Three-dimensional radiation field reconstruction method and device based on deep neural network

Abstract

The invention relates to a three-dimensional radiation field reconstruction method and device based on a deep neural network. Obtaining a data set of three-dimensional radiation field reconstruction according to the obtained feature signal and the target signal; determining neural network hyper-parameters of three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and shielding arrangement characteristics and geometric complexity in the radiation space; training a deep neural network according to the data set and the neural network hyper-parameters to obtain a three-dimensional radiation field reconstruction neural network model for three-dimensional radiation field reconstruction; performing a radiation field reconstruction test of three-dimensional radiation field reconstruction on the three-dimensional radiation field reconstruction neural network model by using a test sample; and after the radiation field reconstruction test is passed, reconstructing the three-dimensional radiation field according to the three-dimensional radiation field reconstruction neural network model, and reconstructing the three-dimensional radiation field reconstructed by the three-dimensional radiation field. According to the invention, the reconstruction of the three-dimensional radiation field can be rapidly carried out through the characteristic signal and the target signal; the three-dimensional radiation field reconstruction neural network model has stability and generalization ability.

Description

Technical Field

[0001] The present invention relates to the field of radiation protection technology, and more specifically, to a three-dimensional radiation field reconstruction method and device based on a deep neural network. Background Art

[0002] If the three-dimensional radiation field in the entire space can be calculated by computer simulation before nuclear workers enter the radiation space to work, and the radiation dose information received by the workers during the simulation process can be given at the same time, the radiation dose of the personnel can be quantitatively given before work, which is conducive to reducing the radiation damage to the workers during the actual operation. The focus of the above process is to establish a three-dimensional radiation field in the personnel working area. Therefore, conducting research on the reconstruction method of the dose distribution of the three-dimensional radiation field can provide a more accurate and practical technical method for the optimization of radiation protection. On the one hand, it can significantly improve the safety level of the nuclear device site, and on the other hand, it can more comprehensively protect the health of nuclear workers. Therefore, the research on the reconstruction method of the radiation field has broad application prospects and urgent needs throughout the life cycle of the operation, inspection, and decommissioning of nuclear equipment. Summary of the invention

[0003] The embodiments of the present application provide a method and device for reconstructing a three-dimensional radiation field based on a deep neural network. In order to have a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary is not a general review, nor is it intended to identify key / important components or describe the scope of protection of these embodiments. Its only purpose is to present some concepts in a simple form as a preface to the detailed description that follows.

[0004] In a first aspect, the present application provides a three-dimensional radiation field reconstruction method based on a deep neural network, the method comprising:

[0005] Obtain characteristic signals and target signals for three-dimensional radiation field reconstruction;

[0006] Obtaining a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal;

[0007] Determining the neural network hyperparameters for the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space;

[0008] Training a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction;

[0009] Using the test sample to perform a radiation field reconstruction test on the three-dimensional radiation field reconstruction neural network model;

[0010] After the radiation field reconstruction test is passed, the three-dimensional radiation field reconstructed by the three-dimensional radiation field is reconstructed according to the three-dimensional radiation field reconstruction neural network model.

[0011] According to a preferred embodiment, the step of acquiring the characteristic signal and the target signal for three-dimensional radiation field reconstruction includes:

[0012] Obtain the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction;

[0013] The particle flux measurement value at the sampling point is used as the characteristic signal of the three-dimensional radiation field reconstruction, and the radiation field parameter distribution is used as the target signal of the three-dimensional radiation field reconstruction.

[0014] According to a preferred embodiment, the deep neural network includes: an interpolation part and a convolutional neural network part; the convolutional neural network part is composed of a U-Net network.

[0015] According to a preferred embodiment, before the deep neural network is trained according to the data set and the neural network hyperparameters to obtain the three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction, the method includes:

[0016] The data set is preprocessed; the preprocessing operation includes: data magnitude transformation operation and standardization operation of different features.

[0017] According to a preferred embodiment, the deep neural network is trained according to the data set and the neural network hyperparameters to obtain the three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction, including:

[0018] Randomly shut down a preset proportion of neurons during training.

