ray tracing algorithm to generate a light transmission attenuation field that changes dynamically with the meteorological field; Through a dynamic texture synthesis algorithm driven by meteorological radar data, the turbulence intensity distribution is modulated in real time to update the parameters of the light transmission attenuation field; The electromagnetic field distribution calculation includes: Construct a coupling model of the mixed potential integral equation and the equivalent dipole method, and divide the high and low frequency calculation domains; Based on the high and low frequency calculation domains, use OpenMP to parallel compute the low frequency electromagnetic field, and accelerate the solution of the high frequency components through cuFFT to generate the full frequency band electromagnetic field distribution; Based on the full - band electromagnetic field distribution, perform real - time mapping operations of electromagnetic field - flow field data through a multi - variable coupling solver, and perform spatio - temporal consistency alignment of multi - physical fields based on the time - step alignment mechanism; The target motion prediction includes: Train an LSTM - VAM encoder using the target historical trajectory data to generate a sequence of motion intention latent variables; Analyze the reward function corresponding to the latent variables through inverse reinforcement learning to construct a Markov decision process based on maximum entropy; Use model predictive control to perform rolling optimization of the Markov decision process to generate trajectory prediction results within a preset time period.

[0012] Preferably, for obtaining simulation data of multi - dimensional physical fields based on a dynamic test scenario, using a method combining hybrid feature analysis and fuzzy inference, extracting multi - dimensional performance indicators from the simulation data and establishing a parameter optimization mapping relationship, including: Call the wavelet packet transform algorithm to perform time - frequency domain decomposition on the simulation data, and extract non - linear feature subsets through kernel principal component analysis to generate a hybrid feature matrix; Define input / output language variables, initialize the membership function of the fuzzy set based on the expert rule base, and construct a fuzzy inference system; Configure a Mamdani inference engine and a centroid defuzzifier, input the hybrid feature matrix into the fuzzy inference system to train the membership function parameters, and generate a mapping relationship from feature indicators to control parameters.

[0013] Preferably, based on the parameter optimization mapping relationship, adopt a hybrid strategy that combines the quantum evolution algorithm and swarm intelligence optimization, and perform multi - objective dynamic optimization of control parameters through an environmental perception weighting mechanism to generate multi - objective dynamic optimization results, including: Use the quantum - bit encoding strategy to initialize the control parameter population, and generate an initial quantum state sequence through chaotic Logistic mapping; Based on the initial quantum state, train and define a bi - objective optimization function: root - mean - square tracking error and control variable change rate, and set the initial range of the quantum gate rotation angle; Initialize the Pareto front archive and set the crowding distance threshold , and establish a non - dominated sorting layer; Perform quantum rotation gate operations on the current population based on the initial range of the quantum gate rotation angle, and the phase update amount is dynamically adjusted by the individual crowding ranking: ; Collapse the quantum state into a classical parameter solution, calculate the bi - objective optimization function value, and perform non - dominated sorting according to the calculation results; Extract the parameter coding patterns of the top 10% individuals based on the non-dominated sorting, and construct a graph structure containing the historical optimization paths; Learn the graph structure parameter correlation features through a graph attention network, generate feature embedding vectors by combining a Transformer encoder, and input the feature embedding vectors into a multi-layer perceptron to generate a dynamic weight distribution; Design a weighting function Optimize the real-time allocation of the target dynamic weight distribution and output the multi-objective dynamic optimization results; ; In the formula, is the weight of the th optimization objective at time , is the feature embedding vector corresponding to the th target, is the real-time environment feature vector, is the vector concatenation operation, is the learnable parameter vector.

[0014] Preferably, a virtual-real data interaction channel is constructed through a digital twin platform, and the multi-objective dynamic optimization results are input into a photon-level simulation engine, and the closed-loop iterative optimization of the control algorithm is completed through real-time verification of the hardware-in-the-loop interface, including: Import the optical CAD model into the digital twin platform and configure the spectral sampling parameters, and initialize the photon tracing parameters; Based on the initialized photon tracing parameters, generate photon tracing data through the Jones matrix and the Monte Carlo algorithm, and output the scattering field distribution to the hardware-in-the-loop interface in combination with the bidirectional reflectance distribution function model; Configure an FPGA module at the hardware-in-the-loop interface for nanosecond-level real-time simulation, and use a timestamp alignment algorithm to compensate for data transmission delays; Perform hardware frequency response calibration through the chirp signal injection method, generate a compensation filter and load it into the FPGA; Construct a virtual-real data two-way channel, use the unscented Kalman filter algorithm for multi-objective state estimation and deviation prediction, establish a deviation compensation function in combination with the diffeomorphic mapping, and generate a control parameter correction amount; Based on the control parameter correction amount, use the covariance matrix adaptation evolution strategy to evolve the control parameters online, and set the iteration period to 20 ms to complete the closed-loop optimization process.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The present invention uses a dual-channel adaptive neural network architecture to perform feature decoupling processing on optoelectronic signals, constructs a device mathematical model by combining a dynamic gating fusion mechanism, accurately processes different types of signal interference, and improves the anti-noise ability of optoelectronic signals; a multi-physical field coupling simulation environment built based on a heterogeneous computing architecture comprehensively considers the interactions of various physical factors such as light transmission, electromagnetic field distribution, and target movement, generates a dynamic test scenario with spatio-temporal correlation, thereby enhancing the simulation accuracy and adaptability.

