A dark matter detection and analysis method and system based on astronomical telescope

Abstract

The present invention provides a dark matter detection and analysis method and system based on an astronomical telescope, which relates to the field of dark matter detection technology, including the acquisition of multi-band astronomical observation data; using a convolutional neural network to perform deep feature extraction on a calibrated multi-band astronomical data set; using large-scale structure simulation software to simulate the distribution of dark matter in the universe and its gravitational lens effect on light, and obtain the prediction result of the potential position of dark matter; using a time-space correlation algorithm to process the prediction result of the potential position of dark matter, and sampling the parameter space of the first estimate by the Markov chain Monte Carlo method; simulating the interaction between dark matter particles and detector materials, and identifying and extracting dark matter signals by a peak detection method; calculating the confidence of dark matter detection, and obtaining dark matter detection analysis results. The present invention extracts dark matter signals from a complex astronomical background, providing in-depth insights into the nature of dark matter and its role in the universe.

Description

Technical Field

[0001] The present invention relates to the field of dark matter detection technology, and in particular to a dark matter detection and analysis method and system based on an astronomical telescope. Background Art

[0002] In current astronomical research, dark matter detection technology has made significant progress. Scientists use advanced detectors, such as high-purity germanium detectors and liquid xenon detectors, to look for direct evidence of dark matter by measuring the elastic scattering of dark matter particles with the detector nuclei. These experiments usually use high-purity detectors to record the signals of the interaction between dark matter particles and atomic nuclei. For example, liquid xenon detectors and high-purity germanium detectors are widely used to detect scattering events of dark matter particles. However, these technologies face extremely high challenges because the interaction between dark matter and ordinary matter is extremely weak, and the detectors must operate in an extremely low-background experimental environment to reduce the interference of background noise, including the control of extremely low background noise, the reduction of the detector energy threshold, and the cooling and readout technology of large-mass detectors.

[0003] Although existing technologies have been able to detect rare events of dark matter particles, these methods usually rely on complex cryogenic systems and deep underground laboratories to reduce the interference of background noise and are very sensitive to environmental changes. In addition, the background noise level of the detector is still one of the main factors limiting the sensitivity of dark matter detection. When detecting the electromagnetic properties of dark matter particles, existing technologies also face the challenge of distinguishing the electromagnetic properties of different dark matter particles. To overcome these limitations, scientists are developing new detection technologies and data analysis methods to improve the ability to identify dark matter events and reduce false alarm rates. Summary of the invention

[0004] The purpose of the present invention is to provide a dark matter detection and analysis method and system based on an astronomical telescope to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, the present application provides a dark matter detection and analysis method based on an astronomical telescope, comprising:

[0006] Collect multi-band astronomical observation data acquired by the Hubble Space Telescope, Chandra X-ray Observatory and Fermi Gamma-ray Space Telescope, where the multi-band astronomical observation data includes optical, infrared, X-ray and gamma-ray bands, pre-process the multi-band astronomical observation data to obtain a calibrated multi-band astronomical data set;

[0007] The convolutional neural network is used to perform deep feature extraction on the calibrated multi-band astronomical data set to obtain the first feature of the dark matter candidate signal. The autoencoder is used to reduce the dimension and enhance the features of the first feature of the dark matter candidate signal to obtain the second feature.

[0008] According to the second feature, large-scale structure simulation software is used to simulate the distribution of dark matter in the universe and its gravitational lensing effect on light, and a generative adversarial network is used to synthesize simulated data of light affected by the distribution of dark matter. The simulated data is compared with the actual observation data to identify the prediction results of the potential location of dark matter;

[0009] The prediction results of the potential location of dark matter are processed by using a spatiotemporal correlation algorithm, and candidate dark matter signals at different time and space scales are analyzed to obtain a candidate event set of dark matter particle annihilation or decay events. The candidate event set is processed using the Bayesian inference method to obtain a first estimate of the dark matter model parameters. The first estimate is sampled in the parameter space and the model is verified using the Markov chain Monte Carlo method to obtain a second estimate of the dark matter model parameters including density distribution, velocity distribution, and energy spectrum.

[0010] Based on the second estimate, a Monte Carlo simulation method is used to simulate the interaction between dark matter particles and detector materials. The simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles produced. Based on the simulation results, a signal response model is constructed to characterize its energy, time, and spatial distribution characteristics. The signal response model is used to identify and extract dark matter signals from the detector readings through a peak detection method.

[0011] Analyze the background noise in the detector, including cosmic rays, environmental radioactivity and other interference sources, determine the baseline noise level of the detector, and obtain the background noise analysis results; combine the identified extracted dark matter signal with the background noise analysis results, calculate the confidence level of dark matter detection, and obtain the dark matter detection analysis results.

[0012] Preferably, the method uses a convolutional neural network to perform deep feature extraction on the calibrated multi-band astronomical data set to obtain a first feature of the dark matter candidate signal, and uses an autoencoder to perform dimensionality reduction and feature enhancement on the first feature of the dark matter candidate signal to obtain a second feature, which includes:

[0013] Based on the calibrated multi-band astronomical data set, after being processed by the convolution layer in the convolutional neural network, a convolution operation is performed using multiple convolution kernels to obtain the first feature of the dark matter candidate signal, where the first feature includes the shape parameters, color changes and spatial distribution characteristics of the light curve extracted by the convolution layer and activation function in the CNN;

[0014] According to the first feature of the dark matter candidate signal, it is processed by the encoding layer of the autoencoder, which includes multiple convolutional layers and pooling layers, and the spatial dimension of the feature map is reduced by downsampling to obtain a compressed representation;

[0015] The compressed representation is processed using the decoding layer of the autoencoder, and the data dimension is changed using the transposed convolution layer and the upsampling layer to obtain the first feature of the processed dark matter candidate signal;

[0016] Based on the processed first feature of the dark matter candidate signal, the second feature of the dark matter candidate signal is obtained after processing by the bottleneck layer and output layer of the autoencoder, where the second feature includes identifying abnormal patterns of light curves, color changes, and spatial distribution characteristics through the nonlinear transformation and feature selection mechanism of the autoencoder.

[0017] Preferably, the compressed representation is processed by using a decoding layer of an autoencoder, and a transposed convolution layer and an upsampling layer are used to change the data dimension to obtain the first feature of the processed dark matter candidate signal, which includes:

[0018] Compressed representation of the output of the encoding layer based on the autoencoder, processed by the transposed convolution layer, using the transposed convolution kernel W Perform convolution operation to obtain upsampled feature maps;

[0019] The upsampled feature map is processed by using a bilinear interpolation method to obtain a feature map processed by an upsampling layer, wherein the feature map processed by the upsampling layer includes an identification spatial resolution;

[0020] The feature map processed by the upsampling layer is input to the decoding layer to reconstruct the feature representation of the original input data and obtain the first feature of the processed dark matter candidate signal, where the original input data is the initial data processed by the encoder.

