Neutron-gamma discrimination method in plastic scintillator based on pulse-shape discrimination

Abstract

The application discloses a plastic scintillator neutron-gamma discrimination method based on a pulse emission cortex, and comprises the following steps: acquiring neutron-gamma pulse signals of a mixed radiation field; establishing a pulse emission cortex model based on the neutron-gamma pulse signals; inputting the filtered neutron-gamma pulse signals into the pulse emission cortex model to generate a firing map containing the firing counts of each sampling point of the pulse signals; extracting the corresponding parts of falling edges and delayed fluorescence information according to the firing map, calculating a discrimination factor; and discriminating neutron-gamma pulse data by using the discrimination factor. The pulse emission cortex model has outstanding performance in extracting internal dynamic information of radiation pulse signals, improves discrimination accuracy, reduces calculation complexity, and enables the method to exhibit excellent neutron-gamma discrimination performance while consuming less time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radiation mixed field pulse signal processing, in particular to a plastic scintillator neutron-gamma discrimination method based on pulse firing cortex. BACKGROUND

[0002] At present, neutron detection technology has made great progress in modern industry in China, such as reactors, medical imaging, geography, nuclear radiation facilities, aerospace, etc. However, due to the interaction between neutrons and the surrounding environment, gamma rays will be accompanied, making it difficult to accurately obtain neutrons. Specifically, when using a radiation detector to detect neutrons, neutron-gamma mixed field pulse signals will be detected at the same time, and it is difficult to distinguish which is neutron and which is gamma ray. In order to solve this problem, Brooks et al. used organic scintillator detectors to find the inherent law of neutron-gamma pulse shape discrimination, which makes it possible to distinguish neutron-gamma by pulse signal. Based on this discovery, pulse shape discrimination (PSD) emerged, thus deriving neutron detectors with neutron-gamma discrimination capability applied in various scenarios, and a large number of neutron-gamma discrimination methods have been proposed, such as zero-crossing comparison method, charge comparison method, and falling edge percentage slope method, etc. These discrimination methods can be divided into time domain discrimination method, frequency domain discrimination method and intelligent discrimination method.

[0003] Generally speaking, intelligent methods are better than time domain and frequency domain methods in discrimination performance because of their better ability to obtain internal dynamic information of pulse signals, but they usually take a long time due to the need for complex matrix calculation and model training. With the rapid development of PSD technology, the requirement for its precision has also been improved, and Liu et al. proposed a discrimination method based on pulse coupled neural network in 2021. This method is particularly sensitive to the dynamic information of pulse signals, and can perform excellent discrimination effect under the same time consumption as the time domain discrimination method, fully demonstrating the significant effect of pulse coupled neural network model in the field of neutron-gamma discrimination. However, this method has a significant disadvantage, which is that it needs too many parameters set manually, which seriously hinders the wide application of the algorithm; and its processing efficiency of the signal is low, which needs a long time to extract useful information from the signal. SUMMARY

[0004] In view of the above problems in the prior art, the present application provides a plastic scintillator neutron-gamma discrimination method based on pulse firing cortex, which applies a more simplified pulse firing cortex model than the pulse coupled neural network model to neutron-gamma discrimination, achieving the advantages of fewer discrimination method parameters, low computational complexity and good discrimination performance.

[0005] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0006] A method for neutron-gamma discrimination in a plastic scintillator based on an impulse emission cortex, comprising the following steps:

[0007] S10, acquiring a neutron-gamma pulse signal of a mixed radiation field, and performing filtering processing;

[0008] S20, establishing an impulse emission cortex model based on the neutron-gamma pulse signal, the impulse emission cortex model configuring three parameters of a membrane potential, an output action potential and a dynamic threshold, the membrane potential of a neuron being calculated by a combination of direct stimulation and synaptic regulation from its adjacent neurons, when the membrane potential of the neuron exceeds its dynamic threshold, the neuron is activated and generates a peak, which further affects its adjacent neurons in the next iteration;

[0009] S30, inputting the filtered neutron-gamma pulse signal into the impulse emission cortex model to generate a firing map containing the firing times of each sampling point of the pulse signal;

[0010] S40, extracting a falling edge and a corresponding part of delayed fluorescence information from the firing map, integrating the firing times of the part, and calculating a discrimination factor;

[0011] S50, discriminating the neutron-gamma pulse data using the discrimination factor.

