Method and system for fast inversion of nuclear radiation field

Abstract

The application provides a kind of nuclear radiation field fast inversion method and system, it is related to nuclear radiation field inversion technical field, and the three-dimensional space model of target nuclear radiation field is constructed and the energy spectrum data of multiple observation positions are obtained in the present application, standard count rate data set is generated after pre-processing;Then by the discrete processing of radiation field, the initial radiation intensity is obtained based on the data between each potential point source and observation position;Further, for any target point, the relative contribution weight of each potential point source and the radiation attenuation factor are aggregated, the field intensity response factor of target point under multiple energy channels is calculated and the field intensity response vector is formed;Finally, the energy fluence mapping model with field intensity response vector as input and energy fluence as output is constructed and trained, the fast and high-precision inversion of energy fluence at any position of nuclear radiation field is realized, and the efficiency and accuracy of multi-energy channel nuclear radiation field inversion in complex medium environment are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of nuclear radiation field inversion technology, specifically to a rapid nuclear radiation field inversion method and system. Background Technology

[0002] Nuclear radiation field inversion technology has significant application value in radiation monitoring, nuclear accident emergency response, and radiation protection. Traditional inversion methods mostly rely on direct measurement and iterative calculations using physical models, which suffers from high computational complexity, slow inversion speed, and insufficient adaptability to radiation attenuation and energy spectrum response in complex media, making it difficult to achieve rapid and high-precision reconstruction of the radiation field. Especially in scenarios with multiple energy channels, non-homogeneous media, and distributed radiation sources, existing methods often fail to effectively integrate detection data with spatial attenuation characteristics, resulting in large errors and poor real-time performance in the inversion results, thus limiting their application in dynamic monitoring and rapid assessment.

[0003] In the prior art, CN115270602A discloses a method and device for rapid inversion of nuclear radiation fields based on artificial intelligence, which specifically includes: calculating radiation field distribution data using a Monte Carlo particle transport program or a deterministic particle transport program; constructing a training set and a test set using the radiation field data, training a neural network, and testing it on the test set; using a radiation measuring instrument for on-site detection, and further training the neural network model after processing the on-site detection data, and using the optimized neural network model to rapidly invert the radiation distribution of the entire region.

[0004] The main problems with the above methods are: they require Monte Carlo or deterministic particle transport procedures to pre-calculate radiation field distribution data to build a training set; once there is a deviation between the actual scenario and the simulation assumptions, the training data will not match the actual measurements, affecting the generalization ability of the inversion model; the neural network in the scheme is mainly trained based on simulated data, and although field detection data is introduced for further optimization, the physical laws of radiation propagation are not explicitly embedded in the model structure, resulting in poor physical consistency of the inversion results in areas with scarce data or complex conditions; and the method for effectively processing energy spectrum data of multiple energy channels is not clearly explained, making it difficult to explain the coupling relationship and attenuation differences between different energy channels, thus limiting its application in energy spectrum-resolved radiation field inversion.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for rapid inversion of nuclear radiation fields to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A rapid inversion method for nuclear radiation fields, comprising the following steps:

[0009] Step 1: Construct a three-dimensional spatial model of the target nuclear radiation field, determine multiple radiation observation locations in the three-dimensional spatial model, synchronously acquire energy spectrum data at each radiation observation location, and preprocess the energy spectrum data to generate a standard count rate dataset.

[0010] Step 2: Discretize the target nuclear radiation field based on the three-dimensional spatial model, define each discretized spatial node as a potential point source, calculate the theoretical count rate contribution of each potential point source to each radiation observation location under each energy channel, and then determine the initial radiation intensity of each potential point source under each energy channel.

[0011] Step 3: For any target point in the target nuclear radiation field, analyze the relative contribution relationship and radiation propagation attenuation relationship of each potential point source relative to the target point. Perform aggregation calculation on the above two relationships to obtain the field strength response factor of the target point in different energy channels. Perform unified characterization on the field strength response factors of different energy channels to generate the field strength response vector corresponding to the target point.

[0012] Step 4: Construct an energy flux mapping model with the field strength response vector as input and the energy flux as output. Select the field strength response vector of a known target point and its corresponding energy flux to train the energy flux mapping model, so that the input and output of the energy flux mapping model form a mapping relationship. Input the field strength response vector of any target point into the trained energy flux mapping model to obtain the energy flux of the target point, and complete the inversion of the target nuclear radiation field.

[0013] Furthermore, the principle underlying the generation of the standard count rate dataset is as follows:

[0014] The energy spectrum data represents the cumulative count of radiation photons collected at the radiation observation location in each energy channel within a set observation time.

[0015] For each radiation observation location, the difference between the cumulative count at each energy channel and the corresponding background cumulative count is calculated. Then, the ratio of this difference to the observation time duration is calculated to determine the count rate at each energy channel. The count rates at each radiation observation location under each energy channel are summarized to construct a standard count rate dataset.

[0016] Specifically, for each radiation observation location, the background cumulative count calculation logic under each energy channel is as follows: during the time interval when the radiation observation location is in a radiation-free environment, the radiation observation location is observed for the same duration as the observation time, and the cumulative count of radiation photons collected at the radiation observation location under each energy channel for the same duration is obtained as the background cumulative count of the radiation observation location under each energy channel.

[0017] Furthermore, the principle for calculating the theoretical count rate contribution of potential point sources to each radiation observation location under each energy channel is as follows:

[0018] The theoretical count rate contribution is the product of the geometric attenuation term and the medium attenuation term; wherein, the calculation logic of the geometric attenuation term is as follows: first calculate the Euclidean distance between the potential point source and the radiation observation location, and obtain the absolute detection efficiency of the potential point source under the energy channel, and calculate the geometric attenuation term based on the absolute detection efficiency and the Euclidean distance;

[0019] The calculation logic for the medium attenuation term is as follows: obtain the straight path from the potential point source to the radiation observation location, then determine the number of media types traversed by the straight path, the path length of the straight path in each medium, and the linear attenuation coefficient of each medium. Based on the linear attenuation coefficient and the path length of the straight path in each medium, calculate the attenuation contribution of each medium to the potential point source. Summate the attenuation contributions of all media to obtain the total attenuation of the radiated photon transmitted between the potential point source and the radiation observation location. Based on the Lambert-Beer law, substitute the negative number of the total attenuation of the medium as the exponent into the exponential function with the natural constant as the base to obtain the medium attenuation term.

[0020] Furthermore, the principle underlying the determination of the initial radiation intensity of each potential point source in each energy channel is as follows:

[0021] Under each energy channel, the count rate at any radiation observation location is affected by all potential point sources. The influence of each potential point source on the radiation observation location is related to the initial radiation intensity of each potential point source and the theoretical count rate contribution of that potential point source to that radiation observation location. The logical relationship among the three is as follows: multiply the initial radiation intensity of each potential point source by the theoretical count rate of each potential point source at that radiation observation location, and sum the results of multiplying the initial radiation intensity of all potential point sources by the theoretical count rate of each potential point source at that radiation observation location to obtain the count rate of that radiation observation location; given the standard count rate dataset and the theoretical count rate contribution of each potential point source to that radiation observation location, the initial radiation intensity of each potential point source under the energy channel can be deduced; perform the above calculation for each energy channel to obtain the initial radiation intensity of each potential point source under each energy channel.