[0021] Preferably, according to the second feature, large-scale structure simulation software is used to simulate the distribution of dark matter in the universe and its gravitational lens effect on light, and a generative adversarial network is used to synthesize simulated data of light affected by the distribution of dark matter, and the simulated data is compared with the actual observation data to identify the prediction results of the potential position of dark matter, including:

[0022] Based on the second characteristic, cosmological principles and historical observational data, after N-body simulation processing, a simulated distribution map of dark matter in the universe is obtained, where the simulated distribution map includes the growth, evolution and aggregation of dark matter halos described by the positions and velocities of particles;

[0023] The gravitational lens effect is simulated on the simulated distribution map by simulation software, thereby obtaining the deflection effect of the light path;

[0024] The light path deflection data obtained by simulating the gravitational lens effect is used as the training target of the generative adversarial network to generate simulated data similar to the real astronomical data, where the training goal of the generative adversarial network is to minimize the difference between the generated data and the real data;

[0025] The structural similarity index is used to compare the simulated data synthesized by the generative adversarial network with the actual observed data to identify the predicted results of the potential location of dark matter.

[0026] Preferably, the prediction results of the potential position of dark matter are processed by using a time-space correlation algorithm, and candidate dark matter signals at different time and space scales are analyzed to obtain a candidate event set of dark matter particle annihilation or decay events, which includes:

[0027] Based on the prediction results, a spatiotemporal correlation graph is constructed, where the spatiotemporal correlation graph includes multiple nodes and edges. The nodes are observation data points at different time points and spatial positions, and the edges are the spatiotemporal correlations between the observation data points. After the graph construction process, the dark matter candidate signals at different time and space scales are analyzed to obtain the correlation strength.

[0028] The clustering algorithm is used to identify the correlation strength, and the nodes belonging to the same dark matter event are grouped together to obtain the collective result, which is recorded as the candidate event set of dark matter particle annihilation or decay events.

[0029] Preferably, the interaction between dark matter particles and detector materials is simulated using a Monte Carlo simulation method according to the second estimated value, wherein the simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles generated, including:

[0030] According to the second estimated value, the physical characteristics of the dark matter particles and the properties of the detector material, the collision probability of the dark matter particles and the atomic nuclei in the detector is obtained through Monte Carlo simulation, wherein the collision probability of the dark matter particles and the atomic nuclei in the detector includes calculating the flux of the dark matter particles and the effective area of ​​the detector and considering the interaction cross section between the dark matter particles and the atomic nuclei;

[0031] Based on the collision probability, the efficiency of energy transfer after the collision between dark matter particles and atomic nuclei in the detector is obtained through calculation and processing of energy transfer efficiency.

[0032] By simulating the energy transfer efficiency of secondary particles, the types and quantities of secondary particles produced after the collision of dark matter particles with atomic nuclei in the detector are obtained.

[0033] Preferably, the combined identification of the extracted dark matter signal and the background noise analysis result, calculating the confidence of the dark matter detection, and obtaining the dark matter detection analysis result, includes:

[0034] According to the characteristics of dark matter signals and the distribution of background noise, the signal strength and noise level are quantified to obtain the signal-noise distinction metric;

[0035] Based on the signal-to-noise ratio, the confidence level of dark matter detection is obtained through calculation and processing;

[0036] The confidence of dark matter detection is analyzed. The analysis process is to determine whether the probability of the current data is less than a preset threshold. If so, the detected signal is considered to be a candidate event for dark matter. If not, the detected signal is not considered to be a candidate event for dark matter, and the dark matter detection analysis results are obtained.

[0037] In a second aspect, the present application also provides a dark matter detection and analysis system based on an astronomical telescope, comprising:

[0038] Collection module: used to collect multi-band astronomical observation data acquired by the Hubble Space Telescope, Chandra X-ray Observatory and Fermi Gamma-ray Space Telescope, where the multi-band astronomical observation data includes optical, infrared, X-ray and gamma-ray bands, and pre-process the multi-band astronomical observation data to obtain a calibrated multi-band astronomical data set;

[0039] Feature extraction module: used to perform deep feature extraction on the calibrated multi-band astronomical data set using a convolutional neural network to obtain the first feature of the dark matter candidate signal, and to perform dimensionality reduction and feature enhancement on the first feature of the dark matter candidate signal through an autoencoder to obtain the second feature;

[0040] Identification module: used to simulate the distribution of dark matter in the universe and its gravitational lensing effect on light using large-scale structure simulation software based on the second feature, and use generative adversarial networks to synthesize simulated data of light affected by the distribution of dark matter, compare the simulated data with the actual observation data, and identify the prediction results of the potential location of dark matter;

[0041] Processing module: used to process the prediction results of the potential position of dark matter using the spatiotemporal correlation algorithm, analyze the candidate dark matter signals at different time and space scales, and obtain the candidate event set of dark matter particle annihilation or decay events; apply the Bayesian inference method to process the candidate event set to obtain the first estimated value of the dark matter model parameters; use the Markov chain Monte Carlo method to sample the parameter space of the first estimated value and verify the model to obtain the second estimated value of the dark matter model parameters including density distribution, velocity distribution and energy spectrum;

[0042] Simulation module: used to simulate the interaction between dark matter particles and detector materials using the Monte Carlo simulation method based on the second estimated value. The simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles generated. Based on the simulation results, a signal response model is constructed to characterize its energy, time, and spatial distribution characteristics. The signal response model is used to identify and extract dark matter signals from the detector readings through the peak detection method.

[0043] Computing module: used to analyze the background noise in the detector, including cosmic rays, environmental radioactivity and other interference sources, determine the baseline noise level of the detector, and obtain the background noise analysis results; combine the identified extracted dark matter signal with the background noise analysis results, calculate the confidence of dark matter detection, and obtain the dark matter detection analysis results.

[0044] In a third aspect, the present application also provides a dark matter detection and analysis device based on an astronomical telescope, comprising:

[0045] Memory for storing computer programs;

[0046] A processor is used to implement the steps of the dark matter detection and analysis method based on an astronomical telescope when executing the computer program.

[0047] In a fourth aspect, the present application further provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned dark matter detection and analysis method based on an astronomical telescope are implemented.

[0048] The beneficial effects of the present invention are:

[0049] The present invention combines feature extraction of multi-band astronomical data, large-scale structure simulation, comparison of simulated data generated by generative adversarial networks, processing of spatiotemporal correlation algorithms, and Bayesian inference and Markov chain Monte Carlo methods. This method not only improves the ability to identify dark matter signals, but also improves the assessment of the confidence level of dark matter detection through multivariate analysis and clustering algorithms. Through this comprehensive approach, dark matter signals can be more effectively extracted from complex astronomical backgrounds, providing deeper insights into the nature of dark matter and its role in the universe.

