A particle similarity waveform identification method, system, device and medium

Abstract

The present invention proposes a particle similarity waveform identification method, system, device and medium, which relates to the field of particle detection. The method comprises: in response to a pulse signal exceeding a predetermined trigger threshold, acquiring and generating a digital waveform D based on the sampling point information before and after the acquired pulse signal. n ; The digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the same waveform as the digital waveform D n Ignition diagram with consistent dimensions, the ignition diagram carries the waveform characteristic information of the amplified pulse signal; the part of the ignition diagram carrying the waveform characteristic information is integrated to obtain the discrimination factor R; the discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain the quality factor value FOM, and the cutoff value L of the gamma and neutron discrimination factors is calculated according to the 3σ principle; based on the quality factor value FOM and the cutoff value L, the discrimination results of gamma and neutron are obtained. This scheme can be used to quickly and accurately discriminate particles with similar waveforms.

Description

Technical Field

[0001] The present invention relates to the field of particle detection, and in particular to a particle similarity waveform identification method, system, equipment and medium. Background Art

[0002] With the rapid development of nuclear technology, particle detection technology has played an important role in national defense, medicine, industry, agriculture, science and technology, and nuclear physics research. However, in the actual particle detection process, due to the existence of inelastic scattering of particles with the surrounding environment and radiation capture of particles, there may be many different particles in the radiation field. Detectors that are sensitive to one particle are often also sensitive to other particles, and it is difficult to distinguish between particles. Therefore, how to identify particles with similar pulse waveforms in a mixed field has become one of the urgent problems to be solved in the field of radiation detection. Summary of the invention

[0003] The object of the present invention is to provide a particle similarity waveform identification method, system, device and medium, which can quickly and accurately perform particle similarity waveform identification.

[0004] The present invention is achieved in that:

[0005] In a first aspect, the present application provides a particle similarity waveform identification method, comprising the following steps:

[0006] In response to the pulse signal exceeding a predetermined trigger threshold, the digital waveform D is generated based on the sampling point information before and after the acquired pulse signal. n ; The above digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the digital waveform D n An ignition diagram with consistent dimension, the above ignition diagram carries waveform characteristic information of the amplified pulse signal; the part of the ignition diagram carrying the waveform characteristic information is integrated to obtain a discrimination factor R; the above discrimination factor R is successively subjected to histogram statistics and Gaussian fitting to obtain a quality factor value FOM, and according to the 3σ principle, a cutoff value L of the gamma and neutron discrimination factors is calculated; based on the quality factor value FOM and the cutoff value L, the discrimination results of gamma and neutron are obtained.

[0007] Further, based on the above scheme, the above digital waveform D n Import the pre-trained RCNN model serially or in parallel, including: n After performing error waveform removal, amplitude normalization and filtering processing in sequence, they are imported into the pre-trained RCNN model in series or in parallel; the mathematical expression of the above filtering processing is: x[k] is the data waveform before filtering, k is the digital waveform index, M is the filter window length, and y[k] is the data waveform after filtering.

[0008] Furthermore, based on the above scheme, the mathematical expression of the above RCNN model is:

[0009]

[0010]

[0011]

[0012]

[0013] Among them, (i, j) is the position coordinate of the internal active neuron, n is the number of iterations of the neuron, and U ij represents the internal activity of neurons, S ij represents external stimulation, β represents the link coefficient, which is used to adjust the intensity of internal neuronal activity; a U represents the attenuation coefficient of the internal activity, V U represents a constant weighting coefficient used to modulate the internal activity characteristics; θ ij represents the dynamic threshold of the neuron, α θ Represents the attenuation coefficient of the neuron dynamic threshold, V θ Represents the amplification factor of the neuron's dynamic threshold; Represents the random dropout weight matrix, which is composed of G ijkl With D ijkl Calculated by element-by-element multiplication; G ijkl represents the Gaussian kernel, which follows the Gaussian distribution; D ijkl Represents a 0-1 random matrix, and the probability of each element appearing 1 also follows a Gaussian distribution.