[0016] The present invention uses a combination of hybrid feature analysis and fuzzy reasoning to extract multi-dimensional performance indicators from simulation data, and performs multi-objective dynamic optimization through a strategy that combines the quantum evolution algorithm and swarm intelligence optimization, improving the real-time performance and accuracy of control parameter optimization; with the help of a digital twin platform and a hardware-in-the-loop interface, real-time verification and closed-loop optimization are realized, greatly improving the response speed and tracking accuracy of optoelectronic tracking devices in complex environments. Brief Description of the Drawings

[0017] Figure 1 It is a flowchart of a real-time simulation method for detecting an optoelectronic tracking device of the present invention. Detailed Embodiments

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] Embodiment 1: Please refer to Figure 1 As shown, a real-time simulation method for detecting an optoelectronic tracking device includes: Performing feature decoupling processing on optoelectronic signals through a dual-channel adaptive neural network architecture, and constructing a device mathematical model based on the processing results using a dynamic gating fusion mechanism; The dual-channel adaptive neural network architecture includes a spatial noise processing channel and a time-varying interference processing channel; Using the spatial noise processing channel to perform non-local mean denoising on optoelectronic signals, retaining high-frequency edge features, and obtaining denoised optoelectronic signals; Inputting the denoised optoelectronic signals into a fourth-order residual network to extract shallow texture features, and using the cross-attention module of SwinTransformer to perform global context enhancement on the shallow texture features to generate a feature matrix; Processing the feature matrix through multi-scale spatial convolution and adaptive entropy weighted fusion to generate a noise distribution map; Process the noise distribution map through a pre-trained conditional generative adversarial network to generate a pixel-level noise compensation signal; Generate a time-frequency spectrum diagram by performing a short-time Fourier transform on the optoelectronic signal using time-varying interference processing; Based on the time-frequency spectrum diagram, extract temporal features through the causal dilated convolutional layer of a probabilistic temporal convolutional network, and combine with a variational autoencoder to probabilistically model the phase jitter of the temporal features to generate interference feature encodings; Use the Kalman smoothing algorithm to post-process the interference feature encodings to generate a time-varying interference suppression compensation signal.

[0020] Furthermore, through a dual-channel adaptive neural network architecture, efficient feature decoupling and optimization processing of the optoelectronic signal are carried out. Combining technologies such as noise removal, time-varying interference suppression, feature enhancement, and noise compensation, precise processing of the optoelectronic signal is achieved. Through the independent processing of spatial noise and time-varying interference, the signal quality can be effectively improved, the edge features and temporal features of the signal can be enhanced, and the device performance can be dynamically optimized, thereby improving the accuracy, adaptability, and robustness of the optoelectronic tracking device.

[0021] Extract features from the pixel-level noise compensation signal and the time-varying interference suppression compensation signal respectively to generate a spatial feature vector and a time-varying feature vector; Based on the spatial feature vector and the time-varying feature vector, construct a dual-channel feature similarity matrix through a multi-layer perceptron and a cosine similarity function; Based on the dual-channel feature similarity matrix, use a policy gradient reinforcement learning network with the weighted tracking error entropy and control efficiency as the reward function to generate dynamic fusion weights; Design a gating function for probabilistic channel selection, maximize the cumulative reward, and output the device mathematical model.

[0022] Furthermore, through the feature extraction and similarity analysis of the noise compensation signal and the interference suppression signal, combined with the reinforcement learning network to optimize the dynamic fusion weights, intelligent fusion and adaptive adjustment of the signal features are achieved. Through the probabilistic channel selection of the gating function, the accuracy and efficiency of signal processing are effectively improved, the tracking error is reduced, and the control performance is optimized, thereby generating an accurate device mathematical model and improving the overall performance and adaptability of the optoelectronic device.

