Alpha-beta-gamma mixed radiation screening method based on radiation response signal

Abstract

The invention discloses an alpha-beta-gamma mixed radiation screening method based on a radiation response signal. The alpha-beta-gamma mixed radiation screening method depends on a CMOS sensor mixed radiation screening system. The hybrid radiation screening system for the CMOS sensor comprises the CMOS sensor, a circuit board, a chip board, a shell and a PC (Personal Computer), the method comprises the following steps: 1, acquiring a frame image containing a radiation response signal; 2, discriminating rays based on the feature difference of the connected regions: identifying and extracting the connected regions corresponding to the radiation response signals in the frame image; analyzing the characteristics of each connected region from multiple dimensions; and based on the difference of different types of rays on at least one or more characteristics, alpha, beta and gamma rays are discriminated. Compared with an existing detection device (or technology), the device has the advantages of being small in external environment interference, low in maintenance cost, high in sensitivity and low in requirements for software and hardware, and the performance of the device is still kept stable in the high-energy radiation environment.

Description

Technical Field

[0001] The present invention relates to the field of nuclear science and technology, in particular to an α-β-γ mixed radiation discrimination method based on radiation response signals. Background Art

[0002] Mixed α, β, and γ radiation discrimination is an important research direction in nuclear science. Its main purpose is to accurately detect and distinguish α, β, and γ rays in complex radioactive environments. It has applications in nuclear accident emergency response, environmental monitoring, nuclear pollution control, nuclear physics experiments, and radioactive material research.

[0003] Currently, the discrimination of mixed α, β, and γ radiation mainly relies on scintillator detectors, semiconductor detectors, gas detectors, and pulse shape discrimination technology. However, these detection devices (or technologies) all have certain limitations.

[0004] For example, the scintillator detector system is complex and sensitive to environmental conditions (sensitivity to environmental conditions is mainly reflected in the fact that temperature, humidity, light and mechanical vibration will affect the performance of the detector, which greatly limits its application range). The manufacturing cost of semiconductor detectors is high and some types require low temperature operation. The gas detector has relatively low sensitivity and relatively high maintenance cost (relatively low sensitivity is mainly reflected in the fact that the detector uses gas as the detection medium. When radiation particles pass through the gas medium, the probability of ionization or excitation with gas molecules is small, resulting in low detection efficiency; the relatively high maintenance cost is mainly reflected in the fact that long-term use may cause sealing failure, and the gas medium needs to be regularly tested and replenished, and the gas medium (usually 3 He) is expensive (about US$2,000 per liter). The detection system based on energy spectrum characteristic peak recognition technology has high requirements for hardware and software (the high hardware requirements are mainly reflected in the need to be equipped with high-performance detectors, high-speed data acquisition systems, low-noise electronic design, high-bandwidth data transmission and storage systems, and high-precision time synchronization trigger systems; the high software requirements are mainly reflected in the need to use complex real-time signal processing algorithms, noise suppression and baseline correction algorithms, and classification models trained with a large amount of calibration data).

[0005] In addition, the above detection equipment (or technology) is limited in its discrimination accuracy in complex radiation fields, especially in high-energy radiation environments, where its performance is significantly reduced.

[0006] In summary, it is of great significance to develop a new α, β, and γ mixed radiation discrimination technology in order to solve the limitations of the above-mentioned detection equipment (or technology). Summary of the Invention

[0007] The present invention aims to overcome the shortcomings of existing technologies by providing a method for discriminating mixed α-β-γ radiation based on radiation response signals. This method addresses the difficulty faced by current detection equipment (or technologies) for discriminating mixed α-β-γ radiation, which struggles to simultaneously achieve low environmental impact, low maintenance costs, high sensitivity, and low software and hardware requirements. It also addresses the significant performance degradation of current detection equipment (or technologies) for discriminating mixed α-β-γ radiation in high-energy radiation environments.

[0008] The technical solution of the present invention is: an α-β-γ mixed radiation discrimination method based on radiation response signals, relying on a CMOS sensor mixed radiation discrimination system;

[0009] A CMOS sensor hybrid radiation discrimination system includes a CMOS sensor, a circuit board, a chip board, a housing, and a PC. The CMOS sensor has no glass encapsulation on its surface and is fixedly mounted on the circuit board. The circuit board is provided with a power supply interface, and the circuit board and the chip board are communicatively connected via a DuPont connector. The chip board is provided with a SOC chip for outputting a frame image containing a radiation response signal. The housing is made of radiation shielding material, and an inner cavity for accommodating the circuit board and the chip board is provided inside the housing. A radiation entrance hole covered by a light-shielding film is provided on the outside of the housing. The circuit board and the chip board are both fixedly mounted in the inner cavity of the housing, and the photosensitive surface of the CMOS sensor is arranged opposite the radiation entrance hole. There is no obstruction between the photosensitive surface of the CMOS sensor and the radiation entrance hole, and the distance between the two is less than 5 mm. The PC is communicatively connected to the chip board via wired or wireless means, and is used to adjust the parameters of the CMOS sensor and store and display frame images containing radiation response signals.

[0010] The screening method is as follows:

[0011] S01, acquiring a frame image containing a radiation response signal: α, β, and γ rays in the environment are incident on the CMOS sensor, causing the CMOS sensor to generate a radiation response signal corresponding to the type of ray. The SOC chip receives the data containing the radiation response signal output by the CMOS sensor and converts it into a continuous frame image and transmits it to the PC;

[0012] S02, distinguishing rays based on differences in connected area features: identifying and extracting connected areas corresponding to each radiation response signal in the frame image; analyzing the features of each connected area from multiple dimensions; distinguishing α, β, and γ rays based on differences in at least one or more features of different types of rays; the features include average pixel value, number of pixels, rectangularity, aspect ratio, compactness, and convexity.

[0013] A further technical solution of the present invention is: the six characteristics are defined as follows:

[0014] I. The average pixel value is used to represent the overall brightness level of the radiation response signal. Its calculation formula is shown in Formula 1.

[0015] Formula 1:

[0016] Where M is the average pixel value of the connected area, P is the number of pixels in the connected area, δ is the screening function used to determine whether the target pixel belongs to the connected area, and the connected area is the area occupied by a radiation response signal in the frame image; (i, j) represents the pixel in the i-th row and j-th column, n and m represent the number of rows and columns of the pixel, respectively, I(i, j) represents the pixel value of the i-th row and j-th column, and T is the pixel value threshold of the connected area pixel;

[0017] II. The number of pixels is the number of pixels in the connected region, which reflects the area of the connected region. Its calculation formula is shown in Formula 2.

