Radiation source identification algorithm and system based on machine learning
By using machine learning-based algorithms in radiation source recognition, data preprocessing and feature engineering, and training and verification models, the problems of low efficiency and different accuracy of radiation source recognition in the existing technology are solved, and higher recognition accuracy and model performance are achieved.
Patent Information
- Application Number
- CN202411505853.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-28
- Publication Date
- 2025-03-18
AI Technical Summary
The prior art has problems in the identification of radiation source with low data processing efficiency, weak model generalization ability and different recognition accuracy, especially in complex environments, it is difficult to effectively distinguish and identify different types of radiation sources.
The radiation source recognition algorithm based on machine learning is used to improve the recognition accuracy of radiation source through data preprocessing (such as outlier processing and standardization), feature engineering (such as energy spectrum analysis and time domain analysis), and machine learning model training and verification.
It improves the recognition accuracy of radiation sources, can more accurately distinguish different types of radiation sources, enhances the training and prediction performance of the model, and reduces the impact of noise and outliers.
Smart Images
Figure CN119646599A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of nuclear safety monitoring and radiation detection, and in particular to a radiation source identification algorithm and system based on machine learning. Background Art
[0002] In the modern nuclear energy industry and its related fields, the detection and identification of radiation sources is crucial. With the popularity of nuclear power plants and the widespread use of radioactive materials, it is necessary to ensure that these radiation sources are properly managed and monitored. Traditional radiation source identification technology relies on sensor detection data to distinguish different radioactive isotopes through physical methods. However, these methods usually have limited data processing capabilities and lack adaptability and accuracy in complex environments.
[0003] With the advancement of technology, machine learning technology has become a powerful tool for data processing and pattern recognition. Through training, machine learning models can identify patterns and features in complex data, thereby providing more accurate and faster identification results. Especially in complex nuclear power scenarios, machine learning can effectively handle a variety of interference factors and improve the accuracy and reliability of radiation source identification.
[0004] At present, although there are some radiation detection and identification solutions based on machine learning on the market, these solutions still face problems such as low data processing efficiency, weak model generalization ability, and different identification accuracy for different radiation sources. Therefore, developing a new radiation source identification algorithm based on machine learning, aiming to provide a technical solution that can effectively distinguish and accurately identify different types of radiation sources, is an important demand for current technological development. Summary of the invention
[0005] In view of the above-mentioned problems, the present invention is proposed.
[0006] Therefore, the technical problem solved by the present invention is: the radiation source data in the existing technical solution may include multiple types of radiation, different energy levels and timestamps, and a method is needed to process multi-dimensional data, as well as how to effectively extract and select features from the radiation source data.
[0007] To solve the above technical problems, the present invention provides the following technical solutions: a radiation source identification algorithm based on machine learning, which comprises the following steps:
[0008] Collect radiation source data and perform preprocessing; establish a machine model based on feature engineering; train and verify the machine model.
[0009] As a preferred solution of the radiation source identification algorithm based on machine learning described in the present invention, the radiation source data includes radiation type, energy level and timestamp.
[0010] The preprocessing includes outlier processing and standardization.
[0011] As a preferred solution of the radiation source identification algorithm based on machine learning described in the present invention, the outlier processing includes extracting data features in the radiation source data, and using the IQR method to eliminate outliers in the data features. The IQR method includes calculating the first quartile and the third quartile, and calculating the quartile range, expressed as (IQR=Q3-Q1), wherein Q3 represents the third quartile, and Q1 represents the first quartile. If the data point in the radiation source data is lower than Q1-1.5×IQR or higher than Q3+1.5×IQR, it is identified as an outlier and the outlier is eliminated.
[0012] The standardization includes calculating the mean value and standard deviation of the data characteristics and standardizing the radiation source data, which is expressed as:
[0013]
[0014] Among them, mean(x) represents the mean value of the data feature, and std(x) represents the standard deviation of the data feature.
[0015] As a preferred solution of the radiation source identification algorithm based on machine learning described in the present invention, wherein: the feature engineering includes energy spectrum analysis and time domain analysis.