[0023] Build a multi-physical field coupling simulation environment based on a heterogeneous computing architecture, introduce the device mathematical model, and perform three-field collaborative evolution through ray transmission modeling, electromagnetic field distribution calculation, and target motion prediction to generate a dynamic test scenario with spatio-temporal correlation; Based on the configuration information of the heterogeneous computing architecture, dynamically identify and initialize the heterogeneous computing architecture through an adaptive resource scheduling algorithm; Based on the initialized heterogeneous computing architecture, a load balancing algorithm is used to predict and allocate various computing resources according to the requirements of multi-physics field coupling simulation; Introduce the mathematical models of each physical field and the interaction rule set, and use an intelligent fusion algorithm to analyze the spatio-temporal correlation parameters to establish a multi-physics field coupling topology architecture; Based on the simulation target constraint conditions and historical data sets, trigger the intelligent optimization algorithm to retrieve the strategy library and generate the parameter configuration tables of each physical field; Input the configuration parameters of each physical field into the established multi-physics field coupling topology architecture, define the data exchange protocol, and establish the time step alignment mechanism of each physical field to form a multi-physics field coupling simulation environment.

[0024] Furthermore, by constructing a multi-physics field coupling simulation environment based on the heterogeneous computing architecture, the three-field collaborative evolution of light transmission, electromagnetic field distribution, and target motion is realized, and a dynamic spatio-temporal correlation test scenario is generated. Through adaptive resource scheduling and load balancing algorithms to optimize the computing resource allocation, combined with intelligent fusion and optimization algorithms, it can efficiently analyze and configure the parameters of each physical field to form an accurate multi-physics field coupling topology architecture. This method not only improves the simulation accuracy and computing efficiency, but also supports the dynamic adjustment and optimization of complex systems through intelligent optimization strategies, significantly enhancing the adaptability and performance of the simulation environment.

[0025] (1) Light transmission modeling Initialize the atmospheric environment parameters to construct a dynamic meteorological field, and use an improved Mie scattering model and a GPU-accelerated Monte Carlo ray tracing algorithm to generate a light transmission attenuation field that changes dynamically with the meteorological field; Through a dynamic texture synthesis algorithm driven by meteorological radar data, modulate the turbulence intensity distribution in real time and update the parameters of the light transmission attenuation field; (2) Electromagnetic field distribution calculation Construct a coupling model of the mixed potential integral equation and the equivalent dipole method, and divide the high- and low-frequency calculation domains; Based on the high- and low-frequency calculation domains, use OpenMP to parallel-compute the low-frequency electromagnetic field and accelerate the solution of the high-frequency components through cuFFT to generate the electromagnetic field distribution in the full frequency band; Based on the electromagnetic field distribution in the full frequency band, perform real-time mapping operations of electromagnetic field-fluid field data through a multi-variable coupling solver, and perform spatio-temporal consistency alignment of multi-physics fields based on the time step alignment mechanism; (3) Target motion prediction Use the target historical trajectory data to train the LSTM-VAM encoder to generate a sequence of latent variables of motion intention; Analyze the reward function corresponding to the latent variables through inverse reinforcement learning to construct a Markov decision process based on maximum entropy; Adopt model predictive control to roll-optimize the Markov decision process and generate trajectory prediction results within a preset time period.

[0026] Furthermore, through the comprehensive optimization of ray transmission modeling, electromagnetic field distribution calculation, and target motion prediction, high-precision, multi-physical field dynamic simulation and prediction are achieved. Ray transmission modeling uses the Monte Carlo ray tracing algorithm with atmospheric environment and GPU acceleration to dynamically update the light attenuation field. Electromagnetic field calculation combines high and low frequency calculations with parallel acceleration technology to accurately simulate the electromagnetic field distribution in the full frequency band and real-time map the electromagnetic field and flow field data. Target motion prediction combines deep learning and reinforcement learning to optimize the trajectory prediction accuracy. Overall, the technology realizes efficient physical field coupling and spatio-temporal consistency alignment, significantly improving the accuracy, real-time performance, and adaptability of the simulation system.