[0014] Furthermore, based on the aforementioned scheme, the above RCNN model includes a reception domain, a modulation domain and a pulse emission domain.

[0015] Furthermore, based on the above scheme, the parameters of the above RCNN model are configured as follows: β = 1.0, V U =1.0, a U =0.1,α θ =0.005, V θ =50.0, G ijkl Obey X~N(0,0,5 2 , 5 2 , 0), size is 9×9, D ijkl The probability of 1 occurring in X~N(0,0,5 2 , 5 2 , 0).

[0016] Further, based on the above scheme, the above discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain the quality factor value FOM, and the boundary value L of the gamma and neutron discrimination factors is calculated according to the 3σ principle, including: performing histogram statistics on the discrimination factor R to obtain a histogram of the discrimination factor R. Performing Gaussian fitting on the above histogram to obtain the corresponding parameter μ a , μ b , σ a , σ b , S, δ a , δ b ; where μ a is the mean of the gamma peak, μ b is the mean of the neutron peak, σ a is the standard deviation of the gamma peak, σ b is the standard deviation of the neutron peak, S is the distance between the gamma peak and the neutron peak, δ a is the width of the half-Gaussian height of the gamma peak, δ b is the width of the half-Gaussian height of the neutron peak, the left Gaussian peak is the gamma peak, and the right Gaussian peak is the neutron peak. The quality factor FOM is calculated. The cutoff value L of the gamma and neutron discrimination factors is calculated.

[0017] Further, based on the aforementioned scheme, the above-mentioned discrimination results of gamma and neutron based on the quality factor value FOM and the cutoff value L are obtained, including: when the discrimination factor R is greater than or equal to the cutoff value L, it is determined to be a neutron; when the discrimination factor R is less than the cutoff value L, it is determined to be a gamma ray.

[0018] In a second aspect, the present application provides a particle similarity waveform identification system, which includes:

[0019] The acquisition module is configured to: in response to the pulse signal exceeding a predetermined trigger threshold, acquire and generate a digital waveform D based on the acquired sampling point information before and after the pulse signal n ; Ignition module, configured to: the above digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the digital waveform D nAn ignition diagram with consistent dimensions, the above ignition diagram carries waveform characteristic information of the amplified pulse signal; an integral module is configured to: perform integral calculation on the part of the ignition diagram carrying the waveform characteristic information to obtain a discrimination factor R; a factor module is configured to: perform histogram statistics and Gaussian fitting on the above discrimination factor R in turn to obtain a quality factor value FOM, and calculate a cutoff value L of the gamma and neutron discrimination factors according to the 3σ principle; a discrimination module is configured to: obtain the discrimination results of gamma and neutron based on the quality factor value FOM and the cutoff value L.

[0020] In a third aspect, the present application provides an electronic device comprising at least one processor, at least one memory and a data bus; wherein: the processor and the memory communicate with each other via the data bus; the memory stores program instructions to be executed by the processor, and the processor calls the program instructions to execute a method as described in any one of the first aspects above.

[0021] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method as described in any one of the above-mentioned first aspects.

[0022] Compared with the prior art, the present invention has at least the following advantages or beneficial effects:

[0023] The present invention proposes a particle similarity waveform identification method, which extracts key information from pulse signals through multiple iterations of the RCNN algorithm, constructs an ignition diagram based on a neuron link structure, and thus quickly and accurately extracts the waveform features of the pulse information, and performs amplification analysis, so as to facilitate the identification of similar pulse shapes while reducing the computational cost and shortening the identification time. That is, it is possible to quickly and reliably identify similar pulse shapes while ensuring that the algorithm complexity is low, which is convenient for practical engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 It is a flow chart of an embodiment of a particle similarity waveform identification method of the present invention;

[0026] Figure 2 Schematic diagram of the RCNN model in one embodiment of the present invention;

[0027] Figure 3 Schematic diagram of the calculation principle of the quality factor in one embodiment of the present invention

[0028] Figure 4 It is a structural block diagram of an embodiment of a particle similarity waveform identification system of the present invention;

[0029] Figure 5 A structural block diagram of an electronic device provided by an embodiment of the present invention.

