A Real-Time X-ray and Gamma-Ray Segmentation System Based on Energy Spectrum-Spatiotemporal Fusion

Abstract

The present application belongs to the technical field of radiation detection and analysis, and specifically relates to an X&gamma ray real-time discrimination system based on energy spectrum-spatiotemporal fusion, which comprises: a double-layer detector, which comprises an upper-layer filtering layer and a lower-layer full-spectrum layer, and is used for receiving and discriminating X rays and gamma rays; a data acquisition module, which synchronously acquires energy spectrum, timing pulse and spatial distribution data; a timing analysis and LSTM network, wherein LSTM is a special recurrent neural network, has significant advantages in processing long-sequence data and capturing long-term dependence, and can make up for the shortcomings of traditional timing analysis methods; and a feature extraction module, which is a bridge connecting original data and a prediction model in timing analysis, is used for calculating the area proportion of K-edge peaks, and determines effective pulses according to the LSTM network output; by adopting the double-layer detector in combination with the LSTM network, 300 ms-level real-time response is realized, the discrimination ability for X rays and gamma rays is improved, and the accuracy and reliability of overall detection are also enhanced.

Description

Technical Field

[0001] This invention belongs to the field of radiation detection and analysis technology, specifically a real-time X&γ ray identification system based on energy spectrum-spatiotemporal fusion. Background Technology

[0002] X-ray security inspection equipment may leak X-rays during operation due to shielding failure or mechanical wear. Its energy range (1-150 keV) significantly overlaps with the gamma rays (10keV-3MeV) released by natural radioactive materials in the environment (such as 137Cs and 226Ra). Traditional methods rely on a single energy threshold or pulse width for detection, resulting in a false alarm rate as high as 30%-40%, which leads to frequent false alarms in radiation safety monitoring systems and affects the normal operation of the equipment.

[0003] Furthermore, X-ray leakage is directional and pulsed, while gamma rays are isotropic and continuous radiation, making it difficult to distinguish between the two in a mixed radiation field using a single parameter. Existing detection techniques, such as the energy spectral threshold method, can distinguish between X-rays and gamma rays by setting a fixed energy threshold (e.g., 50 keV), but low-energy gamma rays (e.g., 59.5 keV in 241 Am) overlap with the K-edge characteristic peak of X-rays (20-80 keV), leading to misjudgment. If the time filtering method is used to distinguish between X-ray pulse width (10-100 μs) and continuous gamma-ray radiation, the response time of commonly used scintillator detectors is >100 ns, making it difficult to distinguish the difference between the two in a very short time.

[0004] Therefore, a real-time X&γ ray discrimination system based on energy spectrum-spatiotemporal fusion is proposed to address the above problems. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a real-time X&γ ray discrimination system based on energy spectrum-spatiotemporal fusion, which solves the problems mentioned in the background technology.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a real-time X&γ ray discrimination system based on energy spectrum-spatiotemporal fusion, the system comprising:

[0009] Dual-layer detector: It consists of an upper filter layer and a lower full-spectrum layer, used to receive and distinguish between X-rays and gamma rays;

[0010] Data acquisition module: synchronously acquires energy spectrum, time-series pulse and spatial distribution data, and ensures data synchronization through timestamp technology;

[0011] Time series analysis and LSTM networks: As a special type of recurrent neural network, LSTM has significant advantages in processing long-sequence data and capturing long-term dependencies, which can make up for the shortcomings of traditional time series analysis methods;

[0012] Feature extraction module: It is the bridge connecting the original data and the prediction model in time series analysis. It is used to calculate the K-edge peak area ratio and determine the valid pulses based on the output of the LSTM network.

[0013] Fusion Decision Module: Based on a dynamic weighted fusion model, it makes a comprehensive judgment by combining energy spectrum weight, peak area ratio, time series weight, and time series matching degree to trigger an X-ray leakage alarm.

[0014] Preferably, the upper filter layer uses 0.2 mm copper foil to shield X-rays with energy <20 keV, retain the K-edge characteristic peak of 20-80 keV, and reduce low-energy gamma-ray interference;

[0015] Lower full-spectrum layer: SiPM array coupled with NaI crystal size is ~50 mm × 50 mm, covering the 1-300 keV energy range, the formula for inverting the leakage X-ray energy spectrum is:

[0016] ;

[0017] In the above formula: For the total spectrum, For the filter layer, the energy spectrum E∈[20,80]keV.

[0018] The upper filter layer directly affects the energy resolution and imaging quality of the detector. Through unique material and structural design, it achieves synchronous acquisition of full-energy spectral information, providing multi-dimensional quantitative analysis capabilities for clinical diagnosis and enhancing detection resolution.

