Intelligent nuclide identification method and system
The nuclide energy spectrum is obtained by a gamma detector, and combined with adaptive nonlinear iterative peak stripping and Bayesian network, the problem of low nuclide identification accuracy in existing technologies is solved, and high-sensitivity and high-resolution nuclide identification is achieved, which is suitable for marine environment monitoring.
Patent Information
- Application Number
- CN202510919409.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-04
- Publication Date
- 2025-09-23
AI Technical Summary
Existing nuclear detection technology has difficulty effectively distinguishing characteristic peaks such as Cs-134 and Cs-137 in marine environments. In addition, high-purity germanium detectors have problems with high power consumption and poor seismic resistance in long-term deep-sea monitoring, resulting in low nuclide identification accuracy.
A gamma detector is used to obtain the nuclide energy spectrum, and background subtraction is performed using an adaptive nonlinear iterative peak-stripping algorithm. Combining fuzzy logic and Bayesian networks, the initial confidence level of the nuclide is determined and converted into probability to identify the existence of the nuclide.
The accuracy of nuclide identification is improved, the false alarm rate is reduced, and it adapts to the needs of low-count-rate marine environment monitoring.
Smart Images
Figure CN120686363A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of nuclear radiation detection and marine environment monitoring, and in particular to an intelligent nuclide identification method and system. Background Art
[0002] In-situ monitoring of radionuclides in the marine environment plays an irreplaceable role in nuclear emergency response and pollution assessment. As a core component of nuclear detection technology, nuclide identification provides critical information such as nuclide type and activity. The accuracy of these identification results is directly related to the scientific nature of environmental risk assessments and emergency decision-making.
[0003] Current mainstream detection technologies face significant technical bottlenecks in practical application. While cost-effective, NaI(Tl) scintillator detectors, with an energy resolution of 7%-10%, cannot effectively distinguish artificial nuclides released after accidents, such as Cs-134 (605keV) and Cs-137 (662keV), whose characteristic peak spacing is less than 50keV, significantly reducing identification accuracy. LaBr3 crystal detectors improve energy resolution to below 3%, but the spontaneous radioactivity of the La-138 isotope within them produces a background count rate as high as 1000cps, severely limiting their application in low-activity radiation fields (<10Bq / L). High-purity germanium (HPGe) semiconductor detectors, while offering exceptional resolution better than 0.2%, inherently rely on liquid nitrogen cooling systems (power consumption >200W) and suffer from insufficient mechanical shock resistance, making them inadequate for the demanding requirements of long-term, continuous deep-sea monitoring.
[0004] These technical limitations make the development of new nuclide identification methods a key issue that urgently needs to be broken through in the current field of nuclear detection. Summary of the Invention
[0005] In order to solve the above technical problems, the present invention discloses an intelligent nuclide identification method, comprising the following steps: A nuclide energy spectrum is acquired based on the actual measurement target area of the gamma detector, and a nuclide characteristic peak is determined based on the nuclide energy spectrum; the peak area of the nuclide characteristic peak and the energy difference between the theoretical energy value and the measured energy value of the nuclide characteristic peak are determined; an initial confidence level of the nuclide is determined based on the peak area and the energy difference; if the initial confidence level of the nuclide is less than or equal to a set threshold, the nuclide is identified based on the initial confidence level of the nuclide, and the existence of the nuclide is determined; if the initial confidence level of the nuclide is greater than the set threshold, the initial confidence level is converted into a probability based on a Bayesian network, and the nuclide is identified based on the probability, and the existence of the nuclide is determined.
[0006] Furthermore, the determining of the peak area of the characteristic peak of the nuclide and the energy difference between the theoretical energy value and the measured energy value of the characteristic peak of the nuclide includes: An adaptive nonlinear iterative peak stripping algorithm is used to perform background subtraction on the nuclide energy spectrum; the peak area and the measured energy value of the nuclide characteristic peak are determined based on the nuclide energy spectrum after background subtraction; the theoretical energy value of the nuclide characteristic peak is obtained, and the energy difference between the theoretical energy value and the measured energy value is calculated.
[0007] Furthermore, the nuclide characteristic peak is a single nuclide characteristic peak; Determining the initial confidence level of the nuclide based on the peak area and the energy difference includes: The energy difference is fuzzified using a trigonometric function and a Gaussian function respectively to determine a first nuclide fuzzy membership of the energy difference; the peak area of the single nuclide characteristic peak is fuzzified using a trapezoidal function to determine a second nuclide fuzzy membership of the nuclide characteristic peak area; a fuzzy rule is determined, and the first nuclide fuzzy membership and the second nuclide fuzzy membership are inferred based on the fuzzy rule to determine an initial nuclide confidence.