[0027] Based on the simulation data of multi-dimensional physical fields obtained from dynamic test scenarios, use a method combining hybrid feature analysis and fuzzy inference to extract multi-dimensional performance indicators from the simulation data and establish a parameter optimization mapping relationship; Call the wavelet packet transform algorithm to decompose the simulation data in the time-frequency domain, and extract the non-linear feature subset through kernel principal component analysis to generate a hybrid feature matrix; Define input / output language variables, initialize the membership function of the fuzzy set based on the expert rule base, and construct a fuzzy inference system; Configure the Mamdani inference engine and the centroid method defuzzifier, input the hybrid feature matrix into the fuzzy inference system to train the membership function parameters, and generate the mapping relationship from feature indicators to control parameters.

[0028] Furthermore, through the simulation data of multi-dimensional physical fields obtained from dynamic test scenarios, combined with hybrid feature analysis and fuzzy inference, the accurate extraction and optimization of system performance are realized. Non-linear features are extracted through wavelet packet transform and kernel principal component analysis to construct a hybrid feature matrix, and the mapping relationship between features and control parameters is modeled and optimized using a fuzzy inference system. The finally generated optimization mapping relationship accurately reflects the complex relationship between multi-dimensional performance indicators and control parameters.

[0029] Based on the parameter optimization mapping relationship, adopt a hybrid strategy that combines quantum evolution algorithm and swarm intelligence optimization, and perform multi-objective dynamic optimization of control parameters through an environmental perception weighting mechanism to generate multi-objective dynamic optimization results; Adopt the quantum bit coding strategy to initialize the control parameter population, and generate the initial quantum state sequence through chaotic Logistic mapping; Based on the initial quantum state, train and define a two-objective optimization function: the root mean square of the tracking error and the change rate of the control quantity, and set the initial range of the quantum gate rotation angle; Initialize the Pareto front archive, set the crowding distance threshold, and establish the non-dominated sorting levels; Perform the quantum rotation gate operation on the current population based on the initial range of the quantum gate rotation angle, and the phase update amount is dynamically adjusted by the individual crowding ranking; Collapse the quantum state into a classical parameter solution, calculate the values of the double-objective optimization function, and perform non-dominated sorting according to the calculation results; Extract the parameter coding patterns of the top 10% individuals based on non-dominated sorting, and construct a graph structure containing the historical optimization paths; Learn the graph structure parameter correlation features through the graph attention network, generate feature embedding vectors by combining the Transformer encoder, and input the feature embedding vectors into the multi-layer perceptron to generate a dynamic weight distribution; Design a weighted function to optimize the real-time allocation of the dynamic weight distribution of the objective, and output the multi-objective dynamic optimization results.

[0030] Furthermore, by fusing the quantum evolution algorithm with the swarm intelligence optimization strategy and combining the environmental perception weighting mechanism, the multi-objective dynamic optimization of the control parameters is realized. Through qubit encoding, chaotic mapping, and quantum rotation operations, the double-objective function (tracking error and control quantity change rate) is optimized, and the best solution is extracted through non-dominated sorting and Pareto front optimization. Combining the graph attention network with the Transformer encoder, the parameter correlation features are dynamically learned, and the dynamic weight allocation of the real-time optimization objective is realized through the weighted function, thereby generating efficient multi-objective optimization results.

[0031] Build a virtual-real data interaction channel through the digital twin platform, input the multi-objective dynamic optimization results into the photon-level simulation engine, and complete the closed-loop iterative optimization of the control algorithm through real-time verification at the hardware-in-the-loop interface; Import the optical CAD model into the digital twin platform and configure the spectral sampling parameters, and initialize the photon tracing parameters; Based on the initialized photon tracing parameters, generate photon tracing data through the Jones matrix and the Monte Carlo algorithm, and combine the bidirectional reflectance distribution function model to output the scattering field distribution to the hardware-in-the-loop interface; Configure the FPGA module at the hardware-in-the-loop interface for nanosecond-level real-time simulation, and use the timestamp alignment algorithm to compensate for the data transmission delay; Perform hardware frequency response calibration through the chirp signal injection method, generate the compensation filter and load it into the FPGA; Build a two-way virtual-real data channel, use the unscented Kalman filter algorithm for multi-objective state estimation and deviation prediction, combine the diffeomorphic mapping to establish a deviation compensation function, and generate the control parameter correction amount; Based on the control parameter correction amount, the covariance matrix adaptation evolution strategy is used to evolve the control parameters online, and the iteration period is set to 20 ms to complete the closed-loop optimization process.

