CN113536679B - Point source dose rate correction method based on artificial neural network - Google Patents

Abstract

The application discloses a point source dose rate correction method based on an artificial neural network, which comprises the following steps: s1, energy spectrum data and corresponding dose rate values of the unmanned aerial vehicle radioactivity monitoring system at different heights are obtained and are respectively used as input parameters and output parameters; s2, dividing part of energy spectrum data at different heights into training data, dividing the other part of energy spectrum data into test data, constructing an artificial neural network model by utilizing input parameters and output parameters in the training data, and importing the training data into the artificial neural network model for training to obtain a trained artificial neural network model; s3, respectively importing the test data into the trained artificial neural network model to obtain an ideal output result, and comparing errors between the ideal output result and the corresponding test data; and if the error is greater than or equal to the set precision expected value, repeating the steps S2 and S3, and if the error is less than the set precision expected value, training the debugged artificial neural network model to be a point source dose rate correction algorithm.

Description

A point source dose rate correction method based on artificial neural network

Technical Field

The present application relates to the technical field of radioactivity monitoring by unmanned aerial vehicles, and in particular, to a point source dose rate correction method based on an artificial neural network.

Background Art

Radioactivity monitoring by drones refers to the use of fixed-wing or rotary-wing drones as platforms, relying on radiation detectors and other airborne radioactivity monitoring payload equipment to conduct aerial radioactivity measurements. It is applicable to the fields of nuclear accidents, environmental radioactivity monitoring, mineral geological exploration, etc. The conventional detectors carried by drone radioactivity monitoring systems obtain the energy spectrum information of nuclides. When conducting radiation evaluation, they can only distinguish the types of different radioactive nuclides, and cannot simultaneously measure the air absorption dose rate of the nuclides, and do not have a multi-parameter evaluation method. If the dose rate is to be measured intuitively, a dose rate meter is required, which greatly increases the difficulty of drone operations, reduces work efficiency, and is not conducive to the use of drone platforms for radiation monitoring. At the same time, when the drone airborne system is operating, in order to intuitively observe the radiation levels of various environments and the location information of each point source and the surrounding radiation levels when a nuclear accident occurs, it is necessary to perform a height correction on the dose rate obtained by the detection, convert the air absorption dose rate of the radioactive source to the same horizontal height, draw a dose rate heat map, and intuitively monitor the radiation level of each location. Figure 1 shows a schematic diagram of the use of drone radiation monitoring system in actual applications. As shown in Figure 1, a sodium iodide detector (thallium excitation) is carried on the drone to form a drone radiation monitoring system. When the drone flies at different altitudes, it detects the dose rate of radioactive point sources on the ground in the air. However, since the flight altitude of the drone changes at this time, the dose rate detected at the same altitude is often inaccurate, so it needs to be corrected.

In order to solve the above problems, researchers at home and abroad have done relevant research work. For aviation radiation monitoring in point source mode, some researchers have used large NaI (Tl) detectors (the detection system consists of 18 4.2L NaI (Tl) crystals, with a total detection volume of 75.6L) to equip large aircraft to perform altitude corrections on the dose rate generated by ground point sources. The purpose is to find the altitude attenuation coefficient in the point source mode. The correction method used is calibrated according to the empirical formula provided by the IAEA. The disadvantages of this method are that the volume of the loaded NaI (Tl) detector is too large to be suitable for small drones, resulting in a limited operating environment; secondly, the conditions of the empirical formula used are relatively idealized, and the actual application scenarios are not considered, so the applicability is poor;

Secondly, some researchers have used the energy spectrum dose rate conversion function (GE function) to realize the conversion between the energy spectrum counts obtained by detection and the corresponding dose rate. This method is relatively cumbersome and requires many steps to implement, which makes it difficult to perform real-time online data processing in the UAV radiation monitoring system; therefore, according to the UAV radioactivity monitoring principle and actual application requirements, it is very necessary to design a radionuclide dose rate height correction scheme for point sources that can be monitored online in real time.

