A multi-radioactive source aerial intelligent search positioning method

Abstract

The application provides a multi-radiation source aerial intelligent search positioning method, which comprises the following steps: 1, constructing a ground radioactivity distribution map based on known radiation sources; 2, training a dose rate distribution map at 1m of the ground of the radiation source based on a convolutional neural network; 3, performing full-coverage search monitoring on a region to be searched; 4, generating a dose rate distribution map at 1m of the ground of the region to be searched; 5, analyzing the dose rate distribution map based on the convolutional neural network; and 6, accurately determining the position of the radiation source. When the types and source strengths of the multi-radiation sources are unknown, the method comprises the following steps: A, performing full-coverage search monitoring on the region to be searched; B, determining and optimizing a monitoring region based on the dose rate distribution map; C, automatically searching based on the optimized monitoring region; and D, determining the specific position of the radiation source based on a parameter estimation method. The application can accurately determine the aerial search positioning of the multi-radiation source and maximize the search efficiency.

Description

Technical Field

[0001] This invention belongs to the field of nuclear safety technology, specifically a multi-radioactive source search and location method based on intelligent optimization algorithms. Background Technology

[0002] With the rapid development of nuclear and radiation technology, radioactive sources are increasingly widely used in industry, agriculture, and scientific research. Various types of radioactive sources are applied in medical diagnosis, industrial flaw detection, container inspection, irradiation breeding, food preservation, and other related industries. As the number of radioactive sources used in various industries increases, although awareness of nuclear safety and radioactive source management systems have developed significantly, the number of uncontrolled radioactive sources may continue to rise. Once an uncontrolled radioactive source becomes uncontrolled, it will pose a significant threat and harm to public safety. The loss, discovery, and recovery of radioactive sources are often extremely difficult. Due to the large potential areas involved and the vast search area, traditional carpet searches using portable monitoring instruments may take several days or even longer, resulting in extremely low efficiency. Lost radioactive sources are highly dangerous, and the longer the search lasts, the greater the radiation hazard they pose. Therefore, rapid and efficient searches should be conducted using unmanned aerial radiation monitoring platforms. Under the constraints of meeting radioactive source search requirements and the specifications of aerial emergency monitoring equipment, intelligent optimization algorithms should be used to scientifically plan radioactive source search routes. Summary of the Invention

[0003] The purpose of this invention is to provide a rapid, intelligent, and efficient search method for unmanned aerial radiation monitoring platforms after a radioactive source is lost. These platforms are equipped with radiation dose rate detectors, navigation and positioning systems, intelligent computing systems, and communication systems. The intelligent search method enables rapid, efficient, and automatic search for radioactive sources. This invention proposes a mission-planning-based aerial radioactive source search method. Based on an estimation of the search area, considering both known and unknown radionuclide types and source strengths, and taking into full account relevant constraints, it constructs relevant optimization algorithms and models to optimize the path for radioactive source search by the unmanned aerial radiation monitoring platform. This maximizes the platform's support capabilities, shortens mission time, and achieves maximum search efficiency.

[0004] This invention relates to a method for searching and locating radioactive sources based on an unmanned aerial radiation monitoring platform and considering two factors: the type of radionuclide and the source intensity of known and unknown radioactive sources.

[0005] Assuming the types of nuclides and source intensities of multiple radioactive sources are known, the specific steps for searching and locating these sources are as follows:

[0006] Step 1: Construct a surface radioactivity distribution map based on known radioactive sources.

[0007] Given a radionuclide, according to the fundamental theory of gamma radiation, the radiation rate it produces at a given distance in air can be expressed as:

[0008]

[0009] In the formula: Γ is the radiation rate at a certain point in the air; A is the activity of the point source; r is the distance between the measurement point of the airborne nuclear radiation detector and the radioactive source; μ is the linear attenuation coefficient of the air; Γ is the γ radiation rate constant, representing the radiation rate per unit activity of the radioactive source at a unit distance.

