Nuclide identification method based on time-sharing energy window enhancement

Abstract

The invention discloses a nuclide identification method based on time-sharing energy window enhancement, and the method comprises the steps: S1, carrying out the time-sharing energy window enhancement: carrying out the time-sharing energy window enhancement based on the segmentation of a time dimension and the dynamic adjustment of an energy window in a preset time-sharing energy window enhancement strategy, sampling a passive sample and an active sample obtained in the movement process of the detector; s2, peak ratio energy spectrum analysis: based on a peak ratio energy spectrum analysis method, completing construction of a cross-domain feature space capable of coding interaction between peak features and background features at the same time; and S3, feature space weighting decision making is carried out, task division is completed in the feature space, different features are endowed with differentiated weights, enhancement of interested features and suppression of redundant information are completed, and radionuclide identification based on feature space weighting is realized. The method is based on energy spectrum feature extraction of multi-space conversion projection and depth migration fusion, and effective identification of radioactive substances in a measurement scene in a moving state is realized.

Description

Technical Field

[0001] The present application belongs to the field of nuclear radiation detection technology, and in particular relates to a nuclide identification method based on time-sharing energy window enhancement. Background Art

[0002] When the relative positions of the detector and the radioactive source are fixed, the source term is typically linearly superimposed proportionally onto a measured background or a simple background model. However, in real measurement environments, a constant background rarely exists. When the detector is in motion, the background signal it acquires will constantly change. Furthermore, the shielding or dense materials surrounding the source and detector will attenuate gamma rays. Because the relative positions of the source and detector constantly change, the effective measurement time is relatively short, and statistical fluctuations caused by insufficient counting become more pronounced, making features easily obscured. The energy spectrum changes dynamically with the measurement environment, making it more difficult to obtain an effective energy spectrum. Traditional identification methods lack strong adaptability, further increasing the difficulty of radionuclide identification. Therefore, it is necessary to construct a classification model that is highly adaptable to background variations to achieve radionuclide identification.

[0003] To address this issue, some researchers choose to pre-process the energy spectrum by background subtraction. Background subtraction methods include peak area background subtraction and full spectrum background subtraction. The peak area background subtraction method estimates the background distribution in a narrow spectrum segment including the full energy peak, and then subtracts the background data channel by channel in the peak area. The full spectrum background subtraction method requires estimating the distribution pattern of the background in the entire measured energy spectrum, and then subtracting the background counts in that channel from the spectrum data channel by channel. SNIP (Sensitive Nonlinear Iterative Peak) is a commonly used full spectrum background subtraction method. In the process of background subtraction, although the background can be calculated and calibrated in a variety of ways, it is always a model built based on theoretical calculations and actual measurement experience, and some useful counts may be mistakenly judged as background counts. Summary of the Invention

[0004] The purpose of this application is to overcome the problems of the prior art and disclose a nuclide identification method based on time-sharing energy window enhancement. The method of this application is based on the energy spectrum feature extraction of multiple spatial transformation projections and depth migration fusion, which realizes the effective identification of radioactive substances in the measurement scene under mobile state.

[0005] The purpose of this application is achieved through the following technical solutions:

[0006] A nuclide identification method based on time-sharing energy window enhancement, the nuclide identification method based on time-sharing energy window enhancement comprising:

[0007] S1: Time-sharing energy window enhancement,

[0008] Based on the time dimension segmentation and dynamic adjustment of the energy window in the preset time-sharing energy window enhancement strategy, the passive samples and active samples obtained during the detector movement are sampled;

[0009] S2: Peak ratio spectrum analysis,

[0010] Based on the peak ratio energy spectrum analysis method, a cross-domain feature space is constructed that can simultaneously encode the interaction between peak features and background features.

[0011] S3: feature space weighted decision,

[0012] The task division is completed in the feature space, and by assigning differentiated weights to different features, the features of interest are enhanced and redundant information is suppressed, thus realizing radionuclide identification based on feature space weighting.

[0013] According to a preferred embodiment, step S1 includes: locating the time coordinates of the radiation source at the time point when the detector is closest to the radiation source position during the movement, and setting the energy window starting position based on the corresponding time coordinates to ensure that the center of the energy window is as close to the radiation source position as possible.

[0014] According to a preferred embodiment, step S1 further includes: constructing a time-sharing vector of the time series detector response data under the time-sharing energy window transformation based on the energy window number and energy window length parameters.