[0018] Formula 2:

[0019] Where P is the number of pixels in the connected area;

[0020] III. Rectangularity is used to characterize the similarity between the shape of the connected region and the ideal rectangle. Its calculation formula is shown in Formula 3.

[0021] Formula 3:

[0022] Among them, R is the rectangular degree of the connected region, A is the area of the connected region, and A MER is the area of the minimum circumscribed rectangle of the connected region;

[0023] IV. The aspect ratio is used to characterize the symmetry and directionality of the connected region. Its calculation formula is shown in Formula 4.

[0024] Formula 4:

[0025] Among them, V is the aspect ratio of the connected area, W and H are the width and height of the minimum circumscribed rectangle of the connected area, respectively. max 、x min are the maximum and minimum coordinates of the connected area in the horizontal direction, y max 、y min are the maximum and minimum coordinates in the vertical direction of the connected area respectively;

[0026] V. Compactness is used to distinguish and identify different shapes of connected regions. It is invariant to translation, scale, and rotation. Its representation formula is shown in Formula 5.

[0027] Formula 5:

[0028] Where O is the compactness of the connected region, S is the boundary length of the connected region, and A is the area of the connected region;

[0029] VI. Convexity is used to describe the shape complexity and edge concavity of connected regions. The smaller the convexity, the more irregular the shape. Its calculation formula is shown in Formula 6.

[0030] Formula 6:

[0031] Where C is the convexity of the connected region, A is the area of the connected region, and F is the sum of the areas of all convex hulls of the connected region.

[0032] A further technical solution of the present invention is: in step S02, a convolutional neural network is used to identify α, β, and γ rays. The specific process is as follows:

[0033] A. Establish a response event template library:

[0034] First, continuous frame images are binarized, and then connected regions are identified and extracted based on the set pixel value threshold of the connected region pixels. The six features of each connected region are analyzed, and the following two operations are performed: ①. Connected regions corresponding to mixed events and edge events are eliminated; ②. The remaining connected regions are marked with corresponding response events, including α, β, and γ response events; the marked α, β, and γ response events are extracted and placed on a pure black background of a specified pixel size, thereby obtaining α, β, and γ response event template libraries respectively;

[0035] B. Data preprocessing:

[0036] Preprocessing the response events in the α, β, and γ response event template libraries includes the following two steps: ① Normalizing the six features contained in the response event to the interval [0, 1] so that the neural network can learn the features more effectively; ② Performing random rotation and / or flipping and / or scaling operations on the response events to increase data diversity and the generalization ability of the model;

[0037] C. Model training:

[0038] The preprocessed data in the α, β, and γ response event template libraries are divided into training, validation, and test sets in a ratio of 7:1:2. The convolutional neural network model is trained using the training set. The model output is calculated through forward propagation, and then the output is compared with the true label to calculate the loss value. The model weights are then updated through the backpropagation algorithm to minimize the loss value. Stochastic gradient descent or Adam optimization algorithms are used to accelerate model convergence and improve training efficiency. Training is stopped when the model loss value stops decreasing for multiple consecutive epochs on the validation set or reaches the predetermined maximum number of training rounds.

[0039] D. Verification and Evaluation:

[0040] The trained model is verified and evaluated using the test set. The evaluation indicators include accuracy, precision, recall, and F1 value. By drawing the confusion matrix, the classification effect of the model on different types of rays is intuitively displayed, and the classification errors and advantage categories of the model are analyzed.

[0041] A further technical solution of the present invention is: under the gain condition of 32dB, the average pixel value distribution of β and γ response events respectively has exclusive intervals. If the average pixel value of the monitored response event falls within the exclusive interval, β and γ radiation can be distinguished.

[0042] A further technical solution of the present invention is: under 9dB or 32dB gain conditions, the compactness of β and γ response events respectively have exclusive intervals. If the compactness of the monitored response event falls within the exclusive interval, β and γ radiation can be distinguished.

[0043] Compared with the prior art, the present invention has the following advantages:

[0044] 1. High Resolution and High Precision: The identification system and method utilize a CMOS APS sensor as a detector. The CMOS APS sensor accurately captures the energy deposition gradient distribution generated by radiation particles across the pixel array, providing the necessary prerequisite for extracting multi-dimensional features of various radiation particles. Compared to traditional detectors, the identification system and method offer significant advantages in resolution and precision.

[0045] 2. High Integration and Real-Time Imaging: The circuit board equipped with the CMOS APS sensor and the chip board equipped with the SOC chip are integrated within the housing, enabling the integration of detection and preliminary signal processing. The chip board can be connected to the PC via a USB data cable or wireless communication, making the identification system more integrated and smaller, making it easier to carry and install. Furthermore, the CMOS APS sensor features real-time imaging capabilities, displaying the distribution and dynamic changes of radiation particles in real time, providing intuitive information for radiation protection and environmental monitoring.

[0046] 3. Strong Radiation Resistance: CMOS APS sensors have a certain tolerance to ionizing radiation and can operate stably for long periods of time in radiation environments. This makes the identification system and method suitable for high-radiation-dose applications, such as reactors and accelerators. Compared to traditional detectors, the identification system and method have significant advantages in radiation resistance.

[0047] 4. Low cost and low power consumption: Compared to high-performance semiconductor detectors and advanced systems based on pulse shape discrimination technology, CMOS APS sensors generally have lower cost and power consumption. This makes the discrimination system and method significantly more cost-effective and energy-efficient, making them more suitable for large-scale deployment and long-term operation.

[0048] 5. Flexibility and Scalability: The identification system and method, based on CMOS APS sensors, offer flexibility and scalability. By adjusting the CMOS APS sensor parameters, signal processing circuitry, and data analysis and processing software, different types of radiation particles can be detected and identified. Furthermore, the identification system can be combined with other detection technologies to form a more comprehensive radiation detection system.

[0049] 6. Compared with existing detection equipment (or technology), the present invention has the advantages of less interference from the external environment, lower maintenance cost, higher sensitivity and lower requirements for software and hardware, and its performance remains stable even in high-energy radiation environment.

