Radiation field dose self-discriminating detection system and method thereof

Abstract

The present application relates to nuclear radiation detection and nuclear electronics technical field, specifically to a kind of radiation field dose self-discrimination detection system and method;It contains double-channel signal acquisition circuit, phase space mapping and manifold curvature entropy analysis module;System utilizes original signal and analog differential signal to construct real-time phase space trajectory, and introduces reference particle attractor model;Its core is to calculate trajectory manifold curvature entropy and carry out hierarchical discrimination: low entropy is determined as single particle, high entropy is determined as accumulation distortion and triggers inverse calculation repair process to deconstruct it;Finally, the energy is updated dose library by phase space closed loop volume integral calculation;The present application uses hardware differentiation to avoid digital quantization noise, ensures that trajectory topological surface is smooth, and significantly improves the measurement accuracy under high dynamic range.

Description

Technical Field

[0001] This invention relates to the fields of nuclear radiation detection and nuclear electronics technology, specifically to a radiation field dose self-discrimination detection system and method. Background Technology

[0002] When dealing with high-throughput radiation field monitoring tasks, the scintillator detection front end responds to randomly incident particle streams and generates high-density photoelectric pulse signals. These signals frequently exhibit waveform overlap and baseline drift within an extremely short time window.

[0003] To obtain radiation dose data, existing processing schemes generally rely on time-domain voltage amplitude analysis or conventional digital differential algorithms to identify pulse events. While such schemes possess basic detection capabilities in low-background environments, in high-count-rate scenarios, due to a lack of in-depth analysis of the signal topology, the system struggles to effectively distinguish between normal single-event events and waveform-coupled stacking distortion events. Furthermore, traditional techniques often employ threshold-based stacking rejection strategies, leading to the loss of a significant number of effective counts, and directly differentiating the digitally sampled signal introduces high-frequency quantization noise, severely limiting the system's energy resolution and measurement accuracy.

[0004] Therefore, how to accurately identify and non-destructively deconstruct the accumulated pulse signal while suppressing noise interference, so as to improve the counting throughput and energy spectrum reconstruction accuracy of the detection system in a high dynamic range, has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to provide a radiation field dose self-discrimination detection system and method to solve the problems mentioned in the background art. Specifically, the technical solution of this invention is as follows:

[0006] A radiation field dose self-discrimination detection method includes:

[0007] A scintillator detection front-end and a dual-channel signal acquisition circuit are configured, wherein the dual-channel signal acquisition circuit includes a raw signal channel and an analog differential channel;

[0008] Initialize a preset set of reference particle phase space trajectories, which records the closed orbit coordinate data of standard particle events in phase space, and serves as the reference particle attractor model;

[0009] In response to the photoelectric pulse signal output by the scintillator detection front end, a dose self-discrimination process is triggered, including:

[0010] Step 1: Based on the original signal channel and the analog differential channel, simultaneously acquire the real-time amplitude signal and the analog differential signal of the photoelectric pulse signal;

[0011] Step 2: Combining the real-time amplitude signal and the analog differential signal, the real-time phase space trajectory of the pulse signal is constructed through phase space mapping processing;

[0012] Step 3: Using the manifold curvature entropy calculation logic, perform topological feature analysis on the real-time phase space trajectory and calculate the trajectory manifold curvature entropy;

[0013] Step 4: Based on the comparison results of the trajectory manifold curvature entropy and the preset distortion threshold, perform hierarchical identification processing: if the value of the trajectory manifold curvature entropy is lower than the preset distortion threshold, it is determined to be a single particle event, and the particle type is identified according to the benchmark particle attractor model; if the value of the trajectory manifold curvature entropy is higher than or equal to the preset distortion threshold, it is determined to be a stacking distortion event, triggering the trajectory repair inverse calculation process to deconstruct the stacking distortion event into multiple independent particle events.

[0014] Step 5: Calculate particle energy and update radiation field dose library based on single-particle events or deconstructed independent particle events through phase space closed-loop volume integration.

[0015] Step 6: Extract geometric micro-perturbation features based on the real-time phase space trajectory, compare them with the preset crystal optical path difference reference dataset, determine the incident ray azimuth angle, and directionally weight the dose accordingly.

[0016] Preferably, the synchronous data acquisition of this system includes:

[0017] The time-domain voltage value of the photoelectric pulse signal is obtained through a high-speed analog-to-digital converter in the original signal channel.

[0018] The photoelectric pulse signal is hardware differentiated using an analog differentiating circuit, and the voltage change rate of the photoelectric pulse signal is obtained through a high-speed analog-to-digital converter of the analog differentiating channel.

[0019] The time-domain voltage value and the voltage change rate value are strictly aligned in time to generate a state vector.

[0020] Preferably, constructing the real-time phase space trajectory of the pulse signal includes:

[0021] A two-dimensional phase plane is established with the time-domain voltage value as the horizontal axis and the voltage change rate as the vertical axis.

[0022] The state vectors of a continuous time series are mapped to a two-dimensional phase plane to generate closed or open geometric curves, which serve as real-time phase space trajectories.

[0023] Preferably, topological feature analysis of real-time phase space trajectories is performed using manifold curvature entropy calculation logic, including:

[0024] Calculate the local geometric curvature of each discrete point on the real-time phase space trajectory;

[0025] The distribution of local geometric curvature along the time axis is statistically analyzed, and a curvature distribution histogram is generated.

