Three-dimensional interstellar medium physical field reconstruction method and device and storage medium

Abstract

The invention relates to a three-dimensional interstellar medium physical field reconstruction method and device and a storage medium, and the method comprises the steps: obtaining three-dimensional interstellar medium physical field simulation data, inputting the three-dimensional interstellar medium physical field simulation data to a three-dimensional photochemical radiation model, generating a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set, constructing a training sample; inputting the training sample into the initial deep learning architecture to obtain a three-dimensional interstellar medium physical field prediction intermediate result; calculating loss based on the difference between the result and a true value three-dimensional interstellar medium physical field data set, and performing gradient reverse conduction to an initial deep learning architecture for iterative training to obtain a final inversion deep learning architecture; and inputting real observation data into the deep learning architecture to obtain a three-dimensional interstellar medium physical field reconstruction result. According to the method and the device, the problem that observation data is difficult to accurately restore to a complex physical state of a real three-dimensional interstellar medium in related technologies is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of astrophysical data processing, and in particular to a method and device for reconstructing a three-dimensional interstellar medium physical field and a storage medium. BACKGROUND

[0002] The density, temperature and chemical structure of the interstellar medium are the core of analyzing the evolution of molecular clouds and the mechanism of star formation, and the reconstruction of the three-dimensional structure thereof is of key significance to astrophysical research. However, in the related art, a method of fitting data by using a simple empirical model cannot accurately restore the observed data to the complex physical state of the real three-dimensional interstellar medium.

[0003] At present, there is no effective solution to the problem that the observed data cannot be accurately restored to the complex physical state of the real three-dimensional interstellar medium in the related art. SUMMARY

[0004] Embodiments of the present application provide a method and device for reconstructing a three-dimensional interstellar medium physical field and a storage medium to at least solve the problem that the observed data cannot be accurately restored to the complex physical state of the real three-dimensional interstellar medium in the related art.

[0005] In a first aspect, embodiments of the present application provide a method for reconstructing a three-dimensional interstellar medium physical field, which comprises:

[0006] obtaining three-dimensional interstellar medium physical field simulation data;

[0007] inputting the three-dimensional interstellar medium physical field simulation data into a preset three-dimensional photochemical radiation model to generate a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set; the true value three-dimensional interstellar medium physical field data set is formed based on the three-dimensional interstellar medium physical field simulation data and further calculated by the three-dimensional photochemical radiation model after being supplemented;

[0008] based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set, constructing a training sample;

[0009] inputting the training sample into an initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field prediction intermediate result; based on the difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set, calculating a loss result;

[0010] conducting iterative training of the loss result by backward propagation of the gradient of the loss result to the initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field inversion deep learning architecture;

[0011] Input the obtained real observation data into the three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field reconstruction result.

[0012] In some embodiments, the loss result is calculated based on a difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set, including:

[0013] A first loss is constructed based on a difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set;

[0014] The three-dimensional interstellar medium physical field prediction intermediate result is input into a preset fast rendering module to obtain a synthetic observation prediction intermediate result; the fast rendering module is an approximate calculation module of the three-dimensional photochemical radiation model;

[0015] A second loss is constructed based on a difference between the synthetic observation prediction intermediate result and the synthetic observation data set;

[0016] The loss result is obtained based on the first loss and the second loss.

[0017] In some embodiments, the loss result is obtained based on the first loss and the second loss, including:

[0018] A third loss is constructed based on the three-dimensional interstellar medium physical field prediction intermediate result and a preset physical constraint;

[0019] The loss result is obtained based on the first loss, the second loss, and the third loss.

[0020] In some embodiments, the three-dimensional interstellar medium physical field simulation data is obtained, including:

[0021] Original three-dimensional interstellar medium physical field simulation data is obtained;

[0022] The original three-dimensional interstellar medium physical field simulation data is resampled to a regular cubic grid to obtain the three-dimensional interstellar medium physical field simulation data.

[0023] In some embodiments, the training sample is constructed based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set, including:

[0024] An original training sample is constructed based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set;

[0025] The original training sample is subjected to multi-view cropping and data enhancement, and is stored in a preset format to obtain the training sample.

[0026] In some embodiments, after obtaining the three-dimensional interstellar medium physical field reconstruction result, further comprising:

[0027] inputting the three-dimensional interstellar medium physical field reconstruction result into a preset radiation transfer module to obtain a synthetic observation result;

[0028] verifying a prediction effect based on the synthetic observation result and the real observation data.

[0029] In some embodiments, verifying the prediction effect based on the synthetic observation result and the real observation data comprises:

[0030] quantitatively comparing the synthetic observation result and the real observation data to obtain the prediction effect; the quantitative comparison includes line intensity error, spectral line type comparison, image structure comparison, and statistical quantity consistency.

[0031] In some embodiments, the true value three-dimensional interstellar medium physical field data set at least includes a three-dimensional density distribution.

[0032] In a second aspect, the embodiments of the present application provide a three-dimensional interstellar medium physical field reconstruction device, the device comprising:

[0033] a three-dimensional data acquisition module, configured to acquire three-dimensional interstellar medium physical field simulation data;

[0034] a data set generation module, configured to input the three-dimensional interstellar medium physical field simulation data into a preset three-dimensional photochemical radiation model to generate a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set; the true value three-dimensional interstellar medium physical field data set is formed based on the three-dimensional interstellar medium physical field simulation data and further calculated and supplemented by the three-dimensional photochemical radiation model;

[0035] a training sample construction module, configured to construct a training sample based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set;

[0036] a loss calculation module, configured to input the training sample into an initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field prediction intermediate result; and calculate a loss result based on a difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set;

[0037] a model training module, configured to conduct back propagation of a gradient of the loss result to the initial three-dimensional interstellar medium physical field inversion deep learning architecture for iterative training to obtain a three-dimensional interstellar medium physical field inversion deep learning architecture;

[0038] The model inference module is configured to input the obtained real observation data into the three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field reconstruction result.

[0039] In a third aspect, an embodiment of the present application provides a storage medium having a computer program stored thereon, where the computer program, when executed by a processor, implements the three-dimensional interstellar medium physical field reconstruction method of the first aspect.