[0019] In a second aspect, the present application provides a three-dimensional radiation field reconstruction device based on a deep neural network, the device comprising:

[0020] A signal acquisition module, used to acquire characteristic signals and target signals for three-dimensional radiation field reconstruction;

[0021] A data determination module, used to obtain a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal;

[0022] A parameter determination module, used to determine the neural network hyperparameters of the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space;

[0023] A model training module, used for training a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction;

[0024] A testing module, used for performing a radiation field reconstruction test of the three-dimensional radiation field reconstruction on the three-dimensional radiation field reconstruction neural network model using a test sample;

[0025] A reconstruction module is used to reconstruct the three-dimensional radiation field reconstructed by the three-dimensional radiation field according to the three-dimensional radiation field reconstruction neural network model after the radiation field reconstruction test passes.

[0026] According to a preferred embodiment, the signal acquisition module is specifically used for:

[0027] Obtain the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction;

[0028] The particle flux measurement value at the sampling point is used as the characteristic signal of the three-dimensional radiation field reconstruction, and the radiation field parameter distribution is used as the target signal of the three-dimensional radiation field reconstruction.

[0029] According to a preferred embodiment, the deep neural network includes: an interpolation part and a convolutional neural network part; the convolutional neural network part is composed of a U-Net network.

[0030] According to a preferred embodiment, before the model training module, it includes:

[0031] The preprocessing module is used to perform preprocessing operations on the data set; the preprocessing operations include: data magnitude transformation operations and standardization operations of different features.

[0032] According to a preferred embodiment, the model training module includes:

[0033] Randomly shut down a preset proportion of neurons during training.

[0034] In a third aspect, the present application provides a computer storage medium, which stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

[0035] In a fourth aspect, the present application provides a terminal, which may include: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0036] The technical solution provided by this application may have the following beneficial effects:

[0037] In the present application, the three-dimensional radiation field reconstruction method based on deep neural network obtains characteristic signals and target signals for three-dimensional radiation field reconstruction; obtains a data set for three-dimensional radiation field reconstruction according to the characteristic signals and target signals; determines the neural network hyperparameters for three-dimensional radiation field reconstruction according to the characteristic signals, target signals, and shielding arrangement characteristics and geometric complexity in the radiation space; trains the deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for three-dimensional radiation field reconstruction; performs a radiation field reconstruction test on the three-dimensional radiation field reconstruction neural network model using a test sample; after the radiation field reconstruction test passes, reconstructs the three-dimensional radiation field for three-dimensional radiation field reconstruction according to the three-dimensional radiation field reconstruction neural network model. The present application can quickly reconstruct the three-dimensional radiation field through characteristic signals and target signals; the three-dimensional radiation field reconstruction neural network model of the present application has stability and generalization ability.

[0038] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0040] Figure 1 It is a flowchart of a three-dimensional radiation field reconstruction method based on a deep neural network provided in an embodiment of the present application;

[0041] Figure 2 It is a schematic diagram of a neural network model for reconstructing a three-dimensional radiation field according to a three-dimensional radiation field reconstruction method based on a deep neural network provided in an embodiment of the present application;

[0042] Figure 3 It is a structural schematic diagram of a three-dimensional radiation field reconstruction device based on a deep neural network provided in an embodiment of the present application;

[0043] Figure 4 It is a structural diagram of a terminal provided in an embodiment of the present application.

[0044] Reference numerals:

[0045] 10000, signal acquisition module, 20000, data determination module, 30000, parameter determination module, 40000, model training module, 50000, testing module, 60000, reconstruction module;

[0046] 1001. Processor, 1002. Communication bus, 1003. User interface, 1004. Network interface, 1005. Memory. DETAILED DESCRIPTION

[0047] The following description and the drawings sufficiently illustrate specific embodiments of the invention to enable those skilled in the art to practice them.

[0048] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are only examples of systems and methods consistent with some aspects of the present invention as detailed in the attached claims.

[0050] In the description of the present invention, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In addition, in the description of the present invention, unless otherwise specified, "plurality" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previously associated objects are in an "or" relationship.

[0051] The following will be combined with the attached Figure 1 and 2 , a three-dimensional radiation field reconstruction method based on deep neural network provided in an embodiment of the present application is introduced in detail.