[0012] Specifically, the impulse emission cortex model is represented as:

[0013]

[0014]

[0015]

[0016] In the formula, n represents an iteration count, represents the membrane potential of a neuron at the ith sampling point of the neutron-gamma pulse signal in the nth iteration, f is a decay constant of the membrane potential, represents an external stimulus, represents a synaptic weight matrix, which controls the connection between the neuron at the ith sampling point of the neutron-gamma pulse signal and its surrounding neurons at the jth sampling point, represents the output action potential of the neuron, i.e. the output pulse, and the convolution of and represents the modulation of the surrounding neurons at the jth sampling point to the central neuron at the ith sampling point of the neutron-gamma pulse signal, is the dynamic threshold of the neuron at the i-th sampling point, g and h are the decay constants of the threshold and the absolute refractory period, respectively, which prevent a neuron that has just been activated from being reactivated immediately.

[0017] Specifically, in the step S40, the discriminant factor is calculated by integrating the number of firings of the falling edge and the corresponding part of the delayed fluorescence information extracted from the firing map.

[0018] Compared with the prior art, the present application has the following beneficial effects:

[0019] The present application improves and optimizes the existing discrimination method by establishing a pulse firing cortex model, effectively simplifies the model parameter configuration, and significantly improves the discrimination performance and reduces the processing time.

[0020] It should be noted that the pulse firing cortex model has been used in the field of image processing in the past, such as image segmentation, image fusion, image denoising, etc., and has shown excellent dynamic information extraction capability in two-dimensional and three-dimensional data processing of images and videos; however, the pulse firing cortex model SCM has rarely been used for processing of one-dimensional signals, and this model and method have never been used for radiation pulse signal processing and particle discrimination. Based on long-term and in-depth research in the field of radiation pulse signal discrimination, the inventors acutely discovered the similarity between the signal to be processed and the application model, and ingeniously applied the pulse firing cortex model to radiation pulse signal discrimination and verified the excellent effect of the method designed in the present application through experiments. The present application uses SCM to extract the dynamic information contained in the neutron-gamma pulse signals of the mixed field of radiation, effectively amplifies the differences between the two types of radiation pulse signals, and utilizes the pulse amplitude of each point of the pulse signal and the decay trend characteristics of the pulse signal covered in the dynamic information to better reflect the differences between the neutron signal and the gamma signal at the falling edge and the afterglow effect positions, thereby achieving the purpose of efficiently discriminating the two types of particles, neutrons and gammas.

[0021] More specifically, in image and video processing work, each neuron in the pulse cortex model network is closely connected with the surrounding 8 neurons, and since the number and position of neurons in SCM are one-to-one corresponding to the pixel points in the image to be processed, SCM considers the influence of the gray values of the 8 surrounding pixel points when processing each pixel point in the image. In the present application, the neutron-gamma pulse data to be processed is a one-dimensional time series signal, and each neuron only needs to be coupled with its previous and next neurons, and only needs to consider the influence of the amplitudes of the previous and next sampling points, greatly reducing the amount of data to be processed. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1Flowchart of the embodiment of the present application.

[0023] Figure 2 Schematic diagram of the pulse emitting cortex model in the embodiment of the present application.

[0024] Figure 3 Schematic diagram of the process of receiving high-intensity stimulation and low-intensity stimulation by the SCM.

[0025] Figure 4 Normalized neutron-gamma pulse waveform chart in the embodiment of the present application.

[0026] Figure 5 Neutron-gamma pulse signal firing chart obtained according to the waveform chart in the embodiment of the present application.

[0027] Figure 6 Schematic diagram of the calculation of the evaluation standard FoM value of the neutron-gamma discrimination.

[0028] Figure 7 Discrimination result scatter plot of the comparison of the four methods in the embodiment of the present application.

[0029] Figure 8 Discrimination factor curve of the comparison of the four methods in the embodiment of the present application.

[0030] Figure 9 is a schematic diagram of the change of the FoM value after changing the parameters in the embodiment of the present application, wherein, Figure 9a is a change chart after changing the iteration number n, Figure 9b is a change chart after changing the threshold decay constant g, Figure 9c is a change chart after changing the membrane potential decay constant f, Figure 9d is a change chart after changing the synaptic weight matrix W, Figure 9e is a change chart after changing the absolute refractory period decay constant h. DETAILED DESCRIPTION

[0031] The present application will be further described below in conjunction with the drawings and embodiments, and the embodiments of the present application include but are not limited to the following embodiments.