[0008] Furthermore, the nuclide characteristic peak is a plurality of nuclide characteristic peaks; Determining the initial confidence level of the nuclide based on the peak area and the energy difference includes: Select the energy difference corresponding to two nuclide characteristic peaks, the energy difference including a first energy difference and a second energy difference; use trigonometric function and Gaussian function to perform fuzzy processing on the first energy difference and the second energy difference respectively, and determine the first nuclide fuzzy membership of the first energy difference and the first nuclide fuzzy membership of the second energy difference; use trapezoidal function to perform fuzzy processing on the peak areas of the multiple nuclide characteristics, and determine the second nuclide fuzzy membership of the peak area; determine fuzzy rules, and based on the fuzzy rules, infer the first nuclide fuzzy membership of the first energy difference, the first nuclide fuzzy membership of the second energy difference, and the second nuclide fuzzy membership to determine the initial confidence of the nuclide.
[0009] Furthermore, if the initial confidence level of the nuclide is less than or equal to a set threshold, identifying the nuclide based on the initial confidence level and determining the existence of the nuclide includes: All the initial confidences of the nuclides are weighted and aggregated to determine a comprehensive confidence; if the value of the comprehensive confidence falls within a high confidence value range, the nuclide is determined to exist.
[0010] Furthermore, if the initial confidence level of the nuclide is greater than a set threshold, converting the initial confidence level into a probability based on a Bayesian network, identifying the nuclide based on the probability, and determining the existence of the nuclide includes: The existence of nuclides, the matching of nuclide characteristic peaks, the background level, the energy difference and the peak area are determined as network nodes of the Bayesian network; the conditional probabilities of the network nodes and the dependencies between the network nodes are determined; based on the initial confidence, the first nuclide fuzzy membership of the energy difference is converted into a first probability, and the second nuclide fuzzy membership of the peak area is converted into a second probability; based on the first probability and the second probability and the dependency, the existence of nuclides is determined.
[0011] Furthermore, determining the nuclide existence according to the first probability, the second probability, and the dependency relationship includes: Obtain a first dependency relationship between the energy difference and the peak position matching, a second dependency relationship between the peak area and the background level, a third dependency relationship between the peak position matching and the nuclide existence, and a fourth dependency relationship between the background level and the peak area; adopt a probability propagation algorithm to calculate the peak position matching probability based on the first probability and the first dependency, and calculate the background level probability based on the second probability and the second dependency; determine the posterior probability of the nuclide existence based on the peak position matching probability and the third dependency, and the background level probability and the fourth dependency; if the posterior probability of the nuclide existence is greater than or equal to the initially set prior probability of the nuclide existence, determine that the nuclide exists.
[0012] Furthermore, the determining of the fuzzy rules includes: Acquire historical detection data, the historical detection data including: historical energy difference, historical peak area and confidence label; cluster the historical energy difference and the historical peak area to determine the cluster center; and automatically generate the fuzzy rule according to the position of the cluster center and the confidence label.
[0013] Furthermore, the determination of the prior probability of the existence of the nuclide includes: Based on the initially set prior probability of the existence of the nuclide, the nuclide is detected. If the number of times the nuclide is detected is greater than or equal to the set number threshold, the prior probability of the existence of the nuclide is increased. Based on the increased prior probability of the existence of the nuclide, the nuclide is continued to be detected. If the number of times the nuclide is detected is less than the set number threshold, the increased prior probability of the existence of the nuclide is restored to the initially set prior probability of the existence of the nuclide.
[0014] According to another aspect of the present disclosure, there is provided an intelligent nuclide identification system, the system comprising: A determination module is used to obtain a nuclide energy spectrum based on the target area measured by the gamma detector, and determine the nuclide characteristic peak based on the nuclide energy spectrum; determine the peak area of the nuclide characteristic peak, and the energy difference between the theoretical energy value and the measured energy value of the nuclide characteristic peak; determine the initial confidence of the nuclide based on the peak area and the energy difference; an identification module is used to identify the nuclide based on the initial confidence of the nuclide if the initial confidence of the nuclide is less than or equal to a set threshold, and determine the existence of the nuclide; if the initial confidence of the nuclide is greater than the set threshold, convert the initial confidence into a probability based on a Bayesian network, identify the nuclide based on the probability, and determine the existence of the nuclide.