[0030] Icons: 1. Acquisition module; 2. Ignition module; 3. Integration module; 4. Factor module; 5. Identification module; 6. Processor; 7. Memory; 8. Data bus. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0032] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0033] Example 1

[0034] Pulse waveform discrimination (PSD) is a method widely used in particle physics, nuclear medicine and other fields. It identifies the type of particles based on the difference in pulse waveforms caused by the different decay times of different particles in organic scintillators. PSD algorithms can be mainly divided into time domain methods, frequency domain methods and intelligent algorithms.

[0035] There are some problems in the identification process of time domain method and frequency domain method, such as poor noise resistance and low identification efficiency. Therefore, seeking a method with high identification efficiency, low computational complexity and good noise resistance has become an urgent problem to be solved in the field. To solve these problems, relevant scholars have proposed a series of intelligent algorithms, such as support vector machine method (SVM), Gaussian mixture model method (GMM), recurrent neural network (RNN), etc. These methods have good noise resistance and excellent identification effect, but they have high computational complexity and long identification time, which cannot meet the needs of practical applications.

[0036] Therefore, an embodiment of the present application provides a particle similarity waveform identification method, which quickly and accurately extracts key features from pulse signals and performs amplified analysis, thereby facilitating the identification of similar pulse shapes while reducing computing costs and shortening identification time, thereby achieving simple and fast identification of particles.

[0037] See also Figure 1 , the particle similarity waveform identification method comprises the following steps:

[0038] Step S101: In response to a pulse signal exceeding a predetermined trigger threshold, acquiring and generating a digital waveform D based on the acquired sampling point information before and after the pulse signal n .

[0039] For example, a radiation detector and a high-speed pulse acquisition card can be used to perform self-triggering digital waveform sampling and storage on similar pulse waveforms of the radiation mixed field. The acquisition card sets a reasonable trigger threshold according to the use requirements. When the pulse signal exceeds the threshold, the digital memory will automatically store the sampling point information before and after the pulse signal to form a digital waveform D n .

[0040] Step S102: convert the digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the digital waveform D n An ignition diagram with consistent dimensions, wherein the ignition diagram carries waveform characteristic information of the amplified pulse signal.

[0041] Among them, the RCNN model (Random-coupled Neural Network) is a neural network with random response characteristics. The unique random response characteristics are: RCNN uses a coupling matrix with random characteristics when working. This coupling matrix represents the random coupling between the central neuron and the adjacent neurons and has a stronger biological basis. Therefore, in the RCNN model, key information in the pulse signal is extracted through multiple iterations to perform digital waveform D n The ignition mapping can improve the subsequent particle identification accuracy. In the RCNN model, the dropout technology can be used, that is, randomly disconnecting the links between some neurons, thereby reducing the network calculation burden and improving the accuracy, noise resistance and biological neural response performance. In some implementations of the present invention, the number of ignitions at each sampling point comes from the joint contribution of the amplitude of the original signal corresponding to the sampling point and the dynamic information of the original signal.

[0042] In a specific implementation of the present invention, when the digital waveform D n Before importing the pre-trained RCNN model serially or in parallel, you can first perform a nPreprocessing includes the removal of erroneous signals (such as flat peaks, pulse accumulation), replication normalization and filtering. The filtering process is mainly to reduce the digital waveform D n The noise in the image can be reduced to improve the recognition efficiency.