[0032] Furthermore, real-time interaction and closed-loop optimization of virtual and real data are realized through the digital twin platform and the photon-level simulation engine. Through the optical CAD model, photon tracing data, and hardware-in-the-loop interface, the accuracy and efficiency of the control algorithm are verified and optimized. Combining the nanosecond-level real-time simulation of the FPGA module and the timestamp alignment algorithm ensures the compensation of data transmission delay. The unscented Kalman filter and diffeomorphic mapping are used for multi-objective state estimation and deviation prediction, and then the control parameters are corrected and online optimization is realized through the covariance matrix adaptation evolution strategy.

[0033] Example 2:

[0034] Application example: Real-time simulation for detecting the working state of optoelectronic tracking equipment In the fields of military, aviation, aerospace, etc., optoelectronic tracking equipment plays a crucial role, and they need to accurately detect and track targets. However, the complex and changeable environmental conditions, such as electromagnetic interference, atmospheric turbulence, and the uncertainty of target movement, pose severe challenges to the performance of optoelectronic tracking equipment. To evaluate and optimize the working state of optoelectronic tracking equipment in these complex environments, we adopt this real-time simulation method based on the dual-channel adaptive neural network architecture and the multi-physical field coupling simulation environment.

[0035] I. Application objectives and requirements (1) Accurately simulate complex environments: Simulate a multi-physical field environment including electromagnetic interference, atmospheric turbulence, and dynamic target movement.

[0036] (2) Real-time process optoelectronic signals: Real-time process optoelectronic signals, including denoising, feature extraction, and interference suppression.

[0037] (3) Dynamically optimize control parameters: According to the simulation results, dynamically optimize the control parameters of optoelectronic tracking equipment to improve the tracking accuracy and response speed.

[0038] (4) Closed-loop iterative optimization: Real-time verify the control algorithm through the hardware-in-the-loop interface to achieve closed-loop iterative optimization.

[0039] II. Application steps 1. Construct a dual-channel adaptive neural network architecture ① Spatial noise processing channel: Perform non-local means denoising on optoelectronic signals to retain high-frequency edge features.

[0040] The denoised signal is input into a fourth-order residual network to extract shallow texture features, and the cross-attention module of the Swin Transformer is used for global context enhancement.

[0041] A noise distribution map is generated through multi-scale spatial convolution and adaptive entropy weighted fusion, and then a pixel-level noise compensation signal is generated through the processing of a conditional generative adversarial network.

[0042] ② Time-varying interference processing channel: The optoelectronic signal is subjected to short-time Fourier transform to generate a time-frequency spectrogram.

[0043] The causal dilated convolution layer of the probabilistic temporal convolutional network is used to extract temporal features, and the variational autoencoder is combined to perform probabilistic modeling on phase jitter.

[0044] The Kalman smoothing algorithm is used for post-processing of the interference feature encoding to generate a time-varying interference suppression compensation signal.

[0045] 2. Construct a multi-physical-field coupling simulation environment (1) Initialize the heterogeneous computing architecture: Based on the configuration information, the heterogeneous computing architecture is dynamically identified and initialized through an adaptive resource scheduling algorithm.

[0046] (2) Allocate computing resources: According to the multi-physical-field coupling simulation requirements, a load balancing algorithm is used to predict and allocate various computing resources.

[0047] (3) Introduce physical field models: Introduce the mathematical models and interaction rule sets of light transmission, electromagnetic field distribution, and target motion prediction.

[0048] (4) Establish a coupling topology architecture: Parse the spatio-temporal correlation parameters through an intelligent fusion algorithm to establish a multi-physical-field coupling topology architecture.

[0049] (5) Generate a dynamic test scenario: Based on the simulation target constraint conditions and historical data sets, generate a dynamic test scenario containing spatio-temporal correlation.

[0050] 3. Light transmission modeling, electromagnetic field distribution calculation, and target motion prediction ① Light transmission modeling: Initialize the atmospheric environment parameters and construct a dynamic meteorological field.

[0051] Adopt an improved Mie scattering model and a GPU-accelerated Monte Carlo ray tracing algorithm to generate a light transmission attenuation field that changes dynamically with the meteorological field.

[0052] Through a meteorological radar data-driven dynamic texture synthesis algorithm, the turbulence intensity distribution is modulated in real time.

[0053] ② Electromagnetic field distribution calculation: Construct a coupling model of the hybrid potential integral equation and the equivalent dipole method, and divide the high- and low-frequency calculation domains.

[0054] Use OpenMP for parallel computing of low-frequency electromagnetic fields and accelerate the solution of high-frequency components through cuFFT.

[0055] Perform real-time mapping operations on electromagnetic field-fluid field data to achieve spatio-temporal consistency alignment of multi-physical fields.