[0015] The embodiments of the present disclosure have the following technical effects: The intelligent nuclide identification method proposed in the present disclosure obtains the nuclide energy spectrum by measuring the target area with a gamma (γ) spectrometer and then determines the nuclide characteristic peak. The initial confidence level of the nuclide is determined by calculating the peak area of the nuclide characteristic peak, the energy difference between the theoretical energy value and the measured energy value. The determined initial confidence level of the nuclide can tolerate energy deviation, avoid prejudgment, and reduce nuclide false alarms. Further, by comparing the initial confidence level of the nuclide with a set threshold, it is determined whether to trigger the Bayesian network. If the initial confidence level of the nuclide is greater than the set threshold, the initial confidence level can be converted into a probability in combination with the Bayesian network, and the nuclide can be identified based on the probability to determine the existence of the nuclide. This can improve the accuracy of nuclide identification in a low-counting marine environment. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0017] Figure 1 is a flow chart of the intelligent nuclide identification method provided by an embodiment of the present disclosure; Figure 2 is a flow chart of a method for fusing fuzzy rules and Bayesian network algorithms provided by an embodiment of the present disclosure; Figure 3 is a flow chart of the adaptive prior probability adjustment method provided by an embodiment of the present disclosure; Figure 4 A schematic structural diagram of an intelligent nuclide identification system provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION
[0018] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are considered to be within the scope of the present invention.
[0019] In-situ monitoring of radionuclides in the marine environment is crucial for nuclear emergency response and pollution assessment. Therefore, nuclide identification, a key research area in nuclear detection, provides qualitative and quantitative information about the nuclides, and the accuracy of the results is crucial.
[0020] Among related technologies, NaI(Tl) scintillators offer cost advantages, but their 7%-10% energy resolution makes it difficult to distinguish the dense gamma peaks (peak spacing <50 keV) of artificial radionuclides (such as Cs-134 and Cs-137) after accidents, resulting in a high rate of false positives. LaBr3 crystals improve resolution to below 3%, but the high background count rate caused by the natural radioactivity of La-138 severely restricts their use in low-activity fields. Semiconductor detectors such as HPGe detectors achieve ultra-high resolution, but the size and power consumption (>200W) associated with their liquid nitrogen cooling systems, as well as their poor seismic resistance, make them unsuitable for long-term deep-sea monitoring. Especially in the specialized application scenario of marine environmental monitoring, balancing detection sensitivity, identification accuracy, and equipment reliability is a major technical challenge facing researchers.
[0021] Based on this, this application proposes an intelligent nuclide identification method and system, which consists of four parts: nuclide energy spectrum measurement and smoothing, background subtraction, peak area acquisition, and fuzzy logic and Bayesian network fusion nuclide identification, so as to achieve high-sensitivity and high-resolution detection and identification of artificial nuclides and natural radioactive nuclides.
[0022] Figure 1 This is a flow chart of the intelligent nuclide identification method provided by the embodiment of the present disclosure. Figure 1 , including the following steps: In step S11 , a nuclide energy spectrum is acquired based on the actual measurement of the target area by the gamma detector, and a nuclide characteristic peak is determined according to the nuclide energy spectrum.
[0023] In the embodiment of the present disclosure, a gamma (γ) spectrometer is a precision instrument for detecting the intensity of γ-ray radiation.
[0024] This disclosure uses an energy-calibrated gamma spectrometer to measure the target area and obtain the nuclide energy spectrum. Furthermore, wavelet transform or Fourier transform can be used to perform noise reduction processing to reduce the impact of noise, avoid peak broadening, and especially preserve the resolution of closely spaced peaks (such as 605 keV for Cs-134 and 662 keV for Cs-137).
[0025] The wavelet transform separates high-frequency noise from low-frequency signals through multi-scale decomposition, and reconstructs the energy spectrum after threshold filtering. The Fourier transform can filter out high-frequency noise components and retain the low-frequency band where the characteristic peaks are located.
[0026] Energy calibration can be understood as using known nuclides (such as Cs-137, Co-60) to calibrate the detector energy channel to ensure that the peak position corresponds linearly to the energy.
[0027] Therefore, the characteristic peaks of the nuclides are determined based on the nuclide energy spectrum after noise reduction. It should be noted that each group of nuclides has a unique set of characteristic peaks.
[0028] In step S12, the peak area of the nuclide characteristic peak and the energy difference between the theoretical energy value and the measured energy value of the nuclide characteristic peak are determined.
[0029] In the disclosed embodiments, the nuclide characteristic peaks can be found using the second derivative or symmetric zero-area peak-finding method. After the nuclide characteristic peaks are identified, the left and right boundaries of the nuclide characteristic peaks are determined, and the peak areas of each characteristic peak are calculated using Gaussian fitting. The peak area is the total count (unit: count) of the nuclide characteristic peak region, ranging from 0 to infinity.
[0030] Furthermore, the measured energy value of the characteristic peak of the nuclide and its corresponding theoretical energy value are obtained based on the nuclide energy spectrum, thereby determining the energy difference between the theoretical energy value and the measured energy value of the characteristic peak of the nuclide. The unit of the energy difference is keV, and the range is 0 to the peak position of the characteristic peak × 1%.
[0031] In step S13, the initial confidence level of the nuclide is determined based on the peak area and energy difference.