[0010] Based on the above formula for radioactivity distribution, and combined with parameters such as the type and activity of each radionuclide, and the height of the nuclear radiation detector, a preliminary surface radioactivity distribution map of each radionuclide can be generated.

[0011] In air, under the same conditions, the relationship between absorbed dose rate and exposure rate can be expressed as:

[0012]

[0013] In the formula: The absorbed dose rate at a point in the air is expressed in μGy / h. The radiation rate at a certain point in the air is expressed in R / h.

[0014] Given the types of radionuclides and their activity distribution, a dose rate distribution map at 1m above the Earth's surface can be obtained.

[0015] Step 2: Train a convolutional neural network to generate a dose rate distribution map at 1m from the surface of the radiation source.

[0016] For radioactive sources with different nuclides and activities, various surface dose rate distribution maps are generated based on the formula in step one. Then, based on the intelligent computing system of the aerial monitoring platform, the generated surface dose rate distribution maps of each radioactive source are input into a two-dimensional convolutional neural network (2D-CNN) through three channels according to RGB features for network training. The 2D-CNN includes three structures: convolutional layers, pooling layers, and fully connected layers. Convolutional layers extract low-level features such as edges, lines, and corners, while higher-level convolutional layers extract higher-level features. The convolutional kernel slides sequentially across the input image from left to right and from top to bottom. Each slide performs a dot product calculation with the input image corresponding to the sliding window position, obtaining a numerical value, thus preserving the spatial features of the input image. Pooling layers are a method for image compression, mainly to accelerate the convergence process of the neural network and improve the stability during training. In the CNN structure, after multiple convolutional layers and sampling layers, one or more fully connected layers are connected. Each neuron in the fully connected layer is fully connected to all neurons in the previous layer. Fully connected layers can integrate local information with class discriminative power from convolutional or sampling layers. Features (color, edges, lines, texture) are extracted from 2D-CNN layers, followed by weighted pooling layers. Pooling layers are used to distribute the deep network to avoid overfitting. After processing by fully connected layers, the output layer provides dose rate distribution maps at 1m from the ground surface for each radiation source, enabling the neural network to recognize the dose rate distribution maps at 1m from the ground surface for the corresponding radiation source.

[0017] Step 3: Conduct a full-coverage search and monitoring of the area to be searched.

[0018] The area to be searched is topologically divided into a grid, and the search begins from the boundary of the area using an airborne radiation monitoring platform. Initially, a coarse-grid airborne nuclear radiation monitoring path is used, based on a "Z" pattern, with the platform maintaining a relatively high flight speed at an altitude of 50 meters. During the monitoring phase, when the dose rate detected by the airborne radiation monitoring platform exceeds a set threshold (0.5 μGy / h), the platform reduces its flight speed to below 10 m / s and monitors at half (or one-third) of its original flight spacing. When the detected value falls below the set threshold, the original flight speed and spacing are restored, repeating this cycle until the entire area to be searched is monitored. During monitoring, the airborne radiation monitoring platform can calculate the distance from any monitoring point to a specific radiation source using the formula from step one. Combined with the platform's flight altitude, the dose rate at the equivalent of 1 meter above the Earth's surface at that monitoring point can be calculated.

[0019] Step 4: Generate a dose rate distribution map at 1m above the ground surface of the area to be searched.

[0020] When an airborne radiation monitoring platform performs regional flight monitoring, the nuclear radiation detectors it carries typically update the ground dose rate monitoring data every certain period of time (usually once per second). The data displays the longitude, latitude, altitude, and dose rate value of the current monitoring point.