[0052] See also Figure 1 and 2 The method of the embodiment of the present application may include the following steps:

[0053] The geometric information, source information and material information of the computational space are modeled, and the Monte Carlo method is used to calculate the radiation field parameters in the geometric space. Although this process can accurately calculate the radiation field parameters in the radiation field, it requires the user to accurately describe the source information in the radiation field. In actual application scenarios, the activity information of the source term is often unknown, so this method has great limitations in practical applications. The direct reconstruction method of the radiation field mainly analyzes and reconstructs the overall radiation field from a limited number of sampling points through calculation methods such as interpolation algorithms and artificial intelligence.

[0054] Therefore, there is a great demand in practical engineering to study the algorithm for reconstructing the three-dimensional radiation field from individual sampling points. In response to these problems, this application proposes to establish a geometric model for common nuclear facility scenarios for testing. The results show that the three-dimensional radiation field reconstruction method proposed in this application and the corresponding deep neural network structure can effectively predict the radiation parameter distribution information in the entire space and effectively solve practical problems. This application can quickly reconstruct the three-dimensional radiation field through individual sampling points to obtain the radiation field distribution in the entire space.

[0055] The method described in this application is an algorithm that uses a neural network to reconstruct a three-dimensional radiation field based on finite sampling values, as follows:

[0056] S100, acquiring characteristic signals and target signals for three-dimensional radiation field reconstruction, including:

[0057] Obtaining the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction; using the particle flux measurement values ​​as characteristic signals for the three-dimensional radiation field reconstruction and the radiation field parameter distribution as target signals for the three-dimensional radiation field reconstruction.

[0058] In actual operation, it is necessary to select the appropriate number and position of detectors according to the actual situation, that is, according to the geometric distribution in the calculated radiation space, and determine the number and position of these detectors as the number and position of sampling points of the radiation field in the space. The sampling point information of each sampling point, that is, the one-dimensional array composed of the particle flux measurement values ​​of the sampling point, is used as the characteristic variable of the deep neural network input proposed later in this application, that is, the characteristic signal.

[0059] In actual operation, the resolution of the radiation field parameters of the three-dimensional radiation field in the required calculation space is determined according to actual needs, and the distribution of the radiation field parameters at the resolution is obtained. The radiation field parameter distribution of the three-dimensional radiation field is essentially the point cloud distribution of the radiation field parameters in the calculation space, which serves as the output part of the deep neural network proposed later in this application. The target signal output by the deep neural network is the distribution of the radiation field parameters at each sampling point position in the entire three-dimensional radiation field space.

[0060] The embodiment of the present application randomly generates a series of source term intensity distributions. After constructing a series of different source term intensity parameters, the radiation field parameter distribution corresponding to different source term intensity parameters in the whole space is simulated by the Monte Carlo method according to the source term intensity distribution and shielding layout characteristics in the space, and the radiation field flux of a limited number of sampling points, i.e., the particle flux measurement value of the sampling point, is extracted from the whole radiation field parameter distribution. The particle flux measurement value of the sampling point is used as the measurement point data input to the deep neural network, and the measurement point data is the characteristic signal. The radiation field parameter distribution under different source term intensity parameters is used as the output of the deep neural network.

[0061] S200, obtaining a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal; a deep neural network requires a certain amount of data sets for training, testing and verification.