[0043] Specifically, MATLAB can be used to identify and present the error waveform, and then the mathematical expression For digital waveform D n Perform amplitude normalization and then use the mathematical expression For digital waveform D n Perform filtering processing, where x[k] is the data waveform before filtering, k is the digital waveform index, M is the filtering window length, and y[k] is the data waveform after filtering.

[0044] Please continue reading Figure 1 , Step S103: Integrate the portion of the ignition diagram that carries the waveform feature information to obtain a discrimination factor R.

[0045] In the RCNN model, a single digital waveform D n After ignition processing is performed at each sampling point, the ignition status at each point (1 represents ignition, 0 represents non-ignition) is integrated and calculated, and finally the single digital waveform D is obtained. n Ignition diagram with consistent dimensions. The value of each position in the ignition diagram represents the number of ignitions at that sampling point. The original digital waveform D n The larger the amplitude is, the larger the ignition value is, and vice versa. Specifically, the rising edge, falling edge and afterglow effect part of the original pulse signal in the ignition diagram can be integrated, and the obtained integral value is the discrimination factor R of the pulse signal.

[0046] Step S104: performing histogram statistics and Gaussian fitting on the discrimination factor R in turn to obtain a quality factor value FOM, and calculating a cutoff value L of the gamma and neutron discrimination factors according to the 3σ principle;

[0047] Step S105: Based on the quality factor value FOM and the cutoff value L, the gamma and neutron discrimination results are obtained.

[0048] After obtaining the discrimination factor R, further analysis can be performed to obtain the corresponding quality factor value FOM and the cutoff value L of the gamma and neutron discrimination factor, so that subsequent analysis and discrimination of similar pulse shapes can be performed based on the quality factor value FOM and the cutoff value L, and discrimination of gamma and neutron can be performed. It is found through actual measurement that, as shown in Table 1, the scheme provided by the present invention is used to discriminate similar pulse shapes, and the quality factor is improved by 74.05% compared with the existing charge comparison method, the quality factor is improved by 107.27% compared with the zero-crossing method, and the quality factor is improved by 14.00% compared with the pulse coupled neural network, and the discrimination efficiency is significantly improved.

[0049]

[0050] Table 1

[0051] Among them, the 3σ principle is also known as the Laida criterion. It first assumes that a set of test data contains only random errors, calculates and processes them to obtain the standard deviation, and determines an interval according to a certain probability. It is believed that any error exceeding this interval is not a random error but a gross error, and the data containing this error should be eliminated. The 3σ principle specifically includes: the probability that the value is distributed in (μ-σ, μ+σ) is 0.6827; the probability that the value is distributed in (μ-2σ, μ+2σ) is 0.9545; the probability that the value is distributed in (μ-3σ, μ+3σ) is 0.9973. Therefore, an energy window of ±3σ of the distribution center can be selected to statistically analyze the PSD value within the energy window, thereby determining the boundary value L of the gamma and neutron discrimination factors.

[0052] In addition, in some specific implementations of the present invention, the identification results of gamma and neutrons are obtained based on the quality factor value FOM and the cutoff value L, including:

[0053] When the discrimination factor R is greater than or equal to the cutoff value L, it is determined to be a neutron; when the discrimination factor R is less than the cutoff value L, it is determined to be a gamma ray. The corresponding mathematical expression is: if Among them, neutron represents neutron and gamma-ray represents gamma ray.

[0054] like Figure 2 As shown, in some specific implementations of the present invention, the discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain a quality factor value FOM, and the boundary value L of the gamma and neutron discrimination factors is calculated according to the 3σ principle, including:

[0055] Performing histogram statistics on the discrimination factor R to obtain a histogram of the discrimination factor R;

[0056] Perform Gaussian fitting on the histogram to obtain the corresponding parameter μ a , μb , σ a , σ b , S, δ a , δ b ; where μ a is the mean of the gamma peak, μ b is the mean of the neutron peak, σ a is the standard deviation of the gamma peak, σ b is the standard deviation of the neutron peak, S is the distance between the gamma peak and the neutron peak, δ a is the width of the half-Gaussian height of the gamma peak, δ b is the width of the half-Gaussian height of the neutron peak, the left Gaussian peak is the gamma peak, and the right Gaussian peak is the neutron peak;

[0057] based on Calculate the quality factor value FOM; and

[0058] based on The cutoff value L of the gamma and neutron discrimination factors is calculated.