[0040] Compared with the related art, the three-dimensional interstellar medium physical field reconstruction method, the device, and the storage medium provided by the embodiments of the present application have the following advantages. The three-dimensional interstellar medium physical field simulation data is obtained. The three-dimensional photochemical radiation model is input into a preset three-dimensional photochemical radiation model to generate a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set. The true value three-dimensional interstellar medium physical field data set is formed based on the three-dimensional interstellar medium physical field simulation data and further calculated by the three-dimensional photochemical radiation model. The training sample is constructed based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set. The training sample is input into an initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field prediction intermediate result. The loss result is calculated based on the difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set. The gradient of the loss result is conducted to the initial three-dimensional interstellar medium physical field inversion deep learning architecture for iterative training to obtain a three-dimensional interstellar medium physical field inversion deep learning architecture. The real observation data obtained is input into the three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field reconstruction result, thereby solving the problem that it is difficult to accurately restore the observation data to the complex physical state of the real three-dimensional interstellar medium in the related art.

[0041] The details of one or more embodiments of the present application are presented in the following drawings and description to make other features, objects, and advantages of the present application more apparent. BRIEF DESCRIPTION OF DRAWINGS

[0042] The accompanying drawings illustrated herein are used to provide further understanding of the present application, constitute a part of the present application, and the illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute improper limitations on the present application. In the drawings:

[0043] Figure 1 FIG. 1 is a hardware structure block diagram of a terminal according to a three-dimensional interstellar medium physical field reconstruction method according to an embodiment of the present application;

[0044] Figure 2 FIG. 2 is a flowchart of a three-dimensional interstellar medium physical field reconstruction method according to an embodiment of the present application;

[0045] Figure 3 is a schematic diagram of a three-dimensional interstellar medium physical field reconstruction system framework according to an embodiment of the present application;

[0046] Figure 4 is a structural block diagram of a three-dimensional interstellar medium physical field reconstruction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is described and explained below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application. In addition, it can be understood that although the efforts made in this development process can be complex and lengthy, some changes in design, manufacture or production and the like based on the technical content disclosed in the present application are only routine technical means for those of ordinary skill in the art related to the content disclosed in the present application, and should not be understood as insufficient disclosure of the content disclosed in the present application.

[0048] In the present application, the phrase "embodiments" means that the specific features, structures or characteristics described in combination with the embodiments can be included in at least one embodiment of the present application. The appearance of this phrase at various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment to other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in the present application can be combined with other embodiments without conflict.

[0049] Unless otherwise defined, technical terms and scientific terms used in the present application shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. Unless otherwise defined, the terms "one" and "a" or "an" used in the present application shall not be limited to singular aspects but can include both singular and plural aspects. The terms "comprising," "including," "containing," and any variations thereof in the present application shall be taken to cover both not-exclusively-inclusive and inclusive aspects; for example, a process, method, system, product, or apparatus including a list of steps or modules (units) is not limited to the listed steps or units, but can further include steps or units not listed or can further include other steps or units inherent to such a process, method, product, or apparatus. The terms "connected," "coupled," and similar terms in the present application are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The term "plurality" in the present application means greater than or equal to two. The term "and / or" describes an associated relationship between associated objects, which means that there can be three relationships, for example, "A and / or B" can mean that A exists alone, A and B exist together, and B exists alone. The terms "first," "second," "third," and the like in the present application are merely to distinguish similar objects, and do not represent a specific order for the objects.

[0050] The method embodiments provided by the present embodiment can be executed in a terminal, a computer, or a similar computing device. Taking a terminal as an example, Figure 1 is a hardware structure block diagram of a terminal according to a three-dimensional interstellar medium physical field reconstruction method of an embodiment of the present application. As shown in Figure 1 , the terminal can include one or more (only one is shown in Figure 1 ) processor 102 (the processor 102 can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the above terminal can also include a transmission device 106 for communication function and an input and output device 108. Those skilled in the art can understand that Figure 1 The structure shown is only schematic, which does not limit the structure of the above terminal. For example, the terminal can include more or fewer components than those shown in Figure 1 , or have a different configuration than that shown in Figure 1 .

[0051] The memory 104 can be used to store computer programs, such as software programs of application software and modules, for example, a computer program corresponding to the three-dimensional interstellar medium physical field reconstruction method in the embodiments of the present application. The processor 102 performs various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the method described above. The memory 104 can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include a memory remotely arranged with respect to the processor 102, which can be connected to the terminal through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0052] The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network can include a wireless network provided by a communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC) which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, RF) module used to communicate with the Internet in a wireless manner.

[0053] The embodiments of the present application provide a three-dimensional interstellar medium physical field reconstruction method, Figure 2 is a flowchart of the three-dimensional interstellar medium physical field reconstruction method according to the embodiments of the present application, as shown in Figure 2 The flowchart includes the following steps:

[0054] In step S201, three-dimensional interstellar medium physical field simulation data is obtained.

[0055] Specifically, the three-dimensional interstellar medium physical field simulation data can be obtained from the STARFORGE project, which is a multi-institutional joint project aimed at developing high-resolution three-dimensional radiation-magnetic hydrodynamic (Radiation-MHD) star formation simulation. Among them, the STARFORGE simulates the process of nebula evolution including gravity, magnetic field, stellar feedback (jet, stellar wind, radiation, supernova explosion), and generates three-dimensional physical fields such as gas density, velocity, temperature, star formation and cluster assembly.

[0056] Step S202, input the three-dimensional interstellar medium physical field simulation data into the preset three-dimensional photochemical radiation model to generate a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set; the true value three-dimensional interstellar medium physical field data set is formed based on the three-dimensional interstellar medium physical field simulation data and further calculated and supplemented by the three-dimensional photochemical radiation model.

[0057] The three-dimensional photochemical radiation model can be a 3D-PDR model, which is a chemical-thermal equilibrium-radiation transfer code suitable for three-dimensional photodissociation region, can process three-dimensional clouds with arbitrary density distribution, and adopts HEALPix ray tracing to calculate far ultraviolet (FUV) radiation attenuation, thermal equilibrium and chemical reaction.