[0016] The energy spectrum analysis includes smoothing the energy spectrum data using moving average or Gaussian smoothing, identifying peaks in the energy spectrum using a local maximum search algorithm, and applying Gaussian fitting to each detected peak, expressed as:
[0017]
[0018] Where μ is the center of the peak and σ is the standard deviation.
[0019] All data points of the energy spectrum are summed to obtain the total count, which is counted according to the number of peaks identified by the peak detection algorithm. The energy spectrum area without obvious radiation peaks is selected, and the average count of the selected background area is calculated as the background level.
[0020] As a preferred solution of the radiation source identification algorithm based on machine learning described in the present invention, the time domain analysis includes differentiating the timestamps of consecutive radiation events, calculating the time interval between two consecutive events, identifying the start and end time points of each event, calculating the time difference from the start to the end, that is, the event duration, setting a fixed time window, counting the number of radiation events in each time window, dividing the total number of events by the length of the time window to obtain the event counting rate per unit time, selecting a time window for high-resolution monitoring, and counting the total number of radiation events in each window.
[0021] As a preferred solution of the radiation source identification algorithm based on machine learning described in the present invention, wherein: the training of the machine model includes training the selected model using a training data set, expressed as,
[0022]
[0023] Among them, w represents the vector, b represents the bias term, and C represents the regularization parameter. represents the loss function, n represents the total number of data points in the training dataset, and x i Represents the feature vector of the i-th training data point, y i represents the true label of the i-th training data point.
[0024] As a preferred solution of the radiation source identification algorithm based on machine learning described in the present invention, the verification includes using a cross-validation method, the data set is divided into k subsets, one subset is set aside as a validation set each time, and the rest are used as training sets, and repeated k times.
[0025] Another object of the present invention is to provide a radiation source identification system based on machine learning, which can solve the problems of existing manual data processing and inefficient identification through automated data preprocessing, efficient feature engineering technology and advanced machine learning models.
[0026] To solve the above technical problems, the present invention provides the following technical solutions: a radiation source identification system based on machine learning, comprising a data acquisition module, a data preprocessing module, a feature engineering module, and a machine learning and verification module.
[0027] The data acquisition module is responsible for collecting data from the radiation source.
[0028] The data preprocessing module is responsible for removing outliers in the data using the IQR method and performing standardization processing.
[0029] The feature engineering module is responsible for smoothing the energy spectrum data by moving average or Gaussian smoothing method, identifying peaks in the energy spectrum using a local maximum search algorithm, and describing each peak using Gaussian fitting.
[0030] The feature engineering module is also responsible for processing timestamp data, calculating the time intervals between consecutive radiation events, performing event duration analysis, and counting the number of radiation events within a fixed time window to obtain the event counting rate per unit time.
[0031] The machine learning and verification module is responsible for training the model using the selected machine learning model and training data set to learn the relationship between the radiation source data and its category.
[0032] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a radiation source identification algorithm based on machine learning as described above when executing the computer program.
[0033] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a radiation source identification algorithm based on machine learning as described above.
[0034] Beneficial effects of the present invention: The present invention improves the recognition accuracy of radiation sources by adopting advanced feature engineering methods (such as energy spectrum analysis and time domain analysis) and machine learning models, and can more accurately distinguish different types of radiation sources and their characteristics. Through outlier processing and standardized preprocessing steps, the present invention improves the quality of data, thereby enhancing the training and prediction performance of the model. The IQR method is used to eliminate outliers and the data features are processed using standardization, so that the model can more effectively learn important patterns in the data and reduce the impact of noise and outliers. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0036] Figure 1 An overall flow chart of a radiation source identification algorithm based on machine learning provided for the first embodiment of the present invention.
[0037] Figure 2 An overall framework diagram of a radiation source identification system based on machine learning provided for the second embodiment of the present invention. DETAILED DESCRIPTION
[0038] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0039] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a radiation source identification algorithm based on machine learning, characterized in that:
[0040] S1: Collect radiation source data and perform preprocessing.
[0041] Radiation source data includes radiation type, energy level, and timestamp.
[0042] The preprocessing includes outlier processing and standardization.