[0050] The present invention is further described below with reference to the figures and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the connection relationship between the components of the CMOS sensor hybrid radiation discrimination system;

[0052] Figure 2 The morphology and pixel value distribution curve of a typical α response event;

[0053] Figure 3 The morphology and pixel value distribution curve of a typical β response event;

[0054] Figure 4 The morphology and pixel value distribution curve of a typical γ response event;

[0055] Figure 5 The distribution diagram of average pixel values of α, β, and γ response events under different gains and integration times;

[0056] Figure 6 The distribution curve of the number of pixels of α, β, and γ response events under different gains and integration times;

[0057] Figure 7 The rectangular distribution curves of α, β, and γ response events under different gains and integration times are shown;

[0058] Figure 8 The aspect ratio distribution curves of α, β, and γ response events under different gains and integration times are shown;

[0059] Figure 9 The compactness distribution curves of α, β, and γ response events under different gains and integration times are shown;

[0060] Figure 10 The convexity distribution curves of α, β, and γ response events under different gains and integration times are shown;

[0061] Figure 11 The bar graphs of the compactness of β and γ response events under different gains and integration times are shown;

[0062] Figure 12 This is a comparison diagram of the average pixel value distribution of α, β, and γ response events;

[0063] Figure 13 This is a comparison diagram of the convexity distribution of α, β, and γ response events;

[0064] Figure 14 This is a comparison diagram of the number of pixels with response events α, β, and γ;

[0065] Figure 15 The comparison diagram of the aspect ratio distribution of α, β, and γ response events;

[0066] Figure 16 This is a comparison diagram of the rectangular degree distribution of α, β, and γ response events;

[0067] Figure 17 Comparison diagram of the compactness distribution of α, β, and γ response events;.

[0068] Legend: CMOS sensor 1; circuit board 2; chip board 3; SOC chip 31; PC 4. DETAILED DESCRIPTION

[0069] Example 1:

[0070] The α-β-γ mixed radiation discrimination method based on radiation response signals relies on the CMOS sensor mixed radiation discrimination system.

[0071] A CMOS sensor hybrid radiation identification system includes a CMOS sensor 1, a circuit board 2, a chip board 3, a housing, and a PC 4. The surface of the CMOS sensor 1 has no glass packaging. The CMOS sensor 1 is fixedly mounted on the circuit board 2. The circuit board 2 is provided with a power supply interface. The circuit board 2 and the chip board 3 are communicated via a DuPont connector. The chip board 3 is provided with a SOC chip 31, which is used to output a frame image containing a radiation response signal. The housing is made of radiation shielding material. The interior of the housing is provided with an inner cavity for accommodating the circuit board and the chip board. The exterior of the housing is provided with a radiation entrance hole covered by a light-shielding film. The light-shielding film is used to block visible light and allow radiation to pass through. The circuit board 2 and the chip board 3 are both fixedly mounted in the inner cavity of the housing. The photosensitive surface of the CMOS sensor 1 is arranged opposite the radiation entrance hole. There is no obstruction between the photosensitive surface of the CMOS sensor 1 and the radiation entrance hole, and the distance between the two is less than 5 mm. The PC 4 is connected to the chip board 3 via wired or wireless communication, and is used to adjust the parameters of the CMOS sensor 1, and store and display frame images containing radiation response signals (the frame images are grayscale images, in which the pixel value of pure black is 0 and the pixel value of pure white is 255).

[0072] The screening method is as follows:

[0073] S01, obtaining a frame image containing a radiation response signal: α, β, and γ rays in the environment are incident on the CMOS sensor, causing the CMOS sensor to generate a radiation response signal corresponding to the type of ray. The SOC chip receives the data containing the radiation response signal output by the CMOS sensor and converts it into a continuous frame image and transmits it to the PC.

[0074] S02, distinguishing rays based on differences in connected area features: identifying and extracting connected areas corresponding to each radiation response signal in the frame image; analyzing the features of each connected area from multiple dimensions; distinguishing α, β, and γ rays based on differences in at least one or more features of different types of rays; the features include average pixel value, number of pixels, rectangularity, aspect ratio, compactness, and convexity.

[0075] The six characteristics are defined as follows:

[0076] I. The mean pixel value (M) is used to represent the overall brightness level of the radiation response signal. Its calculation formula is shown in Formula 1.

[0077] Formula 1:

[0078] Where M is the average pixel value of the connected area, P is the number of pixels in the connected area, δ is the screening function used to determine whether the target pixel belongs to the connected area, and the connected area is the area occupied by a radiation response signal in the frame image; (i, j) represents the pixel in the i-th row and j-th column, n and m represent the number of rows and columns of pixels respectively, I(i, j) represents the pixel value of the i-th row and j-th column pixel (between 0 and 255), and T is the pixel value threshold of the connected area pixel (if the pixel value is lower than the threshold, the pixel does not belong to the connected area, and if the pixel value is higher than or equal to the threshold, the pixel belongs to the connected area).

[0079] II. The number of pixels (P) is the number of pixels in the connected region, which reflects the area of the connected region. Its calculation formula is shown in Formula 2.

[0080] Formula 2:

[0081] Where P is the number of pixels in the connected area.

[0082] III. Rectangularity (R) is used to characterize the similarity between the shape of a connected region and an ideal rectangle. A larger value indicates a higher similarity to a rectangle, while a smaller value indicates a lower similarity. For example, the rectangularity of a circle is approximately 0.785. Its calculation formula is shown in Equation 3.

[0083] Formula 3:

[0084] Among them, R is the rectangular degree of the connected region, A is the area of the connected region, and A MER is the area of the minimum circumscribed rectangle of the connected region.

[0085] IV. The aspect ratio (V) is used to characterize the symmetry and directionality of the connected region. Its calculation formula is shown in Equation 4.

[0086] Formula 4:

[0087] Among them, V is the aspect ratio of the connected area, W and H are the width and height of the minimum circumscribed rectangle of the connected area, respectively. max 、x min are the maximum and minimum coordinates of the connected area in the horizontal direction, y max 、y min are the maximum and minimum vertical coordinates of the connected region, respectively.

[0088] V. Compactness (O) is used to distinguish and identify different shapes of connected regions. It is invariant to translation, scale, and rotation. Its representation formula is shown in Formula 5.

[0089] Formula 5:

[0090] Where O is the compactness of the connected region, S is the boundary length of the connected region, and A is the area of the connected region.

[0091] VI. Convexity (C) is used to describe the shape complexity and edge concavity of connected regions. The smaller the convexity, the more irregular the shape. Its calculation formula is shown in Equation 6.

[0092] Formula 6:

[0093] Where C is the convexity of the connected region, A is the area of the connected region, and F is the sum of the areas of all convex hulls of the connected region.