[0058] The three-dimensional interstellar medium physical field simulation data is input into the three-dimensional photochemical radiation model, on the one hand, a true value three-dimensional interstellar medium physical field data set containing key physical quantities such as chemical abundance distribution (such as H2, C, CO, etc.) and gas temperature distribution is generated; on the other hand, two-dimensional or spectral cube images related to multi-band spectral line emission (such as CO, [C II], [O I] cooling lines) are generated through radiation transfer process, which constitute a synthetic observation data set, providing physical consistency and data dimension matching for subsequent training sample construction and model training.

[0059] It should be noted that the above true value three-dimensional interstellar medium physical field data set is not constructed from zero, but is derived from the original three-dimensional interstellar medium physical field simulation data based on the three-dimensional photochemical radiation model, at least covering the three-dimensional density distribution, and can also include temperature distribution, chemical abundance distribution and far ultraviolet radiation field and other extended physical quantities.

[0060] Step S203, based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set, a training sample is constructed.

[0061] Wherein, based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set generated in step S202, when constructing the training sample, first, the synthetic observation data is taken as the input feature (covering two-dimensional images, spectral cube and other forms of multiple bands or spectral lines, such as cooling line related data of CO, [C II], [O I], etc.), and the corresponding true value three-dimensional physical field is taken as the label (at least containing three-dimensional density distribution, and can also include temperature distribution, chemical abundance distribution, far ultraviolet radiation field and other multi-physical quantity set), forming a training sample pair corresponding to the "input-label" pair.

[0062] Step S204, input the training sample into the initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field prediction intermediate result; based on the difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set, a loss result is calculated.

[0063] Specifically, the training samples are input into the pre-constructed initial three-dimensional interstellar medium physical field inversion deep learning architecture, which can use a deep neural network structure capable of mapping from two-dimensional / 2.5D low-dimensional observation data to high-dimensional three-dimensional physical field, such as 3D U-Net, V-Net, Encoder-Decoder + Skip Connections, Diffusion Model or Schrödinger Bridge inversion model, to output the three-dimensional interstellar medium physical field prediction intermediate results including three-dimensional density distribution, temperature distribution, chemical abundance distribution and other core physical quantities through feature extraction, dimensionality increase and mapping learning of the network.

[0064] Then, based on the numerical difference between the prediction intermediate results and the true value three-dimensional interstellar medium physical field dataset (derived from three-dimensional numerical simulation and 3D-PDR photochemical radiation model, containing density, temperature, chemical abundance and other multiple physical quantities), a loss function (i.e. supervised inversion loss such as L1 or L2 error) is constructed, and finally the loss result is obtained.

[0065] Step S205, the gradient of the loss result is back-propagated to the initial three-dimensional interstellar medium physical field inversion deep learning architecture for iterative training, and the three-dimensional interstellar medium physical field inversion deep learning architecture is obtained.

[0066] After the calculation of the loss result, the gradient of the loss is back-propagated along the network levels of the initial three-dimensional interstellar medium physical field inversion deep learning architecture, and the gradient descent optimization algorithm (such as Adam, SGD, etc.) is used to iteratively update the learnable parameters of the network, such as convolution kernel parameters and fully connected layer weights, continuously reducing the deviation between the model prediction result and the true value three-dimensional interstellar medium physical field dataset in each iteration.

[0067] In this process, by monitoring the loss changes of the training set and the validation set, the prediction accuracy of the model and the degree of satisfaction of the physical constraints, it is judged whether the model converges or not. When the loss value tends to be stable and no longer decreases significantly, and the numerical error of the prediction result reaches the preset standard, the iterative training is stopped, and finally the three-dimensional interstellar medium physical field inversion deep learning architecture with the ability to accurately invert the three-dimensional interstellar medium physical field from two-dimensional / 2.5D observation data is obtained. The architecture can stably output prediction results containing at least three-dimensional density distribution, and optionally temperature distribution, chemical abundance distribution and other core physical quantities, and support uncertainty estimation to improve the explainability and credibility of the results.

[0068] It should be noted that the uncertainty evaluation refers to the model outputting the three-dimensional reconstruction result of interstellar medium while synchronously providing the reliability and uncertainty information of the reconstruction result at each spatial position. The uncertainty mainly comes from three categories: firstly, the physical unavoidable uncertainty caused by incomplete observation information, i.e., the two-dimensional projection cannot uniquely determine the three-dimensional structure (ill-posed inverse problem), and different 3D structures can present similar projection effects; secondly, the uncertainty caused by observation noise and instrument effects, i.e., even if the model is perfect, it is difficult to achieve 100% certainty; thirdly, the uncertainty caused by the approximation of the model and the physical process, because the solving process has approximation, and the machine learning model is not an analytical solution. This uncertainty estimation makes the output no longer a single "unique solution", but a "credible solution space", which can be verified for probability consistency with the real observation, the region with large uncertainty allows larger observation deviation, and the region with small uncertainty requires strict consistency, which can not only enhance the interpretability and credibility of the reconstruction result, but also clearly distinguish the uncertainty caused by the insufficient observation data and the model prediction result, avoid over-interpretation of the three-dimensional structure, and also provide a key basis for the reliability evaluation of the reconstruction result in different spatial regions during the reasoning stage.

[0069] In step S206, the obtained real observation data is input into the three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field reconstruction result.

[0070] In this step, the obtained real observation data (derived from the multi-band spectral lines, two-dimensional projection images and the like captured by a telescope, such as CO, [C II], [O I] and the like cooling line related data) needs to be preprocessed, and the data format is standardized, the scale is unified, and the noise is preliminarily processed according to the format standard of the training sample to ensure that the input data features are consistent with those during model training.

[0071] Subsequently, the preprocessed real observation data is input into the three-dimensional interstellar medium physical field inversion deep learning architecture which has been completed through multiple rounds of iterative training and optimization. The architecture has been trained through a loss function and has stable low-dimensional observation data to high-dimensional three-dimensional physical field mapping capability, and can accurately invert and output a three-dimensional interstellar medium physical field reconstruction result based on the input real observation information. The result at least includes the core physical quantity of three-dimensional density distribution, and can also output temperature distribution, chemical abundance distribution and other key physical parameters according to actual needs, while supporting the output of the uncertainty estimation of each spatial position and the reliability of the reconstruction result in different regions, providing high-precision and interpretable three-dimensional physical field data support for subsequent observation verification and astronomical research.