[0015] According to a preferred embodiment, step S2 includes:

[0016] First, the enhanced time series detector response data is converted to the dynamic energy interval analytical spectrum domain. By obtaining the global energy range and setting the energy interval granularity, the total energy interval is obtained. Based on this, the cumulative amount of photon events in the specific nth interval of the dynamic energy interval analytical spectrum domain for the kth time-sharing vector of the i-th enhanced sample is obtained to characterize the distribution of radioactive energy events with photon energy.

[0017] A cross-domain feature space is constructed by quantifying the relative intensity changes between the complex interactions between peak features and source domain background features during the decay of encoded radionuclides.

[0018] According to a preferred embodiment, the cross-domain feature space is configured to quantify the relative relationship between signals generated during the decay of radionuclides and to capture the intrinsic connections and common features between different domains.

[0019] According to a preferred embodiment, step S3 includes:

[0020] S31: Calculate the distance between the sample to be classified and the training sample according to the distance function, and adjust the distance function to avoid the influence of the denominator being zero and limiting the influence of sample points that are too far away;

[0021] S32: Select K nearest neighbor sample points of the sample to be classified, and assign different weights to these sample points according to the inverse distance weighted interpolation function;

[0022] S33: Using the weighted average method, combined with the category information and corresponding weights of the neighboring sample points, the predicted category of the sample to be classified is calculated to achieve radionuclide identification based on feature space weighting.

[0023] The aforementioned main solution of this application and its further options can be freely combined to form multiple solutions, all of which can be adopted and protected by this application. After understanding the solution of this application, those skilled in the art will understand that there are many combinations based on existing technology and common knowledge, all of which are technical solutions to be protected by this application, and these are not exhaustive here.

[0024] Beneficial effects of this application:

[0025] The method of this application realizes the effective identification of radioactive materials in the measurement scene under mobile state, and provides a new solution to the problem of nuclide identification in radioactive measurement scenes under mobile state in urban environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of the application method;

[0027] Figure 2 It is a flow chart of the time-sharing energy window enhancement of this application. DETAILED DESCRIPTION

[0028] The following describes the embodiments of the present application through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0029] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0030] Example 1

[0031] refer to Figure 1As shown, the present application discloses a nuclide identification method based on time-sharing energy window enhancement, and the nuclide identification method based on time-sharing energy window enhancement includes the following steps.

[0032] Step S1: Time-sharing energy window enhancement: Based on the time dimension segmentation and dynamic adjustment of the energy window in the preset time-sharing energy window enhancement strategy, the passive samples and active samples obtained during the movement of the detector are sampled.

[0033] refer to Figure 2 As shown, the time-sharing energy window enhancement process includes: formalizing time series features, designing a time-sharing energy window enhancement strategy, and generating and characterizing an enhanced dataset. The time series feature formalization process also includes associating labels with position identifiers. The time-sharing energy window enhancement strategy design process includes setting energy window parameters, designing the time-sharing energy window function, and setting the initial energy window position. The enhanced dataset generation and characterization process includes constructing the enhanced dataset, generating a new time series table, and analyzing the dataset representation.

[0034] Specifically, step S1 includes:

[0035] (1) The time coordinates of the radiation source are located by the time point when the detector is closest to the radiation source during its movement, and the starting position of the energy window is set based on the corresponding time coordinates to ensure that the center of the energy window is as close to the radiation source position as possible.

[0036] Time series features (mapping from time index to energy event) X i :

[0037] X i ={(t j ,e j )|j∈{1,2,…,m}}

[0038] Among them, i is the index of the sample, j is the traversal of the time series, that is, the value of j is an integer between 1 and m, t is the time point, and e is the energy value corresponding to the time point.

[0039] Each time series feature sample has a scalar value attribute l i , is used to uniquely identify the time point when the detector is closest to the radiation source during its movement. i =0, then the sample is a passive sample; if l i ≠0, the sample is an active sample.

[0040] (2) Based on the parameters of energy window number and energy window length, a time-sharing vector of the time series detector response data under the time-sharing energy window transformation is constructed.

[0041] Energy window parameter setting: Energy window number w q , energy window length w l , are positive integers, q and l represent the number and length of energy windows respectively.

[0042] The time series feature x i Split into multiple subsequences (i.e. "energy windows"):

[0043] Where, is the time-sharing energy window function, X ′ i Is the output enhancement matrix. Projection matrix X ′ i Contains w q Time-sharing vector The length of each time-sharing vector is w l +1.