[0050] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 Schematic diagram of the process of dark matter detection and analysis method based on astronomical telescope described in an embodiment of the present invention;

[0053] Figure 2 Schematic diagram of the structure of a dark matter detection and analysis system based on an astronomical telescope according to an embodiment of the present invention;

[0054] Figure 3 It is a schematic diagram of the structure of the dark matter detection and analysis equipment based on the astronomical telescope described in an embodiment of the present invention.

[0055] In the figure: 701, collection module; 702, feature extraction module; 703, recognition module; 704, processing module; 705, simulation module; 706, calculation module; 800, dark matter detection and analysis equipment based on astronomical telescope; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0057] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0058] Embodiment 1:

[0059] This embodiment provides a dark matter detection and analysis method based on an astronomical telescope.

[0060] See also Figure 1 , the figure shows that the method includes step S100, step S200, step S300, step S400, step S500 and step S600.

[0061] S100: Collect multi-band astronomical observation data acquired by the Hubble Space Telescope, the Chandra X-ray Observatory and the Fermi Gamma-ray Space Telescope, where the multi-band astronomical observation data include optical, infrared, X-ray and gamma-ray bands, pre-process the multi-band astronomical observation data to obtain a calibrated multi-band astronomical data set.

[0062] It is understandable that in this step, astronomical data in the optical, infrared, X-ray and gamma-ray bands are collected based on the observation capabilities of the Hubble Space Telescope, the Chandra X-ray Observatory and the Fermi Gamma-ray Space Telescope. These data contain radiation information from different types of celestial bodies in the universe, among which traces of dark matter may be hidden. For example, the high-resolution imaging capability of the Hubble Telescope can observe the structure and interaction inside galaxies, while the Chandra X-ray Observatory can detect X-rays emitted by high-temperature plasmas, which may be related to the gravitational effects of dark matter.

[0063] In this step, the preprocessing process includes image correction, dark field and flat field correction, noise removal, wavelength calibration and other steps to eliminate instrument effects and improve data quality. For example, the introduction to NVST spectral data preprocessing mentions the preprocessing process of solar spectral data, including flat field, dark field processing and interference fringe removal, which is crucial to ensure the accuracy and reliability of the data, and can reduce errors and deviations in subsequent analysis. Finally, a calibrated multi-band astronomical data set is obtained, which also provides a basis for subsequent feature extraction and analysis. The calibration process may include absolute and relative calibration of the data to ensure that data from different bands and different telescopes can be compared and analyzed on the same physical scale. The integrity and quality of these data sets directly affect the accuracy of subsequent feature extraction and the confidence of dark matter detection, thus providing astronomers with powerful data processing and analysis capabilities.

[0064] S200, using a convolutional neural network to perform deep feature extraction on the calibrated multi-band astronomical data set to obtain a first feature of the dark matter candidate signal, and using an autoencoder to perform dimensionality reduction and feature enhancement on the first feature of the dark matter candidate signal to obtain a second feature.

[0065] It can be understood that step S200 includes S201, S202, S203 and S204, wherein:

[0066] S201. According to the calibrated multi-band astronomical data set, after being processed by a convolution layer in a convolutional neural network, a convolution operation is performed using multiple convolution kernels to obtain a first feature of the dark matter candidate signal, wherein the first feature includes shape parameters, color changes, and spatial distribution features of the light curve extracted by the convolution layer and activation function in the CNN;

[0067] It should be noted that, first, according to the input multi-band astronomical data set, multiple convolution kernels (for example, K 1 , K 2，… K n ) to perform convolution operation. The calculation of convolution operation can be expressed as:

[0068] F ( x , y )=( f ∗ K )( x , y )+ b In the formula, f is the input data, K is the convolution kernel, ∗ represents the convolution operation, b is the bias term, F ( x , y ) is at position ( x , y )’s feature map.

[0069] The shape parameters of the light curve are extracted through the convolution operation, where the shape parameters reflect the brightness changes of galaxies or other celestial bodies on a time scale, which are related to the gravitational effects of dark matter.

[0070] Secondly, after the convolution operation, nonlinearity is introduced through the activation function (such as ReLU) to enhance the expressiveness of the model. The calculation formula of the ReLU function is:

[0071]

[0072] In the formula, a is the value of the feature map after being processed by the ReLU activation function, z is the result of the convolution operation.

[0073] Finally, through the above convolution operation and activation function processing, feature maps are obtained. These feature maps capture the local pattern and structural information of the input data, such as color changes and spatial distribution characteristics. Color changes provide information about astrophysical properties, such as temperature and composition, which can be used to infer the existence and properties of dark matter. The spatial distribution characteristics describe the distribution of dark matter candidate signals in space, including the distribution density of galaxies and the formation of large-scale structures, which are indirect evidence of the gravitational effect of dark matter.

[0074] S202, according to the first feature of the dark matter candidate signal, the signal is processed by the encoding layer of the autoencoder, which includes multiple convolutional layers and pooling layers, and the spatial dimension of the feature map is reduced by downsampling to obtain a compressed representation;

[0075] It should be noted that according to the first feature, the encoding layer of the autoencoder includes multiple convolutional layers and pooling layers. The calculation of the convolutional layer can be expressed as:

[0076]

[0077] In the formula, H prev is the feature map of the previous layer, W is the weight of the convolution kernel, b is the bias term, s is the activation function, H is the feature map of the current layer;

[0078] The spatial dimension of the feature map is reduced through a pooling layer (such as a maximum pooling layer), where the calculation of the maximum pooling operation can be expressed as:

[0079]

[0080] In the formula, H ( i , j ) is at position ( i , j )’s feature map, P ( x , y ) is the value of the feature map after the pooling operation;

[0081] Then, the height and width of the feature map are reduced through downsampling operations to reduce the spatial size of the data while retaining important feature information. This process can be achieved through convolution operations or pooling operations with a step size greater than 1.

[0082] In summary, in this step, a compressed representation is finally obtained, which includes the key information in the captured original data set. The key information obtained in this step reduces the dimension of the data, helps to reduce the computational burden of subsequent processing, and helps to improve the abstraction level of features. It provides a basis for the reconstruction of the decoding layer, can be used for subsequent dark matter detection and analysis, and also provides a new data compression and feature extraction method for subsequent research.

[0083] S203, using a decoding layer of an autoencoder to process the compressed representation, using a transposed convolution layer and an upsampling layer to change the data dimension, and obtaining a first feature of the processed dark matter candidate signal;

[0084] It should be noted that the decoding layer processing of the autoencoder starts with a compressed representation, which is a low-dimensional data representation generated by the encoding layer, where the goal of the decoding layer is to restore this compressed representation to data similar to the original input data. This process involves using transposed convolutional layers and upsampling layers to increase the spatial dimension of the feature map, thereby enhancing the detail information of the feature map. Among them, the purpose of the transposed convolutional layer (also called the deconvolution layer) is to increase the size of the feature map, thereby gradually restoring the spatial information of the data.