[0059] Exemplarily, a statistical histogram of the discrimination factor R can be drawn using software such as MATLAB, and then Gaussian fitting is performed on the drawn statistical histogram to obtain two Gaussian peaks to obtain corresponding parameters, thereby continuing to calculate the quality factor value FOM and the boundary value L of the gamma and neutron discrimination factors.

[0060] Based on the above scheme, in some embodiments of the present invention, the mathematical expression of the RCNN model is:

[0061]

[0062]

[0063]

[0064]

[0065] Among them, (i, j) is the position coordinate of the internal active neuron, n is the number of iterations of the neuron, and U ij represents the internal activity of neurons, S ij represents external stimulation, β represents the link coefficient, which is used to adjust the intensity of internal neuronal activity; a U represents the attenuation coefficient of the internal activity, V U represents a constant weighting coefficient used to modulate the internal activity characteristics; θ ij represents the dynamic threshold of the neuron, α θ Represents the attenuation coefficient of the neuron dynamic threshold, V θ Represents the amplification factor of the neuron's dynamic threshold; Represents the random dropout weight matrix, which is composed of G ijkl With D ijkl Calculated by element-by-element multiplication; G ijkl represents the Gaussian kernel, which follows a Gaussian distribution; D ijkl Represents a 0-1 random matrix, and the probability of each element appearing 1 also follows a Gaussian distribution.

[0066] Among them, the functions of the parameters involved in the mathematical expression of the RCNN model are as follows:

[0067] β: Link coefficient, a factor used to modulate internal activity, generally takes a value between [0, 1]. The larger the β, the stronger the coupling effect between external stimulation and feedback input, and vice versa;

[0068] V U : constant weighting coefficient, used to amplify the coupling effect of surrounding neurons. In biological neurons, it can be expressed as the inherent potential of the internal activity of neurons, and its value is generally in [0, 10];

[0069] a U : The internal activity attenuation coefficient represents the decay rate of the neuron's internal activity, and is generally between [0, 1]. The larger its value, the greater the internal activity potential U ij There will be a faster decay to the dynamic threshold θ ij Under these conditions, neurons are more difficult to fire.

[0070] α θ : The attenuation coefficient of the dynamic threshold, which is an important parameter, represents the attenuation speed of the dynamic threshold, and is generally taken in the range of [0, 1]. The larger the value, the faster the dynamic threshold attenuation speed, and the internal activity potential U ij It will be easier to exceed the dynamic threshold θ ij Causing neurons to fire.

[0071] V θ : Dynamic threshold amplification factor, used to control the dynamic threshold θ ij The parameter of the amplification effect can be expressed as the neuron dynamic threshold θ in biological neurons. ij The inherent potential of is generally in the range of [0, 100].

[0072] G ijkl : The initial link weight matrix represents the connection strength between the central neuron and the surrounding neurons. In this model, the model follows X~N(0, 0, 5 2 , 5 2, 0), which means: the closer to the central neuron, the greater the connection strength between neurons, and vice versa. This is consistent with biological characteristics. In the standard parameters of this model, its size is 9×9, which determines how many neighboring neurons a central neuron has direct information exchange with. In the case of 9×9, this number is 80. Since the random Dropout process significantly reduces the amount of computation of RCNN, the weight matrix of RCNN can have a larger size, realizing extensive neuron connections and direct information exchange. In fact, the computational burden brought by the extremely large weight matrix of size 20×20 and nearly 400 direct neuron connections is also bearable in RCNN.