[0019] Preferably, the data acquisition module acquires raw data from the dual-layer detector in real time according to a set acquisition frequency, and ensures data synchronization through timestamp technology. At the same time, it performs preliminary processing on the acquired data to improve data quality.

[0020] Preferably, the formula for calculating the K-edge peak area ratio is:

[0021] ;

[0022] in, Inverting the leak X-ray energy spectrum; It is a total spectrum.

[0023] Preferably, the time series analysis and LSTM network include:

[0024] Signal synchronization: Acquiring the high-voltage power supply signal of the X-ray machine. The detector pulse is Construct the cross-correlation function:

[0025] ,

[0026] in, This represents the number of sampling points; The sampling index represents the first sample in the signal sequence. One sampling point; This is the time offset, representing the time delay or offset between two signals; The pulse signal output by the detector is here. The whole thing has been shifted on the timeline. The sequence obtained after sampling points;

[0027] LSTM network: Input layer with a 100-point time window Sequence, hidden layer 32 nodes, output temporal matching degree To capture the causal relationship between X-ray pulses and equipment power-on / off;

[0028] LSTM output threshold It is determined to be a valid pulse.

[0029] Preferably, the dynamic weighted fusion model performs comprehensive calculations based on the energy spectrum weight, peak area ratio, time series weight, and time series matching degree to obtain the discrimination result; it establishes a weight adaptive rule to dynamically adjust the weight according to the background gamma dose rate and the operating frequency of the equipment, so as to ensure that the energy spectrum weight is improved in the high background environment and the time series weight is improved in the pulse-dense scenario.

[0030] Preferably, the criterion formula for the dynamic weighted fusion model is:

[0031] ;

[0032] In the above formula Energy spectral weighting; The percentage of peak area; For time series weights; For time series matching degree; Calibration factor: Based on the temperature sensor, the detector gain drift is dynamically compensated to ensure that the error is <3% under the operating conditions of -30℃ to 60℃.

[0033] Based on the above model, the following weight adaptive rule is established:

[0034] High background environment, gamma dose rate > 1 μSv / h: Increase spectral weighting Suppress time interference;

[0035] In pulse-intensive scenarios with device frequencies >100Hz: Increase timing weights. Enhanced short pulse recognition.

[0036] Preferably, the threshold of the criterion formula of the dynamic weighted fusion model is... X-ray leak alarm triggered.

[0037] Experimental steps:

[0038] The system detector was placed 1 meter from the exit beam of the gamma radiation source at the X-ray machine's exit point. Tests were conducted on the X-ray machine operating alone and on gamma radiation sources of different energy types, as well as on the combined field of the X-ray machine and gamma radiation sources of different energy types operating simultaneously. Simulated temperature and humidity changes were performed (-30℃ → 60℃) with each increase of 5℃, while also simulating pulse-intensive conditions. The tests were repeated 400 times to verify the system's actual performance.

[0039] Experimental data:

[0040] The experimental data obtained from the above experiment are shown in the table below:

[0041] Table 1 Comparison data under different working conditions

[0042]

[0043] Experimental conclusion:

[0044] The real-time X-ray leakage and gamma-ray identification system based on energy spectrum-spatiotemporal feature fusion shows significant advantages in complex radiation fields, especially in the overlapping area of ​​low-energy gamma rays and X-rays and under dynamic operating conditions. The false alarm rate is reduced by 41% and the response speed is improved by 32%, meeting the needs of industrial-grade real-time monitoring.

[0045] Beneficial effects

[0046] Compared with existing technologies, this invention provides a real-time X&γ ray discrimination system based on energy spectrum-spatiotemporal fusion, which has the following beneficial effects:

[0047] 1. This invention, by combining a dual-layer detector with an LSTM network, achieves a real-time response of 300 ms, which improves the ability to distinguish between X-rays and gamma rays, and also enhances the overall accuracy and reliability of detection.

[0048] 2. This invention, through unique material and structural design, enables the system to simultaneously acquire full-energy spectral information, providing multi-dimensional quantitative analysis capabilities for clinical diagnosis and further improving detection resolution and imaging quality.

[0049] 3. This invention overcomes the high misjudgment rate problem caused by traditional methods that rely on a single energy threshold or pulse width detection by integrating energy spectrum K-edge features, device timing synchronization and spatial directionality index.

[0050] 4. This invention introduces a dynamic weighted fusion model, which dynamically adjusts the weights based on the background gamma dose rate and the operating frequency of the equipment. This feature ensures the stability and adaptability of the system under different operating conditions. In particular, it enhances the energy spectrum weights in high background environments and strengthens the short pulse recognition capability in pulse-dense scenarios, effectively reducing the false alarm rate and improving the timeliness of alarms.