[0032] In the embodiment of the present disclosure, the initial confidence of the nuclide may be determined based on the first nuclide fuzzy membership corresponding to the peak area and the second nuclide fuzzy membership corresponding to the energy difference thereof.
[0033] In step S14, if the initial confidence level of the nuclide is less than or equal to the set threshold, the nuclide is identified based on the initial confidence level of the nuclide, and the existence of the nuclide is determined.
[0034] In the embodiment of the present disclosure, the threshold value may be set to 0.5, or the value of the threshold value may be freely set according to actual conditions.
[0035] Different initial confidence levels of nuclides correspond to different levels, and the existence of nuclides can be determined by determining the level corresponding to the initial confidence level of the nuclides.
[0036] For example, corresponding to a high level of initial confidence in the nuclide, the nuclide is determined to be present.
[0037] In step S15, if the initial confidence of the nuclide is greater than the set threshold, the initial confidence of the nuclide is converted into probability based on the Bayesian network, and the nuclide is identified based on the probability to determine the existence of the nuclide.
[0038] In the disclosed embodiment, if the initial confidence level of a nuclide is greater than a set threshold, for example, greater than 0.5, a Bayesian network is triggered. The initial confidence level of the nuclide is converted to a probability based on the Bayesian network, and the nuclide is identified based on the probability and a set prior probability of the nuclide's existence, thereby determining the nuclide's existence.
[0039] The intelligent nuclide identification method proposed in the present disclosure obtains the nuclide energy spectrum by measuring the target area with a gamma spectrometer and then determines the nuclide characteristic peak. The initial confidence of the nuclide is determined by calculating the peak area of the nuclide characteristic peak, the energy difference between the theoretical energy value and the measured energy value. The determined initial confidence of the nuclide can tolerate energy deviation, avoid prejudgment, and reduce nuclide false alarms. Further, by comparing the initial confidence of the nuclide with the set threshold, it is determined whether to trigger the Bayesian network. If the initial confidence of the nuclide is greater than the set threshold, the initial confidence can be converted into a probability in combination with the Bayesian network, and the nuclide can be identified based on the probability to determine the existence of the nuclide, thereby improving the accuracy of nuclide identification in a low-counting marine environment.
[0040] In the disclosed embodiment, a statistics-sensitive nonlinear iterative peak-clipping (SNIP) algorithm may be used to perform background subtraction on the nuclide energy spectrum; based on the nuclide energy spectrum after background subtraction, the peak area and measured energy value of the nuclide characteristic peak are determined; the theoretical energy value of the nuclide characteristic peak is obtained, and the energy difference between the theoretical energy value and the measured energy value is calculated.
[0041] The SNIP algorithm fits the relationship between the peak width and energy of the characteristic peak of the nuclide, and obtains the peak width in real time through this relationship, realizing adaptive transformation width and improving the defect that the fixed peak width cannot accurately deduct the scattering background of the entire spectrum.
[0042] In the embodiment of the present disclosure, if the detected nuclide characteristic peak is a single nuclide characteristic peak, the following implementation method is used to determine the initial confidence level of the nuclide.
[0043] In the present disclosure, a trigonometric function and a Gaussian function can be used to fuzzify the energy difference, respectively, to determine a first nuclide fuzzy membership of the energy difference. A trapezoidal function can be used to fuzzify the peak area of a single nuclide characteristic peak, to determine a second nuclide fuzzy membership of the nuclide characteristic peak area. A fuzzy rule is determined, and based on the fuzzy rule, the first nuclide fuzzy membership and the second nuclide fuzzy membership are inferred to determine an initial nuclide confidence.
[0044] The first nuclide fuzzy membership includes three categories: "small," "medium," and "large." The first nuclide fuzzy membership ranges from 0 to 1. The range of small, medium, and high can be determined by setting a threshold. The corresponding category is determined based on the value of the first nuclide fuzzy membership. It should be noted that the second nuclide fuzzy membership, as well as the initial nuclide confidence and comprehensive confidence, can all be determined according to this method. The second nuclide fuzzy membership includes three categories: "low," "medium," and "high." The initial nuclide confidence includes three categories: "low," "medium," and "high."
[0045] Among them, the fuzzy rules can be the following rules: Rule 1: If the fuzzy membership of the first nuclide of the energy difference is small and the peak area is high, the initial confidence of the nuclide is high.
[0046] Rule 2: If the fuzzy membership of the first nuclide of the energy difference is small and the peak area is low, the initial confidence of the nuclide is medium.
[0047] Rule 3: If the fuzzy membership of the first nuclide of the energy difference is large and the peak area is high, the initial confidence of the nuclide is medium.
[0048] Rule 4: If the fuzzy membership of the first nuclide of the energy difference is large or the peak area is low, the initial confidence of the nuclide is low.