[0085] In this step, the upsampling layer follows the transposed convolution layer to further increase the spatial resolution of the feature map. This process can be achieved through the upsampling operation of nearest neighbor interpolation, bilinear interpolation, or transposed convolution. Among them, the size of the feature map is increased by inserting new pixels into the feature map. These new pixels are usually obtained by interpolating existing pixels, and the first feature of the processed dark matter candidate signal is obtained. At the same time, the first feature retains the key information in the compressed representation, providing a richer and more detailed data basis for subsequent feature selection and pattern recognition.

[0086] S204. Based on the processed first feature of the dark matter candidate signal, the second feature of the dark matter candidate signal is obtained through the bottleneck layer and the output layer of the autoencoder, wherein the second feature includes identifying abnormal patterns of light curves, color changes, and spatial distribution characteristics through the nonlinear transformation and feature selection mechanism of the autoencoder.

[0087] It should be noted that the bottleneck layer processes the features through a nonlinear activation function to enhance the expressiveness of the model. The calculation process of this step can be expressed as:

[0088] Z = s ( W ∗ Y ′+ b )

[0089] In the formula, Y ′ is the upsampled feature map,W is the weight matrix, b is the bias term, s is the activation function, Z is the output feature of the bottleneck layer;

[0090] The output layer converts the output of the bottleneck layer back to the spatial dimension of the original data to generate the final reconstructed data. The calculation process of this step can be expressed as:

[0091]

[0092] In the formula, Z is the output feature of the bottleneck layer, W ′ is the weight matrix of the output layer, b ′ is the bias term, is the reconstructed data, which is as close as possible to the original input data.

[0093] In summary, the processing of the bottleneck layer and the output layer enables the generation of the second features of the dark matter candidate signal. These features include abnormal patterns of the light curve, color changes, and spatial distribution characteristics, which are identified and enhanced through the nonlinear transformation and feature selection mechanism of the autoencoder, providing more accurate and detailed information for the detection and analysis of dark matter.

[0094] It can be understood that step S203 includes S2031, S2032 and S2033, wherein:

[0095] S2031, based on the compressed representation of the output of the encoding layer of the autoencoder, after being processed by the transposed convolution layer, a convolution operation is performed using the transposed convolution kernel W to obtain an upsampled feature map, wherein the calculation formula of the transposed convolution is as follows:

[0096] Among them, the transposed convolution kernel W is obtained through learning, and its size and step size determine the size of the output feature map:

[0097]

[0098] In the formula, Y is the output feature map, X is the input feature map, is the transposed convolution kernel, b is the bias term;

[0099] It should be noted that in the decoding process of the autoencoder, "to reconstruct the feature representation of the original input data" means that the goal of the decoder is to restore the compressed feature representation generated by the encoder to data that is as similar as possible to the original input data. The "original input data" here refers to the initial data that the autoencoder starts to process. For example, in image processing tasks, this may refer to the original image data. In the autoencoder, the encoder layer converts the input data into a low-dimensional feature representation through a series of convolution and pooling operations. This feature representation captures the key information of the input data. Then, the decoder layer needs to restore the original data from this compressed feature representation. This process involves gradually amplifying the output of the encoder (i.e., the compressed feature) through upsampling layers and transposed convolution layers (or deconvolution layers) to increase the spatial dimension of the feature map, and finally reconstructing data similar to the original input data.

[0100] In this process, the upsampling layer is used to increase the spatial resolution of the feature map. It can use different interpolation methods, such as nearest neighbor, bilinear interpolation, etc., to achieve this goal. The transposed convolution layer is used to gradually restore the spatial size of the feature map while reducing the number of channels of the feature map to match the dimension of the original input data. In the context of dark matter detection analysis, features of candidate dark matter signals can be extracted from complex astronomical data, and attempts can be made to reconstruct the original features of these signals for further analysis and verification)

[0101] S2032, using a bilinear interpolation method to process the upsampled feature map to obtain a feature map processed by an upsampling layer, wherein the feature map processed by the upsampling layer includes the recognition spatial resolution, wherein the calculation formula of the upsampling layer is as follows:

[0102]

[0103] In the formula, Y ′ is the feature map after upsampling layer processing, Y is the output feature map. Upsample is an operation in deep learning and image processing used to enlarge the size of the data.

[0104] S2033. Input the feature map processed by the upsampling layer to the decoding layer to reconstruct the feature representation of the original input data to obtain the first feature of the processed dark matter candidate signal, wherein the original input data is the initial data processed by the encoder.

[0105] It should be noted that the main difference between the first feature of the dark matter candidate signal after processing and before processing is the enhancement of features and the restoration of details, as well as the change of data dimension. The following is a specific comparison:

[0106] 1. Feature enhancement: Before processing: The original dark matter candidate signals may be contained in a multi-band astronomical data set. These signals may be relatively weak, mixed with background noise, and difficult to distinguish. After processing: Through the decoding layer of the autoencoder, especially the transposed convolution layer and the upsampling layer, the features of the dark matter candidate signals are enhanced, making these signals more prominent and obvious.

[0107] 2. Detail recovery: Before processing: During the encoding process of the autoencoder, the data may be downsampled, resulting in the loss of some detail information; After processing: The upsampling layer increases the spatial resolution of the feature map, restores the detail information, and makes the characteristics of the dark matter candidate signal clearer and more complete.

[0108] 3. Changes in data dimension: Before processing: the data may be in a compressed low-dimensional representation with low spatial resolution; After processing: through the transposed convolution layer and upsampling layer, the spatial dimension of the data is increased, restoring the original resolution of the data or higher, providing richer spatial information for further analysis.

[0109] 4. Depth of feature extraction: Before processing: feature extraction may be limited to shallow pattern recognition and lack the advanced feature representation capabilities of deep learning models; After processing: using deep learning models, especially CNN's capabilities, deeper feature representations can be extracted, which may be more sensitive and specific in dark matter detection.

[0110] 5. Separation of signal and noise: Before processing: dark matter candidate signals may be mixed with background noise and difficult to distinguish; After processing: through the nonlinear transformation and feature selection mechanism of the autoencoder, the dark matter signal and background noise can be separated more effectively and the purity of the signal can be improved.

[0111] In general, the first features of the processed dark matter candidate signals are enhanced and optimized after being processed by the decoding layer of the autoencoder, making these features more suitable for subsequent dark matter detection and analysis.

[0112] S300. According to the second feature, a large-scale structure simulation software is used to simulate the distribution of dark matter in the universe and its gravitational lens effect on light, and a generative adversarial network is used to synthesize simulated data of light affected by the distribution of dark matter. The simulated data is compared with the actual observation data to identify the predicted result of the potential position of dark matter.