[0032] EMBODIMENT

[0033] As Figure 1 shown, the plastic scintillator neutron-gamma discrimination method based on the pulse emitting cortex includes the following steps:

[0034] S10, obtaining the neutron-gamma pulse signal of the mixed radiation field and performing filtering processing;

[0035] S20, establishing a pulse firing cortex model based on the neutron-gamma pulse signal, the pulse firing cortex model configuring three parameters of membrane potential, output action potential and dynamic threshold, the membrane potential of a neuron being calculated through a combination of direct stimulation and synaptic regulation from its adjacent neurons, the neuron being activated and generating a peak when the membrane potential of the neuron exceeds its dynamic threshold, the peak further affecting its adjacent neurons in the next iteration;

[0036] S30, inputting the filtered neutron-gamma pulse signal into the pulse firing cortex model to generate a firing map containing the firing counts of each sampling point of the pulse signal;

[0037] S40, extracting the corresponding part of falling edge and delayed fluorescence information from the firing map to calculate a discrimination factor; wherein the extracted part contains the part of the firing map that contains important information of the neutron-gamma pulse signal, mainly the part containing the falling edge and delayed fluorescence information, for example, the subsequent experiment selects more than ten points before the peak to more than two hundred points after the peak, and the firing times of this part are integrated when calculating the discrimination factor. For discrete radiation pulse sampling signals, the obtained firing map is also discrete, and the integration of the firing times in this interval can also be regarded as the summation of the firing times in this interval;

[0038] S50, discriminating the neutron-gamma pulse data using the discrimination factor.

[0039] Specifically, the pulse firing cortex model is represented as:

[0040]

[0041]

[0042]

[0043] In the formula, n represents the iteration count, represents the membrane potential of the neuron at the i-th sampling point of the neutron-gamma pulse signal in the n-th iteration, f is the decay constant of the membrane potential, represents external stimulation, represents a synaptic weight matrix, which controls the connection between the neuron at the i-th sampling point of the neutron-gamma pulse signal and its surrounding neurons at the j-th sampling point, represents the output action potential of the neuron, i.e. the output pulse, and the convolution of and represents the modulation of the surrounding neurons at the j-th sampling point to the central neuron at the i-th sampling point of the neutron-gamma pulse signal, is the dynamic threshold of the neuron at the i-th sampling point, and g and h are the decay constants of the threshold and absolute refractory period, respectively. It prevents the newly activated neuron from being immediately reactivated.

[0044] Intuitively speaking, as Figure 2 As shown, there are three important components in the SCM: membrane potential and membrane potential. Output action potential Dynamic threshold They are all closely interconnected; in the final iteration, their individual changes affect the others. The potential of a component is determined by its own potential, and beyond that... Also suffer and external stimuli The impact, Depend on Adjustment, and Directly depends on membrane potential The magnitude relationship between the dynamic threshold and the dynamic threshold.

[0045] Corresponding to the Weber-Fechner law, a psychophysical law describing human visual characteristics, the intensity of human subjective perception. With time matrix It is related that the human visual system is more sensitive to low intensity than to high intensity. Just like the human visual system... Figure 3 The model of pulsed cortical motility (SCM) receiving high-intensity stimulation was drawn in the paper. and low stimulation The results show that for two equal threshold intervals, the processing time for high stimulation Δt1 is much shorter than that for low stimulation Δt2, indicating that the pulse firing cortical model SCM is more sensitive to low intensity than high intensity.

[0046] When using the pulse emission cortex model (SCM) to distinguish neutron-gamma pulse signals, the pulse signal is first processed by the SCM to generate an ignition map, which contains the ignition count for each sampling point of the pulse signal, such as... Figure 4 and Figure 5 As shown, the main difference between neutron-gamma pulse signal shapes lies in the steepness of the fall edge (i.e., the presence or absence of delayed fluorescence). Specifically, the fall edge of gamma rays is steeper than that of neutrons, and delayed fluorescence is a unique characteristic of neutrons. These two differences can be distinguished by SCM and are clearly magnified in the ignition diagram. Neutron-gamma pulse signals can be well distinguished by integrating the corresponding portions of the ignition diagram that contain fall edge and delayed fluorescence information. Compared to gamma rays, neutron pulse signals have a larger integral.

[0047] For ease of comparison, this article provides a brief introduction to three other identification methods for reference.