[0049] In the embodiment of the present disclosure, if the detected nuclide characteristic peak is multiple nuclide characteristic peaks, the following implementation method is used to determine the initial confidence level of the nuclide.
[0050] The energy difference corresponding to two nuclide characteristic peaks is selected, and the energy difference includes a first energy difference and a second energy difference; the first energy difference and the second energy difference are fuzzy-processed respectively by using a trigonometric function and a Gaussian function to determine the fuzzy membership of the first nuclide of the first energy difference and the fuzzy membership of the first nuclide of the second energy difference; the peak area of multiple nuclide characteristics is fuzzy-processed by using a trapezoidal function to determine the fuzzy membership of the second nuclide of the peak area; the fuzzy rule is determined, and the fuzzy membership of the first nuclide of the first energy difference, the fuzzy membership of the first nuclide of the second energy difference and the fuzzy membership of the second nuclide are inferred based on the fuzzy rule to determine the initial confidence of the nuclide.
[0051] The first nuclide fuzzy membership and the second nuclide fuzzy membership are as described in the above embodiment and will not be described again. If the nuclide characteristic peak is determined to be multiple nuclide characteristic peaks, the fuzzy rule may be as follows: Rule 1: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is high, then the initial confidence of the nuclide is high.
[0052] Rule 2: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is low, then the initial confidence of the nuclide is medium.
[0053] Rule 3: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is high, then the initial confidence of the nuclide is medium.
[0054] Rule 4: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is low, then the initial confidence of the nuclide is medium.
[0055] Rule 5: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is high, then the initial confidence of the nuclide is medium.
[0056] Rule 6: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is high, then the initial confidence of the nuclide is medium.
[0057] Rule 7: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is low, then the initial confidence of the nuclide is medium.
[0058] Rule 8: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is high, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is low, then the initial confidence of the nuclide is medium.
[0059] Rule 9: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is high, then the initial confidence of the nuclide is medium.
[0060] Rule 10: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is low, then the initial confidence of the nuclide is medium.
[0061] Rule 11: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is high, then the initial confidence of the nuclide is medium.
[0062] Rule 12: If the fuzzy membership of the first nuclide of the first energy difference is small and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is low, then the initial confidence of the nuclide is medium.
[0063] Rule 13: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is high, then the initial confidence of the nuclide is medium.
[0064] Rule 14: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is high, then the initial confidence of the nuclide is medium.
[0065] Rule 15: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is small and the peak area is low, then the initial confidence of the nuclide is medium.
[0066] Rule 16: If the fuzzy membership of the first nuclide of the first energy difference is large and the peak area is low, and if the fuzzy membership of the first nuclide of the second energy difference is large and the peak area is low, then the initial confidence of the nuclide is medium.
[0067] Furthermore, the initial confidences of all nuclides are weighted and aggregated to determine the comprehensive confidence. If the value of the comprehensive confidence falls within the high confidence range, the nuclide is determined to exist.
[0068] The comprehensive confidence level is between 0 and 1, and the value range of high confidence level can be set according to actual conditions.
[0069] The following embodiments of the present disclosure will convert the initial confidence of a nuclide into a probability based on a Bayesian network if the initial confidence of the nuclide is greater than a set threshold, and identify the nuclide based on the probability to determine the existence of the nuclide.
[0070] Nuclide presence, nuclide characteristic peak position matching, background level, energy difference, and peak area are identified as network nodes in a Bayesian network. The conditional probabilities of the network nodes and the dependencies between them are determined. Based on the initial confidence level, the fuzzy membership of the first nuclide for the energy difference is converted to a first probability, and the fuzzy membership of the second nuclide for the peak area is converted to a second probability. Based on the first and second probabilities and the dependencies, the presence of the nuclide is determined.
[0071] The dependencies between network nodes include: a first dependency between energy difference and peak position matching, a second dependency between peak area and background level, a third dependency between peak position matching and nuclide presence, and a fourth dependency between background level and peak area. Energy difference is understood as having peak position matching as its parent node, nuclide presence as its parent node, peak area as its parent node, and background level as its parent node. Nuclide presence is a binary variable (present / absent). Peak position matching indicates the degree of match between the measured peak and the characteristic peak of the nuclide. Background level reflects the intensity of the ambient radioactivity background. Energy difference and peak area are observed variables.
[0072] Furthermore, a probability propagation algorithm is used to calculate the peak matching probability based on the first probability and the first dependency, and to calculate the background level probability based on the second probability and the second dependency; the posterior probability of the nuclide existence is determined based on the peak matching probability and the third dependency, and the background level probability and the fourth dependency; if the posterior probability of the nuclide existence is greater than or equal to the initially set prior probability of the nuclide existence, the nuclide is determined to exist.