[0073] D ijkl : 0-1 random matrix, used to randomly open and close the value of the initial link weight. 1 in the matrix represents an open channel. The coefficient of the connection between neurons is G ijkl The value of the corresponding position, 0 means that there is no connection between the neuron at this position and the central neuron. The probability of 1 in this matrix follows (0, 0, 5 2 , 5 2 , 0), and randomization is performed in each iteration.

[0074] Based on the above scheme, in some embodiments of the present invention, the RCNN model includes a reception domain, a modulation domain and a pulse emission domain. Figure 3 As shown in Figure 1, the basic functions of RCNN are realized through the mutual influence, mutual restriction and coupling of these three parts. The domain F of neurons receiving external stimulation signals and surrounding neuron stimulation signals ij and L ij , and couple its input modulation part to obtain the internal activity of the neuron U ij When the dynamic threshold θ is exceeded ij When the neuron is fired and outputs a pulse train Y ij At this time, the dynamic threshold θ ij Will rapidly increase V θ To inhibit the next firing of neurons. ij ,θ ij and Y ij The parameters are closely related, and any change in any parameter will affect other parameters. With the continuous input of external stimulus signals, U ij The growth rate of will gradually slow down, and θ ij The growth rate remains unchanged, and eventually θ ij Will exceed U ij And reset the neuron. A neuron is fired a limited number of times when it receives the same intensity stimulus. ij >θ ijThe number of times is called the ignition number. The original signal is input into the RCNN model, and RCNN will output a matrix with the same dimension as the original signal. The element value of the matrix represents the number of ignitions at the corresponding position of the original signal. The matrix is ​​called the ignition map.

[0075] For the RCNN model, in order to improve its discrimination effect, it is necessary to rely on a set of appropriate neural network parameters. Therefore, based on the above scheme, in some embodiments of the present invention, the parameter configuration of the RCNN model is: β = 1.0, V U =1.0, a U =0.1,α θ =0.005, V θ =50.0, G ijkl Obey X~N(0,0,5 2 , 5 2 , 0), size is 9×9, D ijkl The probability of 1 occurring in X~N(0,0,5 2 , 5 2 , 0). Of course, a set of empirical parameters for pulse shape discrimination is given here for example only, and it can be understood that the setting of the parameters is not limited thereto.

[0076] Example 2

[0077] See also Figure 4 , the embodiment of the present application provides a particle similarity waveform identification system, which includes:

[0078] The acquisition module 1 is configured to: in response to the pulse signal exceeding a predetermined trigger threshold, acquire and generate a digital waveform D based on the sampling point information before and after the acquired pulse signal. n ; Ignition module 2, is configured to: the above digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the digital waveform D n An ignition diagram with consistent dimensions, the above ignition diagram carries waveform characteristic information of the amplified pulse signal; an integral module 3 is configured to: perform integral calculation on the part of the ignition diagram carrying the waveform characteristic information to obtain a discrimination factor R; a factor module 4 is configured to: perform histogram statistics and Gaussian fitting on the above discrimination factor R in turn to obtain a quality factor value FOM, and calculate a cutoff value L of the gamma and neutron discrimination factors according to the 3σ principle; a discrimination module 5 is configured to: obtain the discrimination results of gamma and neutron based on the quality factor value FOM and the cutoff value L.

[0079] For the specific implementation process of the above system, please refer to the particle similarity waveform identification method provided in Example 1, which will not be repeated here.

[0080] Example 3

[0081] See also Figure 5 An embodiment of the present application provides an electronic device, which includes at least one processor 6, at least one memory 7 and a data bus 8; wherein: the processor 6 and the memory 7 communicate with each other through the data bus 8; the memory 7 stores program instructions that can be executed by the processor 6, and the processor 6 calls the program instructions to execute a particle similarity waveform identification method.