[0051] 5. This invention establishes a weighted adaptive rule to dynamically compensate for detector gain drift by the temperature sensor, ensuring that the error is less than 3% under extreme temperature conditions, thereby guaranteeing the long-term stable operation of the system. Attached Figure Description

[0052] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0053] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Specific implementation examples are given below.

[0056] like Figure 1 As shown, a real-time X-ray and gamma-ray discrimination system based on energy spectrum-spatiotemporal fusion is described; the specific implementation method is as follows:

[0057] Dual-layer detector: It consists of an upper filter layer and a lower full-spectrum layer, used to receive and distinguish between X-rays and gamma rays;

[0058] Data acquisition module: synchronously acquires energy spectrum, time-series pulse and spatial distribution data, and ensures data synchronization through timestamp technology;

[0059] Time series analysis and LSTM networks: As a special type of recurrent neural network, LSTM has significant advantages in processing long-sequence data and capturing long-term dependencies, which can make up for the shortcomings of traditional time series analysis methods;

[0060] Feature extraction module: It is the bridge connecting the original data and the prediction model in time series analysis. It is used to calculate the K-edge peak area ratio and determine the valid pulses based on the output of the LSTM network.

[0061] Fusion Decision Module: Based on a dynamic weighted fusion model, it makes a comprehensive judgment by combining energy spectrum weight, peak area ratio, time series weight, and time series matching degree to trigger an X-ray leakage alarm.

[0062] The upper filter layer uses 0.2 mm copper foil for shielding against X-rays with energy <20keV, while retaining the K-edge characteristic peak of 20-80keV and reducing interference from low-energy gamma rays;

[0063] Lower full-spectrum layer: SiPM array coupled with NaI crystal size is ~50 mm × 50 mm, covering the 1-300 keV energy range, the formula for inverting the leakage X-ray energy spectrum is:

[0064] ;

[0065] In the above formula: For the total spectrum, For the filter layer, the energy spectrum E∈[20,80]keV.

[0066] The upper filter layer directly affects the energy resolution and imaging quality of the detector. Through unique material and structural design, it achieves synchronous acquisition of full-energy spectral information, providing multi-dimensional quantitative analysis capabilities for clinical diagnosis and enhancing detection resolution.

[0067] Through unique material and structural design, the system is able to simultaneously acquire full-energy spectral information, providing multi-dimensional quantitative analysis capabilities for clinical diagnosis and further improving detection resolution and imaging quality.

[0068] The data acquisition module acquires raw data from the dual-layer detector in real time according to the set acquisition frequency, and ensures the synchronization of the data through timestamp technology. At the same time, it performs preliminary processing on the acquired data to improve the data quality.

[0069] The formula for calculating the area ratio of the K-edge peak is:

[0070] ;

[0071] in, Inverting the leak X-ray energy spectrum; It is a total spectrum;

[0072] The time series analysis and LSTM network include:

[0073] Signal synchronization: Acquiring the high-voltage power supply signal of the X-ray machine. The detector pulse is Construct the cross-correlation function:

[0074] ;

[0075] in, This represents the number of sampling points; The sampling index represents the first sample in the signal sequence. One sampling point; This is the time offset, representing the time delay or offset between two signals; The pulse signal output by the detector is here. The whole thing has been shifted on the timeline. The sequence obtained after sampling points;

[0076] LSTM network: Input layer with a 100-point time window Sequence, hidden layer 32 nodes, output temporal matching degree To capture the causal relationship between X-ray pulses and equipment power-on / off;

[0077] LSTM output threshold This pulse is determined to be valid.

[0078] By combining a dual-layer detector with an LSTM network, a real-time response of 300 ms is achieved, which improves the ability to distinguish between X-rays and gamma rays and enhances the overall accuracy and reliability of detection.

[0079] The dynamic weighted fusion model calculates the discrimination result by comprehensively considering the energy spectrum weight, peak area ratio, time series weight, and time series matching degree. It establishes a weight adaptive rule to dynamically adjust the weight according to the environmental background γ dose rate and the device operating frequency to ensure that the energy spectrum weight is increased in high background environments and the time series weight is increased in pulse-dense scenarios.

[0080] The criterion formula for the dynamic weighted fusion model is:

[0081] ;

[0082] In the above formula Energy spectral weighting; The percentage of peak area; For time series weights; For time series matching degree; Calibration factor: Based on the temperature sensor, the detector gain drift is dynamically compensated to ensure that the error is <3% under the operating conditions of -30℃ to 60℃.