[0021] Once the airborne radiation monitoring platform has completed monitoring of the area to be searched, it generates the locations of each monitoring point along the monitoring flight path and the corresponding dose rates at an equivalent depth of 1 meter on the ground surface. Combined with a GIS map, it initially generates a dose rate distribution map at 1 meter on the ground surface of the area to be searched. To make the dose rate distribution map more accurate, especially near the radiation source, linear interpolation can be used to generate the dose rates at points between the two monitoring lines, thus generating the dose rate distribution map at 1 meter on the ground surface of the area to be searched. Given the locations (x1, y1) and (x2, y2) of corresponding points along the two monitoring flight paths, and the dose rate values ​​d1 and d2, the linear interpolation formula can be expressed as:

[0022]

[0023] In the formula: The dose rate value at any point along either of the two monitoring routes. The location of any point between the two monitoring routes.

[0024] Step 5: Analyze the dose rate situation map based on a convolutional neural network.

[0025] The dose rate distribution maps at 1m below the surface of suspected radioactive source areas are input into a trained convolutional neural network based on features such as color, edge, line, and texture. This network combines these features with those of known radioactive sources to automatically analyze and compare the dose rate distribution maps of suspected radioactive source areas, preliminarily determining the type of each radioactive source. Furthermore, based on known GIS maps and the location of the dose rate distribution maps at 1m below the surface, the location of each radioactive source can also be preliminarily determined.

[0026] Step Six: Precisely determine the location of the radiation source

[0027] At each of the initially determined radioactive source locations (x', y'), the airborne radiation monitoring platform hovers for 30 seconds to accurately measure the dose rate value D1 at 1m above the ground at that point; then, with (x', y') as the center point, a "U"-shaped monitoring pattern with a side length of 50 meters is conducted, and the platform hovers for 30 seconds at each of the four vertices of the "U"-shaped pattern.

[0028] In a point source radiation field, assuming that air attenuation is ignored and the background count rate is constant, the count rate of an airborne nuclear radiation detector can be determined according to the following formula.

[0029]

[0030] In the formula, (x i ,y i (x0, y0) represents the location of the airborne nuclear radiation detector, (x0, y0) represents the location of the radioactive source, M represents the count rate of the airborne nuclear radiation detector at a certain point in the air, A represents the activity of the radioactive source, ε represents the efficiency of the nuclear radiation detector, and r i denoted as b, where b is the distance between the airborne nuclear radiation detector and the radiation source; b is the background count rate; and h is the monitoring altitude of the airborne nuclear radiation detector.

[0031] Since the detector count at the measurement point is proportional to the intensity of the radiation source, the following set of equations can be formed based on the data from the four monitoring points.

[0032]

[0033] Solving the above system of equations yields multiple solutions for the radioactive source: P1 = (x 01 ,y 01 ),P2=(x 02 ,y 02 ),P3=(x 03 ,y 03 ), P4 = (x 04 ,y 04 This allows us to obtain the accurate location of the radiation source.

[0034] x0=(x 01 +x 02 +x 03 +x 04 ) / 4

[0035] y0=(y 01 +yx 02 +y 03 +y 04 ) / 4

[0036] Assuming the types of nuclides and source strengths of multiple radioactive sources are unknown, the specific steps for searching and locating these sources are as follows:

[0037] Step 1: Conduct a full-coverage search and monitoring of the area to be searched.

[0038] The area to be searched is topologically divided into a grid, and the search begins from the boundary of the area using an airborne radiation monitoring platform. Initially, the monitoring path uses a coarse grid (two parallel lines are kept relatively far apart, e.g., 300 meters) in a "Z" shape for airborne nuclear radiation monitoring. During monitoring, the platform maintains a relatively high flight speed (e.g., 50 km / h) at an altitude of 50 meters. During the monitoring phase, when the dose rate detected by the airborne radiation monitoring platform exceeds the set threshold (0.2 μGy / h), the platform reduces its flight speed to below 10 m / s and monitors at half (or one-third) of the original flight spacing. When the detected value falls below the set threshold, the original flight speed and spacing are restored, and this cycle repeats until the entire search area is monitored.