[0026] The information entropy of the curvature distribution histogram is calculated to obtain the trajectory manifold curvature entropy, which characterizes the degree of chaos of the trajectory shape relative to the standard closed loop.

[0027] Preferably, the trigger trajectory repair reverse calculation process includes:

[0028] Locate topological singularities in the real-time phase space trajectory where curvature abruptly changes;

[0029] Using topological singularities as segmentation nodes and leveraging the geometric constraints of the baseline particle attractor model, multiple fitting segmentation is performed on the real-time phase space trajectory.

[0030] Reconstruct the segmented trajectory fragments to generate at least two independent closed regression trajectories, which serve as independent particle events after deconstruction.

[0031] Preferably, the method involves phase space closed-loop volume integration, including:

[0032] Obtain the closed phase space orbits corresponding to the identified single-particle events or independent particle events;

[0033] Calculate the geometric measure enclosed by the closed phase space orbit within the phase space, which is the geometric area in the two-dimensional phase plane and the geometric volume in the three-dimensional phase space;

[0034] The pre-calibrated energy response function is invoked, and the geometric measure is used as the input variable to calculate the corresponding particle deposition energy value, wherein the energy response function defines the numerical mapping relationship between the phase space closed loop area and the deposition energy.

[0035] Preferably, the method further includes a virtual collimation direction discrimination step:

[0036] Extract geometric micro-perturbation feature data on the real-time phase space trajectory;

[0037] The geometric micro-perturbation feature data is compared with a preset crystal optical path difference reference dataset, which contains optical path difference features corresponding to different incident azimuth angles of the incident rays relative to the detector crystal axis.

[0038] Determine the incident azimuth angle of the current incident ray, and perform directional weighting on the radiation field dose library based on the incident azimuth angle.

[0039] A radiation field dose self-discrimination detection system includes:

[0040] The detection front-end module is configured to convert incident rays into photoelectric pulse signals;

[0041] The phase extraction module includes a parallel raw signal channel and an analog differential channel, configured to acquire the real-time amplitude signal and the analog differential signal respectively;

[0042] The topology discrimination engine, configured in the FPGA processing unit, is used to construct real-time phase space trajectories and calculate the trajectory manifold curvature entropy;

[0043] The logic branch unit is configured to execute the following logic: compare the trajectory manifold curvature entropy with a preset distortion threshold. If the entropy value is less than the threshold, activate the single particle recognition logic; if the entropy value is greater than or equal to the threshold, activate the stacking deconstruction logic.

[0044] Stacking deconstruction unit is used to geometrically segment and reconstruct the trajectory based on the baseline particle attractor model when stacking distortion occurs;

[0045] The dose calculation unit is used to calculate energy based on phase space closed-loop volume integral and output dose data.

[0046] Compared with the prior art, the present invention has the following improvements and advantages:

[0047] 1. This invention employs a dual-channel parallel acquisition architecture comprising a raw signal channel and an analog differential channel. It utilizes analog circuitry for hardware differential processing to directly acquire a smooth voltage change rate signal. This design effectively avoids the quantization noise amplification effect caused by directly performing differential operations on discrete digital signals in traditional techniques, ensuring the high smoothness and continuity of the generated phase space trajectory. This provides a high-quality data foundation for subsequent curvature-based fine topology analysis, thereby guaranteeing the system's measurement accuracy and stability within a high dynamic range.

[0048] 2. This invention introduces phase space mapping and manifold curvature entropy calculation logic to transform complex waveform overlaps and baseline fluctuations in the time domain into topological changes in the phase space geometric trajectory (such as knots and bifurcations). By calculating the dimensionless characteristic quantity of trajectory manifold curvature entropy, the system can quantify the degree of chaos of the signal at the global topological level. Compared with traditional methods that rely solely on amplitude thresholds, this technique is naturally immune to baseline drift and can rapidly and accurately identify hidden accumulation events with normal amplitude but abnormal shape, achieving accurate monitoring of high-throughput radiation fields without the need for additional physical shielding.

[0049] 3. This invention proposes a trajectory repair inverse calculation process based on a benchmark particle attractor model, achieving a technological leap from "discarding accumulation" to "resisting accumulation." When accumulation distortion is detected, the system does not discard data but utilizes topological singularity localization and multiple fitting segmentation techniques to deconstruct and reconstruct the complex accumulation signal into multiple independent closed regression trajectories. Experiments demonstrate that this method can significantly reduce the count loss rate under high-flux radiation fields, successfully rescuing effective particle counts from distorted signals, thereby ensuring the accuracy and integrity of energy spectrum measurements.

[0050] 4. This invention utilizes phase-space closed-loop volume integrals instead of traditional time-domain integrals to characterize particle energy. This calculation method naturally cancels out high-frequency noise interference with zero mean, thereby significantly improving energy resolution. Simultaneously, by extracting geometric micro-perturbation features on the phase-space trajectory and comparing them with crystal optical path difference reference data, virtual collimation direction discrimination is achieved. This allows the detection system to determine the azimuth angle of the incident ray and perform dose-direction weighting through software algorithms without adding bulky mechanical structures such as lead collimators, enhancing the system's portability and intelligence. Attached Figure Description

[0051] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0052] Figure 1 This is a flowchart of the method of the present invention;

[0053] Figure 2 This is a structural diagram of the system of the present invention. Detailed Implementation

[0054] Example 1:

[0055] Please see Figure 1 A radiation field dose self-discrimination detection method, comprising:

[0056] The system is equipped with a scintillator detection front-end and a dual-channel signal acquisition circuit, which includes a raw signal channel and an analog differential channel.