[0094] In step S02, a technical route for distinguishing ray types based on feature differences was proposed. However, in practice, the characteristics of the α response event are the most obvious and can be distinguished by a single feature (average pixel value). However, the characteristics of the β and γ response events are relatively similar, differing only in the average pixel value. Due to the limited morphological features studied, there are still certain difficulties in the identification process. To address this problem, a convolutional neural network is used to train the templates of the response events. The convolutional layer extracts local perception features, allowing the network to focus on local details in the image. β and γ response events not only occupy a small area in dark images but also have variable shapes. By designing a specific network structure, more feature information of the response events can be extracted, thereby achieving a distinguishing effect.

[0095] The specific process is as follows:

[0096] A. Establish a response event template library:

[0097] First, the continuous frame images are binarized, and then the connected areas are identified and extracted based on the set pixel value threshold of the connected area pixels; the six features of each connected area (i.e., morphological and structural features) are analyzed, and the following two operations are performed: ①. The connected areas corresponding to mixed events and edge events are eliminated (mixed events refer to the overlap of connected areas excited by multiple types of rays, and edge events refer to the connected areas excited by rays at the edge of the frame image); ②. The remaining connected areas are marked with the corresponding response events, and the response events include α, β, and γ response events; the marked α, β, and γ response events are extracted respectively, and placed on a pure black background of specified pixel size (the template size of the α response event is 30×30 pixels, the template size of the β response event is 15×15 pixels, and the template size of the γ response event is 15×15 pixels), thereby obtaining the α, β, and γ response event template libraries respectively.

[0098] B. Data preprocessing:

[0099] The response events in the α, β, and γ response event template libraries are preprocessed, including the following two steps: ① Normalize the six features contained in the response events to the [0, 1] interval so that the neural network can learn the features more effectively; ② Perform random rotation and / or flipping and / or scaling operations on the response events to increase data diversity and the generalization ability of the model.

[0100] C. Model training:

[0101] The preprocessed data in the α, β, and γ response event template libraries are divided into training set, validation set, and test set in a ratio of 7:1:2; the convolutional neural network model (CNN) is trained using the training set; the model output is calculated through forward propagation, and then the output is compared with the true label to calculate the loss value (the loss function used is the cross-entropy loss function), and then the model weights are updated through the backpropagation algorithm to minimize the loss value; the stochastic gradient descent (SGD) or Adam optimization algorithm is used to accelerate model convergence and improve training efficiency; training is stopped when the model loss value no longer decreases for multiple consecutive epochs on the validation set or reaches the predetermined maximum number of training rounds.

[0102] D. Verification and Evaluation:

[0103] The trained model is verified and evaluated using the test set. The evaluation indicators include accuracy, precision, recall, and F1 value. By drawing the confusion matrix, the classification effect of the model on different types of rays is intuitively displayed, and the classification errors and advantage categories of the model are analyzed.

[0104] Summary of characteristics of single alpha, beta, and gamma response events:

[0105] Figure 2 The following is a typical α response event morphology and pixel value distribution curve. The integration time of the four grids is 45ms, and the only difference is the gain change (9dB, 16dB, 32dB and 63dB for the upper left, upper right, lower left and lower right respectively). Figure 2 As can be seen, an alpha response event appears as a bright white spot on the image. The pixel values of multiple pixels in the center of the white spot are saturated (255), and these saturated pixels are closely connected, forming a bright, uniform, continuous area. This is because alpha particles have high energy and weak penetrating power. When alpha particles interact with the photosensitive surface of the CMOS sensor, the high-density energy deposition causes the pixel values of the response area to quickly reach saturation.

[0106] Figure 3The following is a typical β response event morphology and pixel value distribution curve. The integration time of the four grids is 45ms, and the only difference is the gain change (9dB, 16dB, 32dB and 63dB for the upper left, upper right, lower left and lower right respectively). Figure 3 It can be seen that the beta response event is significantly weaker than that of alpha particles and the response area is smaller. Under low gain conditions, only one pixel in the response signal has a pixel value of more than 200 (the upper left and upper right small pictures of the four-square grid), and only when the gain reaches 32dB and above, the pixel values of multiple pixels reach more than 200. This is because beta particles have a small mass and strong penetrating power, and their ionization value is much smaller than that of alpha particles of the same energy. When beta particles interact with the photosensitive surface of the CMOS sensor, significant scattering occurs, causing their energy to gradually decay along the propagation path, resulting in low signal intensity and energy concentrated in a local area.

[0107] Figure 4 The following is a typical γ response event morphology and pixel value distribution curve. The integration time of the left and right small pictures is 45ms. The only difference is the gain change (the left small picture is 9dB, the right small picture is 32dB). Figure 4 It can be seen that the gamma response event is weaker than the beta response event, but its morphology is similar to that of the beta response event. Furthermore, as the gain increases, the pixel value in the central region of the gamma response event increases more significantly than that of the beta response event. When the gain is 9dB, only one pixel's pixel value exceeds 140 during the response time. At a gain of 32dB, the pixel value of a single pixel reaches 255. This is because gamma particles lack direct ionization and have greater penetrability than beta particles. When gamma particles interact with the photosensitive surface of a CMOS sensor, they primarily generate secondary electrons through the photoelectric effect, the Compton effect, and the electron pair effect. These secondary electrons then trigger a response event in the CMOS sensor. Both gamma and beta response events are triggered by electrons (beta particles are essentially high-speed electron streams), and therefore have similar morphologies.

[0108] Analysis of the average pixel value of a single α, β, or γ response event:

[0109] Figure 5 The distribution curve of average pixel values of α, β, and γ response events under different gains and integration times is shown in the figure. Figure 5 The horizontal axis is the average pixel value, and the vertical axis is the number of response events. Figure 5 It can be seen that the average pixel values of the α, β, and γ response events all show an upward trend with the increase of gain, which is reflected in the rightward shift of the curve peak. The difference lies in the variability of the peak value and peak width.

[0110] For the β response event, when the gain is 9dB and the integration time is 22.5ms or 45ms, the average pixel value is mainly concentrated between 15-30, and the peak value appears at 20; when the gain is 32dB and the integration time is 22.5ms or 45ms, the average pixel value is mainly concentrated between 25-100, and the peak value appears at 52. At this time, the peak value of the distribution curve decreases and the peak width expands.