[0039] Step 2: Determine and optimize the monitoring area based on the dose rate distribution map

[0040] Based on the dose rate values ​​obtained from the airborne radiation monitoring platform, an interpolation method is used to generate a dose rate distribution map at 1m above the ground for each radiation source in the search area. Based on the dose rate distribution map, the approximate location of each radiation source can be preliminarily determined. Within the preliminarily determined radiation source areas, areas exceeding a set threshold are rasterized, with each cell's length and width set to 20 meters.

[0041] Step 3: Perform automatic search based on optimized monitoring area

[0042] In each optimized monitoring area, monitoring begins from the first cell (1,1), and an automatic search and monitoring mode is used to search for and monitor cells one by one starting from a flight altitude of 20 meters. That is, the aerial monitoring platform hovers at the middle position of each cell for 30 seconds to take a measurement. When monitoring the cells in the first column one by one, if the dose rate of a cell changes from high to low, the system returns from the current cell to the previous cell (x', 1). Then, monitoring is carried out one by one from the cells in the corresponding row. Similarly, if the dose rate of a cell in that row changes from high to low, the system returns from the current cell to the previous cell in that row (x', y').

[0043] Step 4: Determine the specific location of the radioactive source based on parameter estimation method

[0044] Using the center point of cell (x', y') as the reference point, a cross-shaped monitoring pattern is established. Monitoring is performed every 5 meters for 30 seconds along the horizontal and vertical lines of the cross, resulting in a total of 8 monitoring points. From these points, the two with the highest dose rate values ​​are selected. Then, using parameter estimation, the precise location of the radiation source can be automatically determined. The specific formula is as follows:

[0045]

[0046] In the formula: D idenoted as , where is the dose rate (uGy / h) of the airborne radiation monitoring platform; 'a' is the background radiation constant; 'c' is a constant determined by the radiation source; '(x0, y0)' is the location of the radiation source; 'r' is the radiation source. i This represents the distance between the airborne nuclear radiation detector and the radiation source.

[0047] The advantages and beneficial effects of this invention are as follows: The method proposed in this invention can accurately determine the location of multiple radioactive sources in the air. This invention can optimize the path for radioactive source searching by unmanned aerial radiation monitoring platforms, maximize the support effectiveness of unmanned aerial radiation monitoring platforms, shorten mission time, maximize search efficiency, and meet the needs of rapid search and location of aerial radioactive sources. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly described below. The accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart of a known method for intelligent aerial search and location of multiple radioactive sources.

[0050] Figure 2 This is a flowchart of an intelligent aerial search and location method for unknown multiple radioactive sources.

[0051] Figure 3 This is a schematic diagram illustrating the estimated surface radioactivity distribution of multiple radioactive sources.

[0052] Figure 4 This is a schematic diagram illustrating the dose rate distribution estimation at 1m from the surface of a multi-radiosource area.

[0053] Figure 5 This is a schematic diagram of a full-coverage search and monitoring system for the area to be searched.

[0054] Figure 6 This is a schematic diagram of the dose rate distribution at 1m from the surface of each radioactive source in the area to be searched.

[0055] Figure 7 This is a schematic diagram of the automatic search function for regional optimization monitoring.

[0056] Figure 8 This is a schematic diagram of a cross-shaped monitoring system. Detailed Implementation

[0057] The implementation data of this invention comes from the comprehensive application practice of unmanned aerial radiation monitoring platforms. Based on the actual needs of the platform in performing aerial radioactive source search missions, two scenarios are distinguished: known and unknown radionuclide types and source intensities. For cases where multiple radionuclide types and source intensities are known, a surface radioactivity distribution map is constructed based on the known radioactive sources. Based on the generated distribution map, a dose rate distribution map at 1m from the surface of the radioactive sources is trained using a convolutional neural network. The unmanned aerial radiation monitoring platform performs a full-coverage search and monitoring of the search area, generating a dose rate distribution map at 1m from the surface of the search area. The dose rate distribution maps are compared and analyzed using a convolutional neural network. Based on the count rate calculation formula, local "loop"-shaped monitoring is used to accurately determine the location of the radioactive sources.