[0044] in,

[0045] Where i is the index of the sample, w q is the number of energy windows, w l is the energy window length, k is the index of the time-sharing vector, and its value is [j ′ ,j ′ +w l ],j ′ is the initial position of the energy window, which is calculated by the following formula:

[0046]

[0047] Where b is a Boolean variable used to determine the energy window positioning benchmark: if it is an active sample, b is true, and the energy window positioning benchmark will be set according to l i and time point t in the time series j′ To calculate the optimal energy window starting position; if it is a passive sample, b is false, and the default energy window layout or other characteristics based on the time series will be used to calculate the starting position. q is the number of energy windows, w l is the energy window length, m is the time series feature X i The dimension, l i is the scalar-valued attribute of the time series feature, t j′ are the time series points obtained based on the calculated initial position of the energy window.

[0048] Step S2: Peak ratio spectrum analysis, based on the peak ratio spectrum analysis method, completes the construction of a cross-domain feature space that can simultaneously encode the interaction between peak features and background features.

[0049] Specifically, step S2 includes:

[0050] (1) First, the enhanced time series detector response data is converted into the dynamic energy interval analytical spectrum domain. By obtaining the global energy range and setting the subdivision granularity of the energy interval, the total energy interval is obtained. Based on this, the cumulative amount of photon events in the specific nth interval of the dynamic energy interval analytical spectrum domain of the kth time-sharing vector of the i-th enhanced sample is obtained to characterize the distribution of radioactive energy events with photon energy.

[0051] The total energy interval s is calculated by the following formula: In the formula, δ is the preset subdivision granularity, The minimum expected global energy range is: is the maximum expected global energy range:

[0052] In the formula, i is the index of the sample, j is the traversal of the time series, and e j is the time series feature sample X i The energy value in .

[0053] The accumulated amount of photon events in the specific nth interval of the kth time-sharing vector of the i-th enhanced sample in the dynamic energy interval analytical spectrum domain

[0054]

[0055] Where Φ represents the energy interval counting mechanism, which traverses the time-sharing vector Each radioactive energy event in the data is counted according to the specific interval to which it belongs; is the minimum expected global energy range, is the maximum expected global energy range; n is the index value of the total energy interval, which is [1, s-1]; δ is the preset subdivision granularity.

[0056] (2) A cross-domain feature space is constructed by quantifying the relative intensity changes between the complex interactions between the peak features and the background features of the source domain during the decay of the encoded radionuclides.

[0057] The cross-domain feature space is configured to quantify the relative relationships between the signals generated during the decay of radionuclides and to capture the intrinsic connections and common features between different domains.

[0058] Definition and area calculation of key areas:

[0059]

[0060] Where a c 、a f 、a u are the areas of the Compton plateau, the full energy peak and the auxiliary peak in the energy interval corresponding to the time-sharing spectrum vector, cl with c r are the left and right boundaries of Compton Flat, f l With f r is the left and right boundary of the full energy peak, a l with a r are the left and right boundaries of the auxiliary peak, The accumulated amount of photon events in the specific nth interval in the dynamic energy interval analytical spectrum domain of the kth time-sharing vector of the i-th enhanced sample.

[0061] Furthermore, we construct ratio features: Where r1 is the ratio of the total energy peak area to the Compton plateau area, and r2 is the ratio of the auxiliary peak area to the total energy peak area.

[0062] Generate cross-domain feature vectors: in, is the time-sharing spectrum vector of the i-th sample in the k-th time-sharing period, and is the ratio characteristic of the i-th sample in the k-th time-sharing period.

[0063] Step S3: Feature space weighted decision-making, completing task division in the feature space (i.e., converting the traditional classification problem into a feature space division task), and by assigning differentiated weights to different features, completing the enhancement of the features of interest and the suppression of redundant information, thus realizing radionuclide identification based on feature space weighting.

[0064] Specifically, step S3 includes:

[0065] (1) Calculate the distance between the sample to be classified and the training sample according to the distance function, and adjust the distance function to avoid the influence of the denominator being zero and limiting the sample points that are too far away.

[0066] The set of training sample feature vectors is C, and the feature matrix C0 corresponding to the sample to be classified S0 is:

[0067]

[0068] Where, is the time-sharing spectrum vector of the sample to be classified in the kth time-sharing period, k is the index of the time-sharing period, w q is the number of windows.

[0069] (2) Select the K nearest neighbor sample points of the sample to be classified, and assign different weights to these sample points according to the inverse distance weighted interpolation function.