[0048] Pulse coupled neural network (PCNN) is a kind of bionic neural network based on Eckhorn cortical model, which is derived from the study of the interaction between cell groups in the cat primary visual cortex, introduced into the field of image processing by Johnson et al. in 1994. Unlike ordinary neural networks, PCNN does not need a pre-training process to form the relationship between input and output data. On the contrary, PCNN works in a way similar to real biological neurons, using the change of action potential when neurons receive stimuli to solve scene analysis problems. In the past few decades, it has developed rapidly in the field of imaging processing, such as feature extraction, image segmentation, pattern recognition, object recognition and image shadow removal. There are three main domains in PCNN, which are receptive domain, modulation domain and pulse generation domain, and the receptive domain is further divided into two parts: link input (LI) and feedback input (FI). FI plays a major role in the action potential of neurons, while LI regulates the action potential of neurons. When the action potential of neurons exceeds its dynamic threshold, the neuron is activated and produces a spike. Its mathematical expression is as follows:

[0049] The mathematical expression is as follows:

[0050]

[0051]

[0052] U ij [n]=F ij [n]{1+βL ij [n]}, (3)

[0053]

[0054]

[0055] where the internal activity of the neuron located at (i,j)U i,j is determined by the feedback input F ij and the link input L ij , which are coupled by a factor called link strength β, n is the iteration count. α F and α L represent the decay time constant of FI and LI, respectively. V F and V L represent the amplification factor of FI and LI, respectively. For the central neuron (i,j) of this position, it is connected with the adjacent neurons (k,l) located through the constant synaptic weight matrix W and M, S ij is the input stimulus, θ ij is the dynamic threshold of the neuron at position (i,j), α θ is the decay time constant of the dynamic threshold, Vθ Y represents the amplification factor for the dynamic threshold. ij It determines whether the neuron located at (i,j) should be activated (U). ij [n]>θ ij [n],Y ij [n]=1)or not(U ij [n]≤θ ij [n],Y ij [n] = 0).

[0056] Charge comparison (CC) is a widely used method in neutron-gamma discrimination due to its efficiency and stability. This method is based on the different ways neutrons and gamma rays interact with the sensitive matter being detected; specifically, the charge ratio of neutrons differs from that of gamma rays. By calculating the charge ratio, Hawkes et al. successfully discriminated between neutron and gamma signals. The formula is as follows:

[0057]

[0058] Q N , represents the integral of the voltage of the slow component of a pulse signal, while Q M This represents the integral of the voltage across the entire signal. Considering the longer decay time of the neutron signal's falling edge and the delayed fluorescence of neutrons, the R-value of neutrons is larger than that of gamma rays.

[0059] In the zero-crossing comparison method (ZC), pulse shape information is determined by an index called the zero-crossing time, which is calculated by converting a neutron or gamma-ray pulse signal into a bipolar pulse. The zero-crossing time is the interval between the start time of the bipolar pulse and its zero-crossing point. Regarding the transformation to a bipolar pulse signal, M. Nakhostin proposed a digital filter whose recursive form is derived by calculating its Z-transfer function. The recursive formula for the bipolar signal can be expressed as follows:

[0060]

[0061] Where x is the pulse signal of neutrons or gamma rays, y represents the filtered pulse, i.e., the bipolar signal, n represents the sample index, and δ and ω are constants.

[0062] δ=e -T / τ (8)

[0063]

[0064] where T is the sampling interval of the pulse signal, and τ = RC represents the shaping time. The time reaching 10% of the pulse maximum is set as the start time, and the first sample below zero after the pulse peak is set as the zero-crossing time. Since the falling edge of the gamma-ray pulse signal is steeper than that of the neutron pulse signal, the zero-crossing time of the gamma-ray is shorter than that of the neutron.

[0065] Evaluation criteria for each discrimination method

[0066] The figure of merit FoM is a standard measure for evaluating the discrimination effect of neutrons and gamma rays. First, the calculation of FoM requires drawing a histogram of the discrimination results, mainly including the neutron count spectrum and the gamma-ray count spectrum. Then, a Gaussian fitting function is used to fit the two sets of data to form a fitting curve, which can be further used to calculate the distance S between the two peaks and the half-height width of the two peaks. Finally, FoM is defined as:

[0067]

[0068] where S represents the interval between the neutron peak and the gamma peak, FWHM γ represents the full width at half maximum of the gamma peak (i.e., the full width at half the peak height), FWHM n is the full width at half maximum of the neutron peak, as shown in Figure 6 . If the discrimination performance is good, S should be larger, the FoM value should be larger, and the discrimination performance should be better.