[0058] For cases where the types and intensities of multiple radioactive sources are unknown, a full-coverage search and monitoring of the area to be searched is performed based on a "Z"-shaped path; the optimized monitoring area is determined based on the dose rate distribution map; the optimized monitoring area is automatically searched; and the specific location of the radioactive source is determined based on parameter estimation methods.

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be systematically described below with reference to the accompanying drawings. The embodiments described in this invention are only some embodiments of the present invention, and not all embodiments.

[0060] Figure 1 , Figure 2 This is a flowchart of the multi-radioactive source aerial intelligent search method of the present invention. The following uses an unmanned aerial radiation monitoring platform as an example to illustrate the implementation steps. Assuming the nuclide types and source strengths of the multiple radioactive sources are known, the specific steps for searching and locating the multiple radioactive sources are as follows:

[0061] Step 1: Initially generate the radioactive plume diffusion pattern based on the diffusion model.

[0062] Assuming the lost radioactive source is 60 Co、 137 Cs, the activity of the point source is 2.2 × 10 13 Bq, 2.1×10 13 Bq, the radiation rate constant is 8.9 × 10 -2 C·m 2 ·kg -1 8.9×10 -2 C·m 2 ·kg -1 The linear attenuation coefficients of the nuclides are 0.69 and 0.7. Based on the fundamental theory of gamma radiation, the radiation rates produced by the two radioactive sources at different distances in the air can be obtained. The surface radioactivity distribution of the two radioactive sources is as follows: Figure 3 As shown.

[0063] In air, under the same conditions, the relationship between absorbed dose rate and exposure rate can be expressed as:

[0064]

[0065] Given the types of radionuclides and their activity distribution, their dose rate distribution at 1m above the Earth's surface can be obtained, such as... Figure 4 As shown.

[0066] Step 2: Train a convolutional neural network to generate a dose rate distribution map at 1m from the surface of the radiation source.

[0067] For the generated radioactive source 60 Co、 137 The dose rate distribution map at 1m above the ground surface of Cs was obtained using an intelligent computing system based on an aerial monitoring platform. The system trained a two-dimensional convolutional neural network by inputting data from different channels based on RGB color and features such as edges, lines, and textures. After training with convolutional layers, pooling layers, and fully connected layers, the neural network was able to identify radiation sources. 60 Co、 137 The ability to map the dose rate distribution at 1m above the ground surface using Cs.

[0068] Step 3: Conduct a full-coverage search and monitoring of the area to be searched.

[0069] The area to be searched is topologically divided into a grid, and the search begins from the boundary of the area using an airborne radiation monitoring platform. Initially, the monitoring path is a coarse grid (two monitoring lines spaced 300 meters apart) using a "Z" shaped airborne nuclear radiation monitoring system. During monitoring, the platform maintains a relatively high flight speed at an altitude of 50 meters. In the monitoring phase, when the dose rate detected by the airborne radiation monitoring platform exceeds a set threshold (0.5 μGy / h), the platform reduces its flight speed to below 10 m / s and maintains a flight spacing of 100 meters. When the detected dose rate falls below the set threshold (0.5 μGy / h), the original flight speed and spacing are restored, repeating this cycle until the entire search area has been monitored. Figure 5 As shown. During the monitoring process, the airborne radiation monitoring platform can calculate the distance from any monitoring point to a certain radiation source according to the formula in step one. Combined with the flight altitude of the monitoring platform, the dose rate value at the equivalent of 1 meter above the ground at that monitoring point can be calculated.

[0070] Step 4: Generate a dose rate distribution map at 1m above the ground surface of the area to be searched.

[0071] Once the airborne radiation monitoring platform has completed monitoring of the area to be searched, it will generate the location of each monitoring point along the monitoring route and the corresponding dose rate at 1m of the equivalent ground surface, and combine it with the GIS map to initially generate a dose rate distribution map at 1m of the ground surface in the area to be searched.