[0056] ③ Target motion prediction: Use the LSTM-VAM encoder to generate a sequence of latent variables of motion intentions.

[0057] Construct a Markov decision process based on maximum entropy through inverse reinforcement learning.

[0058] Use model predictive control to roll and optimize the trajectory prediction results.

[0059] 4. Extract multi-dimensional performance indicators and establish a mapping relationship for parameter optimization (1) Hybrid feature analysis: Call the wavelet packet transform algorithm to decompose the simulation data in the time-frequency domain, and extract the non-linear feature subset through kernel principal component analysis.

[0060] (2) Fuzzy inference: Construct a fuzzy inference system, train the membership function parameters, and generate a mapping relationship from feature indicators to control parameters.

[0061] 5. Multi-objective dynamic optimization of control parameters (1) Initialize the quantum state: Use the quantum bit encoding strategy to initialize the population of control parameters, and generate an initial quantum state sequence through chaotic Logistic mapping.

[0062] (2) Define the optimization function: Define a two-objective optimization function (root mean square of tracking error and change rate of control quantity), and set the initial range of the quantum gate rotation angle.

[0063] (3) Perform quantum rotation gate operations: Based on the initial range of the quantum gate rotation angle, perform quantum rotation gate operations on the current population, and the phase update amount is dynamically adjusted by the individual crowding degree ranking.

[0064] (4) Generate optimization results: Collapse the quantum state into a classical parameter solution, calculate the two-objective optimization function value, and perform non-dominated sorting to output the multi-objective dynamic optimization results.

[0065] 6. Closed-loop iterative optimization (1) Construct a virtual-real data interaction channel: Import the optical CAD model into the digital twin platform and initialize the photon tracing parameters.

[0066] (2) Hardware-in-the-loop Interface: Configure the FPGA module for nanosecond-level real-time simulation, and adopt the timestamp alignment algorithm to compensate for data transmission delay.

[0067] (3) Frequency Response Calibration: Perform hardware frequency response calibration through the chirp signal injection method, generate compensation filtering and load it into the FPGA.

[0068] (4) State Estimation and Deviation Prediction: Adopt the unscented Kalman filter algorithm for multi-target state estimation and deviation prediction, and establish a deviation compensation function in combination with diffeomorphic mapping.

[0069] (5) Online Evolution of Control Parameters: Based on the control parameter correction amount, adopt the covariance matrix adaptation evolution strategy to online evolve the control parameters, and set the iteration period to 20 ms.

[0070] III. Application Effects Through this real-time simulation method, we can accurately simulate the working state of optoelectronic tracking devices in complex environments, process optoelectronic signals in real time, and dynamically optimize control parameters. This not only improves the tracking accuracy and response speed of optoelectronic tracking devices, but also enhances their adaptability and robustness in complex environments. At the same time, through closed-loop iterative optimization, we have achieved continuous optimization of the control algorithm, further improving the overall performance of the device.

[0071] Example 3: The embodiment of the present invention also provides a computer-readable storage medium, on which a program for a real-time simulation method for detecting optoelectronic tracking devices as described in any one of the above is stored. When the program is executed by a processor, it realizes each process of the above real-time simulation method embodiment and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium, such as a read-only memory (Read-Only Memory, abbreviated as ROM), a random access memory (Random ACGess Memory, abbreviated as RAM), a magnetic disk or an optical disc, etc.

[0072] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0073] In the accompanying drawings of the disclosed embodiments of the present invention, only the structures related to the disclosed embodiments of the present invention are involved. Other structures can refer to the general design. Without conflict, the same embodiment and different embodiments of the present invention can be combined with each other.

[0074] The flowcharts shown in the accompanying drawings are only illustrative examples, and do not necessarily include all the contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined or partially merged. Therefore, the actual execution order may be changed according to the actual situation.

[0075] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A real-time simulation method for detecting optoelectronic tracking equipment, characterized in that, Including: Performing feature decoupling processing on optoelectronic signals through a dual-channel adaptive neural network architecture, and constructing a device mathematical model based on the processing results using a dynamic gating fusion mechanism; Building a multi-physical field coupling simulation environment based on a heterogeneous computing architecture, introducing the device mathematical model, and performing three-field collaborative evolution through ray transmission modeling, electromagnetic field distribution calculation, and target motion prediction to generate a dynamic test scene with spatio-temporal correlation; Obtaining simulation data of multi-dimensional physical fields based on the dynamic test scene, and using a method combining hybrid feature analysis and fuzzy inference to extract multi-dimensional performance indicators from the simulation data and establish a parameter optimization mapping relationship; Based on the parameter optimization mapping relationship, adopting a hybrid strategy that combines quantum evolution algorithm and swarm intelligence optimization, and performing multi-objective dynamic optimization of control parameters through an environment perception weighting mechanism to generate a multi-objective dynamic optimization result; Constructing a virtual-real data interaction channel through a digital twin platform, inputting the multi-objective dynamic optimization result into a photon-level simulation engine, and completing the closed-loop iterative optimization of the control algorithm through real-time verification of the hardware-in-the-loop interface.