[0057] Initialize a preset set of reference particle phase space trajectories, which records the closed orbit coordinate data of standard particle events in phase space, and serves as the reference particle attractor model;

[0058] In response to the photoelectric pulse signal output by the scintillator detection front end, a dose self-discrimination process is triggered, including:

[0059] Step 1: Based on the original signal channel and the analog differential channel, simultaneously acquire the real-time amplitude signal and the analog differential signal of the photoelectric pulse signal;

[0060] Step 2: Combining the real-time amplitude signal and the analog differential signal, the real-time phase space trajectory of the pulse signal is constructed through phase space mapping processing;

[0061] Step 3: Using the manifold curvature entropy calculation logic, perform topological feature analysis on the real-time phase space trajectory and calculate the trajectory manifold curvature entropy;

[0062] Step 4: Based on the comparison results of the trajectory manifold curvature entropy and the preset distortion threshold, perform hierarchical identification processing: if the value of the trajectory manifold curvature entropy is lower than the preset distortion threshold, it is determined to be a single particle event, and the particle type is identified according to the benchmark particle attractor model; if the value of the trajectory manifold curvature entropy is higher than or equal to the preset distortion threshold, it is determined to be a stacking distortion event, triggering the trajectory repair inverse calculation process to deconstruct the stacking distortion event into multiple independent particle events.

[0063] Step 5: Calculate particle energy and update radiation field dose library based on single-particle events or deconstructed independent particle events through phase space closed-loop volume integration.

[0064] Step 6: Extract geometric micro-perturbation features based on the real-time phase space trajectory, compare them with the preset crystal optical path difference reference dataset, determine the incident ray azimuth angle, and directionally weight the dose accordingly.

[0065] This embodiment provides a radiation field dose self-discrimination detection system and method, the core of which lies in constructing a real-time phase space topology analysis architecture based on FPGA edge computing; the system performs initialization configuration to establish a set of reference particle phase space trajectories, which is defined as the standard particle such as the neutron in a low dose rate, non-stacking environment. and gamma rays Closed orbital data formed on the phase plane; this set serves as the benchmark particle attractor model. Its mathematical expression is a series of standard state vector sequences, containing neutron attractors with wide phase trajectory loops. and gamma attractors with narrower phase trajectory loops ;

[0066] When the scintillator detector outputs a photoelectric pulse signal, the system does not simply rely on the original voltage signal, but triggers a dual-channel synchronous acquisition step; combined with the real-time amplitude signal... With analog differential signal Physical events are transformed into geometric objects, and the real-time phase space trajectory of the pulse signal is constructed through phase space mapping. ;

[0067] It should be noted that although this embodiment mainly uses a two-dimensional phase plane as an example, the system architecture supports extension to three-dimensional and higher multi-dimensional phase spaces, such as introducing an integral dimension, in which case the trajectory For a three-dimensional closed curve, the subsequent volume integral refers to the calculation of the geometric measure, i.e., the volume, enclosed by this high-dimensional closed trajectory. The two-dimensional area integral is merely a special case of the volume integral in the case of dimensionality reduction. To ensure consistency in terminology, this embodiment defines the phase space closed-loop volume integral as a generalized geometric measure operator. ;

[0068] When phase space dimension hour, It is expressed as an area integral of a planar region; when hour, It is expressed as a volume integral of a spatial region; this abstract definition ensures the rigor of Example 1 as a higher-level concept, covering specific mathematical operations under different dimensions;

[0069] Based on this, the system is quantized using the manifold curvature entropy calculation logic. The geometrical complexity is used to obtain the trajectory manifold curvature entropy. This parameter characterizes the degree of chaos in the trajectory morphology relative to a standard closed loop; to ensure the feasibility of the discrimination logic, a distortion threshold is preset. The configuration logic is as follows: data collection A sample of manifold curvature entropy for a standard single-particle event, where The sample collection environment was a low-background laboratory environment free from interference from other radiation sources, using a standard source. Irradiation, calculation of its statistical distribution, setting ,in The sample mean. For standard deviation, thus covering Normal events; based on and Based on the numerical comparison results, the system executes the hierarchical discrimination logic: in response to The system determines that the trajectory is smooth and closed, confirming it as a single-particle event, and calculates the trajectory and... or The Euclidean distance is used to identify particle types;

[0070] In response to The system determines that the trajectory is knotted or bifurcated, confirming it as a stacking distortion event, and immediately triggers the trajectory repair inverse calculation process to deconstruct the complex stacking distortion event into multiple independent particle events. Finally, based on the finally determined independent particle events, the particle energy is calculated through phase space closed-loop volume integration and the radiation field dose library is updated. To verify the above technical effects, the system's performance was tested.

[0071] Experimental results show that, Under high-flux radiation fields, the count loss rate of this system is only [percentage missing]. Far lower than traditional methods Meanwhile, targeting of Peak, energy resolution remains This verifies the measurement accuracy and stability of the system under high dynamic range.