[0069] The discrimination results are verified by experiments and compared as follows:

[0070] Parameter settings:

[0071] The optimized values of the PCNN parameters are: n = 180, α F = 0.32, α L = 0.356, α θ = 0.08, V F = 0.0005, V L = 0.0005, V θ = 15, M = N = [0.1409, 0.1409]. The integration interval of the ignition map is selected between 10 ns before the pulse signal peak and 120 ns after the pulse signal peak.

[0072] For the charge comparison method, the total component is selected between 10 ns before the pulse signal peak and 90 ns after the pulse signal peak, and the slow component is composed of the interval between 27 ns and 200 ns after the pulse signal peak.

[0073] For the zero-crossing comparison method, T = 1 ns and τ = 72 ns.

[0074] The parameter settings of SCM are: n = 36, f = 0.8, g = 0.704, h = 18.2, The interval between 5 ns before the peak of the pulse signal and 125 ns after the peak of the pulse signal is selected as the integration interval of the ignition map.

[0075] Experimental equipment and conditions: A 241Am-Be isotope neutron source with an average energy of 4.5 MeV was used to generate a mixed field of neutrons and gamma rays. The pulse signal was detected by a plastic scintillator (EJ299-33) and a digital oscilloscope, with a bandwidth of 200 MHz, a sampling rate of 1 GS / s, a pulse duration of 160 ns, and a trigger threshold of 500 mV, corresponding to an energy of about 1.6 MeVee. The collected pulse signal information meets the standard of the Poisson sampling law. A total of 9414 pulse signals were obtained, and their pulse shapes are shown in Figure 4 Obviously, compared with gamma rays, the decay rate of the light emitted by neutrons is significantly slower, which explains why the decay time of neutrons is longer and neutrons have the unique characteristic of delayed fluorescence.

[0076] Before the discrimination process, all neutron-gamma pulse signals need to be filtered by Fourier transform to reduce noise. It is worth noting that there are many different filtering methods in the field of neutron-gamma discrimination, such as wavelet transform, Kalman filtering, and moving average filtering, etc. Different filtering methods plus different discrimination methods may show different discrimination effects. Since the main discussion topic of this paper is not related to the filtering method, we chose the most commonly used filtering method, Fourier transform, to apply to all discrimination methods used in this study, in order to control the experimental variables and compare them under the same conditions as much as possible. Then, the filtered neutron-gamma pulse signals can be distinguished by the aforementioned four methods (CC, ZC, PCNN, SCM), and offline processing is performed on Windows 11 using an AMD 5900X CPU. The results are shown in Figure 7 and Figure 8

[0077] As Figure 7 ​The scatter plots of the neutron-gamma pulse signal discrimination effects of the four methods are shown, which are divided into two groups of points by a crossing line. The points below the line are identified as gamma-ray signals, and the points above the line are identified as neutron signals. The discrimination factors of different discrimination methods are normalized to the range of 0-40, and in the results of the charge comparison method, some signals deviate too far from the neutron count; therefore, we move the range of the charge comparison method from the range of 0-40 to the range of about 10-50, aiming to achieve a universal discrimination line that can span the gap between the neutron group and the gamma-ray group of all discrimination methods. For good discrimination performance, the gamma scatter points and the neutron scatter points should be as far apart from each other as possible, and each group should be concentratedly distributed while showing a clear Gaussian distribution. It is obvious that the performance of PCNN and SCM is significantly better than that of the zero-crossing comparison method and the charge comparison method, and there is a clear gap between the neutron group and the gamma-ray group. In addition, the point group distribution of the neutron-gamma pulse signal count of the two methods is obviously concentrated and maintains the characteristics of Gaussian distribution, with few discrete points between or outside the two groups.

[0078] As shown in Figure 8 The discrimination factor curve of the comparison of the four methods is shown, which is obtained by Gaussian fitting of the discrimination result histogram of each method, for calculating the FoM value of each method (calculated according to the aforementioned formula 10), and also for intuitively viewing the discrimination performance. Good discrimination effect has two characteristics: (i) a wide gap between the two peaks, and (ii) a narrow half-height width of each peak. Since the internal principles of each method are different, the discrimination factors produced by each method for drawing the histogram are different, and therefore the number of boxes in the histogram is also different, and the count in each box is also different, which results in very high Y-axis values for some methods; however, the value of the Y-axis is irrelevant to the discrimination effect. Returning to the analysis of the figure, it can be seen that the discrimination effect of PCNN and SCM is obviously better than that of other methods, with a clear wide gap between the peaks, and a relatively narrow half-height width of each peak.