[0022] Furthermore, the relative contribution relationship of each potential point source to the target point is as follows:

[0023] The relative contribution weight is used to characterize the relative contribution relationship of each potential point source to the target point. The initial radiation intensity of each potential point source under each energy channel is obtained. For any energy channel, the initial radiation intensity of each potential point source under that energy channel is divided by the sum of the initial radiation intensities of all potential point sources under the same energy channel to generate the relative contribution weight of each potential point source to the target point under each energy channel.

[0024] Furthermore, the radiation propagation attenuation relationship between each potential point source and the target point is as follows:

[0025] The radiation attenuation factor is used to characterize the radiation propagation attenuation relationship of each potential point source relative to the target point. The types of media traversed by the straight path from each potential point source to the target point, the path length of the straight path in each medium, and the linear attenuation coefficient of each medium are obtained. Based on the linear attenuation coefficient and the path length of the straight path in each medium, the attenuation contribution of each medium to the potential point source is calculated. The attenuation contributions of all media are summed to obtain the total media attenuation of the radiated photon between the potential point source and the target point. Based on the Lambert-Beer law, the negative number of the total media attenuation is substituted into the exponential function with the natural constant as the base to obtain the radiation attenuation factor.

[0026] Furthermore, the principle underlying the generation of the field intensity response vector corresponding to the target point is as follows:

[0027] The relative contribution relationship and the radiation propagation attenuation relationship are aggregated to obtain the field strength response factor of the target point in different energy channels. The specific calculation logic is as follows: for each energy channel, the reciprocals of the radiation propagation attenuation factors of each potential point source to the target point under that energy channel are weighted and summed, and the weight is the relative contribution weight, so as to obtain the field strength response factor of all potential point sources to the target point under that energy channel.

[0028] The field strength response factors of all potential point sources to the target point under each energy channel are calculated and uniformly characterized to form the field strength response vector of the target point under each energy channel.

[0029] Furthermore, the principle underlying the construction of the energy flux mapping model is as follows:

[0030] Several target points are set in the target nuclear radiation field, and the energy flux of each target point is obtained by simulation. A training set and a validation set are generated in an 8:2 ratio. The field strength response vector of the target point in the training set is used as input and its energy flux is used as output to train the energy flux mapping model. The mean square error loss function is used as the loss function. The model is validated through the validation set until the model is confirmed to have converged. The model at this time is output as the energy flux mapping model.

[0031] The present invention also provides a rapid nuclear radiation field inversion system, the system being used to implement the above-mentioned rapid nuclear radiation field inversion method, specifically including:

[0032] The model building and preprocessing module is used to build a three-dimensional spatial model of the target nuclear radiation field, determine multiple radiation observation locations in the three-dimensional spatial model, acquire energy spectrum data of each radiation observation location simultaneously, and preprocess the energy spectrum data to generate a standard count rate dataset.

[0033] The radiation intensity calculation module is used to discretize the target nuclear radiation field based on a three-dimensional spatial model, define each spatial node after discretization as a potential point source, calculate the theoretical count rate contribution of the potential point source to each radiation observation location under each energy channel, and then determine the initial radiation intensity of each potential point source under each energy channel.

[0034] The radiation attenuation calculation module is used to analyze the relative contribution relationship and radiation propagation attenuation relationship of each potential point source relative to the target point for any target point in the target nuclear radiation field. It aggregates the two relationships to obtain the field strength response factor of the target point in different energy channels, and uniformly represents the field strength response factor of different energy channels to generate the field strength response vector corresponding to the target point.

[0035] The model training and output module is used to construct an energy flux mapping model with the field strength response vector as input and the energy flux as output. The energy flux mapping model is trained by selecting the field strength response vector of a known target point and its corresponding energy flux, so that the input and output of the energy flux mapping model form a mapping relationship. The field strength response vector of any target point is input into the trained energy flux mapping model to obtain the energy flux of the target point, thus completing the inversion of the target nuclear radiation field.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] This invention achieves digital modeling of the spatial structure of the radiation field by establishing a three-dimensional spatial model of the target nuclear radiation field and recording the detector coordinates. Through background subtraction and time normalization, the data more realistically reflects the physical characteristics of the target radiation field itself, providing an accurate spatial reference for subsequent radiation source location and field strength calculation. It acquires and preprocesses energy spectrum data from multiple energy channels, independently calculating the theoretical count rate contribution for each energy channel. Simultaneously, the initial radiation intensity reflects the differences in source strength distribution between different channels, effectively utilizing the energy dimension information of the radiation field. This reflects the characteristics of radiated photons under different energy channels, improving the information dimension and accuracy of the inversion. Traditional inversion methods often rely on purely data-driven or simplified attenuation models, failing to fully consider medium attenuation and energy dependence. This step, by introducing attenuation coefficients and path lengths, more realistically simulates the propagation process of radiation in complex media.

[0038] This invention also normalizes and balances the influence of different point sources on the target point by constructing relative contribution relationships and radiation propagation attenuation relationships, avoiding the dominance of individual strong point sources in the calculation. By weighted aggregation of the contributions of each potential point source to the target point and considering the attenuation of radiation in the medium, a field strength response vector is constructed that can characterize the comprehensive radiation reception of the target point under multiple energy channels, effectively balancing the influence of strong and weak radiation sources and truly reflecting the propagation process of radiation in complex media. Utilizing the nonlinear mapping relationship between the field strength response vector and the energy flux, a neural network model is used to achieve fast and high-precision prediction from radiation response characteristics to energy flux. Taking the field strength response vector as input, which itself integrates information on radiation sources, medium attenuation, and energy channels, the model can learn a physically based mapping relationship, improving the reliability of inversion in data-scarce regions and complex scenarios. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the method flow of an embodiment of the present invention;

[0040] Figure 2 This is a schematic diagram of the curves showing the change of initial radiation intensity and dielectric attenuation factor with energy channel energy value in an embodiment of the present invention;

[0041] Figure 3 This is a schematic diagram of the system modules in an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0043] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0044] Example:

[0045] Please see Figures 1 to 2 The present invention provides a technical solution:

[0046] A rapid inversion method for nuclear radiation fields, comprising the following steps:

[0047] Step 1: Construct a three-dimensional spatial model of the target nuclear radiation field, determine multiple radiation observation locations in the three-dimensional spatial model, synchronously acquire energy spectrum data at each radiation observation location, and preprocess the energy spectrum data to generate a standard count rate dataset.

[0048] In this embodiment, the principle for establishing a three-dimensional spatial model of the target nuclear radiation field is as follows: The target nuclear radiation field area is scanned using a lidar device to obtain point cloud data, which includes the three-dimensional coordinates of each point within the target nuclear radiation field area, used to reconstruct the geometry of the target area; several radiation observation positions are determined within the target nuclear radiation field, and radiation detectors are deployed. The three-dimensional spatial model of the target nuclear radiation field includes the position coordinates of each point within the target nuclear radiation field area.

[0049] The principle underlying the generation of the standard count rate dataset is as follows:

[0050] The energy spectrum data represents the cumulative count of radiation photons collected at the radiation observation location in each energy channel within a set observation time.