[0073] The prior probability of nuclide presence can be 70%, while the prior probability of nuclide absence is 30%. The value of the prior probability of nuclide presence can be freely set based on actual conditions. The peak position match probability is 80% when the nuclide is present, and 20% when it is absent. This probability can also be freely set. The background level probability is 70% when the nuclide is present, and 80% when it is absent. This probability can also be freely set.
[0074] For example, based on the first probability of energy difference and the second probability of peak area, the probability propagation algorithm is used to determine the posterior probability of the nuclide's presence. If the posterior probability of the nuclide's presence (e.g., 0.92, or 92%) is greater than the initially set prior probability of the nuclide's presence (70%), the nuclide is confirmed to be present.
[0075] In this disclosure, the fuzzy rules are determined as follows: Obtain historical test data, including historical energy differences, historical peak areas, and confidence labels. Cluster the historical energy differences and peak areas to determine cluster centers. Automatically generate fuzzy rules based on the locations of the cluster centers and confidence labels.
[0076] The confidence labels are the labels manually annotated in the past, and the clustering is K-means clustering (the default is 3 categories).
[0077] Figure 2 This is a flow chart of the method for integrating fuzzy rules and Bayesian network algorithms provided by the embodiment of the present disclosure. Figure 2 , the energy difference (ΔE), peak area and peak symmetry are determined based on the nuclide energy spectrum. Among them, the peak symmetry can ensure the accuracy of energy calibration and peak position positioning. Through fuzzification processing (for example, membership function calculation), the fuzzy rule base is determined, wherein the fuzzy rule base is automatically generated based on the historical detection data set. Further defuzzification (i.e., area center method) is performed to obtain the initial confidence of the nuclide (i.e., Figure 2 If the confidence level is greater than the set 0.5, the posterior probability of the nuclide existence is further calculated based on the Bayesian network (i.e. Figure 2 The posterior probability in the is used to determine the existence of the nuclide. If the confidence level is less than or equal to 0.5, the posterior probability of the existence of the nuclide is determined based on the confidence level, thus determining the existence of the nuclide.
[0078] In the present disclosure, after the posterior probability is output, the nuclide existence prior probability may be updated according to the adaptive prior. Figure 3 This is a flow chart of the adaptive prior probability adjustment method provided by the embodiment of the present disclosure. Figure 3 , its specific implementation is as follows: Nuclides are detected based on the initially set prior probability of nuclide existence. If the number of times the nuclide is detected is greater than or equal to the set number threshold (for example, 10 times), the prior probability of the nuclide existence is increased (the prior probability is updated from 70% to 90%). Nuclides are continued to be detected based on the increased prior probability of nuclide existence. If the number of times the nuclide is detected is less than the set number threshold, the increased prior probability of the nuclide existence is restored to the initially set prior probability of the nuclide existence.
[0079] Based on Figure 1 The same principle as shown in the method, Figure 4 FIG. 1 shows a schematic structural diagram of an intelligent nuclide identification system provided by an embodiment of the present disclosure, such as Figure 4 As shown, the intelligent nuclide identification system 400 may include: The determination module 401 is used to obtain a nuclide energy spectrum based on the target area measured by the gamma detector, and determine the nuclide characteristic peak based on the nuclide energy spectrum; determine the peak area of the nuclide characteristic peak, and the energy difference between the theoretical energy value and the measured energy value of the nuclide characteristic peak; determine the initial confidence of the nuclide based on the peak area and the energy difference; the identification module 402 is used to identify the nuclide based on the initial confidence of the nuclide if the initial confidence of the nuclide is less than or equal to the set threshold, and determine the existence of the nuclide; if the initial confidence of the nuclide is greater than the set threshold, convert the initial confidence into a probability based on the Bayesian network, identify the nuclide based on the probability, and determine the existence of the nuclide.
[0080] In the present disclosure, the determination module 401 is used to use an adaptive nonlinear iterative peak stripping algorithm to perform background subtraction on the nuclide energy spectrum; determine the peak area and measured energy value of the nuclide characteristic peak based on the nuclide energy spectrum after background subtraction; obtain the theoretical energy value of the nuclide characteristic peak, and calculate the energy difference between the theoretical energy value and the measured energy value.
[0081] In the present disclosure, the nuclide characteristic peak is a single nuclide characteristic peak; the determination module 401 is used to use trigonometric functions and Gaussian functions to fuzzy the energy difference respectively to determine the first nuclide fuzzy membership of the energy difference; use trapezoidal functions to fuzzy the peak area of the single nuclide characteristic peak to determine the second nuclide fuzzy membership of the nuclide characteristic peak area; determine fuzzy rules, and infer the first nuclide fuzzy membership and the second nuclide fuzzy membership based on the fuzzy rules to determine the initial confidence of the nuclide.