[0072] This embodiment transforms the pulse stacking problem, which is difficult to distinguish in the traditional time domain (i.e., waveform overlap), into a geometric topological problem in phase space (i.e., trajectory knotting) by introducing an analog differential channel and phase space mapping. By using manifold curvature entropy as a single criterion, it can quickly distinguish between normal events and stacking events, thereby significantly improving the dynamic range and measurement accuracy of the detection system in high-flux radiation fields without increasing physical shielding.

[0073] Example 2:

[0074] Step 1 synchronous acquisition includes: acquiring the time-domain voltage value of the photoelectric pulse signal through a high-speed analog-to-digital converter in the original signal channel;

[0075] The photoelectric pulse signal is hardware differentiated using an analog differentiating circuit, and the voltage change rate of the photoelectric pulse signal is obtained through a high-speed analog-to-digital converter of the analog differentiating channel.

[0076] The time-domain voltage value and the voltage change rate value are strictly aligned in time to generate a state vector.

[0077] This embodiment provides specific hardware-level limitations for the synchronous acquisition steps, aiming to solve the high-frequency noise problem caused by digital differentiation; the system uses a high-speed analog-to-digital converter (ADC) in the original signal channel at a sampling rate of... Obtain the time-domain voltage value of the photoelectric pulse signal In this embodiment, Set to 500 MSPS derived from the system clock frequency division to ensure that nanosecond-level pulse rising edges can be captured;

[0078] Hardware differentiation processing of photoelectric pulse signals is performed using an analog differentiating circuit. This circuit mainly consists of an operational amplifier and an RC network, and its transfer function is approximately... This design aims to directly obtain a smooth voltage change rate value. Furthermore, it acquires data through a high-speed ADC with an analog differential channel, thereby avoiding traditional digital differential operations. Amplification effect on ADC quantization noise;

[0079] Because there may be differences in transmission delay between analog circuits and ADC channels, the system needs to... and Strict timing alignment is achieved; within the FPGA, a first-in-first-out (FIFO) queue is used to compensate for nanosecond-level delays in the faster channels until the peak response times of the two channels to the same calibration pulse coincide, thereby generating a state vector. :

[0080]

[0081] in For a moment voltage amplitude, For a moment The rate of change of voltage;

[0082] This embodiment employs a hardware differentiation plus dual ADC acquisition architecture, which effectively avoids the amplification effect of differentiation operations on high-frequency noise in digital signal processing. This mechanism ensures that the generated phase space trajectory has sufficient smoothness, providing a high-quality data foundation for subsequent high-precision curvature calculations.

[0083] Example 3:

[0084] Step 2 constructs the real-time phase space trajectory of the pulse signal, including:

[0085] A two-dimensional phase plane is established with the time-domain voltage value as the horizontal axis and the voltage change rate as the vertical axis.

[0086] The state vectors of a continuous time series are mapped to a two-dimensional phase plane to generate closed or open geometric curves, which serve as real-time phase space trajectories.

[0087] This embodiment describes in detail the specific method for trajectory construction; the system establishes a two-dimensional phase plane within the FPGA's storage space. Set the time-domain voltage value The horizontal axis represents the voltage change rate. The vertical axis coordinate;

[0088] The state vector in a continuous time series Mapped sequentially to the two-dimensional phase plane For single-particle events, these discrete points are connected in time sequence to generate a closed geometric curve, i.e., the real-time phase space trajectory. Specifically, for gamma rays, their rapid attenuation leads to... Large value and The decay is rapid. It exhibits a narrower elliptical shape; while for neutrons, due to their slow decay, they contain more slow-decaying components. The value is relatively small and Long duration It exhibits a relatively wide ring-shaped form;

[0089] This embodiment constructs... The phase plane utilizes the physical differences in particle pulse shapes, namely the different attenuation constants, to generate morphological differences in phase space, i.e., different thicknesses, to simplify the complex waveform recognition problem into an intuitive geometric shape recognition problem, which greatly reduces the computational complexity of the algorithm.

[0090] Example 4:

[0091] Step 3 utilizes the manifold curvature entropy calculation logic to perform topological feature analysis on the real-time phase space trajectory, including:

[0092] Calculate the local geometric curvature of each discrete point on the real-time phase space trajectory;

[0093] The distribution of local geometric curvature along the time axis is statistically analyzed, and a curvature distribution histogram is generated.

[0094] The information entropy of the curvature distribution histogram is calculated to obtain the trajectory manifold curvature entropy, which characterizes the degree of chaos of the trajectory shape relative to the standard closed loop.

[0095] This embodiment details the calculation process of the manifold curvature entropy of the core operator; the system calculates the real-time phase space trajectory. Discrete points above Local geometric curvature This embodiment uses the three-point discrete curvature formula; for voltage... With voltage change rate To address the issue of inconsistent dimensions, a normalization factor is first introduced. Perform dimensionless processing on the data to generate standardized variables. ;

[0096] Here, the normalization factor and The standard deviation was not chosen arbitrarily, but was defined as the statistical standard deviation of all standard trajectory data in the baseline particle attractor model, thus ensuring that real-time data and model data are compared on the same scale; the corrected curvature calculation formula is:

[0097]

[0098] in, For the first The curvature of a point Dimensionless variables;

[0099] The difference term appearing in the formula is specifically defined as follows: This represents the first-order central difference. This represents the second-order central difference;