[0051] The energy channels represent multiple discrete energy ranges divided according to the photon energy range. Each channel corresponds to one energy range and is used to record the total number of radiated photons detected within that energy range. The principle of dividing the energy channels is as follows: the lower limit of the energy channel is set as the effective energy threshold of the detector, usually 30 to 50 keV. If it is below the effective energy threshold, it is easily affected by noise interference. The lower limit of the energy channel is determined based on the maximum energy of the radioactive nuclides that may exist in the target nuclear radiation field, and can usually be set to 3000 keV to cover common... For nuclides, this scheme sets the energy channel range to 50 to 3000 keV. The specific channel division strategy is as follows: In the 50 to 200 keV range, which is a low-energy region with significant scattering, a higher resolution is required. A narrower channel is used to accurately capture attenuation changes and scattering characteristics, so the channel width is set to 10 keV. In the 200 to 1500 keV range, which is a region where the characteristic peaks of common nuclides are concentrated, the Compton scattering cross section is the largest in this range, and the change with energy is relatively gradual. The channel width is set to 50 keV to control the total number of channels while ensuring the resolution of characteristic peaks, thus balancing computational efficiency. In the 1500 to 3000 keV range, which is a high-energy region, the attenuation coefficient of high-energy photons is small and the change is relatively gradual. A wider channel is set to improve the counting statistics of a single channel, so the channel width is set to 100 keV.

[0052] For each radiation observation location, the difference between the cumulative count at each energy channel and the corresponding background cumulative count is calculated. Then, the ratio of this difference to the observation time duration is calculated to determine the count rate at each energy channel. The count rates at each radiation observation location under each energy channel are summarized to construct a standard count rate dataset.

[0053] Specifically, for each radiation observation location, the background cumulative count calculation logic under each energy channel is as follows: during the time interval when the radiation observation location is in a radiation-free environment, the radiation observation location is observed for the same duration as the observation time, and the cumulative count of radiation photons collected at the radiation observation location under each energy channel for the same duration is obtained as the background cumulative count of the radiation observation location under each energy channel.

[0054] The purpose of preprocessing the energy spectrum data is to extract signals directly related to the radiation source of the target nuclear radiation field from the raw measurement data, and to reduce or eliminate interference from non-radiation sources. During the data measurement of the nuclear radiation field, the energy spectrum data recorded by the detector includes not only photon counts from radiation sources in the target nuclear radiation field, but also photon counts generated by background radiation from the environment, such as natural radionuclides and cosmic rays. This latter part needs to be removed to obtain the cumulative count generated by the target nuclear radiation field.

[0055] By subtracting the background cumulative count from the energy spectrum data, the actual cumulative count caused by the target nuclear radiation field observed at each radiation observation location is obtained. Dividing the actual cumulative count by the observation time length, the count rate at each radiation observation location in each energy channel is obtained. The specific calculation formula is as follows:

[0056]

[0057]

[0058] in, Indicates the first The radiation observation location at the first The actual cumulative count of each energy channel, An index indicating the location of a radiation observation. Indices representing energy channels. Indicates the first The radiation observed at the [number]th radiation observation location was [number]. The cumulative count of energy channels, Indicates the first The radiation observed at the [number]th radiation observation location was [number]. The baseline cumulative count of each energy channel, Indicates the first The length of observation time at each radiation observation location. Indicates the first The radiation observation location at the first The count rate of each energy channel;

[0059] The observation time may vary at different observation locations, and the cumulative count is affected by the observation time; the longer the time, the larger the count. Therefore, the cumulative count cannot be directly used to compare the radiation intensity at different locations. Instead, the count rate is obtained by dividing the cumulative count by the observation time, thus unifying the data to the same time dimension. The count rate reflects the average number of radiated photons arriving at a specific radiation observation location per unit time under each energy channel; it is an intensity indicator. A higher count rate indicates a greater radiation intensity at the observation location under that energy channel.

[0060] A standard count rate dataset was constructed by combining the count rates of each radiation observation location in each energy channel.

[0061] The standard count rate dataset is a two-dimensional array where the row index is the radiation observation location and the column index is the energy channel. Each data item is... , representing the first The radiation observation location at the first The count rate of the energy channel corresponds to the number of energy channels in the array. Line number The data consists of columns; the standard count rate dataset reflects the spatial variation of radiation intensity by integrating the count rate at different radiation observation locations, and reflects the energy composition of the radiation source by integrating the count rate distribution of the same radiation observation location under different energy channels; each row corresponds to the same radiation observation location, reflecting the radiation intensity of a radiation observation location under different energy channels, and the differences between different rows reflect the spatial non-uniformity of radiation; each column corresponds to the same energy channel, reflecting the intensity distribution of radiation at various observation points in space under that energy channel, and the differences between different columns reflect the differences in the energy composition of the radiation source; after background subtraction and time normalization, the data more realistically reflects the radiation characteristics of the target nuclear radiation field itself, reducing errors caused by environmental interference or differences in measurement conditions.

[0062] By combining the count rates of each radiation observation location with those of each energy channel to construct a standard count rate dataset, measurement data from multiple locations and multiple energy channels can be integrated into a unified dataset, facilitating subsequent retrieval and processing. The nuclear radiation field is a system with a two-dimensional spatial-energy distribution. Using only one energy channel or one observation location cannot fully describe the radiation field. Only by combining data from all observation locations and all energy channels can an accurate data foundation be provided for inverting the nuclear radiation field.

[0063] Step 2: Discretize the target nuclear radiation field based on the three-dimensional spatial model, define each discretized spatial node as a potential point source, calculate the theoretical count rate contribution of each potential point source to each radiation observation location under each energy channel, and then determine the initial radiation intensity of each potential point source under each energy channel.

[0064] In this embodiment, the specific principle of discretizing the target nuclear radiation field is as follows: Based on the constructed three-dimensional spatial model, the target nuclear radiation field region is divided into regular spatial grid cells. The size of the spatial grid cells is set by expert scoring based on the complexity of the actual scene, the computational accuracy requirements, and the detection accuracy of the radiation detector. Typically, the size of the spatial grid cells is between 0.1m and 1.0m. The smaller the size of the spatial grid cells, the higher the spatial resolution, and the more finely the local gradient changes of the nuclear radiation field can be captured. The larger the size of the spatial grid cells, the lower the spatial resolution, which may obscure the detailed structure. This is suitable for scenarios with a gentle radiation field distribution or large-scale evaluation. The nodes of each spatial grid cell are set as potential point source locations. These potential point sources represent a spatial discretization approximation of the target nuclear radiation field region, that is, it is assumed that the nuclear radiation field is composed of these discrete potential point sources.

[0065] The principle for calculating the theoretical count rate contribution of potential point sources to each radiation observation location under each energy channel is as follows:

[0066] The theoretical count rate contribution is the product of the geometric attenuation term and the medium attenuation term; wherein, the calculation logic of the geometric attenuation term is as follows: first calculate the Euclidean distance between the potential point source and the radiation observation location, and obtain the absolute detection efficiency of the potential point source under the energy channel, and calculate the geometric attenuation term based on the absolute detection efficiency and the Euclidean distance;

[0067] The geometric attenuation term describes the natural attenuation of radiation intensity as it propagates uniformly from a point source into space due to increasing distance. This term is only related to distance and absolute detection efficiency. Assuming radiation propagates in a vacuum and neglecting the effects of medium absorption, it reflects the inverse square law of radiation intensity decay with distance. Radiation emitted from a point source diffuses uniformly into space, and the radiant flux per unit area is inversely proportional to the square of the distance. Absolute detection efficiency represents the absolute efficiency of a radiation detector for the full-energy peak of a specific energy channel. It describes the efficiency of a single-energy photon point source that emits that energy channel uniformly into space at a standard distance from the detector. The absolute detection efficiency is the ratio of the full-energy peak count rate recorded by the radiation detector to the actual photon count rate emitted by the monoenergetic photon point source. The standard distance is usually 1m. The absolute detection efficiency is between 0 and 1 and is obtained through Monte Carlo simulation experiments. The absolute detection efficiency must be considered in the geometric attenuation term because the radiation detector is not a perfect sensor that can capture all photons. When a photon reaches the radiation detector, not all photons can produce an event that can be recorded as a full-energy peak, and therefore the count rate generated by the radiation photon cannot be detected. The absolute detection efficiency is used to correct for this situation.