Summary of the invention

In view of the deficiencies in the prior art, the present invention provides a dose rate correction method based on an artificial neural network, which trains key data in the energy spectrum acquired by the detector by adopting a prediction model based on an artificial neural network, thereby achieving prediction.

The present invention provides a point source dose rate correction method based on an artificial neural network, comprising: S1, obtaining dose rate values corresponding to energy spectrum data of an unmanned aerial vehicle radioactivity monitoring system at different altitudes, respectively serving as input parameters and output parameters; S2, dividing part of the energy spectrum data at different altitudes into training data, and the other part into test data, using the input parameters and output parameters in the training data to construct an artificial neural network model, importing the training data into the artificial neural network model for training, and obtaining the trained artificial neural network model; S3, importing the test data into the trained artificial neural network model respectively, obtaining an ideal output result, and comparing the error between the ideal output result and the corresponding test data; if the error is greater than or equal to a set expected accuracy value, repeating S2 and S3; if the error is less than the set expected accuracy value, the trained and debugged artificial neural network model is a point source dose rate correction algorithm.

Furthermore, the method for obtaining the energy spectrum of the UAV radioactivity monitoring system at different altitudes is: using Monte Carlo software to simulate the energy spectrum of the UAV radioactivity monitoring system at different altitudes.

Furthermore, the method for obtaining the input parameters and output parameters in the above-mentioned energy spectrum data at different heights and constructing an artificial neural network model is as follows: normalizing the energy spectrum data at different heights; extracting the number of principal components that mainly contribute to the dose rate deposition after normalization as input parameters; the air absorption dose rate deposited by the radioactive source in the air is the output parameter; the hidden layer is calculated using the above-mentioned input parameters and the above-mentioned output parameters; the structure of the above-mentioned artificial neural network model is input layer-hidden layer-output layer, and the artificial neural network model is constructed using the structure of the above-mentioned artificial neural network model.

Furthermore, the method for extracting the number of principal components that make the main contribution to the dose rate deposition after normalization is: extracting data related to the air deposition dose rate in the energy spectrum data at different heights; obtaining standardized data after standardization of the above-mentioned related data; calculating the correlation coefficients between the above-mentioned standardized data and forming a correlation coefficient matrix; calculating each eigenvalue of the above-mentioned correlation coefficient matrix; using the above-mentioned eigenvalues to calculate the contribution rate and cumulative contribution rate of each of the above-mentioned related data, when the above-mentioned cumulative contribution rate is greater than the threshold, the corresponding minimum number of eigenvalues is used as the number of principal components that make the main contribution, which is the input parameter.

Furthermore, the data related to the air deposition dose rate include: total energy peak count, energy, single escape peak, double escape peak, annihilation peak and corresponding measurement time and height corresponding to each nuclide.

Furthermore, the method for normalizing the above-mentioned related data to obtain the above-mentioned standardized data is to use the following formula:

Among them, x _ij represents the jth index of the i-th energy spectrum; It represents the data after _xij is standardized; It represents the average value of all the above energy spectrum data; is the variance of the data, representing the data difference, and _sj is the standard deviation between the two data represented.

Furthermore, the method for calculating the correlation coefficient between the above standardized data is to use the following formula:

Among them, _rij is the correlation coefficient of the above energy spectrum data _xi and _xj ; _xi is the average value of the i-th energy spectrum data; _xj is the average value of the j-th energy spectrum data; n represents the number of energy spectra used for calculation, and _xkj represents the j-th index of the k-th energy spectrum.

Furthermore, the expected value of the above accuracy is 10 ^-3 .

Furthermore, the above method also includes: if the above error is less than the above expected value of accuracy, using an unmanned aerial vehicle radioactivity monitoring system in an actual scenario to measure the dose rate of the same above radiation source at different heights, and comparing it with the above actual output result to verify the quality of the above artificial neural network model.

The method disclosed in this application solves the problem that the dose verification work in the process of dose rate height correction in the existing point source mode takes a long time and has high manpower and material costs, and can improve the efficiency and quality of dose correction. The result is helpful to analyze the verification results and reduce the verification cost. Using the BPNN artificial neural network can effectively improve the efficiency of the entire dose rate height correction and realize the integrated correction of the nuclide dose rate height correction.

Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a schematic diagram of a UAV radiation monitoring system used in actual applications.

FIG2 is a flow chart of a point source dose rate correction method based on a neural artificial neural network.

Figure 3 is a schematic diagram of the Nal (TI) detector model.

DETAILED DESCRIPTION

The following is a more detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings and examples, so that the scheme of the present invention and its advantages in various aspects can be better understood. However, the specific embodiments and examples described below are only for the purpose of illustration, rather than for limiting the present invention.

The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and/or reference letters in different examples, and this repetition is for the purpose of simplicity and clarity, which itself does not indicate the relationship between the various embodiments and/or settings discussed. In addition, the present invention provides various specific examples of processes and materials, but those of ordinary skill in the art can be aware of the application of other processes and/or the use of other materials.

First of all, we need to explain the origin of BPNN (back propagation neural network algorithm). Artificial Neural Network (ANN) is a research hotspot in the field of artificial intelligence since the 1980s. In the past decade, the research on artificial neural networks has continued to deepen and has made great progress. The back-propagation algorithm (Back-Propagation) as the earliest and most commonly used multi-layer perceptron network training method was proposed as early as the 1970s and 1980s and has been used ever since. The back-propagation artificial neural network (BPNN) with the error back-propagation algorithm as the core is also widely used.

The topological structure of the artificial neural network (BP) model includes an input layer, a hidden layer, and an output layer. The neurons in the input layer are determined by the dimension of the sample attributes, the number of neurons in the output layer is determined by the number of sample classifications, and the number of hidden layers can be determined by the actual situation. In essence, the BP algorithm uses the squared network error objective function and the gradient descent method to calculate the minimum value of the objective function. The basic BP algorithm includes two processes: forward propagation of the signal and back propagation of the error. BPNN (back propagation neural network) has many advantages, such as strong nonlinear mapping capabilities, high self-learning and adaptive capabilities, and strong fault tolerance. Using BPNN can effectively improve the efficiency of the entire dose rate height correction and realize the integrated correction of the nuclide dose rate height correction.

In addition, the MCNP simulation software used in this application needs to be briefly introduced. MCNP software is a software that utilizes the Monte Carlo simulation method. The subject characteristics of nuclear technology lead to the inability to carry out actual measurement or experimental work in many cases, and computer simulation calculations have just shown unique advantages. At the same time, for nuclear reaction problems with complex structures and reaction mechanisms, general numerical methods are difficult to solve, and the MC (Monte Carlo) method can accurately simulate the physical process in practice and solve the problems that were difficult to solve with previous numerical methods. Therefore, this method is widely used in research related to nuclear fields. In the mid-1940s, with the development of atomic energy, the MC method gradually grew up. Its basic idea is a statistical sampling method based on random number selection. Since the traditional empirical method cannot approximate the real physical process and it is difficult to obtain satisfactory results, the Monte Carlo method has significant advantages in solving particle transport problems. Subsequently, people gradually developed a large number of MC simulation programs. The mainstream ones are EGS, MCNP, GEANT4, etc. Among them, MCNP was developed by the Los Alamos National Laboratory in the United States. It is mainly used for the Monte Carlo simulation program of photon, electron, neutron and other transport problems in materials in three-dimensional geometric structures. Its applicable photon energy range is 1E-3MeV-1E5MeV, electron energy range is 1E-3MeV-1E3MeV, and neutron energy range is 1E-11MeV-20MeV. MCNP has a complete program function, rich reaction cross-section data of various substances, various methods to reduce variance, strong versatility and simple use.

FIG2 shows a flow chart of a point source dose rate correction method based on an artificial neural network according to the present invention.

As shown in Figure 2, first obtain the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different altitudes, that is, simulated data. The present application adopts the simulation using MCNP software. Extract the energy spectrum data and dose rate values at different altitudes as input parameters and output parameters. The energy spectrum data at different altitudes described in part are divided into training data, and the other part is divided into test data. It should be noted that the test data must not be less than 50 groups, and the training data must be much larger than the number of test data. For example, in the present application embodiment, the test data is 50 groups and the training data is 450 groups.