[0100] The reason for using central difference instead of forward or backward difference is that central difference has... The truncation error accuracy is high, and it can maintain strict phase alignment between the differential signal and the original amplitude signal, preventing phase space trajectory distortion caused by phase lag. This parameter is derived from the aforementioned dual-channel acquisition data and is physically designed to capture changes in trajectory smoothness. This is because the trajectory of a normal particle is a smooth closed loop, while accumulation events can produce angles or sharp points in the phase trajectory at waveform superposition, leading to... A sudden, pulse-like increase occurred;

[0101] Systematic statistical local geometric curvature The distribution change along the time axis will The range of values ​​is divided into In this embodiment, there are several intervals. The value can be either 64 or 128. This value must be chosen to ensure that the interval width is less than the standard deviation of the curvature distribution. To ensure histogram resolution, count the number of points falling into each interval and normalize them to generate a probability distribution vector. That is, the curvature distribution histogram; finally, the information entropy of this histogram is calculated to obtain the curvature entropy of the trajectory manifold. :

[0102]

[0103] in, For the manifold curvature entropy, For the first The probability of a curvature interval;

[0104] The criterion logic is set as follows: single-particle trajectories are regular, and curvature distribution is concentrated. Low; the accumulation or noise trajectory is chaotic, with multiple abrupt inflection points and divergent curvature distribution. high;

[0105] This embodiment introduces manifold curvature entropy as a dimensionless feature quantity, which can quantify the purity of the signal at the global topology level. Compared with the traditional amplitude thresholding method, this method is not sensitive to baseline drift and can effectively identify hidden distortion events with normal amplitude but abnormal shape, such as slight stacking.

[0106] Example 5:

[0107] Step 4 triggers the trajectory repair reverse calculation process, including:

[0108] Locate topological singularities in the real-time phase space trajectory where curvature abruptly changes;

[0109] Using topological singularities as segmentation nodes and leveraging the geometric constraints of the baseline particle attractor model, multiple fitting segmentation is performed on the real-time phase space trajectory.

[0110] Reconstruct the segmented trajectory fragments to generate at least two independent closed regression trajectories, which serve as independent particle events after deconstruction.

[0111] This embodiment describes in detail the processing logic for accumulation events; in response to being determined to be an accumulation event, the system first scans the curvature sequence. Locate all Exceeding the preset mutation threshold Points; preset mutation threshold The determination logic is as follows: statistical benchmark particle attractor model Find the curvature values ​​of all trajectory points and obtain their maximum curvature. ,set up ,in For the maximum reference curvature, This is a redundancy coefficient, with a value of 1.5; coefficient Based on empirical values ​​derived from a large amount of experimental data, this setting allows for a 50% noise fluctuation tolerance in the baseline trajectory while effectively intercepting over 99% of stacking distortion inflection points, achieving a balance between sensitivity and false alarm rate.

[0112] To accurately distinguish between normal trajectory turns and discontinuous inflections caused by accumulation; these points are marked as topological singularities. This singularity physically corresponds to a geometric inflection point or intersection point resulting from the nonlinear superposition of two pulse signals in the time domain.

[0113] With topological singularity To segment nodes, the system will display the real-time trajectory. Cut into several non-closed trajectory segments. Using the benchmark particle attractor model By constraining the fit on these segments, the system attempts to find the optimal combination of attractors that minimizes the fitting residuals; the fitting residuals are defined here. The calculation formula is:

[0114]

[0115] in, To fit the residuals, For discrete points on the trajectory segment, For discrete points on the attractor model, specifically, This refers to the attractor model point set and the current The closest corresponding point in Euclidean distance;

[0116] This formula quantifies the tightness of the matching by calculating the sum of the minimum Euclidean distances from the fragment point set to the model point set; in order to solve for... Minimize the rotation matrix Translation vector This embodiment employs the Iterative Closest Point (ICP) algorithm based on Singular Value Decomposition (SVD): establishing point pair correspondences and calculating the cross-covariance matrix. ,in Let it be the centroid vector; then... Perform singular value decomposition ,in It is an orthogonal matrix. It is a diagonal matrix of singular values;

[0117] Calculate the rotation matrix With translation vector The process is iterated until convergence. Finally, the missing parts of each trajectory segment are filled in based on the fitting results. These missing parts are usually the parts covered by the falling or rising edges. The specific filling operator is based on rigid registration mapping.

[0118] Next, the system performs the trajectory fragment association clustering step:

[0119] Establish fragment feature matrix ,in, The temporal location of the fragment's centroid. The best attractor subtype identifier vector for matching, such as Representing neutron, Representing Gamma, For fitting residuals;

[0120] Calculate adjacent fragments and spatiotemporal correlation between :

[0121]

[0122] in, These are the weighting coefficients; if Less than the preset association threshold Then determine the fragments With fragments They belong to the same truncated physical particle event and are merged;

[0123] For independent fragments that cannot be merged, they are extended into complete closed curves using the whitening inverse transform operator based on their matching reference particle attractor model.

[0124] Considering the difference in physical dimensions between the horizontal and vertical axes of phase space, direct rotation would lead to numerical mixing errors. Therefore, the data is whitened before registration. To eliminate the influence of dimensions;

[0125] The whitening matrix was constructed using Principal Component Analysis (PCA): the covariance matrix of the baseline particle trajectory point set was calculated. Eigenvalue decomposition ,in The eigenvector matrix, The eigenvalue diagonal matrix is ​​used to construct the whitening transformation matrix. This operation aims to eliminate data correlation and normalize variance; an inverse transformation is then performed to restore the original data.