[0068] The calculation logic for the medium attenuation term is as follows: obtain the straight path from the potential point source to the radiation observation location, then determine the number of media types traversed by the straight path, the path length of the straight path in each medium, and the linear attenuation coefficient of each medium. Based on the linear attenuation coefficient and the path length of the straight path in each medium, calculate the attenuation contribution of each medium to the potential point source. Summate the attenuation contributions of all media to obtain the total attenuation of the radiated photon transmitted between the potential point source and the radiation observation location. Based on the Lambert-Beer law, substitute the negative number of the total attenuation of the medium as the exponent into the exponential function with the natural constant as the base to obtain the medium attenuation term.

[0069] The medium attenuation term describes the intensity attenuation of radiation from a potential point source to the radiation observation location due to the absorption and scattering of photons by the medium it passes through. According to the Lambert-Beer law, radiation intensity attenuates exponentially with the penetration distance in a medium. In nuclear radiation fields, common media include air, concrete, water, and soil. By identifying the types of media along the radiation transmission path and the path length of each medium through a three-dimensional spatial model, the linear attenuation coefficient of each medium is obtained under different energy channels. The attenuation contribution of a certain medium is obtained by multiplying the linear attenuation constant by the path length of a certain medium along the straight path from the radiation observation location to the potential point source. This is the ability of the medium to block photons in that energy channel. The larger the value, the stronger the ability of the medium to block photons in that energy channel, the more obvious the attenuation effect, and the smaller the theoretical count rate. The attenuation contributions of all media along the radiation transmission path are calculated and summed to obtain the total attenuation of radiation along the transmission path due to the medium.

[0070] The formula for calculating the theoretical count rate contribution of potential point sources to each radiation observation location under each energy channel is as follows:

[0071]

[0072] in, Indicates the first The potential point source pairs the first The radiation observation location at the first The theoretical count rate contribution of each energy channel Indicates the first The potential point source is in the first Absolute detection efficiency under each energy channel Indicates the first The first radiation observation location and the first Euclidean distance between potential point sources Represents the geometric attenuation term. Indicates radiation from the first The radiation observation location to the first The index of the media types traversed by the straight path of each potential point source, and , Indicates radiation from the first The radiation observation location to the first The number of media types traversed by the straight-line path of a potential point source. Indicates that the radiation comes from the first The radiation observation location to the first In the straight path of the nth potential point source, the th The medium for the first The linear attenuation coefficient of photons in each energy channel Indicates that the radiation comes from the first The radiation observation location to the first In the straight path of the potential point source, after passing the first... The path length of the medium This represents the medium attenuation term.

[0073] During the propagation of radiated photons from a potential point source to a radiation observation location, they are affected by two main attenuation mechanisms: geometric attenuation and medium attenuation. Geometric attenuation is the attenuation caused by the transmission distance; the radiation intensity decreases inversely with the square of the distance, originating from the spherical diffusion effect during photon propagation. This attenuation is independent of the medium and depends only on the distance and detector efficiency. Medium attenuation is the attenuation caused by absorption and scattering by the medium material. When radiation passes through a medium, it is absorbed or scattered by interactions such as the photoelectric effect and Compton scattering, and its intensity decreases exponentially, depending on the type and thickness of the medium and the photon energy. These two processes occur in series physically: the photon first diffuses with distance and is simultaneously absorbed by the medium along the propagation path. Therefore, the total attenuation effect is the product of the two. The theoretical count rate contribution reflects the attenuation of radiation intensity due to increased distance and medium absorption during the radiation propagation process from the potential point source to the radiation observation location. It is used to convert the initial radiation intensity of the potential point source into the count rate generated by the radiation actually acting on the radiation observation location. Different energy channels have different absolute detection efficiencies and linear attenuation coefficients, so the theoretical count rate contribution varies with energy, reflecting the attenuation differences of photons of different energies, and also reflecting the non-uniformity of the nuclear radiation field in space. The theoretical count rate contribution is proportional to the geometric attenuation term and the medium attenuation term. The larger the theoretical count rate contribution, the better the spatial connectivity between the potential point source and the observation point, the more efficient the radiation propagation, and the less severe the radiation attenuation.

[0074] The principle underlying the determination of the initial radiation intensity of each potential point source in each energy channel is as follows:

[0075] Under each energy channel, the count rate at any radiation observation location is affected by all potential point sources. The influence of each potential point source on the radiation observation location is related to the initial radiation intensity of each potential point source and the theoretical count rate contribution of that potential point source to that radiation observation location. The logical relationship among the three is as follows: multiply the initial radiation intensity of each potential point source by the theoretical count rate of each potential point source at that radiation observation location, and sum the results of multiplying the initial radiation intensity of all potential point sources by the theoretical count rate of each potential point source at that radiation observation location to obtain the count rate of that radiation observation location; given the standard count rate dataset and the theoretical count rate contribution of each potential point source to that radiation observation location, the initial radiation intensity of each potential point source under the energy channel can be deduced; perform the above calculation for each energy channel to obtain the initial radiation intensity of each potential point source under each energy channel.

[0076] The nuclear radiation field is considered to consist of multiple discrete potential point sources, each with different radiation intensities at different energy channels. At any observation location within a single energy channel, the count rate is equal to the sum of the count rates generated at that location after the radiation from all potential point sources has attenuated through spatial propagation. The specific calculation formula is as follows:

[0077]

[0078] in, Indicates the first The potential point source is in the first Initial radiation intensity under each energy channel The index representing the potential point source, and , Indicates the number of potential point sources;

[0079] The above calculation formula reflects the physical relationship between potential point sources and radiation observation locations in the radiation field, specifically including: radiation propagation superposition, energy channel dependence, and comprehensive attenuation during propagation. Radiation propagation superposition indicates that the radiation field consists of multiple discrete potential point sources, each contributing independently to the radiation observation location, with the final count rate being a linear superposition of all potential point source contributions. Energy channel dependence represents the calculation for each energy channel, reflecting the differences in photon propagation and attenuation processes across different energy channels. Comprehensive attenuation during propagation reflects the attenuation caused by distance and medium during the propagation of radiation from potential point sources to the radiation observation location.