2. The real-time simulation method for detecting an optoelectronic tracking device according to claim 1, wherein: The dual-channel adaptive neural network architecture includes a spatial noise processing channel and a time-varying interference processing channel; Using the spatial noise processing channel to perform non-local mean denoising processing on optoelectronic signals, retaining high-frequency edge features, and obtaining denoised optoelectronic signals; Inputting the denoised optoelectronic signals into a fourth-order residual network to extract shallow texture features, and using the cross-attention module of Swin Transformer to enhance the global context of the shallow texture features to generate a feature matrix; Processing the feature matrix through multi-scale spatial convolution and adaptive entropy weighted fusion to generate a noise distribution map; Processing the noise distribution map through a pre-trained conditional generative adversarial network to generate a pixel-level noise compensation signal; Using the time-varying interference processing to perform short-time Fourier transform on optoelectronic signals to generate a time-frequency spectrum map; Based on the time-frequency spectrum map, extracting temporal features through the causal dilated convolution layer of a probabilistic temporal convolutional network; Performing probabilistic modeling on the phase jitter of the temporal features through a variational autoencoder to generate interference feature codes; Performing post-processing on the interference feature codes using a Kalman smoothing algorithm to generate a time-varying interference suppression compensation signal.

3. A real-time simulation method for detecting an optoelectronic tracking device according to claim 2, characterized in that: The steps of constructing a device mathematical model based on the processing results using a dynamic gating fusion mechanism include: Respectively extracting features from the pixel-level noise compensation signal and the time-varying interference suppression compensation signal to generate a spatial feature vector and a time-varying feature vector; Based on the spatial feature vector and the time-varying feature vector, constructing a dual-channel feature similarity matrix through a multi-layer perceptron and a cosine similarity function; Based on the dual-channel feature similarity matrix, using a policy gradient reinforcement learning network with tracking error entropy and control efficiency weighting as the reward function to generate dynamic fusion weights; Design gating function Perform probabilistic channel selection, maximize the cumulative reward, and output the mathematical model of the device ； In the formula, is the action value function, is the channel selection action, is the temperature coefficient.

4. A real-time simulation method for detecting an optoelectronic tracking device according to claim 3, characterized in that: Building a multi-physical field coupling simulation environment based on a heterogeneous computing architecture, including: Based on the configuration information of the heterogeneous computing architecture, dynamically identifying and initializing the heterogeneous computing architecture through an adaptive resource scheduling algorithm; Based on the initialized heterogeneous computing architecture, a load balancing algorithm is used to predict and allocate various computing resources according to the requirements of multi-physical field coupling simulation; Introduce the mathematical models and interaction rule sets of each physical field, and establish a multi-physical field coupling topology architecture by parsing the spatio-temporal correlation parameters through an intelligent fusion algorithm; Based on the simulation target constraint conditions and historical data sets, trigger the intelligent optimization algorithm to retrieve the strategy library and generate the parameter configuration tables of each physical field; Input the configuration parameters of each physical field into the established multi-physical field coupling topology architecture, define the data exchange protocol, and establish a time step alignment mechanism for each physical field to form a multi-physical field coupling simulation environment.

5. A real-time simulation method for detecting an optoelectronic tracking device according to claim 4, wherein: The light transmission modeling includes: Initialize the atmospheric environment parameters to construct a dynamic meteorological field, and use an improved Mie scattering model and a GPU-accelerated Monte Carlo ray tracing algorithm to generate an optical transmission attenuation field that changes dynamically with the meteorological field; Through a dynamic texture synthesis algorithm driven by meteorological radar data, modulate the turbulence intensity distribution in real time and update the parameters of the optical transmission attenuation field; The electromagnetic field distribution calculation includes: Construct a coupling model of the mixed potential integral equation and the equivalent dipole method, and divide the high- and low-frequency calculation domains; Based on the high- and low-frequency calculation domains, use OpenMP to parallel compute the low-frequency electromagnetic field, and accelerate the solution of the high-frequency components through cuFFT to generate the electromagnetic field distribution of the full frequency band; Based on the electromagnetic field distribution of the full frequency band, perform real-time mapping operations of electromagnetic field-fluid field data through a multi-variable coupling solver, and perform spatio-temporal consistency alignment of multi-physical fields based on the time step alignment mechanism; The target motion prediction includes: Use the target historical trajectory data to train the LSTM-VAM encoder to generate a sequence of latent variables of motion intention; Analyze the reward function corresponding to the latent variables through inverse reinforcement learning, and construct a Markov decision process based on maximum entropy; Use model predictive control to roll and optimize the Markov decision process to generate trajectory prediction results within a preset time period.