[0082] For example, to implement:

[0083] In response to the pulse signal exceeding a predetermined trigger threshold, the digital waveform D is generated based on the sampling point information before and after the acquired pulse signal. n ; The digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the same waveform as the digital waveform D n An ignition diagram with consistent dimension, the ignition diagram carries the waveform characteristic information of the amplified pulse signal; the part of the ignition diagram carrying the waveform characteristic information is integrated to obtain the discrimination factor R; the discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain the quality factor value FOM, and according to the 3σ principle, the boundary value L of the gamma and neutron discrimination factors is calculated; based on the quality factor value FOM and the boundary value L, the discrimination results of gamma and neutron are obtained.

[0084] Among them, the memory 7 can be but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.

[0085] The processor 6 may be an integrated circuit chip with signal processing capability. The processor 6 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0086] Understandably, Figure 5 The structure shown is for illustration only. The electronic device may also include Figure 5 More or fewer components as shown, or with Figure 5 Different configurations are shown. Figure 5 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0087] Example 4

[0088] The present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor 6, a particle similarity waveform identification method is implemented. For example, the following is implemented:

[0089] In response to the pulse signal exceeding a predetermined trigger threshold, the digital waveform D is generated based on the sampling point information before and after the acquired pulse signal. n ; The digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the same waveform as the digital waveform D n An ignition diagram with consistent dimension, the ignition diagram carries the waveform characteristic information of the amplified pulse signal; the part of the ignition diagram carrying the waveform characteristic information is integrated to obtain the discrimination factor R; the discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain the quality factor value FOM, and according to the 3σ principle, the boundary value L of the gamma and neutron discrimination factors is calculated; based on the quality factor value FOM and the boundary value L, the discrimination results of gamma and neutron are obtained.

[0090] If the above functions are implemented in the form of software function modules 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 application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes 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 application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0091] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A particle similarity waveform identification method, characterized in that: The following steps are involved: In response to the pulse signal exceeding a predetermined trigger threshold, the digital waveform D is generated based on the sampling point information before and after the acquired pulse signal. n ; The digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the digital waveform D n An ignition diagram of consistent dimensions, the ignition diagram carrying waveform characteristic information of the amplified pulse signal; Integrate the part of the ignition diagram that carries the waveform characteristic information to obtain the discrimination factor R; The discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain a quality factor value FOM, and the cutoff value L of the gamma and neutron discrimination factors is calculated according to the 3σ principle; Based on the quality factor value FOM and the cutoff value L, the discrimination results of gamma and neutron are obtained; The mathematical expression of the RCNN model is: Among them, (i, j) is the position coordinate of the internal active neuron, n is the number of iterations of the neuron, and U ij represents the internal activity of neurons, S ij represents external stimulation, β represents the link coefficient, which is used to adjust the intensity of internal neuronal activity; a U represents the attenuation coefficient of the internal activity, V U represents a constant weighting coefficient used to modulate the internal activity characteristics; θ ij represents the dynamic threshold of the neuron, α θ Represents the attenuation coefficient of the neuron dynamic threshold, V θ Represents the amplification factor of the neuron's dynamic threshold; Represents the random dropout weight matrix, which is composed of G ijkl With D ijkl Calculated by element-by-element multiplication; G ijkl represents the Gaussian kernel, which follows a Gaussian distribution; D ijkl Represents a 0-1 random matrix, and the probability of each element appearing 1 also follows a Gaussian distribution.

2. A particle similarity waveform identification method as claimed in claim 1, characterized in that: The digital waveform D n Import pre-trained RCNN models serially or in parallel, including: In the digital waveform D n After performing error waveform removal, amplitude normalization and filtering processing in sequence, they are imported into the pre-trained RCNN model in series or in parallel; wherein the mathematical expression of the filtering processing is: x[k] is the data waveform before filtering, k is the digital waveform index, M is the filter window length, and y[k] is the data waveform after filtering.