[0080] Indicates the first The radiation observation location at the first The count rate of each energy channel is a known and measurable result, obtainable from standard count rate datasets. It reflects the comprehensive performance of the target nuclear radiation field at that radiation observation location and energy channel. Potential point sources are considered as the sources of radiation, and the initial radiation intensity of each potential point source is the initial value of the radiation affecting each radiation observation location, determining the source and magnitude of the radiation field. The theoretical count rate contribution is a proportionality coefficient reflecting the attenuation of radiation during propagation. If a potential point source releases radiation with an initial radiation intensity, after geometric attenuation and medium absorption, a count rate is ultimately recorded at the radiation observation location. The theoretical count rate contribution reflects the influence of geometric attenuation and medium absorption during this process. Within the detection energy range, the interaction cross-section of radiated photons is very small; a photon will not be deflected, absorbed, or enhanced by the presence of other photons nearby. Therefore, the total count rate measured at a radiation observation location is a linear superposition of the contributions of all potential point sources to that location.

[0081] Table 1 reflects the initial radiation intensity of a potential point source in different energy channels. The intensity gradually increases in the range of 50-200 keV, but remains relatively low, indicating that photons in the low-energy region are easily absorbed and scattered by the medium. Above 1500 keV, the intensity continuously decreases, reflecting the low emission probability of high-energy photons, strong penetrating power but decreased count rate. The intensity does not change monotonically with energy, reaching a peak at approximately 500-600 keV, and then slowly decreasing, reflecting that the characteristic emission energy of nuclides is concentrated in the medium-energy range, and that Compton scattering contributes significantly in the medium-energy range. Note that an energy channel is a range, and the energy value of an energy channel in Table 1 is a specific energy value corresponding to a particular energy channel, used to reflect the energy channel at that point.

[0082] Table 1. Initial radiation intensity as a function of energy channel energy value

[0083] initial radiation intensity Energy channel energy value (keV) 0.012 50 0.025 80 0.048 110 0.102 140 0.156 170 0.21 200 0.275 250 0.32 300 0.38 350 0.425 400 0.46 450 0.49 500 0.52 600 0.51 700 0.48 800 0.445 900 0.4 1000 0.36 1100 0.32 1200 0.28 1300 0.24 1400 0.21 1500 0.18 1700 0.15 1900 0.12 2100 0.095 2300 0.07 2500 0.05 2700 0.035 2900 0.02 3000

[0084] Step 3: For any target point in the target nuclear radiation field, analyze the relative contribution relationship and radiation propagation attenuation relationship of each potential point source relative to the target point. Perform aggregation calculation on the above two relationships to obtain the field strength response factor of the target point in different energy channels. Perform unified characterization on the field strength response factors of different energy channels to generate the field strength response vector corresponding to the target point.

[0085] In this embodiment, the relative contribution relationship of each potential point source to the target point is as follows:

[0086] The relative contribution weight is used to characterize the relative contribution relationship of each potential point source to the target point. The initial radiation intensity of each potential point source under each energy channel is obtained. For any energy channel, the initial radiation intensity of each potential point source under that energy channel is divided by the sum of the initial radiation intensities of all potential point sources under the same energy channel to generate the relative contribution weight of each potential point source to the target point under each energy channel. The specific calculation formula is as follows:

[0087]

[0088] in, Indicates the first The potential point source is in the first The relative contribution weights of each energy channel to the target point This represents a protection parameter to prevent the denominator from being zero.

[0089] The essence of nuclear radiation field inversion is to deduce the known from the unknown. Using data from a finite number of observation points, it infers the entire nuclear radiation field. The target point represents the unknown nuclear radiation field. Relative contribution weights are calculated based on the initial radiation intensity of the point sources. These relative contribution weights reflect the proportion of the radiation intensity of a point source relative to the total intensity of all point sources within a given energy channel. Generally, the greater the intensity of a potential point source, the greater its contribution to the total radiation at other locations, and therefore, the larger its relative contribution weight. The purpose of calculating the relative contribution weights is to quantify the relative importance of each potential point source's radiation contribution to a target point among multiple potential point sources, thus enabling subsequent weighted calculations. The initial radiation intensity of the point source is normalized by dividing by the sum of the initial radiation intensities of all point sources. Within a radiation field, there may be point sources with vastly different intensities. Without normalization, strong sources might dominate the calculation, while the contributions of weak sources would be obscured. Weighted calculations balance the contributions of all point sources to the same scale, facilitating subsequent weighted aggregation. In real-world scenarios, there might be situations where the initial radiation intensity of all point sources within a certain energy channel is extremely low or zero. In such cases, the denominator might approach zero, leading to calculation anomalies. Therefore, a protection parameter is introduced. To avoid division by zero errors, For a very small positive number, we can take... .

[0090] The radiation propagation attenuation relationship between each potential point source and the target point is as follows:

[0091] The radiation attenuation factor characterizes the radiation propagation attenuation relationship between each potential point source and the target point. It obtains the types of media traversed by the straight-line path from each potential point source to the target point, the path length of the straight-line path in each medium, and the linear attenuation coefficient of each medium. Based on the linear attenuation coefficient and the path length of the straight-line path in each medium, it calculates the attenuation contribution of each medium to the potential point source. The attenuation contributions of all media are summed to obtain the total media attenuation of the radiated photon propagating between the potential point source and the target point. Based on Beer-Lambert's law, the negative of the total media attenuation is substituted into an exponential function with the natural constant as the base to obtain the radiation attenuation factor. The specific calculation formula is as follows:

[0092]

[0093] in, Indicates the first The potential point source is in the first Radiation attenuation factor between each energy channel and the target point Indicates from the radiation level A straight path from a potential point source to a target point passes through the index of the medium type. Indicates radiation from the first The number of media types traversed by a straight-line path from a potential point source to a target point. Indicates that the radiation comes from the first In the straight-line path from the potential point source to the target point, the th... The medium for the first The linear attenuation coefficient of photons in each energy channel Indicates that the radiation comes from the first In the straight-line path from the potential point source to the target point, after the first... The path length of the medium.

[0094] The principle for calculating the radiation attenuation factor is the same as that for calculating the medium attenuation term in the theoretical count rate contribution. However, the radiation attenuation factor here refers to the relationship between a potential point source and any target point, while the medium attenuation term refers to the relationship between a potential point source and the radiation observation location. The attenuation law of radiation in a medium is universal and does not change the calculation principle due to the different locations of the "radiation observation location" and "any target point".

[0095] Table 2 reflects the variation of radiation attenuation factor under different energy channels. In the low energy range of 50~200keV, the photoelectric effect dominates and the attenuation factor is small. In the medium energy range of 200~1500keV, Compton scattering dominates and the attenuation factor gradually increases. When the energy is greater than 15000keV, high-energy photons have a stronger ability to penetrate matter and the probability of being absorbed or scattered by the medium is significantly reduced. The radiation attenuation factor approaches 1, indicating that the radiation is almost unimpeded by the medium under this energy channel and can penetrate the medium to propagate to the target point with almost no loss.