[0062] S300, according to the characteristic signal, the target signal, and the shielding layout characteristics and geometric complexity in the calculated radiation space, determine the neural network hyperparameters for the reconstruction of the three-dimensional radiation field; the neural network hyperparameters include the number of U-Net networks in the deep neural network, the number of layers, the size and number of filters, the activation function, the learning rate, the weight decay coefficient, the batch size, the network structure, the mean square error, the grid search, and the model quantization, pruning or the use of lightweight network architecture and other settings. In view of the geometric complexity and shielding layout characteristics of the radiation space, the U-Net network architecture can effectively capture and process the spatial relationship of three-dimensional data by virtue of its encoder-decoder structure and skip connection. On this basis, the initial network layer number is set and the layer number is adjusted according to the training effect, the filter size is 3x3, and the number is doubled by 32 step by step, and Leaky ReLU is selected as the activation function according to the radiation field characteristics, i.e., the shielding layout characteristics. The Adam optimizer is introduced, and the cosine annealing learning rate decay strategy is adopted. That is, in the subsequent deep neural network training process, the learning rate gradually decreases according to the law of the cosine function, from the initial learning rate to the minimum learning rate of 1e-6, and the cycle is set to 10 epochs. In addition, the dropout technology and L2 regularization technology are applied, and the weight decay coefficient is 1e-4 to prevent overfitting and improve the generalization ability of the following three-dimensional radiation field reconstruction neural network model. Under the condition of limited computing resources, the batch size and network structure are optimized to ensure the computational efficiency while maintaining the performance of the following three-dimensional radiation field reconstruction neural network model. The mean square error is used as a performance indicator to evaluate the following three-dimensional radiation field reconstruction neural network model to measure the prediction accuracy of the following three-dimensional radiation field reconstruction neural network model. Through systematic parameter adjustment methods such as grid search, the generalization ability of the following three-dimensional radiation field reconstruction neural network model is verified, and the selection of neural network hyperparameters and their corresponding performance are recorded in detail to provide a reference for the subsequent improvement of the three-dimensional radiation field reconstruction neural network model. Finally, the real-time application requirements of the following three-dimensional radiation field reconstruction neural network model are considered to ensure its robustness and efficiency under different input variations. The model can be further quantized, pruned or a lightweight network architecture can be used to reduce the complexity of the following three-dimensional radiation field reconstruction neural network model, improve the reasoning speed, and meet the needs of real-time applications. The different input variations refer to the situation where different characteristic signals are input and different target signals are output.

[0063] Before the S400, it includes:

[0064] The data set is preprocessed; the preprocessing operation includes: data magnitude transformation operation and standardization operation of different features.

[0065] The data set needs to be preprocessed for standardization. For the distribution of radiation field parameters and the particle flux measurements at the sampling points in the data set, this application uses the Z-Score standardization method to transform the different feature ranges in the data set to a distribution with a mean of 0 and a variance of 1, which can significantly accelerate the convergence speed of the deep neural network and the accuracy of the prediction.

[0066] The embodiment of the present application preprocesses the data set before training through S400, which can transform the radiation field parameter distribution and the particle flux measurement value of the sampling point from several orders of magnitude different to the same order of magnitude, and standardize the different features in the data set, so that the deep neural network of the present application can better learn the fitting relationship between the two, which can significantly accelerate the convergence speed of the deep neural network and the accuracy of prediction.

[0067] S400, training a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction, including:

[0068] During the training process, a preset proportion of neurons is randomly turned off to avoid overfitting of the three-dimensional radiation field reconstruction neural network model.

[0069] In the embodiment of the present application, the deep neural network is trained using the preprocessed data set and the neural network hyperparameters; Figure 2 As shown, the deep neural network includes: an interpolation part and a convolutional neural network part; wherein the interpolation part is mainly based on the radiation parameter measurement value of each sampling point in the space, that is, the particle flux measurement value of the sampling point, and the radiation parameter value of all sampling points in the space is obtained by interpolation calculation, thereby determining the radiation field parameter distribution. The convolutional neural network part is composed of a U-Net network, which is used to reconstruct the shielding layout features in the space and has good flexibility and generalization. The U-Net network consists of two parts: a contraction path and an expansion path. A series of convolution and pooling operations are used in the contraction path to gradually reduce the spatial size and depth of the shielding layout feature map; the expansion path uses a series of deconvolution and splicing operations to gradually increase the spatial size and depth of the shielding layout feature map.

[0070] In this application, between the convolution layer and the deconvolution layer of the deep neural network, there is a batch normalization layer after each convolution layer to normalize the features and improve the stability and generalization ability of the model. Finally, the shielding arrangement feature map is passed through the activation function to output the inversion model, that is, the three-dimensional radiation field reconstruction neural network model.

[0071] S500, performing a radiation field reconstruction test of the three-dimensional radiation field reconstruction on the three-dimensional radiation field reconstruction neural network model using a test sample.

[0072] S600, after the radiation field reconstruction test is passed, reconstruct the three-dimensional radiation field reconstructed by the three-dimensional radiation field according to the three-dimensional radiation field reconstruction neural network model, thereby realizing the prediction of the three-dimensional radiation field.