[0079] In order to further distinguish the performance of PCNN and SCM, objective evaluation criteria are needed. First, the FoM value of each method is calculated for evaluating the discrimination effect, and second, the discrimination time of each method is calculated for evaluating the complexity and real feasibility of the method to achieve discrimination. In the experiment, the discrimination time for processing 9414 pulse signals, and the FoM value of each discrimination method are shown in Table 1 below.

[0080] Table 1 Discrimination time and FoM value of different discrimination methods

[0081] Discrimination method PCNN Charge comparison method Zero-crossing comparison method SCM Discrimination time 2.22s 1.84s 0.39s 0.54s Discrimination effect (FoM value) 1.709 1.533 1.097 2.304

[0082] According to the results shown in Table 1, compared with the use of Figure 7 and 8The subjective evaluation results are consistent, and the discrimination performance of PCNN and SCM is obviously better than that of the zero-crossing comparison method and the charge comparison method, and the FoM value is obviously larger. The discrimination effect of SCM is better than that of other discrimination methods. Compared with PCNN, the FoM value is increased by 34.81%, and the discrimination time is shortened by 75.67%. Compared with the charge comparison method, the FoM value is increased by 50.29%, and the discrimination time is shortened by 70.65%. Compared with the zero-crossing comparison method, the FoM value is increased by 110.02%, and the discrimination time is increased by 38.4%. The outstanding discrimination performance of SCM is due to its excellent ability to identify the internal dynamic information of the pulse signal, which is the most important ability to distinguish neutrons and gamma rays. Compared with PCNN, SCM improves the discrimination accuracy and reduces the computational complexity. This excellent performance is due to the improvement of the accuracy and computational ability of SCM, and the number of iterations of SCM is less than that of PCNN, which has a great influence on the time consumption of discrimination. Although the PCNN model performs well when applied to two-dimensional image processing, its complex design requires more iteration counts to capture the information contained in the data fed back to it, while the SCM as a simplified model requires much fewer iterations, especially when it needs to process one-dimensional matrices, such as pulse signals. This makes the SCM-based neutron-gamma discrimination method applicable to many occasions that require accurate discrimination of neutrons and gammas, and the shorter discrimination time proves the application potential of the method in real-time discrimination. The short discrimination time of SCM is at the same level as the real-time discrimination method such as the zero-crossing comparison method, making it possible to realize high-quality real-time discrimination.

[0083] Regarding the parameters of the pulse firing cortex model SCM in the application, the number of SCM parameters is much less than that of PCNN, but these parameters still affect the discrimination effect. Therefore, this part discusses the experimental discussion of the parameter setting of SCM. Based on the experimental parameter setting described above, the experimental conditions at this time are designed as follows: change one parameter while keeping the other parameters at the set value; record the data, and plot the FoM value change curve according to the data as shown in FIG. 9. Among them, Figure 9a is the change graph after changing the iteration number n, Figure 9b is the change graph after changing the threshold decay constant g, Figure 9c is the change graph after changing the membrane potential decay constant f, Figure 9d is the change graph after changing the synaptic weight matrix is the change graph after changing, Figure 9e is the change graph after changing the absolute refractory period decay constant h.