[0072] The steps S201 to S206 described above, by obtaining three-dimensional interstellar medium physical field simulation data, inputting the three-dimensional interstellar medium physical field simulation data into a preset three-dimensional photochemical radiation model to generate a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set further calculated based on the simulation data, constructing a training sample based on the two types of data sets, inputting the training sample into an initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a predicted intermediate result and calculate a loss result, then conducting gradient backpropagation of the loss result to the initial architecture for iterative training to obtain an optimized inversion deep learning architecture, and finally inputting real observation data to obtain a three-dimensional interstellar medium physical field reconstruction result, realize deep integration of real physical simulation, photochemical radiation calculation and machine learning inverse problem modeling, break through the limitations of traditional methods relying on idealized assumptions or simple empirical fitting, solve the technical problem of accurately restoring the real three-dimensional interstellar medium complex physical state from observation data, achieve a three-dimensional interstellar medium physical field reconstruction effect with high precision, physical self-consistency, and interpretability, and have good generalization ability and scalability, which can adapt to observation data of different wavebands and different instruments, and provide strong technical support for astrophysical research such as molecular cloud evolution and star formation mechanism.

[0073] In some embodiments, the loss result is calculated based on a difference between the three-dimensional interstellar medium physical field predicted intermediate result and the true value three-dimensional interstellar medium physical field data set, including:

[0074] A first loss is constructed based on a difference between the three-dimensional interstellar medium physical field predicted intermediate result and the true value three-dimensional interstellar medium physical field data set.

[0075] The three-dimensional interstellar medium physical field predicted intermediate result is input into a preset fast rendering module to obtain a synthetic observation predicted intermediate result; the fast rendering module is an approximate calculation module of the three-dimensional photochemical radiation model.

[0076] A second loss is constructed based on a difference between the synthetic observation predicted intermediate result and the synthetic observation data set.

[0077] The loss result is obtained based on the first loss and the second loss.

[0078] Firstly, the first loss (i.e., supervised inversion loss) is constructed. The core role of the first loss is to enable the model to quickly learn the basic mapping relationship from the observed data to the three-dimensional physical field, ensuring that the prediction result is close in value to the true physical state. Specifically, after inputting the training sample into the initial three-dimensional interstellar medium physical field inversion deep learning architecture, the prediction intermediate result containing the three-dimensional density distribution and possibly the temperature distribution, chemical abundance distribution, and other core physical quantities is obtained. The intermediate result is compared with the true value three-dimensional interstellar medium physical field dataset (based on three-dimensional numerical simulation, further calculated and supplemented by the 3D-PDR photochemical radiation model, covering the true value set of multiple physical quantities such as density, temperature, and chemical abundance) pixel by pixel or voxel by voxel. The difference between the two is quantified using L1 error, L2 error, and other loss calculation methods to construct the first loss, providing a basic supervision signal for model training.

[0079] Secondly, the second loss (i.e., observation consistency constraint loss) is constructed. In order to avoid the model only pursuing numerical fitting and deviating from the actual observation scenario, observation consistency constraint needs to be introduced, which relies on a pre-set fast rendering module. The fast rendering module is an approximate calculation module of the three-dimensional photochemical radiation model (such as 3D-PDR), which has the core advantage of significantly reducing computational complexity while ensuring main physical consistency. Specifically, it can be achieved by weighted integration along the line of sight direction of the predicted three-dimensional physical field, emission approximation calculation based on physical empirical formula, table lookup or interpolation radiation calculation, or using a pre-trained proxy model to approximate the three-dimensional photochemical radiation model. After inputting the above-mentioned three-dimensional interstellar medium physical field prediction intermediate result into the fast rendering module, the synthesized observation prediction intermediate result is generated, which simulates the two-dimensional projection image or spectral line cube data in the real observation scenario. Then, the synthesized observation prediction intermediate result is compared with the corresponding synthesized observation dataset (standard observation data generated by the complete radiation transfer process of the three-dimensional true value physical field), and the difference in key observation features such as spectral line intensity, image structure, and spectral line type is quantified, and then the second loss is constructed to ensure that the three-dimensional physical field predicted by the model can correspond to the result that meets the real observation law in reverse.

[0080] Finally, the final loss result is obtained by weighted fusion of the first loss and the second loss. By configuring reasonable adjustable weights for the first loss and the second loss, the importance of numerical fitting accuracy and observation consistency constraint can be balanced according to the training requirements. For example, the weight of the first loss can be appropriately increased in the early stage of training to enable the model to quickly grasp the basic structural features of the three-dimensional physical field, and the weight of the second loss can be increased in the later stage of training to strengthen the matching degree of the model prediction result and the observation data. The loss result obtained by weighted summation of the two contains both numerical level error constraints and observation level physical consistency requirements, providing a comprehensive and reliable guidance signal for the subsequent iteration and optimization of model parameters.

[0081] The above steps ensure that the model prediction conforms to the real physical state at the numerical level by constructing a first loss based on the difference between the intermediate result of the three-dimensional interstellar medium physical field prediction and the true value three-dimensional interstellar medium physical field dataset, and then inputting the prediction intermediate result into a fast rendering module of a three-dimensional photochemical radiation model approximation calculation module to obtain a synthetic observation prediction intermediate result, constructing a second loss based on the difference between the result and the synthetic observation dataset, and finally combining the first loss and the second loss to obtain a loss result, which realizes double constraints on the model prediction, guarantees the numerical accuracy of the three-dimensional physical field prediction, and avoids the model from deviating from the actual observation law by only pursuing numerical fitting through the observation consistency constraint, so that the mapping relationship learned by the model from two-dimensional / 2.5D observation data to a three-dimensional physical field conforms to both the numerical true value and the real observation scene spectrum, image and other characteristics, significantly improves the reliability and physical self-consistency of the model inversion, and lays a solid foundation for subsequent accurate reconstruction of the three-dimensional interstellar medium physical field.

[0082] In some embodiments, the loss result is obtained based on the first loss and the second loss, including:

[0083] A third loss is constructed based on the three-dimensional interstellar medium physical field prediction intermediate result and a preset physical constraint;

[0084] The loss result is obtained based on the first loss, the second loss and the third loss.