An artificial neural network model, namely a BPNN algorithm model, is constructed using the input parameters and output parameters in the training data, and the training data are respectively imported into the artificial neural network model for training, and the trained artificial neural network model is obtained.

Then import the above test data into the trained artificial neural network model respectively, and get the ideal output result. Compare the ideal output at the same height with the corresponding test data. If the error is less than the expected value of accuracy, the artificial neural network training model is the required point source air dose rate correction algorithm. If the error is greater than or equal to the expected value of accuracy, the artificial neural network training model does not meet the requirements, and it is necessary to repeat the above steps to continue training the artificial neural network model. In addition, if the error is less than the expected value of accuracy, the energy spectrum data measured on site by the drone airborne radiation monitoring system is required to verify the correction algorithm.

Each step will be described in detail below. This application uses MCNP to simulate the radioactivity of ground point sources in the air at different altitudes. The first step is to select the appropriate UAV flight altitude range and the energy range of the nuclide, and select the appropriate airborne detection system.

In the second step, MCNP simulation is used to obtain the energy spectrum data of the UAV radioactivity monitoring system at different altitudes. The energy spectrum data is divided into test data and training data. The data related to the air deposition dose rate in the training data is extracted, and all the information data related to the air deposition dose rate are normalized, and the number of principal components that contribute to the dose rate deposition is selected as the input parameter. The air absorption dose rate deposited by the radioactive source in the air is used as the output parameter.

The third step is to calculate the number of hidden layer nodes according to the above input parameters and output parameters. Then, an artificial neural network model is constructed. The training data is respectively imported into the artificial neural network model constructed above to perform model training. Then, the trained artificial neural network model is obtained.

The method of constructing an artificial neural network model is as follows: the energy spectrum data at different heights are normalized. The number of principal components that contribute to the dose rate deposition after normalization is extracted as an input parameter. The air absorption dose rate deposited by the radioactive source in the air is the output parameter; the number of hidden layer nodes is calculated using the input parameters and the output parameters. The structure of the artificial neural network model is input parameters-number of hidden layer nodes-output parameters, and the artificial neural network model is constructed using the structure of the artificial neural network model.

The method for extracting the number of principal components that contribute mainly to the dose rate deposition after normalization is to extract data related to the air deposition dose rate from the energy spectrum data at different heights. Standardized data is obtained after the related data is standardized. The correlation coefficients between the standardized data are calculated and a correlation coefficient matrix is formed. Each eigenvalue of the correlation coefficient matrix is calculated; the contribution rate and cumulative contribution rate of each of the related data are calculated using each eigenvalue, and the minimum number of eigenvalues corresponding to the cumulative contribution rate exceeds the threshold is determined as the number of principal components that contribute mainly.

The data related to the air deposition dose rate include the total energy peak count, energy, single escape peak, double escape peak, annihilation peak and the corresponding measurement time and height of each nuclide.

The method of performing standardization processing on the relevant data to obtain the standardized data is to use the following formula:

Among them, x _ij represents the jth index of the i-th energy spectrum; It represents the data after _xij is standardized; It represents the average value of all the energy spectrum data; is the variance of the data, representing the data variability, _sj is the standard deviation between the two data represented, and n represents the number of energy spectra used for calculation.

The mean and standard deviation can be calculated according to the following formula:

The meaning of each term in the formula is the same as above.

The method for calculating the correlation coefficient between the standardized data is to use the following formula:

Among them, _rij is the correlation coefficient of the energy spectrum data _xi and _xj ; _xi is the average value of the i-th energy spectrum data; _xj is the average value of the j-th energy spectrum data; n represents the number of energy spectra used for calculation, and _xkj represents the j-th index of the k-th energy spectrum.