[0072] A linear interpolation method is used to generate the dose rate at each point between the two monitoring lines, thereby generating a dose rate distribution map at 1m above the ground surface of the area to be searched, as shown below. Figure 6 As shown. Given the positions (x1, y1) and (x2, y2) of corresponding points on two monitoring routes, and the dose rate values ​​d1 and d2, the linear interpolation formula can be expressed as:

[0073]

[0074] Step 5: Analyze the dose rate distribution map using a convolutional neural network.

[0075] The dose rate distribution maps at 1m below the surface of suspected radioactive source areas are input into a trained convolutional neural network based on features such as color, edge, line, and texture. This network combines these features with those of known radioactive sources to automatically analyze and compare the dose rate distribution maps of suspected radioactive source areas, preliminarily determining the type of each radioactive source. Furthermore, based on known GIS maps and the location of the dose rate distribution maps at 1m below the surface, the location of each radioactive source can also be preliminarily determined.

[0076] Step Six: Precisely determine the location of the radiation source

[0077] Assuming the initially identified radioactive source 60 The coordinates of Co are (x', y'). The airborne radiation monitoring platform hovers at this location for 30 seconds to accurately determine the dose rate value D1 at 1m above the ground. Then, a "U"-shaped monitoring with a side length of 50 meters is carried out with (x', y') as the center point, and the platform hovers at the four vertices of the "U" shape for 30 seconds. 60 The activity of the Co source is 2.2 × 10⁻⁶. 13 Bq, assuming negligible air attenuation, a background count rate of 100 CPS, and a nuclear radiation detector efficiency of 1, the coordinates of the four hovering monitoring points in the "U" shape are (28500, 30450), (28500, 30500), (28450, 30500), and (28450, 30450), respectively, with count rates of 9800 CPS, 9650 CPS, 9500 CPS, and 9700 CPS. The monitoring height of the airborne nuclear radiation detector is 50 meters. Substitute these values ​​into the following system of equations.

[0078]

[0079] Multiple coordinate positions of the radioactive source can be obtained (28479, 30470), (28477, 30469), (28480, 30471), and (28481, 30474), and then the accurate position of the radioactive source (28479, 30471) can be obtained.

[0080] Assuming the types of nuclides and source strengths of multiple radioactive sources are unknown, the specific steps for searching and locating these sources are as follows:

[0081] Step 1: Conduct a full-coverage search and monitoring of the area to be searched.

[0082] The area to be searched is topologically divided into a grid, and the search begins from the boundary of the area using an airborne radiation monitoring platform. Initially, the monitoring path uses a coarse grid (two monitoring lines spaced 300 meters apart) in a "Z" shape for airborne nuclear radiation monitoring. During monitoring, the platform conducts comprehensive monitoring of the area to be searched at a flight altitude of 50 meters and a flight speed of 50 kilometers per hour (for example). During monitoring, when the airborne radiation monitoring platform detects a dose rate value at 1 meter below the ground surface exceeding a set threshold, the platform's flight speed is reduced to below 10 meters per second, and monitoring is conducted at a flight interval of 100 meters. When the dose rate value at 1 meter below the ground surface falls below the set threshold again, the initial flight speed and flight interval are restored, repeating this cycle until the entire area to be searched is monitored.

[0083] Step 2: Determine and optimize the monitoring area based on the dose rate distribution map

[0084] Based on the dose rate values ​​at various locations monitored by the airborne radiation monitoring platform, an interpolation method was used to generate a dose rate distribution map at 1m above the ground surface of the area to be searched. Based on the surface dose rate distribution map, the approximate locations of each radiation source can be preliminarily determined. Within the preliminarily determined radiation source areas, areas exceeding a set threshold of 1 μGy / h were rasterized, with each cell's length and width set to 20 meters.