[0113] It can be understood that step S300 includes S301, S302, S303 and S304, wherein:

[0114] S301. Based on the second characteristic, cosmological principles and historical observation data, a simulated distribution map of dark matter in the universe is obtained through N-body simulation processing, wherein the simulated distribution map includes the growth, evolution and aggregation of dark matter halos described by the positions and velocities of particles;

[0115] S302, simulating the gravitational lens effect on the simulated distribution map through simulation software, thereby obtaining the deflection effect of the light path, and the calculation formula is as follows:

[0116]

[0117] In the formula, dth is the deflection angle of the light, G is the gravitational constant, c is the speed of light, r (x′) is the dark matter density at position x′, |x−x′| is the distance between the light source position x and the dark matter position x′, is a volume element in three-dimensional space;

[0118] It should be noted that the gravitational lensing effect is calculated and processed on the simulated dark matter distribution map using large-scale structure simulation software, such as Gadget-2 or RAMSES. This process involves simulating the deflection effect of the gravitational field of dark matter on the light path passing through or near these structures. The simulation software can predict the path changes of light when passing through the dark matter distribution area, thereby obtaining a detailed distribution map describing the gravitational lensing effect. The distribution map reveals how light is bent by the gravitational field of dark matter, forming a unique light deflection pattern, which can be compared with actual astronomical observation data to identify the existence and distribution of dark matter.

[0119] It should be noted that in this process, the simulation software uses a computationally intensive method, namely N-body simulation technology, which is used to simulate the interaction and movement of a large number of particles (in this case, dark matter particles) under the action of gravity. In this way, a detailed dark matter distribution model of the universe is generated, which reflects the aggregation of dark matter at different scales, from galaxy clusters to individual galaxies, and even to smaller scales. In this step, the key to N-body simulation is to solve Newton's equations of motion, which describe how each particle is affected by the gravity of all other particles, and is obtained through an iterative process, usually using special algorithms to improve computational efficiency, such as fast Fourier transform (FFT) and tree structure methods, which allow simulations to handle millions of particles in a reasonable time. During the simulation, the position and velocity of each particle are updated to reflect the cumulative effect of gravitational interactions. As the simulation proceeds, particles will gather to form larger structures, such as galaxy clusters and superclusters.

[0120] It can be understood that this formula accumulates the dark matter density in all dark matter distribution areas in an integral form, and calculates the total deflection angle of light when passing through these dark matter. Its meaning is to predict the deflection of light based on the distribution of dark matter, so as to detect the existence of dark matter in astronomical observations.

[0121] S303, using the light path deflection data obtained by simulating the gravitational lens effect as a training target for a generative adversarial network, and generating simulated data similar to real astronomical data, wherein the training target of the generative adversarial network is to minimize the difference between the generated data and the real data;

[0122] It should be noted that the generator network accepts random noise z as input, and generates simulated data G(z) through a series of transposed convolutional layers and activation functions. This process can be expressed as:

[0123] G(z)=Activation(W T ∗z+ b )

[0124] Where W T is the transposed convolution kernel, b is the bias term, Activation is the activation function, z is the random noise accepted by the synthesizer network as input, and G(z) is the effect of dark matter distribution on light;

[0125] S304. Using the structural similarity index, the simulated data synthesized by the generative adversarial network is compared with the actual observed data to identify the predicted results of the potential location of dark matter.

[0126] It should be noted that based on the comparison results, areas that are significantly different from the actual data are identified. These areas are potential locations of dark matter. This process involves pattern recognition and anomaly detection algorithms, such as DBSCAN or Isolation Forest, which are used to identify abnormal patterns or potential dark matter signals from large amounts of data.

[0127] S400. Use the space-time correlation algorithm to process the prediction results of the potential position of dark matter, analyze the dark matter candidate signals at different time and space scales, and obtain a candidate event set of dark matter particle annihilation or decay events; apply the Bayesian inference method to process the candidate event set to obtain the first estimated values ​​of the dark matter model parameters; use the Markov chain Monte Carlo method to sample the parameter space of the first estimated value and verify the model to obtain the second estimated values ​​of the dark matter model parameters including density distribution, velocity distribution and energy spectrum.

[0128] It can be understood that step S400 includes S401 and S402, wherein:

[0129] S401. Based on the prediction results, a spatiotemporal correlation graph is constructed, wherein the spatiotemporal correlation graph includes a plurality of nodes and edges, wherein the nodes are observation data points at different time points and spatial positions, and the edges are the spatiotemporal correlations between the observation data points; after the graph construction process, dark matter candidate signals at different time and space scales are analyzed to obtain correlation strengths;

[0130] It can be understood that these nodes are data points obtained through dark matter detection experiments, representing events where dark matter particles interact with detectors, where the connections (edges) of the nodes represent the spatiotemporal correlations between these data points, that is, whether they may belong to different observational manifestations of the same dark matter event. In this step, it involves calculating the strength of the association between nodes in the graph, which can be achieved by measuring the similarity or correlation between nodes. For example, if two nodes are close in time and space, and their observational characteristics (such as energy deposition patterns) are similar, the strength of the association between them may be high. )

[0131] S402. Using a clustering algorithm to identify the correlation strength, the nodes belonging to the same dark matter event are grouped together to obtain a group result, which is recorded as a candidate event set for dark matter particle annihilation or decay events.

[0132] It is understandable that in this process, some basic concepts in graph theory, such as nodes (observation data points) and edges (space-time correlations), as well as clustering algorithms in machine learning are used to process and analyze data. The combination of these methods can extract information about dark matter from complex astronomical data and predict possible dark matter events, which can more effectively identify potential signals of dark matter from a large amount of astronomical observation data.

[0133] It should be noted that in this embodiment, the Bayesian inference method is used to process the candidate event set to obtain the first estimated value of the dark matter model parameter, update the belief in the parameter, and obtain a more accurate parameter estimate. According to the first estimated value, the Markov Chain Monte Carlo (MCMC) method is used to sample the parameter space and verify the model to obtain the second estimated value of the dark matter, where the calculation process of MCMC involves constructing a Markov chain that performs a random walk in the parameter space to meet the detailed balance condition. This process can be expressed as:

[0134]

[0135] In the formula, x t It is t The parameter status of the step, Q ( x t ∣ xt+1 ) is the proposed distribution, indicating that from the current state x t To the new state x t+1 Through MCMC, we can obtain the posterior distribution of parameters, thus making a more accurate estimate of the dark matter model parameters.

[0136] S500. Based on the second estimated value, use a Monte Carlo simulation method to simulate the interaction between dark matter particles and detector materials. The simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles generated. Based on the simulation results, construct a signal response model to characterize its energy, time, and spatial distribution characteristics. Use the signal response model to identify and extract dark matter signals from the detector readings through a peak detection method.