[0111] For β response events, when the gain is 9dB and the integration time is 22.5ms or 45ms, the average pixel value is mainly concentrated between 15-30, and the peak value appears at 20; when the gain is 63dB and the integration time is 22.5ms or 45ms, the average pixel value is mainly concentrated between 50-130, and the peak value appears at 100. At this time, the peak value of the distribution curve decreases and the peak width expands.

[0112] For the γ response event, when the gain is 9dB and the integration time is 22.5ms or 45ms, the average pixel value is mainly concentrated between 13.5-14.5, and the peak value appears at 14. When the gain is 32dB and the integration time is 22.5ms or 45ms, the average pixel value is mainly concentrated between 13-25, and the peak value appears at 17. At this time, the peak value of the distribution curve decreases and the peak width expands.

[0113] Analysis of experimental results: Increasing gain significantly enhances the average pixel value of the three response events, with the magnitude of the enhancement for β and γ response events being significantly greater than that for α response events. This is because increasing gain leads to an exponential increase in event intensity. α particles have strong ionization ability and a more pronounced response event, resulting in a more concentrated distribution of response events at high gain (as evidenced by a narrowing of the peak width of the distribution curve). However, due to the lower energy deposition of β and γ particles, increasing gain amplifies the originally weaker events. Combined with interparticle scattering and mutual influence, the response events are more widely distributed at high gain (as evidenced by a widening of the peak width of the distribution curve). In contrast, integration time primarily affects the cumulative amount of response event sampling and has no significant effect on the distribution of average pixel values.

[0114] Analysis of the number of pixels of single α, β, and γ response events:

[0115] Figure 6 The distribution curve of the number of pixels of α, β, and γ response events under different gains and integration times is shown in the figure. Figure 6 The horizontal axis is the number of pixels, and the vertical axis is the number of response events. Figure 6 It can be seen that the number of pixels included in the α response event increases significantly with the increase of gain, while the number of pixels included in the β and γ response events remains relatively stable with the change of gain.

[0116] For the α response event, at a gain of 9dB and an integration time of 22.5ms or 45ms, the number of pixels included in the response event is mainly concentrated between 50-75, with the peak occurring at 60. When the gain reaches 63dB and the integration time is 22.5ms or 45ms, the number of pixels included in the response event is mainly concentrated between 115-140, with the peak occurring at 130. At this time, the peak of the distribution curve becomes lower and the peak width expands.

[0117] For the β response event, at a gain of 9dB, 16dB, 32dB, or 63dB and an integration time of 22.5ms or 45ms, the number of pixels contained in the response event is mainly concentrated between 8 and 9, and the peak also appears between 8 and 9. Under various gains or integration times, the peak value and peak width of the distribution curve do not change significantly.

[0118] For the γ response event, at a gain of 9dB or 32dB and an integration time of 22.5ms or 45ms, the number of pixels contained in the response event is mainly concentrated between 7.5 and 8.5, with a peak value of 8. Under various gains or integration times, the peak value and peak width of the distribution curve do not change significantly.

[0119] Analysis of experimental results: Increasing gain significantly increases the number of pixels responding to α events, while having no significant effect on the number of pixels responding to β and γ events. This is because α particles have a relatively strong ionization ability, resulting in a larger number of pixels affected by these events, affecting a larger number of surrounding pixels. Therefore, increasing gain tends to amplify the increase in the number of pixels. β and γ particles, on the other hand, have relatively weak ionization abilities and affect fewer surrounding pixels. Therefore, increasing gain makes it less likely to increase the number of pixels. In contrast, integration time has no significant effect on the number of pixels responding to these events.

[0120] Rectangularity analysis of single α, β, and γ response events:

[0121] Figure 7 It is the rectangular distribution curve of α, β, and γ response events under different gains and integration times. Figure 7 The horizontal axis is the rectangular degree, and the vertical axis is the number of response events. Figure 7 It can be seen that the rectangularity of the α response event decreases with the increase of gain, and under high gain conditions (63dB), there is only one peak in the distribution curve; the rectangularity of the β and γ response events is stable with the change of gain, and there are multiple peaks in the distribution curve under various gain conditions.

[0122] For the α response event, at gains of 9dB and 16dB and integration times of 22.5ms or 45ms, the rectangularity is mainly concentrated between 0.77 and 0.98, with multiple peaks appearing around 0.89; at a gain of 63dB and integration times of 22.5ms or 45ms, the rectangularity shrinks to between 0.7-0.91, with peaks appearing around 0.8. At this time, the peak of the distribution curve decreases (compared to a gain of 32dB and an integration time of 22.5ms or 45ms), but the peak width does not change significantly.

[0123] For the β response event, at a gain of 9dB, 16dB, 32dB, or 63dB and an integration time of 22.5ms or 45ms, the rectangularity is mainly concentrated between 0.4 and 0.56, with peaks appearing at 0.4, 0.5, and 0.56. Under various gains or integration times, the peak value and peak width of the distribution curve do not change significantly.

[0124] For the γ response event, at a gain of 9dB or 32dB and an integration time of 22.5ms or 45ms, the rectangularity is mainly concentrated between 0.4-0.56, with peaks at 0.4, 0.5 and 0.56. Under various gains or integration times, the peak value and peak width of the distribution curve do not change significantly.

[0125] Experimental Results Analysis: Increasing gain makes the α response event's shape in the dark image more circular, which is also confirmed by the aforementioned rectangularity variation. This is because increasing gain enhances the CMOS sensor's characterization efficiency for ionized particles, making the sensor more sensitive to ionizing radiation and reducing shape distortion caused by ionized particle trajectories or internal inhomogeneities in the sensor pixel structure. As gain changes, the rectangularity distributions of the β and γ response events exhibit stability, with multiple peaks appearing in the distribution curves, indicating concavities or irregularities in the β and γ response event shapes. This is due to the complexity of particle interactions, scattering, and the high-resolution pixel structure of the CMOS sensor. In contrast, integration time has no significant effect on the rectangularity of the response events.

[0126] Aspect ratio analysis of single α, β, and γ response events:

[0127] Figure 8 The aspect ratio distribution curve of α, β, and γ response events under different gains and integration times is shown in the figure. Figure 8 The horizontal axis is the aspect ratio, and the vertical axis is the number of response events. Figure 8 It can be seen that the aspect ratio distribution of α response events is the most concentrated (the range is the narrowest), the aspect ratio distribution of β response events is the loosest (the range is the widest), and the aspect ratio distribution range of γ response events is between α response events and β response events (the range is medium).