[0082] In the present disclosure, the nuclide characteristic peaks are multiple nuclide characteristic peaks; the determination module 401 is used to select the energy difference corresponding to two nuclide characteristic peaks, and the energy difference includes a first energy difference and a second energy difference; the first energy difference and the second energy difference are fuzzy processed respectively by using a trigonometric function and a Gaussian function to determine the first nuclide fuzzy membership of the first energy difference and the first nuclide fuzzy membership of the second energy difference; the peak area of the multiple nuclide characteristics is fuzzy processed by using a trapezoidal function to determine the second nuclide fuzzy membership of the peak area; a fuzzy rule is determined, and based on the fuzzy rule, the first nuclide fuzzy membership of the first energy difference, the first nuclide fuzzy membership of the second energy difference, and the second nuclide fuzzy membership are inferred to determine the initial confidence of the nuclide.
[0083] In the present disclosure, the determination module 401 is used to perform weighted aggregation on all the initial confidences of the nuclides to determine a comprehensive confidence; if the value of the comprehensive confidence belongs to a high confidence value range, the nuclide is determined to exist.
[0084] In the present disclosure, the determination module 401 is used to determine the existence of nuclides, the matching of nuclide characteristic peaks, the background level, the energy difference and the peak area as network nodes of the Bayesian network; determine the conditional probability of the network nodes, and the dependency relationship between the network nodes; according to the initial confidence, convert the first nuclide fuzzy membership of the energy difference into a first probability, and convert the second nuclide fuzzy membership of the peak area into a second probability; determine the existence of nuclides according to the first probability and the second probability and the dependency relationship.
[0085] In the present disclosure, the determination module 401 is used to obtain a first dependency relationship between the energy difference and the peak position matching, a second dependency relationship between the peak area and the background level, a third dependency relationship between the peak position matching and the nuclide existence, and a fourth dependency relationship between the background level and the peak area; a probability propagation algorithm is used to calculate the peak position matching probability based on the first probability and the first dependency, and to calculate the background level probability based on the second probability and the second dependency; a posterior probability of the nuclide existence is determined based on the peak position matching probability and the third dependency, and the background level probability and the fourth dependency; if the posterior probability of the nuclide existence is greater than or equal to the initially set prior probability of the nuclide existence, the nuclide is determined to exist.
[0086] In the present disclosure, the determination module 401 is used to obtain historical detection data, which includes: historical energy difference, historical peak area and confidence label; clustering the historical energy difference and the historical peak area to determine the cluster center; and automatically generating the fuzzy rule based on the position of the cluster center and the confidence label.
[0087] In the present disclosure, the determination module 401 is used to detect nuclides based on the initially set prior probability of nuclide existence, and if the number of times the nuclide is detected is greater than or equal to the set number threshold, the prior probability of the nuclide existence is increased; based on the increased prior probability of the nuclide existence, the nuclide is continued to be detected, and if the number of times the nuclide is detected is less than the set number threshold, the increased prior probability of the nuclide existence is restored to the initially set prior probability of the nuclide existence.
[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.
Claims
1. An intelligent nuclide identification method, characterized in that: The method comprises: Acquire a nuclide energy spectrum based on the actual measurement of the target area by the gamma detector, and determine the nuclide characteristic peak according to the nuclide energy spectrum; Determining the peak area of the characteristic peak of the nuclide, and the energy difference between the theoretical energy value and the measured energy value of the characteristic peak of the nuclide; determining an initial confidence level of the nuclide based on the peak area and the energy difference; If the initial confidence level of the nuclide is less than or equal to a set threshold, identifying the nuclide based on the initial confidence level of the nuclide and determining the existence of the nuclide; If the initial confidence of the nuclide is greater than a set threshold, the initial confidence is converted into a probability based on a Bayesian network, and the nuclide is identified based on the probability to determine the existence of the nuclide.
2. The intelligent nuclide identification method according to claim 1, characterized in that: Determining the peak area of the nuclide characteristic peak and the energy difference between the theoretical energy value and the measured energy value of the nuclide characteristic peak includes: Adopting an adaptive nonlinear iterative peak stripping algorithm to perform background subtraction on the nuclide energy spectrum; Determining the peak area and measured energy value of the characteristic peak of the nuclide based on the nuclide energy spectrum after background subtraction; Obtain a theoretical energy value of the nuclide characteristic peak, and calculate an energy difference between the theoretical energy value and the measured energy value.
3. The intelligent nuclide identification method according to claim 1, characterized in that: The nuclide characteristic peak is a single nuclide characteristic peak; Determining the initial confidence level of the nuclide based on the peak area and the energy difference includes: Performing fuzzy processing on the energy difference using a trigonometric function and a Gaussian function respectively to determine the fuzzy membership of the first nuclide of the energy difference; Performing fuzzy processing on the peak area of the single nuclide characteristic peak using a trapezoidal function to determine the second nuclide fuzzy membership of the nuclide characteristic peak area; A fuzzy rule is determined, and the first nuclide fuzzy membership and the second nuclide fuzzy membership are inferred based on the fuzzy rule to determine an initial nuclide confidence.