[0137] It can be understood that step S500 includes S501, S502 and S503, wherein:

[0138] S501. According to the second estimated value, the physical properties of the dark matter particles and the properties of the detector material, the collision probability of the dark matter particles and the atomic nuclei in the detector is obtained through Monte Carlo simulation processing, wherein the collision probability of the dark matter particles and the atomic nuclei in the detector includes calculating the flux of the dark matter particles and the effective area of ​​the detector and considering the interaction cross section between the dark matter particles and the atomic nuclei, and the calculation formula is as follows:

[0139] P collision= s ⋅ N ⋅ v

[0140] In the formula, s is the interaction cross section between dark matter particles and atomic nuclei, N is the number of nuclei in the detector, v is the speed of dark matter particles, P collision is the collision probability;

[0141] S502. Based on the collision probability, the energy transfer efficiency is calculated and processed to obtain the efficiency of energy transfer after the dark matter particle collides with the atomic nucleus in the detector. The calculation formula is as follows:

[0142]

[0143] In the formula, is the energy deposited in the detector, is the initial energy of the dark matter particle, is the energy transfer efficiency;

[0144] S503. By simulating the energy transfer efficiency produced by secondary particles, the types and quantities of secondary particles produced after the dark matter particles collide with the atomic nuclei in the detector are obtained.

[0145] It should be noted that this process involves simulating the type and quantity of particles (such as electrons and gamma rays) produced after the collision, which can be expressed by the following formula:

[0146] N secondaries= N collision⋅ or production

[0147] In the formula, N collision is the number of collisions, or production is the average efficiency of producing secondary particles per collision, N secondaries refers to the types and number of secondary particles;

[0148] In summary, in the process of simulating the interaction between dark matter particles and detector materials, the simulation results include not only the types and numbers of secondary particles produced after the collision between dark matter particles and atomic nuclei in the detector, but also the probability of these interactions and the efficiency of energy transfer. The three steps of "collision probability", "energy transfer efficiency" and "production of secondary particles" are interrelated and combined to provide information for the simulation results. Each step provides an understanding of different aspects of the overall interaction process, and ultimately together constitutes a comprehensive description of the behavior of dark matter particles in the detector. In this process, simulating the interaction between dark matter particles and detector materials is an indispensable part of the experimental design, which not only helps to improve the scientific return of the experiment, but also ensures the safety and effectiveness of the experiment.

[0149] In this step, a signal response model is constructed based on the results of the three-step simulation. This model is used to describe the characteristics of the signal generated when the dark matter particle interacts with the detector. These characteristics include the energy distribution of the signal (the size or intensity of the signal), the time distribution (the time order of the signal occurrence), and the spatial distribution (the location of the signal in the detector). Among them, the construction of the signal response model may involve statistical analysis methods, such as maximum likelihood estimation or Bayesian methods, to determine the parameters of the signal characteristics, that is, to show what kind of signal the dark matter particle is most likely to produce.

[0150] It is understood that signal response models can help understand the general characteristics of dark matter signals and help predict the behavior of signals under specific experimental conditions. For example, the model may show that dark matter signals tend to be more concentrated in a specific area of ​​the detector, or peak within a specific time window. By analyzing the simulated data, the statistical distribution of the signal is determined, such as whether it conforms to the Poisson distribution, Gaussian distribution, or other known statistical models. Then once you have the signal response model, you can apply it to the actual detector data. The signal response model can guide the design of data analysis algorithms, such as peak detection algorithms, to identify those signals that are consistent with the model predictions. Peak detection is a commonly used signal processing technique used to identify significant changes or peaks in the data. In dark matter detection, this method is used to extract possible dark matter signals from the detector readings.

[0151] In general, this step describes a process from simulation to practical application, in which the simulation results are used to build a model, which is further used to analyze actual data and extract dark matter signals. This process is iterative and may require multiple adjustments and optimizations of model parameters to more accurately identify dark matter signals.

[0152] S600, analyzing the background noise in the detector, including cosmic rays, environmental radioactivity and other interference sources, determining the baseline noise level of the detector, and obtaining background noise analysis results; combining the identified extracted dark matter signal with the background noise analysis results, calculating the confidence of dark matter detection, and obtaining dark matter detection analysis results.

[0153] It is understood that in this step, this usually involves statistical analysis, such as hypothesis testing and calculation of confidence intervals, to assess whether the observed signal exceeds the expectation of random noise, thereby determining the confidence of dark matter detection. Through multivariate analysis and clustering algorithms, such as DBSCAN or spectral clustering, pattern recognition and anomaly detection are performed on the data of different detectors to obtain the detection confidence analysis results. This method can improve the ability to identify dark matter signals, especially when there is complex background noise in the data.

[0154] It should be noted that step S600 includes S601, S602 and S603, wherein:

[0155] S601. According to the characteristics of the dark matter signal and the distribution of the background noise, the signal strength and the noise level are quantified to obtain a signal-noise distinction metric, the calculation formula of which is as follows:

[0156]

[0157] In the formula, S is the average signal strength,N is the average intensity of noise, SNR is the signal-to-noise ratio;

[0158] It can be understood that the larger the ratio, the more likely the signal is to be distinguished from the background noise.

[0159] S602. Based on the signal-to-noise ratio, the confidence level is calculated and processed to obtain the confidence level of dark matter detection. The calculation formula is as follows:

[0160]

[0161] Where: is the probability of observing the current data under the null hypothesis (i.e., no dark matter signal), P is the probability, The hypothesis of no effect or no difference;

[0162] From this formula, we can see that if this probability is very small, it indicates that the observed data is unlikely to be due to random fluctuations alone, thus providing evidence against the null hypothesis, i.e. p − value The smaller it is, the less likely it is that the observed signal is just a fluctuation of background noise.

[0163] S603. Analyze the confidence of dark matter detection. The analysis process is to determine whether the probability of the current data is less than a preset threshold. If so, the detected signal is considered to be a candidate event for dark matter. If not, the detected signal is not considered to be a candidate event for dark matter, thereby obtaining the dark matter detection analysis result.

[0164] This step involves setting a confidence threshold, which is usually determined based on the experimental design and the requirements for statistical significance. p − value If it is less than a preset threshold (such as 0.05), the detected signal is considered statistically significant and can be considered a candidate event for dark matter. The formula is as follows:

[0165]

[0166] Therefore, through these steps, the dark matter signal can be identified from the detector readings, and the confidence of the detection can be calculated to obtain the analysis results of dark matter detection.