[0083] Based on the above model, the following weight adaptive rule is established:

[0084] High background environment, gamma dose rate > 1 μSv / h: Increase spectral weighting Suppress time interference;

[0085] In pulse-intensive scenarios with device frequencies >100Hz: Increase timing weights. Enhance short pulse recognition;

[0086] By establishing a weighted adaptive rule, the temperature sensor dynamically compensates for detector gain drift, ensuring that the error is less than 3% under extreme temperature conditions, thereby guaranteeing the long-term stable operation of the system.

[0087] The threshold value of the criterion formula for the dynamic weighted fusion model is obtained. Triggered an X-ray leak alarm;

[0088] By introducing a dynamic weighted fusion model, the weights are dynamically adjusted according to the background gamma dose rate and the operating frequency of the equipment. This feature ensures the stability and adaptability of the system under different operating conditions. In particular, it enhances the energy spectrum weights in high background environments and strengthens the short pulse recognition capability in pulse-dense scenarios, effectively reducing the false alarm rate and improving the timeliness of alarms.

[0089] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A real-time X&γ ray discrimination system based on energy spectrum-spatiotemporal fusion, characterized in that, The system includes: Dual-layer detector: It consists of an upper filter layer and a lower full-spectrum layer, used to receive and distinguish between X-rays and gamma rays; The upper filter layer uses 0.2mm copper foil to shield X-rays with energy <20keV, retain the K-edge characteristic peak of 20-80keV, and reduce low-energy gamma-ray interference; Lower full-spectrum layer: SiPM array coupled with NaI crystal size is ~50mm×50mm, covering the 1-300keV energy range, the formula for inverting the leakage X-ray energy spectrum is: ； In the above formula: For the total spectrum, For the filter layer, the energy spectrum E∈[20,80]keV; Data acquisition module: synchronously acquires energy spectrum, time-series pulse and spatial distribution data, and ensures data synchronization through timestamp technology; Time series analysis and LSTM networks: including: Signal synchronization: Acquiring the high-voltage power supply signal of the X-ray machine. The detector pulse is Construct the cross-correlation function: ； in, This represents the number of sampling points; The sampling index represents the first sample in the signal sequence. One sampling point; This is the time offset, representing the time delay or offset between two signals; The pulse signal output by the detector is here. The whole thing has been shifted on the timeline. The sequence obtained after sampling points; LSTM network: Input layer with a 100-point time window Sequence, hidden layer 32 nodes, output temporal matching degree To capture the causal relationship between X-ray pulses and equipment power-on / off; LSTM output threshold This pulse is determined to be valid. Feature extraction module: It is the bridge connecting the original data and the prediction model in time series analysis. It is used to calculate the K-edge peak area ratio and determine the valid pulses based on the output of the LSTM network. Fusion Decision Module: Based on a dynamic weighted fusion model, it makes a comprehensive judgment by combining energy spectrum weight, peak area ratio, time series weight, and time series matching degree to trigger an X-ray leakage alarm; The dynamic weighted fusion model calculates the discrimination result by comprehensively considering the energy spectrum weight, peak area ratio, time series weight, and time series matching degree. It establishes a weight adaptive rule to dynamically adjust the weight according to the environmental background γ dose rate and the device operating frequency to ensure that the energy spectrum weight is increased in high background environments and the time series weight is increased in pulse-dense scenarios. The criterion formula for the dynamic weighted fusion model is: ； In the above formula Energy spectral weighting; The percentage of peak area; For time series weights; For time series matching degree; Calibration factor: Based on dynamic compensation of detector gain drift by temperature sensor, ensuring error <3% under operating conditions of -30℃ to 60℃; Based on the above model, the following adaptive weighting rule is established: High background environment, gamma dose rate > 1 μSv / h: Increase spectral weighting Suppress time interference; In pulse-intensive scenarios with device frequencies >100Hz: Increase timing weights. Enhance short pulse recognition.

2. The X&Gamma ray real-time discrimination system based on energy spectrum-spatiotemporal fusion according to claim 1, characterized in that, The data acquisition module acquires raw data from the dual-layer detector in real time according to the set acquisition frequency, and ensures the synchronization of the data through timestamp technology. At the same time, it performs preliminary processing on the acquired data to improve the data quality.

3. The X&G real-time discrimination system based on energy spectrum-spatiotemporal fusion according to claim 1, characterized in that, The formula for calculating the K-edge peak area ratio is as follows: ； in, Inverting the leak X-ray energy spectrum; It is a total spectrum.

4. The X&γ ray real-time discrimination system based on energy spectrum-spatiotemporal fusion according to claim 1, characterized in that, The threshold value of the criterion formula for the dynamic weighted fusion model X-ray leak alarm triggered.

Images