[0126]

[0127] in, For the reconstruction point, For whitening transformation, Let be a rotation matrix. It is a translation vector. These are the coordinates of the template point in the baseline particle attractor model;

[0128] This operator maps standard template data back to the real-time trajectory space to fill the gap, generating at least two independent closed regression trajectories. These reconstructed closed orbits represent the deconstructed independent particle events;

[0129] This embodiment achieves true anti-stacking rather than simple stacking rejection; through topological segmentation and reconstruction, the system can salvage multiple valid particle counts from a distorted stacking signal, thereby significantly reducing the system's count loss rate at high dose rates and ensuring the accuracy of energy spectrum measurements.

[0130] Example 6:

[0131] Step 5 involves phase space closed-loop volume integration, including:

[0132] Obtain the closed phase space orbits corresponding to the identified single-particle events or independent particle events;

[0133] Calculate the geometric measure enclosed by a closed phase space orbit in phase space, which is the geometric area in the two-dimensional phase plane and the geometric volume in the three-dimensional phase space;

[0134] The pre-calibrated energy response function is invoked, and the geometric measure is used as the input variable to calculate the corresponding particle deposition energy value. The energy response function defines the numerical mapping relationship between the closed loop area of ​​phase space and the deposition energy.

[0135] This embodiment describes the energy calculation method; the system obtains the original trajectory of a single-particle event. Or the reconstruction trajectory after stacking and deconstruction Using the discrete form of Green's theorem, calculate the geometric area enclosed by the closed phase space orbit within the phase plane. :

[0136]

[0137] in, Let be the closed-loop area of ​​the phase space. The coordinates of the trajectory points, The number of points on the trajectory;

[0138] The calculation is based on ordered coordinate points constituting a closed trajectory; furthermore, to support a broader definition of the phase space closed-loop volume integral in Embodiment 1, this embodiment also discloses a three-dimensional phase space implementation; the system introduces a third-dimensional state variable, namely the pulse integral value. Construct a three-dimensional state vector It must be noted that, in order to conform to the dual-channel hardware architecture of Embodiment 1, this pulse integral value It does not originate from a separate hardware channel, but rather is acquired by the digital signal processing unit within the FPGA from the original signal channel. The sequence is obtained by real-time trapezoidal numerical integration, i.e. ,in The sampling period is [period value]; at this time, the closed trajectory forms a closed loop in three-dimensional space, and the system calculates the phase space cone volume of this closed loop relative to the centroid of the trajectory. As a representation of energy:

[0139]

[0140] in, For phase space volume, For state vectors, It is the centroid vector;

[0141] In this formula, the reference vector The geometric centroid selected as the trajectory This avoids the problem of the cross product being zero due to using the origin vector; coefficients Based on the principles of solid geometry, that is, from the origin Trajectory centroid two adjacent points on the trajectory The volume of the infinitesimal tetrahedron formed by these four vertices; therefore, Essentially, it is the volume of the generalized cone formed by the closed trajectory relative to the center of mass;

[0142] The system calls the pre-calibrated energy response function. or The corresponding particle deposition energy value is calculated; this function is derived from multi-point calibration fitting using standard radioactive sources such as Cs-137 and Co-60; for the three-dimensional phase space volume integral, the energy response function is... Specifically, it can be expressed as follows:

[0143]

[0144] Including the introduction of index and In order to change the dimensions of volume Dimensionality reduction to linear scale To conform to the physical linear laws of energy deposition; coefficient as well as To pass through three or more standard radioactive sources with different energies, such as The calibration coefficients are obtained by calibration and least squares regression; regarding the ambiguity that may arise from the volume integral terminology in Example 1, it is clarified here: and Both refer to the generalized concept of phase space closed-loop volume integral in different dimensional subspaces. and Both of these are specific mathematical projections in the equation, and are essentially calculations of the geometric measure enclosed by the closed trajectory.

[0145] This embodiment utilizes phase space closed-loop geometric measures, namely area. or volume Instead of the traditional time-domain integral area, i.e., the Q value, to characterize energy, it has better noise immunity; because the phase space integral naturally cancels out the high-frequency noise interference of zero mean, the energy resolution is significantly improved.

[0146] Example 7:

[0147] According to Embodiment 1, a radiation field dose self-discrimination detection system and method are provided, characterized in that the method further includes a virtual collimation direction discrimination step:

[0148] Extract geometric micro-perturbation feature data on the real-time phase space trajectory;

[0149] The geometric micro-perturbation feature data is compared with the preset crystal optical path difference reference dataset, which contains optical path difference features corresponding to different incident azimuth angles of the incident rays relative to the detector crystal axis.

[0150] Determine the incident azimuth angle of the current incident ray, and perform directional weighting on the radiation field dose library based on the incident azimuth angle.

[0151] This embodiment adds a virtual collimation direction discrimination step, utilizing crystal anisotropy or optical path difference to achieve direction sensing; the system extracts real-time phase space trajectory. Geometric micro-perturbation feature data To quantify this feature, the system first calculates the discrete curvature sequence of each point on the trajectory. The discrete Fourier transform is then applied to extract the amplitude spectrum of a specific high-frequency band as a feature vector.