[0096] Table 2. Variation of Radiation Attenuation Factor with Energy Channel Energy Value

[0097] Radiation attenuation factor Energy channel energy value (keV) 0.082 50 0.135 80 0.21 110 0.285 140 0.352 170 0.418 200 0.51 250 0.595 300 0.668 350 0.72 400 0.768 450 0.81 500 0.862 600 0.902 700 0.928 800 0.948 900 0.96 1000 0.97 1100 0.976 1200 0.982 1300 0.986 1400 0.99 1500 0.992 1700 0.994 1900 0.995 2100 0.996 2300 0.997 2500 0.998 2700 0.998 2900 0.999 3000

[0098] The principle underlying the generation of the field intensity response vector corresponding to the target point is as follows:

[0099] The relative contribution relationship and the radiation propagation attenuation relationship are aggregated to obtain the field strength response factor of the target point in different energy channels. The specific calculation logic is as follows: For each energy channel, the reciprocals of the radiation propagation attenuation factors of each potential point source in that energy channel to the target point are weighted and summed, with the weights being the relative contribution weights. This yields the field strength response factor of all potential point sources in that energy channel to the target point. The specific calculation formula is as follows:

[0100]

[0101] in, This indicates that the target point is located within the nuclear radiation field. Field strength response factor of each energy channel. Indicates the index of the energy channel, and , Indicates the number of energy channels;

[0102] The purpose of calculating the field strength response factor is to expand from a limited "radiation observation location" to "any target point," thereby enabling rapid inversion of the energy fluence at any location in the nuclear radiation field. The field strength response factor is a dimensionless index that reflects the combined radiation contribution of all potential point sources emitting radiation to any target point under various energy channels. It comprehensively considers the superimposed influence of various potential point sources on the target point. It can be regarded as the first The normalized value of the radiation intensity of a potential point source directly affects the radiation situation at the target point. Furthermore, during radiation transmission, attenuation occurs due to medium loss. This attenuation was quantified; the calculated field strength response factor did not directly reflect the effect of geometric attenuation, but rather... Indirect manifestation, In It is obtained by inversion using data from known radiation observation locations. It is itself a calibrated intensity value, reflecting the radiation intensity of a potential point source at that location without medium attenuation. Therefore, when calculating the field strength response factor, it is calculated... , obtained the The radiation contribution of a potential point source to a target point is directly proportional to the initial radiation intensity of the potential point source and inversely proportional to the radiation attenuation factor. This reflects that the greater the initial radiation intensity, the higher the radiation contribution to the target point. The smaller the radiation attenuation factor, the less the radiation is attenuated by the medium during propagation, and the higher the radiation contribution to the target point.

[0103] The field strength response factors of all potential point sources to the target point under each energy channel are calculated and uniformly characterized to form the field strength response vector of the target point under each energy channel, specifically:

[0104]

[0105] in, This represents the field strength response vector at the target point.

[0106] The field strength response factor is a single value calculated for a specific energy channel, taking into account the radiation contribution and attenuation effect of all potential point sources to the target point. Its function is to reflect the comprehensive radiation contribution of all potential point sources to the target point under a given energy channel, considering the point source intensity weights and the effect of medium attenuation. Each energy channel has an independent field strength response factor. Different energy channels have different radiation attenuation behaviors and source strength distributions; therefore, the field strength response factor is calculated separately for each energy channel to more accurately reflect the energy spectrum characteristics. The field strength response vector is a vector formed by arranging the field strength response factors of all energy channels in sequence. It uniformly represents the radiation response characteristics of the target point under multiple energy channels, integrating the information of all energy channels to form a structured, high-dimensional feature representation, facilitating subsequent model processing. It not only reflects the spatial radiation reception of the target point but also embodies the energy distribution characteristics.

[0107] Step 4: Construct an energy flux mapping model with the field strength response vector as input and the energy flux as output. Select the field strength response vector of a known target point and its corresponding energy flux to train the energy flux mapping model, so that the input and output of the energy flux mapping model form a mapping relationship. Input the field strength response vector of any target point into the trained energy flux mapping model to obtain the energy flux of the target point, and complete the inversion of the target nuclear radiation field.

[0108] In this embodiment, the principle upon which the energy flux mapping model is constructed is as follows:

[0109] Several target points are set in the target nuclear radiation field, and the energy flux of each target point is obtained by simulation. A training set and a validation set are generated in an 8:2 ratio. The field strength response vector of the target point in the training set is used as input and its energy flux is used as output to train the energy flux mapping model. The mean square error loss function is used as the loss function. The model is validated through the validation set until the model is confirmed to have converged. The model at this time is output as the energy flux mapping model.

[0110] In this embodiment, the energy flux mapping model employs a regression structure of a multi-layer fully connected neural network to map the field intensity response vector of a target point to the energy flux scalar of that target point, thereby achieving rapid inversion of the energy flux at any target point. The model input is the field intensity response vector, the dimension of which is equal to the number of energy channels E; to facilitate providing a specific numerical value, this embodiment takes the number of energy channels E as 128, meaning that the input for each target point is a 128-dimensional vector. The model output is a single numerical value, representing the energy flux of the target point within a preset energy range. The unit system used is the same as that in the simulation output, and the units are not modified within the model.

[0111] To improve numerical stability and enhance cross-scenario generalization ability, this embodiment applies a fixed numerical scale to both input and output: the 128-dimensional input vector is first standardized dimension-by-dimensionally, with the standardized parameters obtained statistically from the training set and fixed after training; the output energy flux is processed using a logarithmic regression method, i.e., the training labels are first logarithmically transformed before being fed into the loss calculation, and after model training, the prediction results are inversely transformed during the inference phase to obtain the final energy flux value. This processing addresses the training instability caused by energy flux potentially spanning multiple orders of magnitude at different spatial locations, without altering the intrinsic correspondence between physical quantities, as the input vector still originates from the weighted aggregation result of the contribution weights and attenuation factors of each potential point source in step 3.

[0112] In terms of network structure, this embodiment explicitly sets the energy injection mapping model as a series structure of "input layer - feature upscaling layer - residual feature extraction layer - regression output layer", and the dimensions, activation methods and regularization parameters of each layer are fixed as follows: The input layer receives a 128-dimensional input vector and then enters the first fully connected layer. The output dimension of the first fully connected layer is set to 512, and the activation function is ReLU. After the first layer, batch normalization is applied to stabilize the gradient, and Dropout is applied to suppress overfitting. The Dropout rate is set to 0.10. Then, the second fully connected layer is entered. The second layer maps the 512-dimensional model to the first 512-dimensional model, and also uses ReLU and batch normalization, and sets the Dropout rate to 0.10, so that the model can maintain its expressive power while being robust to simulation noise and small perturbations of the response factor. Then, the residual feature extraction structure is introduced. The residual structure contains two residual blocks, each of which is a “two-layer fully connected short-circuit connection”. The first layer of each residual block maps the 512-dimensional data to the 256-dimensional data and applies ReLU. The second layer maps the 256-dimensional data back to the 512-dimensional data and retains the linear output without applying ReLU. The output is then added to the residual block input and then processed by ReLU to obtain the residual block output. After the summed output of each residual block, a Dropout layer is added with a dropout rate of 0.05 to improve generalization without significantly weakening the residual representation.

[0113] After the residual module, the model enters a dimensionality reduction layer. This layer is a fully connected layer that maps the 512 dimensions to 128 dimensions, employing ReLU and batch normalization, but without Dropout to avoid excessive random discarding of the fine-grained information required for the regression output. Finally, the model enters the regression output layer, which is a fully connected layer that maps the 128 dimensions to 1 dimension, outputting the predicted energy fluence value on a logarithmic scale. This output layer does not use an activation function to meet the requirement of continuous value output for the regression task. At this point, the model's hierarchy, connectivity, and specific dimensions of each layer are determined, and this structure corresponds one-to-one with the 128-dimensional response factor vector generated in step 3, ensuring a clear and traceable intrinsic link between "energy channel response factors in the input vector—network nonlinear mapping—energy fluence scalar".