[0043] The outlier processing includes extracting data features from the radiation source data and using the IQR method to eliminate outliers in the data features. The IQR method includes calculating the first quartile and the third quartile, and calculating the quartile range, which is expressed as (IQR=Q3-Q1), where Q3 represents the third quartile and Q1 represents the first quartile. If the data point in the radiation source data is lower than Q1-1.5×IQR or higher than Q3+1.5×IQR, it is identified as an outlier and the outlier is eliminated.
[0044] Standardization includes calculating the mean and standard deviation of data features and standardizing the radiation source data, expressed as,
[0045]
[0046] Among them, mean(x) represents the mean value of the data feature, and std(x) represents the standard deviation of the data feature.
[0047] S2: Build a machine model based on feature engineering.
[0048] Feature engineering includes energy spectrum analysis and time domain analysis.
[0049] The energy spectrum analysis includes smoothing the energy spectrum data using moving average or Gaussian smoothing, identifying peaks in the energy spectrum using a local maximum search algorithm, and applying Gaussian fitting to each detected peak, expressed as:
[0050]
[0051] Where μ is the center of the peak and σ is the standard deviation.
[0052] All data points of the energy spectrum are summed to obtain the total count, which is counted according to the number of peaks identified by the peak detection algorithm. The energy spectrum area without obvious radiation peaks is selected, and the average count of the selected background area is calculated as the background level.
[0053] Time domain analysis includes differentiating the timestamps of consecutive radiation events, calculating the time interval between two consecutive events, identifying the start and end time points for each event, calculating the time difference from the start to the end, which is the event duration, setting a fixed time window, counting the number of radiation events in each time window, dividing the total number of events by the length of the time window to obtain the event counting rate per unit time, selecting a time window for high-resolution monitoring, and counting the total number of radiation events in each window.
[0054] S3: Train and validate the machine model.
[0055] Training the machine model includes training the selected model using the training dataset, expressed as,
[0056]
[0057] Among them, w represents the vector, b represents the bias term, and C represents the regularization parameter. represents the loss function, n represents the total number of data points in the training dataset, and x i Represents the feature vector of the i-th training data point, y i represents the true label of the i-th training data point.
[0058] The validation includes,using the cross-validation method, the data set is divided into k subsets, each time one subset is left out as the validation set, and the rest are used as the training set, which is repeated k times.
[0059] The model performance is evaluated using indicators such as confusion matrix, precision, recall rate, F1 score, etc. These indicators can fully reflect the model's ability to recognize different categories.
[0060] Parameter optimization:
[0061] Grid search technique is used to optimize the model's hyperparameters.
[0062] Use the performance on the validation set as a metric to select the optimal parameters.
[0063] Example 2, reference Figure 2, which is an embodiment of the present invention, provides a system for radiation source identification algorithm based on machine learning, characterized in that it includes a data acquisition module 100, a data preprocessing module 200, a feature engineering module 300, and a machine learning and verification module 400.
[0064] The data acquisition module 100 is responsible for collecting data from the radiation source.
[0065] The data preprocessing module 200 is responsible for removing outliers in the data using the IQR method and performing standardization processing.
[0066] The feature engineering module 300 is responsible for smoothing the energy spectrum data by moving average or Gaussian smoothing method, identifying peaks in the energy spectrum using a local maximum search algorithm, and describing each peak using Gaussian fitting.
[0067] The feature engineering module 300 is also responsible for processing timestamp data, calculating the time intervals between consecutive radiation events, performing event duration analysis, and counting the number of radiation events within a fixed time window to obtain the event counting rate per unit time.
[0068] The machine learning and verification module 400 is responsible for training the model using the selected machine learning model and training data set to learn the relationship between the radiation source data and its category.
[0069] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.
[0070] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.
[0071] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.
[0072] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0073] Embodiment 3: In this embodiment, in order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments. This embodiment experiments the existing traditional method and the method of this embodiment respectively.
[0074] To complete this experiment, we will now present specific simulation data and model evaluation results, which will help us better understand the entire process and its effects.
[0075] Data experiment 1.