[0152]

[0153] in, For feature vectors, For spectral coefficients, For high-frequency index set;

[0154] Here, collection The mathematical boundary is defined as ,in The total number of sampling points corresponds to the nanosecond-level geometric jitter caused by the difference in the scattering path of photons within the crystal. Specifically, due to the anisotropic optical transmission characteristics of CLYC crystals, rays with different incident angles will cause slight changes in the critical condition for total internal reflection and the dispersion of collection time at the crystal interface. This difference in optical transport at the physical level is directly mapped to unique high-frequency texture features on the phase space trajectory, rather than random electronic noise.

[0155] Physically, this feature manifests as high-frequency texture jitter superimposed on the trajectory. It is usually due to the complex propagation path of photons in the crystal when rays are incident from the side or back of the detector, which causes a slight change in the light collection time dispersion and leaves traces on the phase diagram.

[0156] Will Compared with the preset crystal optical path difference reference dataset Perform comparison; dataset Construct as containing A lookup table for standard feature templates corresponding to discrete angles. The process of constructing this dataset is as follows: the detector is fixed on a precision rotating stage, and within the horizontal plane... For step-size rotation, standard point source illumination is used at each angle to collect accumulated trajectory features, thereby establishing a mapping library between angles and feature vectors; The system finds the best match by calculating the Euclidean distance between the input features and each template, thereby determining the incident azimuth angle.

[0157]

[0158] in, For optimal indexing, For template vector, To estimate the azimuth angle;

[0159] Based on the determined incident azimuth angle Based on this angle, the radiation field dose library is directionally weighted. The weighting calculation follows the following angle response correction formula, which aims to compensate for the decrease in crystal detection efficiency caused by different incident angles:

[0160]

[0161] in The normalized angular response function has the following fitting model:

[0162]

[0163] coefficients in the above formula The detection efficiency was measured using a standard radioactive source at multiple known incident angles, and the results were obtained by fitting the measurement data using the least squares method, while satisfying the constraints. ; here This represents the corrected true dose, which affects detection efficiency when lateral incidence is involved. At that time, the calculated weighting factor This correctly compensates for the dose loss;

[0164] coefficient and Derived from pre-calibration experiments: Measurements were taken of the detector at... positive radian and lateral response value in radians Solving the system of equations yields the following results. For example, if the detection mission primarily focuses on forward radiation, the weight of laterally incident particles can be reduced by adjusting the coefficient; further experimental verification shows that the method of this embodiment is effective for... The radioactive source undergoes directional discrimination testing. to Within the incident range, the system is sensitive to the incident azimuth angle. The inversion average error is less than This effectively verified the directional resolution capability of the virtual collimation technology.

[0165] Example 8:

[0166] Please see Figure 2 A radiation field dose self-discrimination detection system, comprising:

[0167] The detection front-end module is configured to convert incident rays into photoelectric pulse signals;

[0168] The phase extraction module includes a parallel raw signal channel and an analog differential channel, configured to acquire the real-time amplitude signal and the analog differential signal respectively;

[0169] The topology discrimination engine, configured in the FPGA processing unit, is used to construct real-time phase space trajectories and calculate the trajectory manifold curvature entropy;

[0170] The logic branch unit is configured to execute the following logic: compare the trajectory manifold curvature entropy with a preset distortion threshold. If the entropy value is less than the threshold, activate the single particle recognition logic; if the entropy value is greater than or equal to the threshold, activate the stacking deconstruction logic.

[0171] Stacking deconstruction unit is used to geometrically segment and reconstruct the trajectory based on the baseline particle attractor model when stacking distortion occurs;

[0172] The dose calculation unit is used to calculate energy based on phase space closed-loop volume integral and output dose data.

[0173] This embodiment discloses a hardware system architecture for implementing the above method; the detection front-end module uses a combination device derived from a CLYC scintillator coupled photomultiplier tube; the phase extraction module has physically separated traces designed on the circuit board, wherein the original signal channel includes a low-noise amplifier and a 14-bit 500MSPS ADC, and the analog differential channel includes a differentiating circuit composed of a high-speed operational amplifier and an ADC of the same specification; the topology discrimination engine is configured in an FPGA processing unit such as a Xilinx Kintex-7, and its internal logic includes a dual-port RAM for caching data and a system for real-time curvature calculation. Entropy The pipelined computing unit; the logic branch unit uses the state machine inside the FPGA according to... The value switches the data flow direction;

[0174] The stacking deconstruction unit employs parallel iterative nearest point (ICP) hardware-accelerated logic based on singular value decomposition (SVD) to segment the trajectory; this logic unit contains a dedicated systolic array for processing. The SVD decomposition of the covariance matrix avoids the convergence instability problem of the nonlinear gradient descent algorithm in FPGA; the dose calculation unit is used to calculate energy based on phase space closed-loop volume integral and output dose data, which is finally transmitted to the host computer through UART or Ethernet interface;

[0175] This embodiment embeds complex topology analysis algorithms into FPGA hardware logic, achieving nanosecond-level real-time processing capabilities. The system architecture meets the application requirements of high-throughput and real-time scenarios such as nuclear emergency monitoring and high-energy physics experiments.