3. A particle similarity waveform identification method as claimed in claim 1, characterized in that: The RCNN model includes a reception domain, a modulation domain and a pulse emission domain.

4. A particle similarity waveform identification method as claimed in claim 1, characterized in that: The parameter configuration of the RCNN model is: β = 1.0, V U =1.0, a U =0.1,α θ =0.005, V θ =50.0, G ijkl Obey X~N(0,0,5 2 , 5 2 , 0), size is 9×9, D ijkl The probability of 1 occurring in X~N(0,0,5 2 , 5 2 , 0).

5. A particle similarity waveform identification method as claimed in claim 1, characterized in that: The discrimination factor R is subjected to histogram statistics and Gaussian fitting in turn to obtain the quality factor value FOM, and the boundary value L of the gamma and neutron discrimination factors is calculated according to the 3σ principle, including: Performing histogram statistics on the discrimination factor R to obtain a histogram of the discrimination factor R; Perform Gaussian fitting on the histogram to obtain the corresponding parameter μ a , μ b , σ a , σ b , S, δ a , δ b ; where μ a is the mean of the gamma peak, μ b is the mean of the neutron peak, σ a is the standard deviation of the gamma peak, σ b is the standard deviation of the neutron peak, S is the distance between the gamma peak and the neutron peak, δ a is the width of the half-Gaussian height of the gamma peak, δ b is the width of the half-Gaussian height of the neutron peak, the left Gaussian peak is the gamma peak, and the right Gaussian peak is the neutron peak; based on Calculate the quality factor value FOM; and based on The cutoff value L of the gamma and neutron discrimination factors is calculated.

6. A particle similarity waveform identification method as claimed in claim 5, characterized in that: The gamma and neutron discrimination results are obtained based on the quality factor value FOM and the cutoff value L, including: When the discrimination factor R is greater than or equal to the cutoff value L, it is determined to be a neutron; when the discrimination factor R is less than the cutoff value L, it is determined to be a gamma ray.

7. A particle similarity waveform identification system, characterized in that: include: The acquisition module is configured to: in response to the pulse signal exceeding a predetermined trigger threshold, acquire and generate a digital waveform D based on the acquired sampling point information before and after the pulse signal n ; The ignition module is configured to: convert the digital waveform D n Import the pre-trained RCNN model serially or in parallel to obtain the digital waveform D n An ignition diagram of consistent dimensions, the ignition diagram carrying waveform characteristic information of the amplified pulse signal; The mathematical expression of the RCNN model is: Among them, (i, j) is the position coordinate of the internal active neuron, n is the number of iterations of the neuron, and U ij represents the internal activity of neurons, S ij represents external stimulation, β represents the link coefficient, which is used to adjust the intensity of internal neuronal activity; a U represents the attenuation coefficient of the internal activity, V U represents a constant weighting coefficient used to modulate the internal activity characteristics; θ ij represents the dynamic threshold of the neuron, α θ Represents the attenuation coefficient of the neuron dynamic threshold, V θ Represents the amplification factor of the neuron's dynamic threshold; Represents the random dropout weight matrix, which is composed of G ijkl With D ijkl Calculated by element-by-element multiplication; G ijkl represents the Gaussian kernel, which follows a Gaussian distribution; D ijkl Represents a 0-1 random matrix, and the probability of each element appearing 1 also follows a Gaussian distribution; The integration module is configured to: perform integration calculation on the part of the ignition diagram that carries the waveform characteristic information to obtain a discrimination factor R; The factor module is configured to: perform histogram statistics and Gaussian fitting on the discrimination factor R in sequence to obtain a quality factor value FOM, and calculate a cutoff value L of the gamma and neutron discrimination factors according to the 3σ principle; The identification module is configured to obtain the identification results of gamma and neutron based on the quality factor value FOM and the cutoff value L.

8. An electronic device, characterized in that: It includes at least one processor, at least one memory and a data bus; wherein: the processor and the memory communicate with each other through the data bus; the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images