[0076] Here is an example of 5 randomly generated data points from each type (gamma rays and neutron rays):
[0077] Type A (Gamma Ray) Sample
[0078] Energy (KeV) Count Total Time (seconds) type) 540.7 210 12.5 Gamma 489.3 198 23.1 Gamma 510.2 215 35.7 Gamma 470.1 205 45.2 Gamma 530.5 190 59.9 Gamma
[0079] Type B (Neutron Beam) Sample
[0080] Energy (KeV) Count Total Time (seconds) type) 310.4 160 5.4 Neutron 299.6 145 19.8 Neutron 280.3 155 30.2 Neutron 320.2 140 49.5 Neutron 305.7 150 58.1 Neutron
[0081] Data preprocessing
[0082] Assuming we have applied outlier handling and normalization, here is the normalized data after processing:
[0083] Processed sample (part)
[0084]
[0085] Model training and evaluation
[0086] Assume that a support vector machine (SVM) was used for training and 5-fold cross validation was applied. Here are the simulated evaluation results:
[0087] Cross validation results
[0088] -Fold 1: Accuracy = 95%
[0089] -Fold 2: Accuracy = 96%
[0090] -Fold3: Accuracy = 94%
[0091] -Fold4: Accuracy = 97%
[0092] -Fold5: Accuracy = 95%
[0093] Average accuracy
[0094] - Average accuracy = 95.4%
[0095] Test set results
[0096] Assume that the evaluation results on the independent test set are as follows:
[0097] -Accuracy = 96%
[0098] -Recall = 95%
[0099] - Accuracy = 97%
[0100] These results show that the proposed machine learning model is able to distinguish and identify different types of emitters very effectively. In this way, the model demonstrates the high performance that is possible in real-world scenarios, proving that the method has practical application value for emitter identification.
[0101] Experiment 2: (This experiment introduces more radiation sources, but the code is basically the same as Experiment 1, except that the data features are different)
[0102] Data Generation and Feature Definition
[0103] Data characteristics
[0104] 1. Type A (Gamma Ray):
[0105] -Energy level (E A ): Normal distribution, mean value (μ = 500) KeV, standard deviation (σ = 70) KeV
[0106] -Count total (C A ): Poisson distribution, mean value (λ=220)
[0107] - Background radiation (B A ): Uniform distribution, ranging from 10 to 50cps
[0108] 2. Type B (neutron beam):
[0109] -Energy level (E B ): Normal distribution, mean value (μ=300)KeV, standard deviation (σ=70)KeV
[0110] -Count total (C B ): Poisson distribution, mean value (λ=180)
[0111] - Background radiation (B B ): Uniform distribution, ranging from 20 to 60cps
[0112] 3. Type C (Beta rays):
[0113] -Energy level (E C ): Normal distribution, mean value (μ = 200) KeV, standard deviation (σ = 50) KeV
[0114] -Count total (C C ): Poisson distribution, mean (λ = 160)
[0115] - Background radiation (B C ): Uniform distribution, ranging from 15 to 55cps
[0116] Data preprocessing
[0117] 1. Outlier processing:
[0118] An IQR method was used for energy, total counts, and background radiation to handle outliers.
[0119] Determine (Q1) and (Q3) for each feature, calculate (IQR) and remove features with a value below (Q1-1.5
[0120] ×IQR) or above (Q3+1.5×IQR).
[0121] 2. Data standardization and normalization:
[0122] - Z-score normalization was applied to energy level and total number of counts.
[0123] -Apply min-max normalization to background radiation:
[0124] Feature Engineering
[0125] The energy level, total number of counts, and normalized background radiation are used as components of the eigenvector. The eigenvector is in the form of ([E normalized ,C normalized ,B normalized ]).
[0126] Model training and validation
[0127] Support vector machines (SVMs) using linear kernels may need to use kernel techniques such as radial basis function (RBF) kernels to increase the nonlinear discrimination ability of the model, considering the diversity of data characteristics.
[0128] Cross Validation
[0129] A 10-fold cross validation was used to ensure that each data subset was used as a test set to verify the generalization ability of the model on different samples.