[0073] In summary, in the radiation protection scenario, the particle flux measurement values ​​and radiation field parameter distribution of each sampling point of the present application are the result of the combined influence of the source term intensity distribution and the shielding layout characteristics in the radiation field on the sampling point, and have strong nonlinear characteristics. The three-dimensional radiation field reconstruction neural network model of the present application is established based on a deep neural network and trained using a data set composed of different source term examples. It can quickly reconstruct the three-dimensional radiation field directly based on the particle flux measurement values ​​and radiation field parameter distribution of the sampling point.

[0074] In the present application, the three-dimensional radiation field reconstruction method based on deep neural network obtains characteristic signals and target signals for three-dimensional radiation field reconstruction; obtains a data set for three-dimensional radiation field reconstruction according to the characteristic signals and target signals; determines the neural network hyperparameters for three-dimensional radiation field reconstruction according to the characteristic signals, target signals, and shielding arrangement characteristics and geometric complexity in the radiation space; trains the deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for three-dimensional radiation field reconstruction; performs a radiation field reconstruction test on the three-dimensional radiation field reconstruction neural network model using a test sample; after the radiation field reconstruction test passes, reconstructs the three-dimensional radiation field for three-dimensional radiation field reconstruction according to the three-dimensional radiation field reconstruction neural network model. The present application can quickly reconstruct the three-dimensional radiation field through characteristic signals and target signals; the three-dimensional radiation field reconstruction neural network model of the present application has stability and generalization ability.

[0075] The following are embodiments of the device of the present invention, which can be used to implement the method embodiments of the present invention. For details not disclosed in the device embodiments of the present invention, please refer to the method embodiments of the present invention.

[0076] See also Figure 3 , which shows a schematic diagram of the structure of a three-dimensional radiation field reconstruction device based on a deep neural network provided by an exemplary embodiment of the present invention. The device includes: a signal acquisition module 10000, a data determination module 20000, a parameter determination module 30000, a model training module 40000, a test module 50000 and a reconstruction module 60000.

[0077] The signal acquisition module 10000 is used to acquire characteristic signals and target signals for three-dimensional radiation field reconstruction;

[0078] A data determination module 20000 is used to obtain a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal;

[0079] A parameter determination module 30000, for determining the neural network hyperparameters for the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space;

[0080] A model training module 40000 is used to train a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction;

[0081] A testing module 50000 is used to perform a radiation field reconstruction test of the three-dimensional radiation field reconstruction on the three-dimensional radiation field reconstruction neural network model using a test sample;

[0082] The reconstruction module 60000 is used to reconstruct the three-dimensional radiation field reconstructed by the three-dimensional radiation field according to the three-dimensional radiation field reconstruction neural network model after the radiation field reconstruction test passes.

[0083] According to a preferred implementation, the signal acquisition module 10000 is specifically used for:

[0084] Obtain the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction;

[0085] The particle flux measurement value at the sampling point is used as the characteristic signal of the three-dimensional radiation field reconstruction, and the radiation field parameter distribution is used as the target signal of the three-dimensional radiation field reconstruction.

[0086] According to a preferred embodiment, the deep neural network includes: an interpolation part and a convolutional neural network part; the convolutional neural network part is composed of a U-Net network.

[0087] According to a preferred embodiment, before the model training module 40000, it includes:

[0088] The preprocessing module is used to perform preprocessing operations on the data set; the preprocessing operations include: data magnitude transformation operations and standardization operations of different features.

[0089] According to a preferred embodiment, the model training module 40000 includes:

[0090] Randomly shut down a preset proportion of neurons during training.

[0091] It should be noted that the three-dimensional radiation field reconstruction device based on a deep neural network provided in the above embodiment only uses the division of the above-mentioned functional modules as an example when executing the three-dimensional radiation field reconstruction method based on a deep neural network. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the three-dimensional radiation field reconstruction device based on a deep neural network provided in the above embodiment and the three-dimensional radiation field reconstruction method based on a deep neural network embodiment belong to the same concept. The implementation process is detailed in the method embodiment, which will not be repeated here.