[0085] The process of obtaining a loss result based on the first loss and the second loss further introduces a regularization mechanism of a physical constraint dimension. By constructing a third loss and fusing it with the first two, a comprehensive loss result is formed, which has numerical accuracy, observation consistency and physical reasonableness. Specifically, first, a third loss (i.e., a physical constraint regularization loss) is constructed. The core purpose is to avoid the model outputting a three-dimensional physical field result that violates the basic physical laws in the process of pursuing numerical fitting and observation matching, and to ensure the physical self-consistency of the reconstruction result. The preset physical constraint is based on the inherent physical properties of interstellar medium and covers multiple key rules: density non-negative constraint, because the density of matter cannot be negative, the density value of all spatial positions in the predicted intermediate result must be greater than or equal to 0; temperature range constraint, according to the astronomical physical knowledge, the gas temperature is limited to a reasonable interval to avoid extreme temperature values far beyond the actual interstellar environment; chemical abundance normalization constraint, to ensure that the abundance ratio of various chemical species (such as H2, C, CO, etc.) conforms to the stoichiometric law, and the total sum is within a reasonable range; in addition, constraints such as energy conservation and mass conservation can be included to prevent the model output from generating energy out of thin air or violating the mass conservation, which is against the physical nature. When constructing the third loss, a special physical constraint penalty function Φ is designed to quantitatively punish the part of the predicted intermediate result that violates the above constraints. The more serious the constraint, the greater the punishment value, and finally the third loss is formed.

[0086] Subsequently, the final loss result is obtained by weighted fusion based on the first loss, the second loss and the third loss. By configuring adjustable weights α, β, γ (α>0, β>0, γ>0) for the three loss terms, the importance of each constraint dimension can be flexibly balanced according to the actual needs of model training. For example, in the early stage of training, the weight of α can be appropriately increased to let the model prioritize the basic numerical characteristics of the three-dimensional physical field; as the training progresses, the weight of β can be increased to strengthen the observation consistency matching; at the same time, a certain weight of γ is retained to continuously constrain the physical reasonableness, avoiding the model from appearing physical violation in the later stage. The weighted sum of the three forms the total loss result (wherein, is the first loss, is the second loss, is the third loss), which contains not only the numerical difference between the predicted value and the true value, but also the consistency requirement on the observation level and the rule constraint on the physical level, providing a comprehensive and scientific guidance signal for model parameter iteration optimization, ensuring that the trained model can accurately invert the three-dimensional physical field and meet the physical law requirements of the real astronomical scene.

[0087] The above steps realize threefold collaborative constraints on model training by constructing a third loss based on the intermediate results of three-dimensional interstellar medium physical field prediction and preset physical constraints (such as density non-negativity, reasonable temperature interval, chemical abundance normalization, mass conservation, and other rules conforming to the inherent physical properties of interstellar medium), and then weighting and fusing the third loss with the first loss for ensuring numerical fitting accuracy and the second loss for strengthening observation consistency to obtain a final loss result. The threefold collaborative constraints ensure that the prediction result is numerically consistent with the true value three-dimensional physical field through the first loss, enables the model output to reversely correspond to the spectral line and image features of the real observation through the second loss, and avoids unreasonable results that violate basic physical laws through the third loss. The physical knowledge is integrated into the model optimization process in the form of soft constraints, which significantly improves the physical self-consistency, prediction reliability and explainability of the three-dimensional interstellar medium physical field inversion, and enhances the robustness of the model in the scene of sparse or noisy observation data, thereby providing comprehensive and scientific optimization guidance for subsequent accurate and reasonable three-dimensional interstellar medium structure reconstruction.

[0088] In some embodiments, the obtaining of the three-dimensional interstellar medium physical field simulation data comprises:

[0089] obtaining original three-dimensional interstellar medium physical field simulation data;

[0090] resampling the original three-dimensional interstellar medium physical field simulation data to a regular cubic grid to obtain the three-dimensional interstellar medium physical field simulation data.

[0091] First, the original three-dimensional interstellar medium physical field simulation data is obtained, which is derived from the high-resolution three-dimensional radiation-magnetohydrodynamic star formation simulation plan jointly conducted by multiple institutions such as the STARFORGE project, covering complex physical processes such as gravity, magnetic field, stellar feedback (jet, stellar wind, radiation, supernova explosion), and containing key three-dimensional physical information such as gas density field, velocity field, temperature distribution, chemical abundance, and stellar formation and cluster assembly. The original output format of the data may be unstructured grid or adaptive mesh refinement (AMR) format, and the grid resolution is dynamically adjusted according to the intensity of the physical process.

[0092] Subsequently, the original simulation data is resampled to map the irregularly distributed original data to a preset regular cubic grid (such as 128³), while maintaining the conservation of physical quantities and the integrity of the data during the resampling process, eliminating the interference caused by inconsistent grids in the original data, and finally obtaining three-dimensional interstellar medium physical field simulation data with regular structure, unified dimension, and direct input to the subsequent three-dimensional photochemical radiation model.

[0093] Through the above steps, the standardization and regularization processing of the original analog data is realized, the subsequent processing interference caused by the inconsistent grid in the original data is eliminated, the processed data can adapt to the input requirements of the three-dimensional photochemical radiation model, and the structure unified, dimensionally specified and physically complete basic data support is provided for subsequent photochemical calculation, training sample construction and model training, which ensures the smooth progress and data consistency of the entire three-dimensional reconstruction process.

[0094] In some embodiments, based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set, a training sample is constructed, including:

[0095] Based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set, an original training sample is constructed;

[0096] The original training sample is subjected to multi-view cropping and data enhancement, and is stored in a preset format to obtain the training sample.

[0097] First, the synthetic observation data set (containing CO, [C II], [O I] and other multi-band spectral line images or spectral line cubes) is taken as the input feature, and the true value three-dimensional interstellar medium physical field data set (containing density, temperature, chemical abundance and other core physical quantities) is taken as the corresponding label, to construct an original training sample with one-to-one correspondence between input and label, ensuring the physical consistency and mapping relevance of the sample.