The correlation coefficients are combined into a correlation coefficient matrix R

Calculate the correlation coefficient matrix and sort the obtained eigenvalues and eigenvectors, that is, solve the characteristic equation:

|λI-R|＝0

Where I is the unit matrix of the corresponding dimension. The above formula is used to obtain the eigenvalues λ ₁ , λ ₂ , λ ₃ , ..., λ ₁₇ and their corresponding eigenvectors, and they are sorted by size.

Next, the contribution rate and cumulative contribution value of each component are calculated to screen the principal components. The calculation formula for the contribution rate α _l of the lth principal component is:

The calculation formula for the cumulative contribution rate G _k of k principal components is:

When the cumulative contribution rate is greater than 80%, the corresponding minimum k value, that is, the number of principal components, is also the input information.

The fourth step is to test the trained artificial neural network model. The test data are imported into the trained artificial neural network model to obtain the ideal output result, and the error between it and the corresponding test data is compared. If the error is less than the set expected accuracy value, the trained network model is the required correction method.

The method for comparing the error between the ideal output result and the corresponding test data is based on the following formula:

Wherein, _mi represents the ideal output result obtained by the trained artificial neural network model, o _i represents the actual output result obtained by the trained artificial neural network model, and n represents the number of samples.

According to experience, the expected value of model accuracy is generally set at 10 ^-3 .

In addition, part of the input layer, output layer, and hidden layer information extracted from the MCNP software can be used to train the artificial neural network model, and the other part can be used to test the recognition ability of the BN network model before training the network model.

In addition, as shown in Figure 2, the trained artificial neural network is used to verify the data samples of the radioactive source at a certain distance whose mass is measured using the UAV radioactivity monitoring system to verify the training quality of the artificial neural network model.

Example

For ease of understanding, the present application discloses in detail Example 1 in which the present method is applied for correction.

Step 1: Select a suitable airborne detection system. The present invention uses a NaI (T1) detector system, which is composed of a 2×2 inch NaI (T1) detector, a multi-channel amplitude analyzer and a matching MAESTRO software. The detector energy range is 30keV to 3MeV, the energy resolution is 7.7% (Cs-137, 662KeV), and the multi-channel amplitude analyzer is 1024 channels. It can be seen from the detection principle of the NaI (T1) detector that the gamma particles and the detector mainly interact in the crystal part. Figure 3 is a model diagram of the NaI (T1) detector, and its crystal geometric dimensions and surrounding coating material structure are shown in the figure, wherein the aluminum shell thickness is 2.5mm, the glass material is _SiO2 , the thickness is 2mm, and the reflective layer material is MgO, the thickness is 0.5mm. The model used in the MCNP simulation process is also modeled based on the above data.

Step 2: According to the method provided by the present invention, the simulated selected UAV flight altitude range is 0 meters to 100 meters, and a height value is taken at an interval of 0.2 meters between 0-100 meters. A total of 500 sets of altitude data are selected, and 500 sets of simulation data can be obtained accordingly; the corresponding energy information is expressed as the energy of the incident photon, which is set to represent the energy values of different nuclides, including: Am-241 (59.5KeV), Cs-137 (662KeV), Co-60 (1173KeV and 1332KeV), K-40 (1460KeV).

Step 3: According to the method provided by the present invention, the detector model selected for MCNP simulation: the point source is at 0 meters on the ground, and the detector of the UAV radioactivity monitoring system is directly above the point source, and the distance takes the above distance interval (as shown in Figure 1). The simulated source term is isotropic, and the emission weight is 1. Use the corresponding surface card and gate element card to establish the model, and use the F8 card to obtain the energy spectrum of the sodium iodide crystal detector. According to the method provided by the present invention, in the MCNP simulation process, the authenticity of the simulated energy spectrum is related to the number of emitted particles, and the number of emitted particles can be set to 1×10 ¹¹ for simulation.

Step 4: Use 450 sets of data from the 500 sets of data obtained by simulation as training data to train the BP artificial neural network, and use the remaining 50 sets of data as test data to test the recognition ability of the BP artificial neural network.