[0070] The set of K sample points closest to the sample to be classified

[0071]

[0072] In the formula, C is the set of training sample feature vectors, C i is the training sample feature vector, i is the sample index, and the function rank(d(C0,C i )) represents the distance d(C0,C i ) are arranged in ascending order, and K is the number of sample points closest to the sample to be classified.

[0073] (3) Using the weighted average method, the category information and corresponding weights of the neighboring sample points are combined to calculate the predicted category of the sample to be classified, and radionuclide identification based on feature space weighting is realized.

[0074] Among them, the sample to be classified C0 and the sample point C i Distance between:

[0075] In the formula, C0 is the sample to be classified, C i is the nearest neighbor sample point, and the distance metric is L p norm, β is the index of the matrix dimension, for The βth dimension of for The β-th dimension of , p is the representation of the norm.

[0076] Get the classification function based on the weighted sample points in the local neighborhood Get the predicted category of the sample C0 to be classified:

[0077]

[0078] In the formula, C0 is the sample to be classified, C i is the nearest sample point, Represents the set of K nearest neighbor sample points of the target point C0, y i is the true label, and w(·) is the inverse distance difference function:

[0079]

[0080] Among them, d(C0,C i ) is the sample to be classified C0 and the sample point C i The distance between them, q is the normalized power, which can adjust the influence of the distance on the weight.

[0081] The method of this application realizes the effective identification of radioactive materials in the measurement scene under mobile state, and provides a new solution to the problem of nuclide identification in radioactive measurement scenes under mobile state in urban environment.

[0082] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A nuclide identification method based on time-sharing energy window enhancement, characterized in that: The nuclide identification method based on time-sharing energy window enhancement includes: S1: Time-sharing energy window enhancement, Based on the time dimension segmentation and dynamic adjustment of the energy window in the preset time-sharing energy window enhancement strategy, the passive samples and active samples obtained during the detector movement are sampled; S2: Peak ratio spectrum analysis, Based on the peak ratio energy spectrum analysis method, a cross-domain feature space is constructed that can simultaneously encode the interaction between peak features and background features. S3: feature space weighted decision, The task division is completed in the feature space, and by assigning differentiated weights to different features, the features of interest are enhanced and redundant information is suppressed, thus realizing radionuclide identification based on feature space weighting.

2. The nuclide identification method based on time-sharing energy window enhancement according to claim 1, characterized in that: Step S1 includes: locating the time coordinates of the radiation source by the detector at the time point when the detector is closest to the radiation source position during the movement, and setting the energy window starting position based on the corresponding time coordinates to ensure that the center of the energy window is as close to the radiation source position as possible.

3. The nuclide identification method based on time-sharing energy window enhancement according to claim 2, characterized in that: Step S1 also includes: constructing a time-sharing vector of the time series detector response data under the time-sharing energy window transformation based on the energy window number and energy window length parameters.

4. The nuclide identification method based on time-sharing energy window enhancement according to claim 3, characterized in that: Step S2 includes: First, the enhanced time series detector response data is converted to the dynamic energy interval analytical spectrum domain. By obtaining the global energy range and setting the energy interval granularity, the total energy interval is obtained. Based on this, the cumulative amount of photon events in the specific nth interval of the dynamic energy interval analytical spectrum domain for the kth time-sharing vector of the i-th enhanced sample is obtained to characterize the distribution of radioactive energy events with photon energy. A cross-domain feature space is constructed by quantifying the relative intensity changes between the complex interactions between peak features and source domain background features during the decay of encoded radionuclides.

5. The nuclide identification method based on time-sharing energy window enhancement according to claim 4, characterized in that: The cross-domain feature space is configured to quantify the relative relationships between signals generated during the decay of radionuclides and to capture the intrinsic connections and common features between different domains.

6. The nuclide identification method based on time-sharing energy window enhancement according to claim 1, characterized in that: Step S3 includes: S31: Calculate the distance between the sample to be classified and the training sample according to the distance function, and adjust the distance function to avoid the influence of the denominator being zero and limiting the influence of sample points that are too far away; S32: Select K nearest neighbor sample points of the sample to be classified, and assign different weights to these sample points according to the inverse distance weighted interpolation function; S33: Using the weighted average method, combined with the category information and corresponding weights of the neighboring sample points, the predicted category of the sample to be classified is calculated to achieve radionuclide identification based on feature space weighting.

Images