[0085] Step 3: Perform automatic search based on optimized monitoring area

[0086] In each optimized monitoring area, monitoring begins from the first cell (1,1), using an automatic search and monitoring mode starting at a flight altitude of 20 meters. Specifically, the aerial monitoring platform hovers for 30 seconds at the center of each cell and takes a measurement. While monitoring the cells in the first column, if the dose rate of the 5th cell changes from high to low, the system returns to the previous cell (4,1). Then, monitoring begins from the cells in the 4th row corresponding to that cell. Similarly, if the dose rate of the 6th cell in that row changes from high to low, the system returns to the previous cell in that row (4,5), and so on. Figure 7 As shown.

[0087] Step 4: Determine the specific location of the radioactive source based on parameter estimation method

[0088] Using the center point of cell (4,5) as the center, perform a cross-shaped monitoring, such as... Figure 8 As shown. Monitoring was conducted every 5 meters for 30 seconds along the two horizontal and vertical lines of the "+" shape, resulting in a total of 8 monitoring points. From these points, the two with the highest dose rate values ​​(-5, 0, 12 cGy / h) and (0, 5, 13 cGy / h) were selected. Then, using the parameter estimation method, the exact location of the radiation source was determined to be (-2.1, 3.2) in cell (4, 5).

[0089] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes will be obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for intelligent aerial search and positioning of multiple radiation sources, characterized in that: Given the types of nuclides and source intensities of multiple radioactive sources, the steps for searching and locating these sources are as follows: Step 1: Construct a surface radioactivity distribution map based on known radioactive sources; including: estimating the radiation rate produced by the radioactive source at a given distance in the air and generating a dose rate distribution map of the radioactive source at 1m on the surface; Step 2: Train the dose rate distribution map at 1m from the surface of the radioactive source based on the convolutional neural network; including: training the neural network on the generated dose rate distribution maps of each radioactive source surface and enabling the neural network to recognize the dose rate distribution map at 1m from the surface of the radioactive source; Step 3: Conduct full-coverage search and monitoring of the area to be searched; including: coarse-grid aerial nuclear radiation monitoring based on a "Z" pattern; when the radiation exceeds the set threshold, the monitoring platform will monitor at low speed and narrow spacing; and when the radiation falls below the set threshold, the monitoring platform will resume its original flight speed and spacing. Step 4: Generate a dose rate distribution map at 1m from the surface of the area to be searched; including: based on the dose rate values ​​obtained from monitoring at each location, using interpolation to generate a dose rate distribution map at 1m from the surface of each radiation source in the area to be searched. Step 5: Analyze the dose rate distribution map based on a convolutional neural network; including: inputting the distribution map into a trained convolutional neural network to achieve species identification and location determination; Step 6: Accurately determine the location of the radioactive sources; including: performing "loop"-shaped monitoring of the preliminary location of each radioactive source and obtaining the accurate location of the radioactive source based on the count rate calculation formula.

2. The method for intelligent aerial search and positioning of multiple radiation sources according to claim 1, characterized in that: In step 1, given a radionuclide, according to the fundamental theory of gamma radiation, the radiation rate produced at a given distance in air is expressed as: In the formula: Γ is the radiation rate at a certain point in the air; A is the activity of the point source; r is the distance between the measurement point of the airborne nuclear radiation detector and the radioactive source; μ is the linear attenuation coefficient of the air; Γ is the γ radiation rate constant, which represents the radiation rate per unit activity of the radioactive source at a unit distance. In air, under the same conditions, the relationship between absorbed dose rate and exposure rate is expressed as: In the formula: The absorbed dose rate at a point in the air is expressed in μGy / h. The radiation rate at a certain point in the air is expressed in R / h.

3. The method for intelligent aerial search and positioning of multiple radioactive sources according to claim 1, characterized in that: In step 2, the generated surface dose rate distribution maps of each radiation source are input into a two-dimensional convolutional neural network (2D-CNN) through three channels based on RGB features for network training. The 2D-CNN includes three structures: convolutional layers, pooling layers, and fully connected layers. The convolutional layers extract features by sliding the convolutional kernel sequentially across the input image from left to right and from top to bottom. Each time the kernel slides, it performs a dot product calculation on the input image corresponding to the sliding window position, obtaining a numerical value that preserves the spatial features of the input image. The pooling layers compress the image, accelerating the convergence process of the neural network and improving the stability during training. In the CNN structure, after multiple convolutional and sampling layers, one or more fully connected layers are connected. Each neuron in the fully connected layer is fully connected to all neurons in the previous layer. The fully connected layers integrate local information with class distinctions from the convolutional or sampling layers.