[0176] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A radiation field dose self-discrimination detection method, characterized in that, include: A scintillator detection front-end and a dual-channel signal acquisition circuit are configured, wherein the dual-channel signal acquisition circuit includes a raw signal channel and an analog differential channel; Initialize a preset set of reference particle phase space trajectories, which records the closed orbit coordinate data of standard particle events in phase space, and serves as the reference particle attractor model; In response to the photoelectric pulse signal output by the scintillator detection front end, a dose self-discrimination process is triggered, including: Step 1: Based on the original signal channel and the analog differential channel, simultaneously acquire the real-time amplitude signal and the analog differential signal of the photoelectric pulse signal; Step 2: Combining the real-time amplitude signal and the analog differential signal, the real-time phase space trajectory of the pulse signal is constructed through phase space mapping processing; Step 3: Using the manifold curvature entropy calculation logic, perform topological feature analysis on the real-time phase space trajectory and calculate the trajectory manifold curvature entropy; Step 4: Based on the comparison results of the trajectory manifold curvature entropy and the preset distortion threshold, perform hierarchical identification processing: if the value of the trajectory manifold curvature entropy is lower than the preset distortion threshold, it is determined to be a single particle event, and the particle type is identified according to the benchmark particle attractor model; if the value of the trajectory manifold curvature entropy is higher than or equal to the preset distortion threshold, it is determined to be a stacking distortion event, triggering the trajectory repair inverse calculation process to deconstruct the stacking distortion event into multiple independent particle events. Step 5: Calculate particle energy and update radiation field dose library based on single-particle events or deconstructed independent particle events through phase space closed-loop volume integration. Step 6: Extract geometric micro-perturbation features based on the real-time phase space trajectory, compare them with the preset crystal optical path difference reference dataset, determine the incident ray azimuth angle, and directionally weight the dose accordingly; Step 3, which utilizes the manifold curvature entropy calculation logic to perform topological feature analysis on the real-time phase space trajectory, includes: Calculate the local geometric curvature of each discrete point on the real-time phase space trajectory; The distribution of local geometric curvature along the time axis is statistically analyzed, and a curvature distribution histogram is generated. The information entropy of the curvature distribution histogram is calculated to obtain the trajectory manifold curvature entropy, which characterizes the degree of chaos of the trajectory shape relative to the standard closed loop. Step 4, the trigger trajectory repair reverse calculation process, includes: Locate topological singularities in the real-time phase space trajectory where curvature abruptly changes; Using topological singularities as segmentation nodes and leveraging the geometric constraints of the baseline particle attractor model, multiple fitting segmentation is performed on the real-time phase space trajectory. Reconstruct the segmented trajectory fragments to generate at least two independent closed regression trajectories, which serve as independent particle events after deconstruction.

2. The radiation field dose self-discrimination detection method according to claim 1, characterized in that, The synchronous acquisition mentioned in step 1 includes: The time-domain voltage value of the photoelectric pulse signal is obtained through a high-speed analog-to-digital converter in the original signal channel. The photoelectric pulse signal is hardware differentiated using an analog differentiating circuit, and the voltage change rate of the photoelectric pulse signal is obtained through a high-speed analog-to-digital converter of the analog differentiating channel. The time-domain voltage value and the voltage change rate value are strictly aligned in time to generate a state vector.

3. The radiation field dose self-discrimination detection method according to claim 2, characterized in that, Step 2, which involves constructing the real-time phase space trajectory of the pulse signal, includes: A two-dimensional phase plane is established with the time-domain voltage value as the horizontal axis and the voltage change rate as the vertical axis. The state vectors of a continuous time series are mapped to a two-dimensional phase plane to generate closed or open geometric curves, which serve as real-time phase space trajectories.

4. The radiation field dose self-discrimination detection method according to claim 1, characterized in that, Step 5, which involves volume integration through a closed loop in phase space, includes: Obtain the closed phase space orbits corresponding to the identified single-particle events or independent particle events; Calculate the geometric measure enclosed by the closed phase space orbit within the phase space, which is the geometric area in the two-dimensional phase plane and the geometric volume in the three-dimensional phase space; The pre-calibrated energy response function is invoked, and the geometric measure is used as the input variable to calculate the corresponding particle deposition energy value, wherein the energy response function defines the numerical mapping relationship between the phase space closed loop area and the deposition energy.

5. The radiation field dose self-discrimination detection method according to claim 1, characterized in that, The method also includes a virtual collimation direction discrimination step: Extract geometric micro-perturbation feature data on the real-time phase space trajectory; The geometric micro-perturbation feature data is compared with a preset crystal optical path difference reference dataset, which contains optical path difference features corresponding to different incident azimuth angles of the incident rays relative to the detector crystal axis. Determine the incident azimuth angle of the current incident ray, and perform directional weighting on the radiation field dose library based on the incident azimuth angle.

6. A radiation field dose self-discrimination detection system, characterized in that, The system is configured to perform the method of any one of claims 1-5, the system comprising: The detection front-end module is configured to convert incident rays into photoelectric pulse signals; The phase extraction module includes a parallel raw signal channel and an analog differential channel, configured to acquire the real-time amplitude signal and the analog differential signal respectively; The topology discrimination engine, configured in the FPGA processing unit, is used to construct real-time phase space trajectories and calculate the trajectory manifold curvature entropy; The logic branch unit is configured to execute the following logic: compare the trajectory manifold curvature entropy with a preset distortion threshold. If the entropy value is less than the threshold, activate the single particle recognition logic; if the entropy value is greater than or equal to the threshold, activate the stacking deconstruction logic. Stacking deconstruction unit is used to geometrically segment and reconstruct the trajectory based on the baseline particle attractor model when stacking distortion occurs; The dose calculation unit is used to calculate energy based on phase space closed-loop volume integral and output dose data.

Images