[0084] It can be seen that by selecting appropriate parameters, the FoM value can be easily stabilized between 1.5 and 2.3, indicating that SCM does not rely much on parameter setting to achieve good discrimination performance. For example, Figure 9aAs shown, for the iteration count n, the FoM value variation curve exhibits obvious periodic fluctuations, with larger peaks when the number of iterations is low. This is because, as the number of iterations increases, the general action potential of each neuron initially rises and then tends to stabilize. This means that the action potential initially depends more on the external stimulus and then becomes more dependent on the neuron's action potential in the previous iteration. In neutron-gamma discrimination, the information contained in the external stimulus is more important, and its modulation of the potential in the previous iteration should not be too large; therefore, a smaller iteration count value should be used. Figure 9b As shown, the decay constant g of the threshold determines the decay rate of the threshold and affects the activation frequency of neurons. The FoM value variation curve also exhibits a periodic fluctuation pattern, and its value can be selected at the peak of the curve. Figure 9c As shown, for the decay constant f of the membrane potential, there is an optimal range of values ​​around 0.8. f controls the decay rate of the neuron's membrane potential. If its value is too large, the contribution from the previous iteration will be too small; conversely, if its value is too small, its contribution will be too large. Therefore, finding an optimal range in the middle is entirely reasonable. Figure 9d As shown, for the synaptic weight matrix It controls the connections between the central neuron and its neighboring neurons, when When the value is too far from 0.44, the FoM value curve shows an overall downward trend. The value of S significantly influences the contribution of the membrane potential U of neighboring neurons, which determines the proportion of external stimuli in U. As mentioned earlier, the contribution of external stimulus S is crucial for neutron-gamma discrimination; it needs to dominate in U. Although modulation of neighboring neurons is also important for information retrieval in the SCM, the modulation level should not be too high to exceed that of external stimuli, nor too low to negate the modulation effect. Therefore, performance is good when the value is in the middle. Finally, as Figure 9e As shown, for the absolute refractory period decay constant h, the FoM value curve rises sharply in the first step and then stabilizes after 18. This parameter determines the dynamic threshold. This characteristic means that activated neurons will not be immediately reactivated, and their values ​​should be chosen to be larger within a reasonable range.

[0085] Generally, the FoM value variation curves of many parameters show periodic fluctuations within a reasonable range, and when the parameters are selected too extreme and beyond the reasonable range, the discrimination performance will be very poor. The external reason is that different parts of the SCM are closely connected, and if the characteristics of a single component change too far, it will affect other parts, so that others adjust the part to work normally, thus causing periodic fluctuations. However, when the parameter value is too extreme and cannot maintain the connection and mutual modulation relationship between different parts of the SCM, the discrimination performance of the SCM will decrease sharply. By using the above parameter variation mode, the appropriate parameters of the SCM can be selected to achieve good neutron-gamma discrimination performance.

[0086] The above embodiments are only preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any changes made by using the design principles of the present application and on this basis without creative labor shall belong to the protection scope of the present application.

Claims

1. A neutron-gamma discrimination method for plastic scintillators based on pulsed emission cortex, characterized in that, The method comprises the following steps: S10, acquiring a neutron-gamma pulse signal of a mixed field of radiation and performing filtering processing; S20, establishing a pulse firing cortex model based on the neutron-gamma pulse signal, the pulse firing cortex model being configured with three parameters of a membrane potential, an output action potential and a dynamic threshold value, a membrane potential of a neuron being calculated through a combination of direct stimulation and synaptic regulation from adjacent neurons of the neuron, the neuron being activated and generating a peak value when the membrane potential of the neuron exceeds a dynamic threshold value of the neuron, the peak value further affecting adjacent neurons of the neuron in a next iteration; S30, inputting the neutron-gamma pulse signal after the filtering processing into the pulse firing cortex model to generate a firing map containing a firing number of each sampling point of the pulse signal; S40, extracting a corresponding part of falling edges and delayed fluorescence information from the firing map, integrating the firing number of the part, and calculating a discrimination factor; S50, discriminating neutron-gamma pulse data using the discrimination factor.

2. The method of claim 1, wherein the plastic scintillator is a pulse emitting cortical based plastic scintillator. The pulse firing cortex model is represented as: where n denotes the iteration count, denotes the membrane potential of a neuron located at the i-th sample point of the neutron-gamma pulse signal at the n-th iteration, f is a decay constant for the membrane potential, denotes an external stimulus, denotes a synaptic weight matrix controlling the connections between a neuron located at the i-th sample point of the neutron-gamma pulse signal and its surrounding neurons located at the j-th sample point, represents the output action potential, i.e. the output pulse, of a neuron, and denotes the modulation of a neuron located at the j-th sample point by a neuron located at the i-th sample point of the neutron-gamma pulse signal, is the dynamic threshold value of a neuron at the i-th sample point, g and h are decay constants for the threshold value and the absolute refractory period, respectively, which prevents a neuron that has just been activated from being reactivated immediately.

3. The method of claim 2, wherein the plastic scintillator is a pulse emitting cortical based plastic scintillator. In the step S40, the discrimination factor is calculated by integrating the firing number of the corresponding part of the falling edges and the delayed fluorescence information extracted from the firing map.

Images