[0130] Results Evaluation
[0131] 1. Cross-validation results: Display the accuracy of each fold and calculate the average accuracy.
[0132] 2. Confusion Matrix: Using Confusion on an Independent Test Set
[0133] The matrix shows in detail the recognition accuracy of the model for each type of radiation source.
[0134] 3. Performance metrics: Precision, recall, and F1 score are displayed to evaluate the performance of the model in classifying each radiation type.
[0135] Data Sample
[0136] The following table provides five randomly selected sample data points from three radiation source types to illustrate the characteristics and distribution of the data.
[0137] Sample Data
[0138] type Energy (KeV) Count Total Background radiation (cps) Time (seconds) Gamma 520 215 30 25.5 Neutron 290 175 45 15.3 Beta 210 160 25 40.2 Gamma 480 230 35 32.1 Neutron 310 190 55 18.4
[0139] Data preprocessing and normalization results
[0140] After applying IQR outlier processing and Z-score normalization (energy and total counts) and min-max normalization (background radiation) to the above data, it is assumed that the normalized data obtained is as follows:
[0141] Standardized sample data
[0142]
[0143] Model training and evaluation
[0144] Using support vector machine (SVM) training and 10-fold cross validation, the following hypothetical evaluation results are provided:
[0145] Cross validation results
[0146] An example of the accuracy of each fold is as follows:
[0147] -Fold 1: 94%
[0148] -Fold 2: 92%
[0149] -Fold 3: 93%
[0150] -Fold 4: 95%
[0151] -Fold 5: 91%
[0152] -Fold 6: 94%
[0153] -Fold 7: 93%
[0154] -Fold 8: 92%
[0155] -Fold 9: 95%
[0156] -Fold 10: 94%
[0157] Average accuracy: 93.3%
[0158] Test set results
[0159] Assume that the evaluation results on the independent test set are as follows:
[0160] -Accuracy: 94%
[0161] -Recall (Gamma): 93%
[0162] -Recall (Neutron): 95%
[0163] -Recall (Beta): 92%
[0164] - Accuracy (Gamma): 94%
[0165] -Accuracy (Neutron): 95%
[0166] -Accuracy (Beta): 93%
[0167] These results show the good performance of the model in environments with high variability and different background radiation, ensuring the accuracy and reliability of radiation source identification. In this way, this experiment successfully demonstrated the application potential of machine learning technology in radiation monitoring in complex environments.
[0168] Summary of the 2 experiments:
[0169] Experiment 1: Basic radiation source identification
[0170] Features
[0171] Type of radiation: There are only two basic types of radiation involved, gamma rays and neutron rays.
[0172] Data processing: The standardization process is relatively simple, mainly including outlier processing and Z-score standardization.
[0173] Feature Engineering: Focuses on basic features such as energy and total counts, without involving complex or context-dependent features.
[0174] Application Scenario
[0175] It is suitable for scenarios with relatively stable environments and a single type of radiation source, such as basic nuclear facility monitoring or radiation detection in laboratory environments.
[0176] Ideal for education and primary research, providing basic radiation identification capabilities.
[0177] Experiment 2: Identification of multiple types of radiation sources in complex environments
[0178] Features
[0179] Data type: includes three radiation source types (gamma rays, neutron rays and beta rays), and introduces environmental variation factors such as background radiation.
[0180] Data processing: More complex data preprocessing, including min-max normalization of background radiation to adapt to environmental changes.
[0181] Feature engineering: Features include environmental factors, which more comprehensively describe the situation of monitoring data and are used to cope with environmental changes and complex data characteristics.
[0182] Application Scenario
[0183] It is suitable for scenarios with complex or changing environmental conditions, such as radiation monitoring at a nuclear accident site in the field or radiation safety management in industrial sites.
[0184] It is particularly suitable for advanced research and industrial applications that need to consider the influence of multiple radiation sources and environmental factors, and can provide more accurate monitoring and identification results.
[0185] Comprehensive comparison
[0186] Experiment 1 focuses more on basic radiation identification, is simple to operate and focuses on education and use by beginners. Experiment 2, on the other hand, is aimed at multi-factor radiation monitoring in complex environments, with more complex technical processing and stronger adaptability, and is more suitable for advanced applications and complex situations in actual operations.