[0098] Subsequently, the original training sample is subjected to multi-view cropping processing, and is projected and cropped in different directions along the x, y and z axes to generate sub-samples of multiple scales and perspectives, enriching the perspective diversity of the sample and strengthening the understanding ability of the model to different scale features. At the same time, targeted data enhancement operations are carried out, including geometric transformation (rotation, mirroring, flipping), observation simulation (adding common astronomical observation noises such as photon shot noise, dark noise, readout noise), resolution perturbation, numerical scaling, etc., effectively improving the generalization ability and robustness of the model and reducing the risk of overfitting.

[0099] Finally, the samples after cropping and enhancement processing are uniformly stored in a preset standard format such as HDF5, which can efficiently carry massive complex physical data and retain complete meta information. When storing, the key metadata such as the perspective parameters, band type, simulation time, physical variable field corresponding to the sample are attached, and finally the training sample with regular structure, sufficient diversity, complete physical information and adaptive model training input requirements is obtained.

[0100] The above steps achieve the standardization, diversification and high-quality construction of the training samples, ensure the physical consistency and mapping correlation of the sample "input-label", expand the learning boundary of the model through multi-view expansion and data enhancement, reduce the risk of overfitting, and ensure the adaptability of the sample to the subsequent machine learning model training input through unified format and complete meta information. The above steps provide a solid data support for the stable mapping relationship from two-dimensional / 2.5D observation data to three-dimensional physical field for efficient learning of the model, and significantly improve the efficiency of model training and the reliability of final inversion.

[0101] In some embodiments, after obtaining the three-dimensional interstellar medium physical field reconstruction result, the method further comprises:

[0102] inputting the three-dimensional interstellar medium physical field reconstruction result into a preset radiation transfer module to obtain a synthetic observation result;

[0103] verifying the prediction effect based on the synthetic observation result and the real observation data.

[0104] Specifically, the reconstruction result containing three-dimensional density distribution, temperature distribution, chemical abundance distribution and other core physical quantities is first input into a preset radiation transfer module. The module solves the radiation transfer equation through ray tracing along the preset line-of-sight direction, Monte Carlo method, LVG approximation or escape probability model, fully considers absorption, emission and velocity Doppler effect, and generates synthetic observation results consistent with the dimension and format of the real observation data.

[0105] Subsequently, the synthetic observation result is compared with the real observation data (derived from multi-band spectral lines, two-dimensional projection images and the like) in all directions. The reliability and physical self-consistency of the reconstruction result are verified through line intensity error analysis, spectral line feature comparison, image structure consistency verification and statistical quantity (such as mean, variance, correlation coefficient, etc.) matching degree evaluation. If the verification result meets the preset accuracy standard, the three-dimensional reconstruction effect is confirmed to be effective. If there is a deviation, it can be fed back to the model training link for parameter adjustment, forming a closed loop of "reconstruction-verification-optimization", and further ensuring the accuracy and reliability of the three-dimensional interstellar medium physical field reconstruction result.

[0106] The above steps realize the closed loop verification mechanism of "reconstruction-verification", which not only objectively and accurately evaluates the reliability and physical self-consistency of the three-dimensional reconstruction result, but also timely discovers the deviation between the reconstruction result and the real observation, provides feedback for model parameter optimization or data preprocessing improvement, effectively avoids the problem of unreliable reconstruction result caused by no verification, further ensures the accuracy and scientificity of the three-dimensional interstellar medium physical field reconstruction, and makes the reconstruction result more consistent with the actual scene and physical law of astronomical observation, providing reliable data support for astrophysical research such as molecular cloud evolution and star formation mechanism.

[0107] In some embodiments, verifying the prediction effect based on the synthetic observation result and the real observation data comprises:

[0108] quantitatively comparing the synthetic observation result and the real observation data to obtain the prediction effect; the quantitative comparison comprises line intensity error, spectral line type comparison, image structure comparison, and statistical quantity consistency.

[0109] The verification of the prediction effect based on the synthetic observation result and the real observation data specifically comprises a full-azimuth quantitative comparison between the synthetic observation result obtained by the radiation transfer module (generated by solving the radiation transfer equation through ray tracing, Monte Carlo method, etc.) from the three-dimensional interstellar medium physical field reconstruction result and the real observation data derived from the multi-band spectral lines and two-dimensional projection images of the telescope, and then comprehensively determining the prediction effect.

[0110] The quantitative comparison covers the verification of multi-dimensional core observation characteristics: line intensity error analysis, which quantifies the matching degree of the radiation intensity by calculating the absolute error and relative error of both on the intensity of key spectral lines (such as CO, [C II], [O I], etc.); spectral line type comparison, which compares the key parameters of spectral line profile shape, peak position, and half-width one by one to verify the accuracy of the reduction of physical processes such as velocity Doppler effect; image structure comparison, which analyzes the consistency of the synthetic observation and the real observation in spatial structure and morphological distribution through edge detection and feature extraction to ensure the reliability of the spatial features of the three-dimensional reconstruction; statistical quantity consistency evaluation, which calculates the statistical parameters such as mean, variance, and correlation coefficient of both to verify the stability and reasonableness of the prediction result from the overall data distribution level; through the comprehensive consideration of multi-dimensional quantitative indicators, the accuracy and physical self-consistency of the three-dimensional interstellar medium physical field reconstruction result are fully and objectively verified.

[0111] The above steps achieve multi-dimensional and objective effect verification of the three-dimensional reconstruction result, which not only accurately evaluates the degree of fit between the reconstruction result and the real observation in numerical characteristics, physical process restoration, and spatial structure presentation, but also fully explores the sources of deviation between the two, provides clear feedback for model parameter optimization and data preprocessing improvement, effectively avoids the problem of unreliable reconstruction results caused by no verification, further strengthens the accuracy, physical self-consistency, and scientific reliability of the three-dimensional interstellar medium physical field reconstruction, and provides reliable data support for molecular cloud evolution and star formation mechanism analysis in astrophysics research.

[0112] In some embodiments, the ground truth three-dimensional interstellar medium physical field data set at least includes a three-dimensional density distribution.