4. The intelligent nuclide identification method according to claim 1, characterized in that: The nuclide characteristic peak is a plurality of nuclide characteristic peaks; Determining the initial confidence level of the nuclide based on the peak area and the energy difference includes: Selecting an energy difference corresponding to two nuclide characteristic peaks, the energy difference including a first energy difference and a second energy difference; Performing fuzzy processing on the first energy difference and the second energy difference using a trigonometric function and a Gaussian function respectively, to determine a first nuclide fuzzy membership degree of the first energy difference and a first nuclide fuzzy membership degree of the second energy difference; Performing fuzzy processing on the peak areas of the plurality of nuclide features using a trapezoidal function to determine a second nuclide fuzzy membership degree of the peak area; A fuzzy rule is determined, and based on the fuzzy rule, the first nuclide fuzzy membership of the first energy difference, the first nuclide fuzzy membership of the second energy difference, and the second nuclide fuzzy membership are inferred to determine an initial nuclide confidence.
5. The intelligent nuclide identification method according to claim 1, characterized in that: If the initial confidence level of the nuclide is less than or equal to a set threshold, identifying the nuclide based on the initial confidence level and determining the existence of the nuclide includes: Performing weighted aggregation on all the initial confidences of the nuclides to determine a comprehensive confidence; If the value of the comprehensive confidence falls within the high confidence value range, the nuclide is determined to exist.
6. The intelligent nuclide identification method according to claim 1, characterized in that: If the initial confidence level of the nuclide is greater than a set threshold, converting the initial confidence level into a probability based on a Bayesian network, identifying the nuclide based on the probability, and determining the existence of the nuclide includes: Determining nuclide existence, nuclide characteristic peak position matching, background level, the energy difference and the peak area as network nodes of the Bayesian network; Determining the conditional probabilities of the network nodes and the dependencies between the network nodes; According to the initial confidence level, converting the first nuclide fuzzy membership of the energy difference into a first probability, and converting the second nuclide fuzzy membership of the peak area into a second probability; The nuclide existence is determined according to the first probability, the second probability and the dependency relationship.
7. The intelligent nuclide identification method according to claim 6, characterized in that: The determining of the nuclide existence according to the first probability, the second probability and the dependency relationship includes: Obtaining a first dependency relationship between the energy difference and the peak position match, a second dependency relationship between the peak area and the background level, a third dependency relationship between the peak position match and nuclide presence, and a fourth dependency relationship between the background level and the peak area; Using a probability propagation algorithm, calculate the peak matching probability according to the first probability and the first dependency, and calculate the background level probability according to the second probability and the second dependency; Determining a posterior probability of nuclide existence based on the peak position matching probability and the third dependency, and the background level probability and the fourth dependency; If the posterior probability of the existence of the nuclide is greater than or equal to the initially set prior probability of the existence of the nuclide, it is determined that the nuclide exists.
8. The intelligent nuclide identification method according to claim 3 or 4, characterized in that: The determining of fuzzy rules comprises: Acquiring historical detection data, wherein the historical detection data includes: historical energy difference, historical peak area, and confidence label; Clustering the historical energy differences and the historical peak areas to determine cluster centers; The fuzzy rules are automatically generated according to the positions of the cluster centers and the confidence labels.
9. The intelligent nuclide identification method according to claim 7, characterized in that: The determination of the prior probability of the existence of the nuclide includes: Detecting a nuclide based on the initially set prior probability of the existence of the nuclide, and if the number of times the nuclide is detected is greater than or equal to a set number threshold, increasing the prior probability of the existence of the nuclide; The nuclides are continuously detected based on the improved prior probability of the existence of the nuclides. If the number of times the nuclides are detected is less than the set number threshold, the improved prior probability of the existence of the nuclides is restored to the initially set prior probability of the existence of the nuclides.
10. An intelligent nuclide identification system, characterized in that: The system comprises: a determination module, configured to obtain a nuclide energy spectrum based on the target area measured by the gamma detector, and determine a nuclide characteristic peak based on the nuclide energy spectrum; determine a peak area of the nuclide characteristic peak, and an energy difference between a theoretical energy value and a measured energy value of the nuclide characteristic peak; and determine an initial confidence level of the nuclide based on the peak area and the energy difference; The identification module is used to identify the nuclide based on the initial confidence level of the nuclide and determine the existence of the nuclide if the initial confidence level of the nuclide is less than or equal to the set threshold; if the initial confidence level of the nuclide is greater than the set threshold, convert the initial confidence level into a probability based on a Bayesian network, identify the nuclide based on the probability, and determine the existence of the nuclide.