[0187] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A radiation source identification algorithm based on machine learning, characterized in that: include: Collect radiation source data and perform preprocessing; Build machine models based on feature engineering; Train and validate machine models.
2. The radiation source identification algorithm based on machine learning as claimed in claim 1, characterized in that: The radiation source data includes radiation type, energy level and time stamp; The preprocessing includes outlier processing and standardization.
3. The radiation source identification algorithm based on machine learning as claimed in claim 2, characterized in that: The outlier processing includes extracting data features from the radiation source data, and using the IQR method to remove outliers from the data features. The IQR method includes calculating the first quartile and the third quartile, and calculating the quartile range, which is expressed as (IQR=Q3-Q1), wherein Q3 represents the third quartile, and Q1 represents the first quartile. If a data point in the radiation source data is lower than Q1-1.5×IQR or higher than Q3+1.5×IQR, it is identified as an outlier and the outlier is removed; The standardization includes calculating the mean value and standard deviation of the data characteristics and standardizing the radiation source data, which is expressed as: Among them, meanx represents the mean value of the data feature, and stdx represents the standard deviation of the data feature.
4. The radiation source identification algorithm based on machine learning as claimed in claim 3, characterized in that: The feature engineering includes energy spectrum analysis and time domain analysis; The energy spectrum analysis includes smoothing the energy spectrum data using moving average or Gaussian smoothing, identifying peaks in the energy spectrum using a local maximum search algorithm, and applying Gaussian fitting to each detected peak, expressed as: Where μ is the center of the peak and σ is the standard deviation; All data points of the energy spectrum are summed to obtain the total count, which is counted according to the number of peaks identified by the peak detection algorithm. The energy spectrum area without obvious radiation peaks is selected, and the average count of the selected background area is calculated as the background level.
5. The radiation source identification algorithm based on machine learning as claimed in claim 4, characterized in that: The time domain analysis includes differentiating the timestamps of consecutive radiation events, calculating the time interval between two consecutive events, identifying the start and end time points of each event, calculating the time difference from the start to the end, that is, the event duration, setting a fixed time window, counting the number of radiation events in each time window, dividing the total number of events by the length of the time window to obtain the event counting rate per unit time, selecting a time window for high-resolution monitoring, and counting the total number of radiation events in each window.
6. The radiation source identification algorithm based on machine learning as claimed in claim 5, characterized in that: The training of the machine model includes training the selected model using a training data set, which is expressed as: Among them, w represents a vector, b represents a bias term, C represents a regularization parameter, represents the loss function, n represents the total number of data points in the training dataset, and x i Represents the feature vector of the i-th training data point, y i represents the true label of the i-th training data point.
7. The radiation source identification algorithm based on machine learning as claimed in claim 6, characterized in that: The validation includes using a cross-validation method, where the data set is divided into k subsets, one subset is set aside each time as a validation set, and the rest are used as training sets, and this is repeated k times.
8. A system using a radiation source identification algorithm based on machine learning as claimed in any one of claims 1 to 7, characterized in that: It includes a data collection module (100), a data preprocessing module (200), a feature engineering module (300), and a machine learning and verification module (400); The data acquisition module (100) is responsible for collecting data from the radiation source; The data preprocessing module (200) is responsible for removing outliers in the data using the IQR method and performing standardization processing; The feature engineering module (300) is responsible for smoothing the energy spectrum data by moving average or Gaussian smoothing method, identifying peaks in the energy spectrum by using a local maximum search algorithm, and describing each peak by using Gaussian fitting; The feature engineering module (300) is also responsible for processing timestamp data, calculating the time intervals between consecutive radiation events, performing event duration analysis, and counting the number of radiation events within a fixed time window, thereby obtaining an event count rate per unit time; The machine learning and verification module (400) is responsible for training the model using the selected machine learning model and training data set to learn the relationship between the radiation source data and its category.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a radiation source identification algorithm based on machine learning described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a radiation source identification algorithm based on machine learning as described in any one of claims 1 to 7 are implemented.