[0092] In the present application, the three-dimensional radiation field reconstruction device based on the deep neural network obtains the characteristic signal and target signal of the three-dimensional radiation field reconstruction; obtains the data set of the three-dimensional radiation field reconstruction according to the characteristic signal and the target signal; determines the neural network hyperparameters of the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space; trains the deep neural network according to the data set and the neural network hyperparameters to obtain the three-dimensional radiation field reconstruction neural network model of the three-dimensional radiation field reconstruction; performs the radiation field reconstruction test of the three-dimensional radiation field reconstruction on the three-dimensional radiation field reconstruction neural network model using the test sample; after the radiation field reconstruction test passes, reconstructs the three-dimensional radiation field of the three-dimensional radiation field reconstruction according to the three-dimensional radiation field reconstruction neural network model. The present application can quickly reconstruct the three-dimensional radiation field through the characteristic signal and the target signal; the three-dimensional radiation field reconstruction neural network model of the present application has stability and generalization ability.

[0093] The present invention also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implements the three-dimensional radiation field reconstruction method based on a deep neural network provided in the above-mentioned method embodiments.

[0094] The present invention also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the three-dimensional radiation field reconstruction method based on a deep neural network of each of the above-mentioned method embodiments.

[0095] See also Figure 4 , which is a schematic diagram of the structure of a terminal according to an embodiment of the present application. The terminal may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.

[0096] The communication bus 1002 is used to realize the connection and communication between these components.

[0097] The user interface 1003 may include a display screen (Display) and a camera (Camera), and the optional user interface 1003 may also include a standard wired interface and a wireless interface.

[0098] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0099] Among them, the processor 1001 may include one or more processing cores. The processor 1001 uses various interfaces and lines to connect various parts in the entire terminal, and executes various functions of the terminal and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 1005, and calling data stored in the memory 1005. Optionally, the processor 1001 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 1001 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 1001, and it can be implemented separately through a chip.

[0100] Among them, the memory 1005 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 1005 may optionally be at least one storage device located away from the aforementioned processor 1001. As Figure 4 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a three-dimensional radiation field reconstruction application based on a deep neural network.

[0101] exist Figure 4 In the terminal shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 1001 can be used to call the three-dimensional radiation field reconstruction application based on the deep neural network stored in the memory 1005, and specifically perform the following operations:

[0102] Obtain characteristic signals and target signals for three-dimensional radiation field reconstruction;

[0103] Obtaining a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal;

[0104] Determining the neural network hyperparameters for the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space;

[0105] The deep neural network is trained according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction; the deep neural network includes: an interpolation part and a convolutional neural network part; the convolutional neural network part is composed of a U-Net network;

[0106] Using the test sample to perform a radiation field reconstruction test on the three-dimensional radiation field reconstruction neural network model;

[0107] After the radiation field reconstruction test is passed, the three-dimensional radiation field reconstructed by the three-dimensional radiation field is reconstructed according to the three-dimensional radiation field reconstruction neural network model.

[0108] In one embodiment, when the processor 1001 executes the method of acquiring the characteristic signal and the target signal for three-dimensional radiation field reconstruction, the processor 1001 specifically performs the following operations:

[0109] Obtain the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction;

[0110] The particle flux measurement value at the sampling point is used as the characteristic signal of the three-dimensional radiation field reconstruction, and the radiation field parameter distribution is used as the target signal of the three-dimensional radiation field reconstruction.

[0111] In one embodiment, before executing the step of training the deep neural network according to the data set and the neural network hyperparameters to obtain the three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction, the processor 1001 specifically performs the following operations:

[0112] The data set is preprocessed; the preprocessing operation includes: data magnitude transformation operation and standardization operation of different features.

[0113] In one embodiment, when the processor 1001 performs the training of the deep neural network according to the data set and the neural network hyperparameters to obtain the three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction, the processor 1001 specifically performs the following operations:

[0114] Randomly shut down a preset proportion of neurons during training.

[0115] In the present application, the method and device for three-dimensional radiation field reconstruction based on deep neural network obtains characteristic signals and target signals for three-dimensional radiation field reconstruction; obtains a data set for three-dimensional radiation field reconstruction according to the characteristic signals and target signals; determines the neural network hyperparameters for three-dimensional radiation field reconstruction according to the characteristic signals, target signals, and shielding arrangement characteristics and geometric complexity in the radiation space; trains the deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for three-dimensional radiation field reconstruction; performs a radiation field reconstruction test on the three-dimensional radiation field reconstruction neural network model using a test sample; after the radiation field reconstruction test passes, reconstructs the three-dimensional radiation field for three-dimensional radiation field reconstruction according to the three-dimensional radiation field reconstruction neural network model. The present application can quickly reconstruct the three-dimensional radiation field through the characteristic signals and target signals; the three-dimensional radiation field reconstruction neural network model of the present application has stability and generalization ability.