4. The method for intelligent aerial search and positioning of multiple radiation sources according to claim 1, characterized in that: In step 3, the area to be searched is topologically divided into a gridded area, and the search begins from the boundary of the area to be searched using an airborne radiation monitoring platform. In the initial stage, the monitoring path is based on a "Z" shape for coarse-grid airborne nuclear radiation monitoring, and the platform flies at an altitude of 50 meters during the monitoring process. In the monitoring stage, when the dose rate value detected by the airborne radiation monitoring platform exceeds the set threshold during the search, the monitoring platform reduces its flight speed to below 10 meters per second and monitors at 1 / 2 or 1 / 3 of the original flight spacing. When the detected value is lower than the set threshold, the original flight speed and flight spacing are restored, and the cycle is repeated until the area to be searched is completely monitored.

5. The method for intelligent aerial search and positioning of multiple radiation sources according to claim 1, characterized in that: In step 4, when the airborne radiation monitoring platform performs regional flight monitoring, the nuclear radiation detector on board updates the ground dose rate monitoring data at regular intervals. The data displays the longitude, latitude, altitude, and dose rate value of the current monitoring point. Once the airborne radiation monitoring platform has completed monitoring of the area to be searched, it will generate the location of each monitoring point along the monitoring route and the corresponding dose rate at 1m of the equivalent ground surface. Combined with a GIS map, it will initially generate a dose rate distribution map at 1m of the ground surface in the area to be searched. Linear interpolation will then be used to generate the dose rate at each point between the two monitoring lines, thus generating the dose rate distribution map at 1m of the ground surface in the area to be searched. Given the locations (x1, y1) and (x2, y2) of corresponding points along the two monitoring routes, and the dose rate values ​​d1 and d2, the linear interpolation formula is as follows: In the formula: The dose rate value at any point along the two monitoring routes. The location of any point between the two monitoring routes.

6. The method for intelligent aerial search and positioning of multiple radiation sources according to claim 1, characterized in that: In step 6, at each of the initially determined radioactive source locations (x', y'), the airborne radiation monitoring platform hovers and monitors for 30 seconds to measure the dose rate value D1 at 1m above the ground at that point; then, with (x', y') as the center point, a "U"-shaped monitoring with a side length of 50 meters is carried out, and the platform hovers and monitors for 30 seconds at the four vertices of the "U" shape. In a point source radiation field, neglecting air attenuation, the background count rate is constant. The count rate of an airborne nuclear radiation detector is determined according to the following formula. In the formula, (x i ,y i (x0, y0) represents the location of the airborne nuclear radiation detector, (x0, y0) represents the location of the radioactive source, M represents the count rate of the airborne nuclear radiation detector at a certain point in the air, A represents the activity of the radioactive source, ε represents the efficiency of the nuclear radiation detector, and r i denoted as , where b is the distance between the airborne nuclear radiation detector and the radiation source, b is the background count rate, and h is the monitoring altitude of the airborne nuclear radiation detector. Since the detector count at the measurement point is proportional to the intensity of the radiation source, the following set of equations is formed based on the data from the four monitoring points. Solving the above system of equations yields multiple solutions for the radioactive source: P1 = (x 01 ,y 01 ),P2=(x 02 ,y 02 ),P3=(x 03 ,y 03 ), P4 = (x 04 ,y 04 This allows us to obtain the accurate location of the radiation source; x0=(x 01 +x 02 +x 03 +x 04 ) / 4 y0=(y 01 +yx 02 +and 03 +and 04 ) / 4。

Images