[0167] Embodiment 2:

[0168] like Figure 2 As shown, this embodiment provides a dark matter detection and analysis system based on an astronomical telescope, see Figure 2 The system comprises:

[0169] Collection module 701: used to collect multi-band astronomical observation data acquired by the Hubble Space Telescope, the Chandra X-ray Observatory and the Fermi Gamma-ray Space Telescope, wherein the multi-band astronomical observation data includes optical, infrared, X-ray and gamma-ray bands, and pre-process the multi-band astronomical observation data to obtain a calibrated multi-band astronomical data set;

[0170] Feature extraction module 702: used to perform deep feature extraction on the calibrated multi-band astronomical data set using a convolutional neural network to obtain a first feature of the dark matter candidate signal, and perform dimensionality reduction and feature enhancement on the first feature of the dark matter candidate signal through an autoencoder to obtain a second feature;

[0171] Identification module 703: used to simulate the distribution of dark matter in the universe and its gravitational lens effect on light according to the second feature using large-scale structure simulation software, and synthesize light simulation data affected by the distribution of dark matter using a generative adversarial network, compare the simulation data with the actual observation data, and identify the prediction result of the potential position of dark matter;

[0172] Processing module 704: used to process the prediction results of the potential position of dark matter by using the spatiotemporal correlation algorithm, analyze the candidate dark matter signals at different time and space scales, and obtain a candidate event set of dark matter particle annihilation or decay events; apply the Bayesian inference method to process the candidate event set to obtain a first estimate of the dark matter model parameters; perform parameter space sampling and model verification on the first estimate by the Markov chain Monte Carlo method to obtain a second estimate of the dark matter model parameters including density distribution, velocity distribution and energy spectrum;

[0173] Simulation module 705: used to simulate the interaction between dark matter particles and detector materials using a Monte Carlo simulation method according to the second estimated value, wherein the simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles generated; based on the simulation results, a signal response model is constructed to characterize its energy, time, and space distribution characteristics, and the signal response model is used to identify and extract dark matter signals from the detector readings through a peak detection method;

[0174] Calculation module 706: used to analyze the background noise in the detector, including cosmic rays, environmental radioactivity and other interference sources, determine the baseline noise level of the detector, and obtain background noise analysis results; combine the identified extracted dark matter signal with the background noise analysis results, calculate the confidence of dark matter detection, and obtain dark matter detection analysis results.

[0175] Specifically, the feature extraction module 702 includes:

[0176] The first processing unit is used to obtain the first feature of the dark matter candidate signal by performing convolution operation using multiple convolution kernels according to the calibrated multi-band astronomical data set after processing by the convolution layer in the convolutional neural network, wherein the first feature includes the shape parameters, color change and spatial distribution characteristics of the light curve extracted by the convolution layer and activation function in the CNN;

[0177] The second processing unit is used for processing the first feature of the dark matter candidate signal through the encoding layer of the autoencoder, which includes multiple convolutional layers and pooling layers, and reducing the spatial dimension of the feature map by downsampling to obtain a compressed representation;

[0178] The third processing unit is used to process the compressed representation using the decoding layer of the autoencoder, and change the data dimension using the transposed convolution layer and the upsampling layer to obtain the first feature of the processed dark matter candidate signal;

[0179] The fourth processing unit is used to obtain the second feature of the dark matter candidate signal based on the processed first feature of the dark matter candidate signal through the bottleneck layer and the output layer of the autoencoder, wherein the second feature includes identifying abnormal patterns of light curves, color changes, and spatial distribution characteristics through the nonlinear transformation and feature selection mechanism of the autoencoder.

[0180] Specifically, the third processing unit includes:

[0181] Operation unit: used for the compressed representation of the output of the encoding layer based on the autoencoder, processed by the transposed convolution layer, using the transposed convolution kernel W Perform a convolution operation to obtain the upsampled feature map, where the calculation formula for the transposed convolution is as follows:

[0182]

[0183] In the formula, Y is the output feature map, X is the input feature map, is the transposed convolution kernel, b is the bias term;

[0184] The fifth processing unit is used to process the upsampled feature map using a bilinear interpolation method to obtain a feature map processed by an upsampling layer, wherein the feature map processed by the upsampling layer includes an identification spatial resolution, wherein the calculation formula of the upsampling layer is as follows:

[0185]

[0186] In the formula, Y ′ is the feature map after upsampling layer processing, Yis the output feature map. Upsample is an operation in deep learning and image processing used to enlarge the size of the data.

[0187] Reconstruction unit: used to input the feature map processed by the upsampling layer to the decoding layer to reconstruct the feature representation of the original input data and obtain the first feature of the processed dark matter candidate signal, wherein the original input data is the initial data processed by the encoder.

[0188] Specifically, the identification module 703 includes:

[0189] Based on the second characteristic, cosmological principles and historical observational data, after N-body simulation processing, a simulated distribution map of dark matter in the universe is obtained, where the simulated distribution map includes the growth, evolution and aggregation of dark matter halos described by the positions and velocities of particles;

[0190] The gravitational lens effect is simulated on the simulated distribution map through simulation software to obtain the deflection effect of the light path. The calculation formula is as follows:

[0191]

[0192] In the formula, dth is the deflection angle of the light, G is the gravitational constant, c is the speed of light, r (x′) is the dark matter density at position x′, |x−x′| is the distance between the light source position x and the dark matter position x′, is a volume element in three-dimensional space;

[0193] The light path deflection data obtained by simulating the gravitational lens effect is used as the training target of the generative adversarial network to generate simulated data similar to the real astronomical data, where the training goal of the generative adversarial network is to minimize the difference between the generated data and the real data;

[0194] The structural similarity index is used to compare the simulated data synthesized by the generative adversarial network with the actual observed data to identify the predicted results of the potential location of dark matter.

[0195] Specifically, the processing module 704 includes:

[0196] Based on the prediction results, a spatiotemporal correlation graph is constructed, where the spatiotemporal correlation graph includes multiple nodes and edges. The nodes are observation data points at different time points and spatial positions, and the edges are the spatiotemporal correlations between the observation data points. After the graph construction process, the dark matter candidate signals at different time and space scales are analyzed to obtain the correlation strength.

[0197] The clustering algorithm is used to identify the correlation strength, and the nodes belonging to the same dark matter event are grouped together to obtain the collective result, which is recorded as the candidate event set of dark matter particle annihilation or decay events.

[0198] Specifically, the simulation module 705 includes:

[0199] According to the second estimate, the physical characteristics of dark matter particles and the properties of the detector material, the collision probability of dark matter particles and atomic nuclei in the detector is obtained through Monte Carlo simulation. The collision probability of dark matter particles and atomic nuclei in the detector includes the calculation of the flux of dark matter particles and the effective area of ​​the detector and the consideration of the interaction cross section between dark matter particles and atomic nuclei. The calculation formula is as follows:

[0200] P collision= s ⋅ N ⋅ v

[0201] In the formula, s is the interaction cross section between dark matter particles and atomic nuclei, N is the number of nuclei in the detector, v is the speed of dark matter particles, P collision is the collision probability;

[0202] Based on the collision probability, after calculating the energy transfer efficiency, the efficiency of energy transfer after the dark matter particle collides with the atomic nucleus in the detector is obtained. The calculation formula is as follows:

[0203]

[0204] In the formula, is the energy deposited in the detector, is the initial energy of the dark matter particle, is the energy transfer efficiency;

[0205] By simulating the energy transfer efficiency of secondary particles, the types and quantities of secondary particles produced after the collision of dark matter particles with atomic nuclei in the detector are obtained.