[0116] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0117] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A three-dimensional radiation field reconstruction method based on deep neural network, characterized in that: The following steps are involved: Obtain characteristic signals and target signals for three-dimensional radiation field reconstruction; Obtaining a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal; Determining the neural network hyperparameters for the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space; Training a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction; Using the test sample to perform a radiation field reconstruction test on the three-dimensional radiation field reconstruction neural network model; After the radiation field reconstruction test is passed, the three-dimensional radiation field reconstructed by the three-dimensional radiation field is reconstructed according to the three-dimensional radiation field reconstruction neural network model.

2. The three-dimensional radiation field reconstruction method based on deep neural network according to claim 1 is characterized in that: The step of obtaining the characteristic signal and the target signal for three-dimensional radiation field reconstruction includes: Obtain the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction; The particle flux measurement value at the sampling point is used as the characteristic signal of the three-dimensional radiation field reconstruction, and the radiation field parameter distribution is used as the target signal of the three-dimensional radiation field reconstruction.

3. The three-dimensional radiation field reconstruction method based on deep neural network according to claim 1 is characterized in that: The deep neural network includes: an interpolation part and a convolutional neural network part; the convolutional neural network part is composed of a U-Net network.

4. The three-dimensional radiation field reconstruction method based on deep neural network according to claim 1 is characterized in that: Before the deep neural network is trained according to the data set and the neural network hyperparameters to obtain the three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction, the method includes: The data set is preprocessed; the preprocessing operation includes: data magnitude transformation operation and standardization operation of different features.

5. The three-dimensional radiation field reconstruction method based on deep neural network according to claim 1 is characterized in that: The step of training a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction includes: Randomly shut down a preset proportion of neurons during training.

6. A three-dimensional radiation field reconstruction device based on a deep neural network, the device is used to execute the three-dimensional radiation field reconstruction method based on a deep neural network according to any one of claims 1 to 5, characterized in that: include: A signal acquisition module, used to acquire characteristic signals and target signals for three-dimensional radiation field reconstruction; A data determination module, used to obtain a data set for reconstructing the three-dimensional radiation field according to the characteristic signal and the target signal; A parameter determination module, used to determine the neural network hyperparameters of the three-dimensional radiation field reconstruction according to the characteristic signal, the target signal, and the shielding arrangement characteristics and geometric complexity in the radiation space; A model training module, used for training a deep neural network according to the data set and the neural network hyperparameters to obtain a three-dimensional radiation field reconstruction neural network model for the three-dimensional radiation field reconstruction; A testing module, used for performing a radiation field reconstruction test of the three-dimensional radiation field reconstruction on the three-dimensional radiation field reconstruction neural network model using a test sample; A reconstruction module is used to reconstruct the three-dimensional radiation field reconstructed by the three-dimensional radiation field according to the three-dimensional radiation field reconstruction neural network model after the radiation field reconstruction test passes.

7. The three-dimensional radiation field reconstruction device based on deep neural network according to claim 6, characterized in that: The signal acquisition module is specifically used for: Obtain the particle flux measurement values ​​and radiation field parameter distribution of the sampling points for three-dimensional radiation field reconstruction; The particle flux measurement value at the sampling point is used as the characteristic signal of the three-dimensional radiation field reconstruction, and the radiation field parameter distribution is used as the target signal of the three-dimensional radiation field reconstruction.

8. The three-dimensional radiation field reconstruction device based on deep neural network according to claim 6, characterized in that: The deep neural network includes: an interpolation part and a convolutional neural network part; the convolutional neural network part is composed of a U-Net network.

9. The three-dimensional radiation field reconstruction device based on deep neural network according to claim 6, characterized in that: Prior to the model training module, including: The preprocessing module is used to perform preprocessing operations on the data set; the preprocessing operations include: data magnitude transformation operations and standardization operations of different features.

10. The three-dimensional radiation field reconstruction device based on deep neural network according to claim 6, characterized in that: The model training module includes: Randomly shut down a preset proportion of neurons during training.

Images