[0113] The true value three-dimensional interstellar medium physical field data set at least includes a three-dimensional density distribution, which is a core physical quantity. The data is based on the original density field data generated by high-resolution three-dimensional radiation-magnetohydrodynamic star formation simulation such as STARFORGE, and is formed after further calculation and supplement by a 3D-PDR photochemical radiation model. The data can accurately reflect the material distribution characteristics of interstellar medium in three-dimensional space. In addition, the data set can further include far ultraviolet radiation field, three-dimensional temperature distribution, chemical abundance distribution and other multi-dimensional physical quantities according to actual application requirements, forming a complete multi-physical quantity set. This design takes three-dimensional density distribution as the basic essential physical quantity, ensures the core supporting role of the true value data set, and provides the most critical inversion target benchmark for the machine learning model. The extensible multi-physical quantity enriches the label dimension of model training, so that the model can learn more comprehensive interstellar medium physical property correlations. The data set has both the physical reality of the original simulation and the supplement completeness of the photochemical calculation, providing high-precision, physically self-consistent label data for training sample construction, ensuring the accuracy of model inversion from the source, and laying a solid foundation for subsequent collaborative inversion of multi-physical quantity three-dimensional field, so that the final three-dimensional interstellar medium physical field reconstruction result is more consistent with the complex physical state of the real astronomical scene.

[0114] Figure 3 is a three-dimensional interstellar medium physical field reconstruction system framework schematic diagram according to the embodiments of the application. The diagram shows the complete technical process: first, input the three-dimensional interstellar medium physical field simulation data into the three-dimensional photochemical radiation model to generate the true value three-dimensional interstellar medium physical field data set and the synthetic observation data set, which are combined to construct the training sample; then the training sample is used to train the three-dimensional interstellar medium physical field inversion deep learning architecture. After the architecture training is completed, the real observation data is input into the architecture, and the three-dimensional interstellar medium physical field reconstruction result is finally obtained through the inference link, realizing the whole-link physical field reconstruction from simulation data generation, model training to real observation inversion.

[0115] The training data set constructed by the embodiments of the present application is derived from a real physical process (gas dynamics, radiation feedback, chemical evolution), so the learned model has stronger physical robustness, rather than just fitting by experience. With the help of the machine learning model, the ability to directly output three-dimensional structure from two-dimensional observation is realized, greatly improving the efficiency of three-dimensional reconstruction, which is suitable for large-scale nebula, star cluster or galaxy environment; the forward rendering verification mechanism is introduced, which can automatically evaluate the consistency between the reconstruction result and the real observation, improve the result reliability, and can be used for scientific research and engineering application; the system can be extended to observation data of different wave bands and different instruments, has good applicability and universality, and provides a powerful tool for interstellar structure research, star formation mechanism exploration and astronomical observation interpretation; compared with the traditional method, the embodiments of the present application have three advantages of “physical simulation of reality”, “machine learning efficient inversion” and “real observation verification”, which fills the gap of the prior art in the mapping from observation to three-dimensional structure.

[0116] Among them, the observation data of different wave bands and different instruments means that the method realizes the decoupling of physical modeling and observation representation, so that the same three-dimensional inversion framework can be flexibly adapted to different wave band characteristics and instrument responses without redesigning the model structure for specific observation equipment; specifically, different wave band observations correspond to different physical processes, and the 3D-PDR model itself can output multiple spectral lines and radiation intensities of different wave bands, so that the synthesized observation data naturally has multi-wave band properties, and at the same time, in view of the differences of different instruments in spatial resolution, spectral resolution, observation noise, wave band coverage, etc., different beam convolution, different spectral channel sampling, different noise model, etc. can be added in the synthesis observation generation stage to adapt, so that the same three-dimensional physical field can be “rendered” as different styles of observation data such as ALMA, JWST, single antenna telescope, etc. Finally, the model learns is the invariant physical relationship of “observation to three-dimensional structure”, rather than the exclusive characteristics of a specific instrument, so as to realize the unified input and efficient processing of observation data from different telescopes and different wave bands.

[0117] The embodiments also provide a three-dimensional interstellar medium physical field reconstruction device, which is used to realize the above-mentioned embodiments and preferred embodiments, and will not be described again. As used below, the terms “module”, “unit”, “sub-unit” and the like can be a combination of software and / or hardware that realizes a predetermined function. Although the device described in the following embodiments is preferably realized in software, hardware, or a combination of software and hardware is also possible and contemplated.

[0118] Figure 4 is a structural block diagram of a three-dimensional interstellar medium physical field reconstruction device according to the embodiments of the present application, as Figure 4 shown, the device comprises:

[0119] The three-dimensional data acquisition module 41 is configured to acquire three-dimensional interstellar medium physical field simulation data.

[0120] The data set generation module 42 is configured to input the three-dimensional interstellar medium physical field simulation data into a preset three-dimensional photochemical radiation model to generate a synthetic observation data set and a true value three-dimensional interstellar medium physical field data set; the true value three-dimensional interstellar medium physical field data set is formed based on the three-dimensional interstellar medium physical field simulation data and further calculated and supplemented by the three-dimensional photochemical radiation model.

[0121] The training sample construction module 43 is configured to construct a training sample based on the synthetic observation data set and the true value three-dimensional interstellar medium physical field data set.

[0122] The loss calculation module 44 is configured to input the training sample into an initial three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field prediction intermediate result; and calculate a loss result based on a difference between the three-dimensional interstellar medium physical field prediction intermediate result and the true value three-dimensional interstellar medium physical field data set.

[0123] The model training module 45 is configured to conduct back propagation of the loss result to the initial three-dimensional interstellar medium physical field inversion deep learning architecture for iterative training to obtain a three-dimensional interstellar medium physical field inversion deep learning architecture.

[0124] The model inference module 46 is configured to input the acquired real observation data into the three-dimensional interstellar medium physical field inversion deep learning architecture to obtain a three-dimensional interstellar medium physical field reconstruction result.

[0125] It should be noted that each of the above modules can be a functional module or a program module, and can be implemented by software or hardware. For the modules implemented by hardware, each of the above modules can be located in the same processor; or each of the above modules can also be located in different processors in any combination. The specific examples in the embodiment can refer to the examples described in the above embodiments and optional implementation manners, and will not be described herein.