[0128] For the α response event, at a gain of 9dB, 16dB, 32dB, or 63dB and an integration time of 22.5ms or 45ms, the aspect ratio is concentrated between 0.85 and 1.1. As the gain increases, the peak value gradually approaches 1, showing a more obvious concentration trend.

[0129] For the β response event, at gains of 9dB, 16dB, 32dB, or 63dB, and integration times of 22.5ms or 45ms, the aspect ratio is concentrated between 0.75 and 1.32, with a peak at 1. Smaller peaks also appear at 0.8 and 1.25. The peak value and width of the distribution curve do not change significantly at various gains or integration times.

[0130] For the gamma response event, at gains of 9dB or 32dB and integration times of 22.5ms or 45ms, the aspect ratio is concentrated between 0.78 and 1.4, with a peak at 1. Relatively small peaks also appear at 0.8 and 1.25. The peak value and width of the distribution curve do not change significantly at various gains or integration times.

[0131] Analysis of experimental results: Increasing gain causes α response events to adopt a more symmetrical or circular shape in dark images, which is also confirmed by the aspect ratio variation pattern described above. As gain changes, the aspect ratio distributions of β and γ response events remain stable, with multiple peaks appearing in the distribution curves. This indicates that there are irregularities in the β and γ response events, but their overall geometric characteristics exhibit a certain degree of symmetry or uniformity. This is due to the scattering of particles in the CMOS sensor, which leads to irregularities in the response event morphology. However, this irregularity is not sufficient to affect the overall aspect ratio distribution. In contrast, integration time has no significant effect on the aspect ratio of the response events.

[0132] Compactness analysis of single α, β, and γ response events:

[0133] Figure 9 The distribution curve of the compactness of α, β, and γ response events under different gains and integration times is shown in the figure. Figure 9 The horizontal axis in is the compactness, and the vertical axis is the number of response events. As can be seen from Figure 9, the compactness of the α response event increases with the increase of gain, and under high gain conditions, there is only one peak in the distribution curve. The compactness of the β and γ response events is stable with the change of gain, and there are multiple peaks in the distribution curve under various gain conditions.

[0134] For α response events, at a gain of 9dB and an integration time of 22.5ms or 45ms, the tightness is concentrated between 10-11; at a gain of 63dB and an integration time of 22.5ms or 45ms, the tightness is concentrated between 10.5-11.5, with a clear peak near 11.

[0135] For both β and γ response events, the compactness distribution is relatively loose under various gain conditions and integration times of 22.5ms or 45ms, mainly distributed between 12 and 15, with multiple peaks within the distribution range. In addition, a small number of γ response events have a compactness distribution of 16.

[0136] Analysis of experimental results: The increase in gain makes the α response event more symmetrical or circular in the dark image, which is also confirmed by the above-mentioned compactness change law. Figure 9 After rounding down the compactness of β and γ response events, the compactness histograms of β and γ response events under different gains and integration times are drawn, as shown in Figure 2. Figure 11 shown. Figure 11 The numbers above each column in the table are the number of response events. Figure 11 As can be seen in the figure, under 32dB gain conditions, the compactness of β and γ response events differs. The compactness of β response events is concentrated between 14 and 15, while the compactness of γ response events is concentrated between 15 and 16. This phenomenon indicates that the morphology of γ response events is flatter than that of β. This is because gamma particles have stronger penetrating power, and the scattering effect they induce in the CMOS sensor makes the propagation path more complex, resulting in the flattening of the morphology. In contrast, the integration time has no significant effect on the compactness of the response events.

[0137] Convexity analysis of single α, β, and γ response events:

[0138] Figure 10 It is the convexity distribution curve of α, β, and γ response events under different gains and integration times. Figure 10 The horizontal axis is the convexity, and the vertical axis is the number of response events. Figure 10 It can be seen that the convexity of the α response event shows a slow upward trend with increasing gain, but the peak value and peak width of the distribution curve remain relatively stable. The convexity of the β and γ response events remains relatively stable with changes in gain.

[0139] For the α response event, at a gain of 9 dB and an integration time of 22.5 ms or 45 ms, the convexity is concentrated between 0.85 and 0.87, with the only peak occurring at 0.86. At a gain of 63 dB and an integration time of 22.5 ms or 45 ms, the convexity is concentrated between 0.89 and 0.91, with the only peak occurring at 0.9.

[0140] For the beta response event, the convexity distribution remains stable within a relatively fixed range under various gain conditions and with either a 22.5ms or 45ms integration time. Its convexity is concentrated between 0.55 and 0.66, with multiple peaks, most notably at 0.55 and 0.58. When the gain is increased to 63dB, the peak at 0.55 decreases, while the peak at 0.58 increases.

[0141] For the gamma response event, the convexity distribution remains stable within a relatively fixed range under various gain conditions and with either a 22.5ms or 45ms integration time. Its convexity is concentrated between 0.52 and 0.56, with multiple peaks, most notably at 0.55 and 0.58. When the gain is increased to 32dB, the peak at 0.55 becomes significantly larger.

[0142] Experimental Results Analysis: Increasing gain generally increases the convexity of α, β, and γ response events, leading to more regularized event morphologies. The most pronounced change is observed in the α response event. This is due to the stronger ionization power of α particles compared to β and γ particles, which makes the convexity increase more pronounced with increased response gain. However, the convexity changes in the β and γ response events require higher gain levels to manifest. In contrast, integration time has no significant effect on the compactness of the response events.

[0143] Comparison of the average pixel value distribution of α, β, and γ response events:

[0144] Figure 12 The following figure compares the average pixel value distributions of α, β, and γ response events. The horizontal axis represents the average pixel value, and the vertical axis represents the number of response events. The left panel shows a comparison of α and β response events, the middle panel shows a comparison of α and γ response events, and the right panel shows a comparison of β and γ response events. The left panel shows that as the gain increases, the average pixel value distribution curves of α and β response events gradually converge. When the gain reaches 63dB, the curves overlap significantly within the average pixel value range of 112-135. The middle panel shows that the average pixel value distribution curves of α and γ response events maintain a certain separation at all gain levels, indicating that before the gain reaches 32dB, analyzing the differences in average pixel values can effectively identify α radiation in a mixed radiation field of the three. For β and γ response events, the average pixel value distributions show some similarity, but they can be distinguished within specific ranges. As shown in the right panel, at a gain of 32dB, average pixel values below 20 are exclusively due to γ response events, while average pixel values above 70 are exclusively due to β response events.