Step 5: According to the method provided by the present invention, the energy spectrum obtained by MCNP is first normalized, and then the energy spectrum features are extracted to obtain the total energy peak count and energy corresponding to the Am-241 nuclide in the energy spectrum; the total energy peak count and energy corresponding to the Cs-137 nuclide; the total energy peak count and energy corresponding to the 1173KeV peak of Co-60; the total energy peak count, energy, single escape peak and double escape peak corresponding to the 1332KeV peak of Co-60; the total energy peak count, energy, single escape peak and double escape peak corresponding to the 1460KeV peak of K-40; plus the annihilation peak in the energy spectrum, the corresponding measurement time and height, a total of 17 data related to the air deposition dose rate of the nuclide. Due to the large amount of energy spectrum information, the extracted energy spectrum data is multidimensional data, and the extracted data needs to be reduced in dimension to select the data with the largest contribution to the corresponding dose rate.

Step 6: According to the method provided by the present invention, 450 sets of data are selected as training sets, and then the 17 data selected in step 5 are used as variables to construct a 450×17 data matrix X.

Wherein, i=1, 2, 3, ..., 450, j=1, 2, 3, ..., 17. i represents 450 data of the training set, and j represents 17 energy spectrum eigenvalue data extracted.

Step 7: Process the data from step 6 to obtain a standardized matrix. The purpose is to eliminate the differences in dimensionality and magnitude between the data and obtain standardized data. The data in the matrix It can be calculated by the following formula:

Where x _ij represents the jth index of the i-th energy spectrum; It represents the data after _xij is standardized; It represents the average value of all data; is the variance of the data, representing the data difference, and _sj is the standard deviation between the two data represented.

In the above formula, n represents the number of energy spectra used for calculation, n=450.

In order to obtain the contribution value of each component, the matrix R (correlation coefficient matrix) is calculated:

The calculation formula of r _ij is:

In the above formula, _rij is the correlation coefficient between the original variables _xi and _xj ; _xi is the i-th energy spectrum; _xj is the j-th energy spectrum eigenvalue; n represents the number of energy spectra used for calculation.

Calculate the eigenvalues and eigenvectors obtained from the correlation coefficient matrix and sort them, that is, solve the characteristic equation:

|λI-R|＝0

Where I is the unit matrix of the corresponding dimension. The above formula is used to obtain the eigenvalues λ ₁ , λ ₂ , λ ₃ , ..., λ ₁₇ and their corresponding eigenvectors, and they are sorted by size.

Next, the contribution rate and cumulative contribution value of each component are calculated to screen the principal components. The calculation formula for the contribution rate α _l of the lth principal component is:

The calculation formula for the cumulative contribution rate G _k of k principal components is:

It can be calculated from the above formula and the embodiment provided by the present invention that when the number of the main components is equal to 12, the cumulative contribution rate is greater than 80%. It can be considered that these twelve data are the main components of this embodiment.

The 12 main components are the energies of the rays emitted by the four radiation sources Am-241 (59.5KeV), Cs-137 (662KeV), Co-60 (1173KeV and 1332KeV), and K-40 (1460KeV) in the simulation, the counts of the full-energy peaks corresponding to the five energies in the energy spectrum, and the time and height during measurement.

Step 8: Apply the 12 characteristic parameters obtained in step 7 to the input parameters of the artificial neural network. At the same time, the output parameters are the deposition air absorption dose rate of each nuclide in the air, which is a total of 4 output parameters. Therefore, the number of input nodes used is N = 12, and the number of output nodes is M = 4. The number of hidden layer nodes Q can be determined by the empirical formula:

The optimal number of nodes selected by this method is 10. Therefore, the structure of the BPNN in the present invention is 12-10-4.

Step 9: Import the 450 sets of training data after random grouping into the artificial neural network model to train the model;

Step 10: Import the test data into the trained artificial neural network model and compare the error MSE between the output results obtained by the model and the corresponding test data: calculated according to formula 1, where n = 450;

Step 11: Compare the test error MSE obtained from training with the expected value of model accuracy. If the MSE is less than 10 ^-3 , the model training is considered complete. If the MSE is greater than the expected value of accuracy, the network parameters need to be modified until the MSE is less than the expected value of accuracy and the training is completed.