[0114] Regarding the construction of training data, this embodiment is consistent with the description in step 4, that is, several target points are sampled uniformly or in sections in the three-dimensional model of the target radiation field, and each target point is simulated to obtain the real energy flux label. At the same time, the field intensity response vector of the target point is calculated as the input feature using the process of steps 1 to 3.

[0115] To ensure that the digital parameters are clear and reproducible, this embodiment sets the total number of target points to 50,000, of which 40,000 are used as the training set and 10,000 as the validation set, with a ratio of 8:2. The spatial sampling of the target points adopts a combination of "voxel uniform sampling and boundary densification sampling". 70% of the points come from global uniform sampling, and 30% of the points come from densification sampling near the interface of the shielding object, the wall of the device, or the interface of the medium, in order to cover the strong gradient region caused by the attenuation change, so that the model is more sensitive to occlusion and medium changes in the actual scene.

[0116] To further enhance generalization, this embodiment adds a small noise perturbation to the input vector of the training set as data augmentation. The perturbation method is to multiply each dimension of the input by a random scaling factor centered at 1, with a standard deviation of 0.01. The perturbed data is then pruned to ensure that the values ​​do not have non-physical negative values ​​or abnormal extreme values. This augmentation is only used on the training set, and the validation set is not augmented. It is used to objectively evaluate the model's fitting ability.

[0117] Regarding training hyperparameters and optimization strategies, this embodiment uses the Adam optimizer for training. The initial learning rate is set to 0.001, the first momentum coefficient is set to 0.90, the second momentum coefficient is set to 0.999, and the numerical stability term is set to 0.00000001. The batch size is set to 256, and all training samples are traversed once in each training epoch, with a maximum of 200 training epochs. To avoid overfitting and accelerate convergence, this embodiment uses weight decay as a parameter regularization, with a weight decay coefficient set to 0.0001, and enables an early stopping strategy. When the validation set loss does not decrease for 20 consecutive epochs, training stops and rolls back to the weight of the epoch with the lowest validation set loss. A piecewise decay strategy is used for the learning rate. When the validation set loss does not decrease for 10 consecutive epochs, the learning rate is multiplied by 0.50, and the minimum learning rate lower limit is set to 0.00001 to ensure that the regression accuracy can still be refined in the later stages of training without oscillation. Gradient clipping is enabled during training to improve numerical stability. The gradient clipping threshold is set to 1.0, and the gradient norm is scaled proportionally when it exceeds the threshold. Single-precision floating-point calculations are sufficient to meet the accuracy requirements during training, and the same precision can be deployed directly during the inference phase to achieve faster speed.

[0118] Regarding the setting of the loss function, this embodiment uses mean squared error as the main loss to measure the difference between the logarithmic energy flux predicted by the model and the simulated logarithmic energy flux, and takes the average of the mean squared error within a batch as the optimization objective. At the same time, in order to suppress the destruction of training by a small number of simulated outliers or strong noise points, this embodiment introduces a robust error pruning strategy, that is, when the error of a single sample exceeds a preset threshold, its contribution is constrained to an upper limit. The threshold is set to 4.0, which corresponds to the upper limit tolerance range of the energy flux multiplier error on the logarithmic scale, thereby avoiding extreme samples from dominating the gradient direction.

[0119] In addition to the main loss, this embodiment incorporates a parameter norm penalty term into the loss to achieve weight smoothing. The penalty strength is consistent with the aforementioned weight decay, ensuring that the model maintains continuity and physical plausibility for unseen spatial locations while fitting the training set sufficiently. The loss function consists of a regression error term and a parameter regularization term. The regression error term uses mean squared error, and the parameter regularization term uses weight decay.

[0120] Once the model training is complete and convergence is confirmed through the validation set, the field strength response vector obtained in step 3 for any target point can be input into the energy flux mapping model. The model outputs a logarithmic scale prediction value, which is then inversely transformed to obtain the energy flux. This allows for the rapid inversion of the energy flux within the nuclear radiation field without repeatedly performing high-overhead simulations.

[0121] Since each dimension of the input vector corresponds to the field strength response factor under a specific energy channel, and these response factors are jointly determined by the initial radiation intensity of the potential point source, the medium attenuation, and the relative contribution weight, the model learns the mapping relationship from the multi-channel response mode to the energy flux. This mapping naturally reflects the intrinsic relationship between nuclear radiation in space propagation and energy spectrum response, making the inversion results consistent and interpretable within the scenario.

[0122] Please see Figure 3 The present invention also provides a rapid nuclear radiation field inversion system, the system being used to implement the above-mentioned rapid nuclear radiation field inversion method, specifically including:

[0123] The model building and preprocessing module is used to build a three-dimensional spatial model of the target nuclear radiation field, determine multiple radiation observation locations in the three-dimensional spatial model, acquire energy spectrum data of each radiation observation location simultaneously, and preprocess the energy spectrum data to generate a standard count rate dataset.

[0124] The radiation intensity calculation module is used to discretize the target nuclear radiation field based on a three-dimensional spatial model, define each spatial node after discretization as a potential point source, calculate the theoretical count rate contribution of the potential point source to each radiation observation location under each energy channel, and then determine the initial radiation intensity of each potential point source under each energy channel.

[0125] The radiation attenuation calculation module is used to analyze the relative contribution relationship and radiation propagation attenuation relationship of each potential point source relative to the target point for any target point in the target nuclear radiation field. It aggregates the two relationships to obtain the field strength response factor of the target point in different energy channels, and uniformly represents the field strength response factor of different energy channels to generate the field strength response vector corresponding to the target point.

[0126] The model training and output module is used to construct an energy flux mapping model with the field strength response vector as input and the energy flux as output. The energy flux mapping model is trained by selecting the field strength response vector of a known target point and its corresponding energy flux, so that the input and output of the energy flux mapping model form a mapping relationship. The field strength response vector of any target point is input into the trained energy flux mapping model to obtain the energy flux of the target point, thus completing the inversion of the target nuclear radiation field.

[0127] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0128] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0129] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0130] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for rapid inversion of nuclear radiation fields, characterized in that, The specific steps include: Step 1: Construct a three-dimensional spatial model of the target nuclear radiation field, determine multiple radiation observation locations in the three-dimensional spatial model, synchronously acquire energy spectrum data at each radiation observation location, and preprocess the energy spectrum data to generate a standard count rate dataset. Step 2: Discretize the target nuclear radiation field based on the three-dimensional spatial model, define each discretized spatial node as a potential point source, calculate the theoretical count rate contribution of each potential point source to each radiation observation location under each energy channel, and then determine the initial radiation intensity of each potential point source under each energy channel. Step 3: For any target point in the target nuclear radiation field, analyze the relative contribution relationship and radiation propagation attenuation relationship of each potential point source relative to the target point. Perform aggregation calculation on the above two relationships to obtain the field strength response factor of the target point in different energy channels. Perform unified characterization on the field strength response factors of different energy channels to generate the field strength response vector corresponding to the target point. Step 4: Construct an energy flux mapping model with the field strength response vector as input and the energy flux as output. Select the field strength response vector of a known target point and its corresponding energy flux to train the energy flux mapping model, so that the input and output of the energy flux mapping model form a mapping relationship. Input the field strength response vector of any target point into the trained energy flux mapping model to obtain the energy flux of the target point, and complete the inversion of the target nuclear radiation field.