6. A real-time simulation method for detecting an optoelectronic tracking device according to claim 5, characterized in that: The simulation data of multi-dimensional physical fields are obtained based on the dynamic test scenario, and a method combining hybrid feature analysis and fuzzy inference is used to extract multi-dimensional performance indicators from the simulation data and establish a parameter optimization mapping relationship, including: Call the wavelet packet transform algorithm to decompose the simulation data in the time-frequency domain, and extract the non-linear feature subset through kernel principal component analysis to generate a hybrid feature matrix; Define the input / output language variables, initialize the membership function of the fuzzy set based on the expert rule base, and construct a fuzzy inference system; Configure the Mamdani inference engine and the centroid method defuzzifier, input the hybrid feature matrix into the fuzzy inference system to train the membership function parameters, and generate the mapping relationship from the feature index to the control parameter.

7. A real-time simulation method for detecting an optoelectronic tracking device according to claim 6, characterized in that: Based on the parameter optimization mapping relationship, adopt a hybrid strategy that combines the quantum evolution algorithm and swarm intelligence optimization, and perform multi-objective dynamic optimization of the control parameters through the environmental perception weighting mechanism to generate multi-objective dynamic optimization results, including: Use the quantum bit coding strategy to initialize the control parameter population, and generate an initial quantum state sequence through chaotic Logistic mapping; Define a bi-objective optimization function based on the training of the initial quantum state: the root mean square of the tracking error and the change rate of the control quantity, and set the initial range of the quantum gate rotation angle; Initialize the Pareto front archive and set the crowding distance threshold , and establish the non-dominated sorting levels; Perform the quantum rotation gate operation on the current population based on the initial range of the quantum gate rotation angle, and the phase update amount is dynamically adjusted by the individual crowding degree ranking: ； Collapse the quantum state into a classical parameter solution, calculate the bi-objective optimization function value, and perform non-dominated sorting according to the calculation results; Extract the parameter encoding patterns of the top 10% individuals based on the non-dominated sorting, and construct a graph structure containing the historical optimization path; Learn the graph structure parameter correlation features through a graph attention network, generate a feature embedding vector in combination with a Transformer encoder, and input the feature embedding vector into a multi-layer perceptron to generate a dynamic weight distribution; Design weighted function Optimize the real-time allocation of the target dynamic weight distribution and output the multi-target dynamic optimization result; ； wherein, is the weight of the th optimization objective at time , is the feature embedding vector corresponding to the th objective, is the real-time environment feature vector, is the vector concatenation operation, is the learnable parameter vector.

8. A real-time simulation method for detecting an optoelectronic tracking device according to claim 7, characterized in that: Build a virtual-real data interaction channel through the digital twin platform, input the multi-objective dynamic optimization result into the photon-level simulation engine, and complete the closed-loop iterative optimization of the control algorithm through real-time verification at the hardware-in-the-loop interface, including: Import the optical CAD model in the digital twin platform and configure the spectral sampling parameters, and initialize the photon tracing parameters; Based on the initialized photon tracing parameters, generate photon tracing data through the Jones matrix and the Monte Carlo algorithm, and output the scattering field distribution to the hardware-in-the-loop interface in combination with the bidirectional reflectance distribution function model; Configure the FPGA module at the hardware-in-the-loop interface for nanosecond-level real-time simulation, and use the timestamp alignment algorithm to compensate for the data transmission delay; Perform hardware frequency response calibration through the chirp signal injection method, generate a compensation filter and load it into the FPGA; Build a virtual-real data two-way channel, use the unscented Kalman filter algorithm for multi-objective state estimation and deviation prediction, establish a deviation compensation function in combination with the diffeomorphic mapping, and generate a control parameter correction amount; Based on the control parameter correction amount, use the covariance matrix adaptation evolution strategy to evolve the control parameters online, set the iteration period to 20ms, and complete the closed-loop optimization process.

Images