Step 12: According to the method provided by the present invention, the trained artificial neural network model is finally used to verify the data samples of three artificial radiation sources, Am-241, Cs-137, and Co-60, measured at 1 meter using the drone radioactivity monitoring system to verify the quality of network training.

Obviously, the above embodiments are merely examples for clearly explaining the present invention, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims (7)

1. A point source dose rate correction method based on artificial neural network, characterized by comprising: S1, obtaining the energy spectrum data and corresponding dose rate values of the UAV radioactivity monitoring system at different altitudes as input parameters and output parameters respectively; S2, dividing part of the energy spectrum data at different heights into training data, and dividing the other part into test data, using the input parameters and output parameters in the training data to build an artificial neural network model, importing the training data into the artificial neural network model for training, and obtaining a trained artificial neural network model; The method of obtaining input parameters and output parameters in energy spectrum data at different heights and constructing an artificial neural network model is as follows: a, normalize the energy spectrum data at different heights; b. Extracting the number of principal components that contribute mainly to the dose rate deposition after normalization as an input parameter, the method is as follows: extracting data related to the air deposition dose rate from the energy spectrum data at different heights; obtaining standardized data after normalization of the related data; calculating the correlation coefficients between the standardized data and forming a correlation coefficient matrix; calculating each eigenvalue of the correlation coefficient matrix; using the eigenvalues to calculate the contribution rate and cumulative contribution rate of each of the related data, when the cumulative contribution rate is greater than a threshold, the corresponding minimum number of eigenvalues is used as the number of principal components that contribute mainly, which is the input parameter; c. The air absorbed dose rate deposited by the radiation source in the air is the output parameter; d. calculating a hidden layer using the input parameters and the output parameters; e. The structure of the artificial neural network model is input parameter-hidden layer-output parameter, and the artificial neural network model is constructed using the structure of the artificial neural network model; S3, importing the test data into the trained artificial neural network model respectively to obtain an ideal output result, and comparing the error between the ideal output result and the corresponding test data; If the error is greater than or equal to the set expected value of accuracy, S2 and S3 are repeated. If the error is less than the set expected value of accuracy, the trained and debugged artificial neural network model is a point source dose rate correction algorithm. 2. The point source dose rate correction method based on artificial neural network as described in claim 1 is characterized in that the method for obtaining the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different altitudes is: using Monte Carlo software to simulate the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different altitudes. 3. The point source dose rate correction method based on artificial neural network according to claim 2, characterized in that the data related to the air deposition dose rate includes: The total energy peak count, energy, single escape peak, double escape peak, annihilation peak and corresponding measurement time and height of each nuclide. 4. The point source dose rate correction method based on artificial neural network according to claim 3 is characterized in that the method of normalizing the relevant data to obtain the standardized data is to use the following formula: Among them, x _ij represents the jth index of the i-th energy spectrum; It represents the data after _xij is standardized; represents the average value of all the energy spectrum data; _sj represents the standard deviation between the two data. 5. The point source dose rate correction method based on artificial neural network according to claim 4, characterized in that the method for calculating the correlation coefficient between the standardized data is to use the following formula: Wherein, r _ij is the correlation coefficient between the energy spectrum data x _i and x _j ; is the average value of the i-th energy spectrum data; is the average value of the j-th energy spectrum data; n represents the number of energy spectra used for calculation, x _kj represents the j-th index of the k-th energy spectrum, and x _ki represents the ith index of the k-th energy spectrum. 6 . The point source dose rate correction method based on artificial neural network according to claim 5 , wherein the expected accuracy value is 10 ⁻³ . 7. The point source dose rate correction method based on artificial neural network according to claim 1, characterized in that it also includes: If the error is less than the expected value of accuracy, the dose rate of the same radiation source at different altitudes is measured using a UAV radioactivity monitoring system in an actual scenario and compared with the actual output results to verify the quality of the artificial neural network model.