[0145] Comparison of convexity distribution of α, β, and γ response events:

[0146] Figure 13 The following figure compares the convexity distributions of α, β, and γ response events. The horizontal axis represents convexity, and the vertical axis represents the number of response events. The left panel compares α and β response events, the middle panel compares α and γ response events, and the right panel compares β and γ response events. As can be seen from the left and middle panels, the convexity of the α response event is significantly higher than that of the β and γ response events. The curves for the α response event maintain a certain distance from those for the β and γ response events at all gain levels, and the distance between the curves for the α response event and the γ response event is even more significant. This indicates that based on the convexity characteristics of the response events, α rays can be effectively identified in the mixed radiation field of the three. For the β and γ response events, the convexity distributions show some similarity. As shown in the right panel, the curves corresponding to the β and γ response events overlap significantly under all gain conditions, but the β response events are significantly more distributed than the γ response events in the region where the convexity exceeds 0.6.

[0147] Comparison of pixel number distribution of α, β, and γ response events:

[0148] Figure 14 The following figure compares the distribution of pixel numbers for α, β, and γ response events. The horizontal axis represents the number of pixels, and the vertical axis represents the number of response events. The left panel compares α and β response events, the middle panel compares α and γ response events, and the right panel compares β and γ response events. As can be seen from the left and middle panels, the α response event contains significantly more pixels than the β and γ response events. The curve for the α response event maintains a certain distance from the curves for the β and γ response events at all gain levels, and the separation between the curves for the α response event and the γ response event is even more pronounced. This indicates that the difference in the number of pixels in the response event can effectively distinguish α rays from the mixed radiation field of the three. In contrast, the number of pixels in the β and γ response events does not show significant difference. As shown in the inset on the right, the distribution curves of β and γ response events overlap significantly under all gain conditions and are mainly concentrated between 7 and 12 pixels. However, the number of γ response event pixels at 6 pixels is greater than that of β response events, and the number of γ response event pixels at 9 pixels is less than that of β response events.

[0149] Comparison of aspect ratio distribution of α, β, and γ response events:

[0150] Figure 15The figure below shows a comparison of the aspect ratio distributions of α, β, and γ response events. The horizontal axis represents the aspect ratio, and the vertical axis represents the number of response events. The left panel shows a comparison of α and β response events, the middle panel shows a comparison of α and γ response events, and the right panel shows a comparison of β and γ response events. The left and middle panels show that the distribution peaks of the α, β, and γ response events overlap significantly near an aspect ratio of 1. Furthermore, the α response event's peak is closer to 1 and gradually approaches 1 as the gain increases, while the β and γ response event peaks remain relatively stable. After the gain exceeds 32 dB, there is no significant overlap of the distribution peaks except for the aspect ratio of 1. This indicates that, after reaching a certain gain level, analyzing the differences in the aspect ratios of the response events can effectively identify α radiation in a mixed radiation field of the three. In contrast, in the right panel, all the distribution peaks of the β and γ response events show significant overlap, and some γ response events are distributed in the region with an aspect ratio greater than 1.3. Although there are some differences between the two events, they do not constitute obvious distinguishing features as a whole. Therefore, it is very difficult to distinguish between β and γ response events based solely on the distribution of aspect ratios.

[0151] Comparison of rectangular degree distribution of α, β, and γ response events:

[0152] Figure 16 This figure shows a comparison of the rectangularity distributions of α, β, and γ response events. The horizontal axis represents rectangularity, and the vertical axis represents the number of response events. The left panel shows a comparison of α and β response events, the middle panel shows a comparison of α and γ response events, and the right panel shows a comparison of β and γ response events. The left and middle panels show that the rectangularity of α response events is significantly higher than that of β and γ response events. The curves for α response events maintain a certain distance from those for β and γ response events at all gain levels, and the distance between the curves for α response events and γ response events is even more significant. This indicates that the rectangularity characteristics of response events can effectively identify α radiation in a mixed radiation field of the three. In contrast, the rectangularity of β and γ response events exhibits some similarity. As shown in the right panel, the curves for β and γ response events overlap significantly in the rectangularity regions of 0.4 and 0.5. Furthermore, compared to β response events, γ response events are more numerous near rectangularity 0.36 and 0.53, while being significantly less numerous near rectangularity 0.56.

[0153] Comparison of the compactness distribution of α, β, and γ response events:

[0154] Figure 17The following figure compares the compactness distribution of α, β, and γ response events. The horizontal axis represents compactness, and the vertical axis represents the number of response events. The left panel compares α and β response events, the middle panel compares α and γ response events, and the right panel compares β and γ response events. As can be seen from the left and middle panels, the compactness of the α response event is significantly lower than that of the β and γ response events. Furthermore, the α response event curve gradually approaches the β and γ response event curves as the gain increases. When the gain exceeds 32dB, the α response event curve overlaps slightly with the β response event curve, but maintains a certain distance from the γ response event curve. This indicates that at a certain gain level, the distribution of event compactness can effectively identify α radiation in a mixed radiation field of the three. In contrast, the compactness characteristics of the β and γ response events are similar within a certain range (intersecting in the region of 14-15), but can be distinguished within specific intervals. Referring to the small figure on the right, under all gain conditions, the compactness of 15±0.1 is exclusive to γ response events, the compactness of 15.9-16.3 is exclusive to γ response events, and the compactness of 11-11.3 is exclusive to β response events.

[0155] Comparative summary of α, β, and γ response event characteristics:

[0156] In summary, at all gain levels, the distributions of average pixel value, pixel count, rectangularity, convexity, and compactness for α response events differ significantly from those for β and γ response events. Therefore, these features can be used to distinguish α radiation from mixed radiation fields. The characteristics of β and γ response events are relatively similar, and can be distinguished by the average pixel value and compactness within specific intervals. However, overall, the distributions of pixel count, rectangularity, convexity, and compactness for β and γ response events do not show significant differences. Furthermore, the aspect ratios of all three response events are primarily concentrated at 1, and within a specific interval near 1, their distinguishing features are effective for α response events.