2. The method for rapid inversion of nuclear radiation fields according to claim 1, characterized in that: The principle underlying the generation of the standard count rate dataset in step 1 is as follows: The energy spectrum data represents the cumulative count of radiation photons collected at the radiation observation location in each energy channel within a set observation time. For each radiation observation location, the difference between the cumulative count at each energy channel and the corresponding background cumulative count is calculated. Then, the ratio of this difference to the observation time duration is calculated to determine the count rate at each energy channel. The count rates at each radiation observation location under each energy channel are summarized to construct a standard count rate dataset. Specifically, for each radiation observation location, the background cumulative count calculation logic under each energy channel is as follows: during the time interval when the radiation observation location is in a radiation-free environment, the radiation observation location is observed for the same duration as the observation time, and the cumulative count of radiation photons collected at the radiation observation location under each energy channel for the same duration is obtained as the background cumulative count of the radiation observation location under each energy channel.

3. The rapid inversion method for nuclear radiation fields according to claim 2, characterized in that: The principle behind calculating the theoretical count rate contribution of potential point sources to each radiation observation location under each energy channel in step 2 is as follows: The theoretical count rate contribution is the product of the geometric attenuation term and the medium attenuation term; wherein, the calculation logic of the geometric attenuation term is as follows: first calculate the Euclidean distance between the potential point source and the radiation observation location, and obtain the absolute detection efficiency of the potential point source under the energy channel, and calculate the geometric attenuation term based on the absolute detection efficiency and the Euclidean distance; The calculation logic for the medium attenuation term is as follows: obtain the straight path from the potential point source to the radiation observation location, then determine the number of media types traversed by the straight path, the path length of the straight path in each medium, and the linear attenuation coefficient of each medium. Based on the linear attenuation coefficient and the path length of the straight path in each medium, calculate the attenuation contribution of each medium to the potential point source. Summate the attenuation contributions of all media to obtain the total attenuation of the radiated photon transmitted between the potential point source and the radiation observation location. Based on the Lambert-Beer law, substitute the negative number of the total attenuation of the medium as the exponent into the exponential function with the natural constant as the base to obtain the medium attenuation term.

4. The rapid inversion method for nuclear radiation fields according to claim 3, characterized in that: The principle underlying the determination of the initial radiation intensity of each potential point source in each energy channel is as follows: Under each energy channel, the count rate at any radiation observation location is affected by all potential point sources. The influence of each potential point source on the radiation observation location is related to the initial radiation intensity of each potential point source and the theoretical count rate contribution of that potential point source to the radiation observation location. The logical relationship among the three is as follows: multiply the initial radiation intensity of each potential point source by the theoretical count rate of each potential point source at the radiation observation location, and sum the results of multiplying the initial radiation intensity of all potential point sources by the theoretical count rate of each potential point source at the radiation observation location to obtain the count rate of the radiation observation location; given the standard count rate dataset and the theoretical count rate contribution of each potential point source to the radiation observation location, the initial radiation intensity of each potential point source under the energy channel can be deduced. The above calculations are performed for each energy channel to obtain the initial radiation intensity of each potential point source in each energy channel.

5. The method for rapid inversion of nuclear radiation fields according to claim 4, characterized in that: The relative contribution relationship of each potential point source to the target point in step 3 is as follows: The relative contribution weight is used to characterize the relative contribution relationship of each potential point source to the target point. The initial radiation intensity of each potential point source under each energy channel is obtained. For any energy channel, the initial radiation intensity of each potential point source under that energy channel is divided by the sum of the initial radiation intensities of all potential point sources under the same energy channel to generate the relative contribution weight of each potential point source to the target point under each energy channel.

6. The method for rapid inversion of nuclear radiation fields according to claim 5, characterized in that: The radiation propagation attenuation relationship between each potential point source and the target point in step 3 is as follows: The radiation attenuation factor is used to characterize the radiation propagation attenuation relationship of each potential point source relative to the target point. The types of media traversed by the straight path from each potential point source to the target point, the path length of the straight path in each medium, and the linear attenuation coefficient of each medium are obtained. Based on the linear attenuation coefficient and the path length of the straight path in each medium, the attenuation contribution of each medium to the potential point source is calculated. The attenuation contributions of all media are summed to obtain the total media attenuation of the radiated photon between the potential point source and the target point. Based on the Lambert-Beer law, the negative number of the total media attenuation is substituted into the exponential function with the natural constant as the base to obtain the radiation attenuation factor.

7. The method for rapid inversion of nuclear radiation fields according to claim 6, characterized in that: The principle underlying the generation of the field strength response vector corresponding to the target point in step 3 is as follows: The relative contribution relationship and the radiation propagation attenuation relationship are aggregated to obtain the field strength response factor of the target point in different energy channels. The specific calculation logic is as follows: for each energy channel, the reciprocals of the radiation propagation attenuation factors of each potential point source to the target point under that energy channel are weighted and summed, and the weight is the relative contribution weight, so as to obtain the field strength response factor of all potential point sources to the target point under that energy channel. The field strength response factors of all potential point sources to the target point under each energy channel are calculated and uniformly characterized to form the field strength response vector of the target point under each energy channel.

8. The method for rapid inversion of nuclear radiation fields according to claim 1, characterized in that: The principle underlying the construction of the energy flux mapping model in step 4 is as follows: Several target points are set in the target nuclear radiation field, and the energy flux of each target point is obtained by simulation. A training set and a validation set are generated in an 8:2 ratio. The field strength response vector of the target point in the training set is used as input and its energy flux is used as output to train the energy flux mapping model. The mean square error loss function is used as the loss function. The model is validated through the validation set until the model is confirmed to have converged. The model at this time is output as the energy flux mapping model.

9. A rapid nuclear radiation field inversion system, characterized in that: The system is used to implement the rapid inversion method for nuclear radiation fields according to any one of claims 1-8, specifically including: The model building and preprocessing module is used to build a three-dimensional spatial model of the target nuclear radiation field, determine multiple radiation observation locations in the three-dimensional spatial model, acquire energy spectrum data of each radiation observation location simultaneously, and preprocess the energy spectrum data to generate a standard count rate dataset. The radiation intensity calculation module is used to discretize the target nuclear radiation field based on a three-dimensional spatial model, define each spatial node after discretization as a potential point source, calculate the theoretical count rate contribution of the potential point source to each radiation observation location under each energy channel, and then determine the initial radiation intensity of each potential point source under each energy channel. The radiation attenuation calculation module is used to analyze the relative contribution relationship and radiation propagation attenuation relationship of each potential point source relative to the target point for any target point in the target nuclear radiation field. It aggregates the two relationships to obtain the field strength response factor of the target point in different energy channels, and uniformly represents the field strength response factor of different energy channels to generate the field strength response vector corresponding to the target point. The model training and output module is used to construct an energy flux mapping model with the field strength response vector as input and the energy flux as output. The energy flux mapping model is trained by selecting the field strength response vector of a known target point and its corresponding energy flux, so that the input and output of the energy flux mapping model form a mapping relationship. The field strength response vector of any target point is input into the trained energy flux mapping model to obtain the energy flux of the target point, thus completing the inversion of the target nuclear radiation field.

Images