[0126] In addition, in combination with the three-dimensional interstellar medium physical field reconstruction method in the above embodiment, an embodiment of the present application can provide a storage medium for implementation. The storage medium has a computer program stored thereon; and the computer program is executed by a processor to implement any one of the three-dimensional interstellar medium physical field reconstruction methods in the above embodiments.

[0127] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0128] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0129] Those skilled in the art should understand that any combination of the technical features of the above-mentioned embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above-mentioned embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0130] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for reconstructing the physical field of a three-dimensional interstellar medium, characterized in that, include: Acquire three-dimensional interstellar medium physical field simulation data; The simulated data of the three-dimensional interstellar medium physical field is input into a preset three-dimensional photochemical radiation model to generate a synthetic observation dataset and a true three-dimensional interstellar medium physical field dataset. The true three-dimensional interstellar medium physical field dataset is formed by further calculation and supplementation based on the simulated data of the three-dimensional interstellar medium physical field through the three-dimensional photochemical radiation model. Training samples are constructed based on the synthetic observation dataset and the true three-dimensional interstellar medium physical field dataset; The training samples are input into the initial three-dimensional interstellar medium physics field inversion deep learning architecture to obtain intermediate results of three-dimensional interstellar medium physics field prediction. Based on the difference between the intermediate results of the three-dimensional interstellar medium physics field prediction and the true three-dimensional interstellar medium physics field dataset, the loss result is calculated. The gradient of the loss result is backpropagated to the initial three-dimensional interstellar medium physics field inversion deep learning architecture for iterative training, to obtain the three-dimensional interstellar medium physics field inversion deep learning architecture. The acquired real observation data is input into the three-dimensional interstellar medium physics field inversion deep learning architecture to obtain the three-dimensional interstellar medium physics field reconstruction results.

2. The method for reconstructing the physical field of a three-dimensional interstellar medium according to claim 1, characterized in that, The loss result is calculated based on the difference between the intermediate predicted results of the three-dimensional interstellar medium physics field and the true three-dimensional interstellar medium physics field dataset, including: Based on the difference between the intermediate results of the predicted three-dimensional interstellar medium physical field and the true three-dimensional interstellar medium physical field dataset, a first loss is constructed. The intermediate results of the three-dimensional interstellar medium physical field prediction are input into a preset fast rendering module to obtain the synthetic observation prediction intermediate results; the fast rendering module is the approximate calculation module of the three-dimensional photochemical radiation model. A second loss is constructed based on the difference between the intermediate results of the synthetic observation prediction and the synthetic observation dataset. The loss result is obtained based on the first loss and the second loss.

3. The method for reconstructing the physical field of a three-dimensional interstellar medium according to claim 2, characterized in that, The process of obtaining the loss result based on the first loss and the second loss includes: Based on the intermediate results of the three-dimensional interstellar medium physical field prediction and the preset physical constraints, a third loss is constructed; The loss result is obtained based on the first loss, the second loss, and the third loss.

4. The method for reconstructing the physical field of a three-dimensional interstellar medium according to claim 1, characterized in that, The acquisition of three-dimensional interstellar medium physical field simulation data includes: Obtain raw three-dimensional interstellar medium physical field simulation data; The original three-dimensional interstellar medium physical field simulation data is resampled to a regular cubic mesh to obtain the three-dimensional interstellar medium physical field simulation data.

5. The method for reconstructing the physical field of a three-dimensional interstellar medium according to claim 1, characterized in that, The construction of training samples based on the synthetic observation dataset and the ground truth three-dimensional interstellar medium physical field dataset includes: Based on the synthetic observation dataset and the true three-dimensional interstellar medium physical field dataset, original training samples are constructed; The original training samples are cropped and augmented from multiple perspectives, and then stored in a preset format to obtain the training samples.

6. The method for reconstructing the physical field of a three-dimensional interstellar medium according to claim 1, characterized in that, After obtaining the three-dimensional interstellar medium physical field reconstruction results, the following is also included: The reconstructed physical field of the three-dimensional interstellar medium is input into a preset radiation transfer module to obtain the synthetic observation results; The prediction effect is verified based on the synthetic observation results and the real observation data.

7. The method for reconstructing the physical field of a three-dimensional interstellar medium according to claim 6, characterized in that, The verification of the prediction effect based on the synthetic observation results and the real observation data includes: The synthetic observation results and the real observation data are quantitatively compared to obtain the prediction effect; the quantitative comparison includes line intensity error, spectral line type comparison, image structure comparison and statistical consistency.

8. The method for reconstructing the physical field of a three-dimensional interstellar medium according to any one of claims 1 to 7, characterized in that, The true three-dimensional interstellar medium physics dataset includes at least a three-dimensional density distribution.

9. A three-dimensional interstellar medium physical field reconstruction device, characterized in that, The device includes: The three-dimensional data acquisition module is used to acquire three-dimensional interstellar medium physical field simulation data; The dataset generation module is used to input the three-dimensional interstellar medium physical field simulation data into a preset three-dimensional photochemical radiation model to generate a synthetic observation dataset and a true three-dimensional interstellar medium physical field dataset; the true three-dimensional interstellar medium physical field dataset is formed based on the three-dimensional interstellar medium physical field simulation data and further calculated and supplemented by the three-dimensional photochemical radiation model. The training sample construction module is used to construct training samples based on the synthetic observation dataset and the true three-dimensional interstellar medium physical field dataset; The loss calculation module is used to input the training samples into the initial three-dimensional interstellar medium physics field inversion deep learning architecture to obtain intermediate prediction results of the three-dimensional interstellar medium physics field; and to calculate the loss result based on the difference between the intermediate prediction results of the three-dimensional interstellar medium physics field and the ground truth three-dimensional interstellar medium physics field dataset. The model training module is used to backpropagate the gradient of the loss result to the initial three-dimensional interstellar medium physics field inversion deep learning architecture for iterative training, so as to obtain the three-dimensional interstellar medium physics field inversion deep learning architecture. The model inference module is used to input the acquired real observation data into the three-dimensional interstellar medium physical field inversion deep learning architecture to obtain the three-dimensional interstellar medium physical field reconstruction result.

10. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the three-dimensional interstellar medium physics field reconstruction method according to any one of claims 1 to 8 when it is run.

Images