Claims

1. An α-β-γ mixed radiation discrimination method based on radiation response signals, relying on a CMOS sensor mixed radiation discrimination system; A CMOS sensor hybrid radiation discrimination system includes a CMOS sensor, a circuit board, a chip board, a housing, and a PC. The CMOS sensor has no glass encapsulation on its surface and is fixedly mounted on the circuit board. The circuit board is provided with a power supply interface, and the circuit board and the chip board are communicatively connected via a DuPont connector. The chip board is provided with a SOC chip for outputting a frame image containing a radiation response signal. The housing is made of radiation shielding material, and an inner cavity for accommodating the circuit board and the chip board is provided inside the housing. A radiation entrance hole covered by a light-shielding film is provided on the outside of the housing. The circuit board and the chip board are both fixedly mounted in the inner cavity of the housing, and the photosensitive surface of the CMOS sensor is arranged opposite the radiation entrance hole. There is no obstruction between the photosensitive surface of the CMOS sensor and the radiation entrance hole, and the distance between the two is less than 5 mm. The PC is communicatively connected to the chip board via wired or wireless means, and is used to adjust the parameters of the CMOS sensor and store and display frame images containing radiation response signals. Its characteristics are: The screening method is as follows: S01, acquiring a frame image containing a radiation response signal: α, β, and γ rays in the environment are incident on the CMOS sensor, causing the CMOS sensor to generate a radiation response signal corresponding to the type of ray. The SOC chip receives the data containing the radiation response signal output by the CMOS sensor and converts it into a continuous frame image and transmits it to the PC; S02, distinguishing rays based on differences in connected area features: identifying and extracting connected areas corresponding to each radiation response signal in the frame image; analyzing the features of each connected area from multiple dimensions; distinguishing α, β, and γ rays based on differences in at least one or more features of different types of rays; the features include average pixel value, number of pixels, rectangularity, aspect ratio, compactness, and convexity.

2. The α-β-γ mixed radiation discrimination method based on radiation response signals according to claim 1, characterized in that: The six characteristics are defined as follows: I. The average pixel value is used to represent the overall brightness level of the radiation response signal. Its calculation formula is shown in Formula 1. Formula 1: Where M is the average pixel value of the connected area, P is the number of pixels in the connected area, δ is the screening function used to determine whether the target pixel belongs to the connected area, and the connected area is the area occupied by a radiation response signal in the frame image; (i, j) represents the pixel in the i-th row and j-th column, n and m represent the number of rows and columns of pixels respectively, I(i, j) represents the pixel value of the pixel in the i-th row and j-th column, and T is the pixel value threshold of the connected area pixel; II. The number of pixels is the number of pixels in the connected region, which reflects the area of the connected region. Its calculation formula is shown in Formula 2. Formula 2: Where P is the number of pixels in the connected area; III. Rectangularity is used to characterize the similarity between the shape of the connected region and the ideal rectangle. Its calculation formula is shown in Formula 3. Formula 3: Among them, R is the rectangular degree of the connected region, A is the area of the connected region, and A MER is the area of the minimum circumscribed rectangle of the connected region; IV. The aspect ratio is used to characterize the symmetry and directionality of the connected region. Its calculation formula is shown in Formula 4. Formula 4: Among them, V is the aspect ratio of the connected area, W and H are the width and height of the minimum circumscribed rectangle of the connected area, respectively. max 、x min are the maximum and minimum coordinates of the connected area in the horizontal direction, y max 、y min are the maximum and minimum coordinates in the vertical direction of the connected area respectively; V. Compactness is used to distinguish and identify different shapes of connected regions. It is invariant to translation, scale, and rotation. Its representation formula is shown in Formula 5. Formula 5: Where O is the compactness of the connected region, S is the boundary length of the connected region, and A is the area of the connected region; VI. Convexity is used to describe the shape complexity and edge concavity of connected regions. The smaller the convexity, the more irregular the shape. Its calculation formula is shown in Formula 6. Formula 6: Where C is the convexity of the connected region, A is the area of the connected region, and F is the sum of the areas of all convex hulls of the connected region.

3. The α-β-γ mixed radiation discrimination method based on radiation response signals according to claim 2, characterized in that: In step S02, a convolutional neural network is used to identify α, β, and γ rays. The specific process is as follows: A. Establish a response event template library: First, continuous frame images are binarized, and then connected regions are identified and extracted based on the set pixel value threshold of the connected region. The six features of each connected region are analyzed and the following two operations are performed:

1. Connected regions corresponding to mixed events and edge events are eliminated; 2. The remaining connected regions are marked with the corresponding response events, which include α, β, and γ response events. Extract the marked α, β, and γ response events respectively, and place them on a pure black background of a specified pixel size, thereby obtaining the α, β, and γ response event template libraries respectively; B. Data preprocessing: Preprocessing the response events in the α, β, and γ response event template libraries includes the following two steps: ① Normalizing the six features contained in the response event to the [0, 1] interval so that the neural network can learn the features more effectively; ② Perform random rotation and / or flipping and / or scaling operations on the response events to increase data diversity and model generalization ability; C. Model training: The preprocessed data in the α, β, and γ response event template libraries are divided into training, validation, and test sets in a ratio of 7:1:

2. The convolutional neural network model is trained using the training set. The model output is calculated through forward propagation, and then the output is compared with the true label to calculate the loss value. The model weights are then updated through the backpropagation algorithm to minimize the loss value. Stochastic gradient descent or Adam optimization algorithms are used to accelerate model convergence and improve training efficiency. Training is stopped when the model loss value stops decreasing for multiple consecutive epochs on the validation set or reaches the predetermined maximum number of training rounds. D. Verification and Evaluation: The trained model is verified and evaluated using the test set. The evaluation indicators include accuracy, precision, recall, and F1 value. By drawing the confusion matrix, the classification effect of the model on different types of rays is intuitively displayed, and the classification errors and advantage categories of the model are analyzed.

4. The α-β-γ mixed radiation discrimination method based on radiation response signal according to claim 3 is characterized in that: Under the gain conditions of 9-63dB, the average pixel value, number of pixels, rectangularity, aspect ratio, compactness and convexity of the α response event are different from those of the β and γ response events, based on which α radiation can be distinguished in the mixed radiation field.

5. The α-β-γ mixed radiation discrimination method based on radiation response signals according to claim 3, characterized in that: Under the gain condition of 32dB, the average pixel value distributions of β and γ response events have exclusive intervals respectively. If the average pixel value of the monitored response event falls within the exclusive interval, β and γ radiation can be distinguished.

6. The α-β-γ mixed radiation discrimination method based on radiation response signals according to claim 4, characterized in that: Under the conditions of 9dB or 32dB gain, the compactness of β and γ response events respectively have exclusive intervals. If the compactness of the monitored response event falls within the exclusive interval, β and γ radiation can be distinguished.

Images