[0206] Specifically, the calculation module 706 includes:

[0207] According to the characteristics of dark matter signals and the distribution of background noise, the signal intensity and noise level are quantified to obtain the signal-noise distinction metric, which is calculated as follows:

[0208]

[0209] In the formula, Sis the average signal strength, N is the average intensity of noise, SNR is the signal-to-noise ratio;

[0210] Based on the signal-to-noise ratio, the confidence level of dark matter detection is calculated and processed, and the calculation formula is as follows:

[0211]

[0212] In the formula, is the probability of observing the current data under the null hypothesis (i.e., no dark matter signal), P is the probability, The hypothesis of no effect or no difference;

[0213] The confidence of dark matter detection is analyzed. The analysis process is to determine whether the probability of the current data is less than a preset threshold. If so, the detected signal is considered to be a candidate event for dark matter. If not, the detected signal is not considered to be a candidate event for dark matter, and the dark matter detection analysis results are obtained.

[0214] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0215] Embodiment 3:

[0216] Corresponding to the above method embodiment, this embodiment also provides a dark matter detection and analysis device based on an astronomical telescope. The dark matter detection and analysis device based on an astronomical telescope described below and the dark matter detection and analysis method based on an astronomical telescope described above can refer to each other.

[0217] Figure 3 FIG. 8 is a block diagram of a dark matter detection and analysis device 800 based on an astronomical telescope according to an exemplary embodiment. Figure 3 As shown, the dark matter detection and analysis device 800 based on an astronomical telescope includes: a processor 801 and a memory 802. The dark matter detection and analysis device 800 based on an astronomical telescope also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0218] The processor 801 is used to control the overall operation of the astronomical telescope-based dark matter detection and analysis device 800 to complete all or part of the steps in the above-mentioned astronomical telescope-based dark matter detection and analysis method. The memory 802 is used to store various types of data to support the operation of the astronomical telescope-based dark matter detection and analysis device 800, and these data may include, for example, instructions for any application or method operating on the astronomical telescope-based dark matter detection and analysis device 800, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be a keyboard, a mouse or a button, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the dark matter detection and analysis device 800 based on the astronomical telescope and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: a Wi-Fi module, a Bluetooth module or an NFC module.

[0219] In an exemplary embodiment, the dark matter detection and analysis device 800 based on an astronomical telescope can be implemented by one or more application specific integrated circuits (Application Specific Integrated Circuit, referred to as ASIC), digital signal processors (Digital Signal Processing Device, referred to as DSPD), programmable logic devices (Programmable Logic Device, referred to as PLD), field programmable gate arrays (Field Programmable Gate Array, referred to as FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned dark matter detection and analysis method based on an astronomical telescope.

[0220] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned dark matter detection and analysis method based on an astronomical telescope are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the dark matter detection and analysis device 800 based on an astronomical telescope to complete the above-mentioned dark matter detection and analysis method based on an astronomical telescope.

[0221] Embodiment 4:

[0222] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the dark matter detection and analysis method based on an astronomical telescope described above can refer to each other.

[0223] A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, the steps of the dark matter detection and analysis method based on an astronomical telescope of the above method embodiment are implemented.

[0224] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0225] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0226] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A dark matter detection and analysis method based on an astronomical telescope, characterized in that: include: Collect multi-band astronomical observation data acquired by the Hubble Space Telescope, Chandra X-ray Observatory and Fermi Gamma-ray Space Telescope, where the multi-band astronomical observation data includes optical, infrared, X-ray and gamma-ray bands, pre-process the multi-band astronomical observation data to obtain a calibrated multi-band astronomical data set; The convolutional neural network is used to perform deep feature extraction on the calibrated multi-band astronomical data set to obtain the first feature of the dark matter candidate signal. The autoencoder is used to reduce the dimension and enhance the features of the first feature of the dark matter candidate signal to obtain the second feature. According to the second feature, large-scale structure simulation software is used to simulate the distribution of dark matter in the universe and its gravitational lensing effect on light, and a generative adversarial network is used to synthesize simulated data of light affected by the distribution of dark matter. The simulated data is compared with the actual observation data to identify the prediction results of the potential location of dark matter; The prediction results of the potential location of dark matter are processed by using a spatiotemporal correlation algorithm, and candidate dark matter signals at different time and space scales are analyzed to obtain a candidate event set of dark matter particle annihilation or decay events. The candidate event set is processed using the Bayesian inference method to obtain a first estimate of the dark matter model parameters. The first estimate is sampled in the parameter space and the model is verified using the Markov chain Monte Carlo method to obtain a second estimate of the dark matter model parameters including density distribution, velocity distribution, and energy spectrum. Based on the second estimate, a Monte Carlo simulation method is used to simulate the interaction between dark matter particles and detector materials. The simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles produced. Based on the simulation results, a signal response model is constructed to characterize its energy, time, and spatial distribution characteristics. The signal response model is used to identify and extract dark matter signals from the detector readings through a peak detection method. Analyze the background noise in the detector, including cosmic rays, environmental radioactivity and other interference sources, determine the baseline noise level of the detector, and obtain background noise analysis results; Combine the extracted dark matter signal with the background noise analysis results to calculate the confidence of dark matter detection and obtain the dark matter detection analysis results; The method uses a convolutional neural network to perform deep feature extraction on the calibrated multi-band astronomical data set to obtain a first feature of the dark matter candidate signal, and uses an autoencoder to perform dimensionality reduction and feature enhancement on the first feature of the dark matter candidate signal to obtain a second feature, which includes: Based on the calibrated multi-band astronomical data set, after being processed by the convolution layer in the convolutional neural network, a convolution operation is performed using multiple convolution kernels to obtain the first feature of the dark matter candidate signal, where the first feature includes the shape parameters, color changes and spatial distribution characteristics of the light curve extracted by the convolution layer and activation function in the CNN; According to the first feature of the dark matter candidate signal, it is processed by the encoding layer of the autoencoder, which includes multiple convolutional layers and pooling layers, and the spatial dimension of the feature map is reduced by downsampling to obtain a compressed representation; The compressed representation is processed using the decoding layer of the autoencoder, and the data dimension is changed using the transposed convolution layer and the upsampling layer to obtain the first feature of the processed dark matter candidate signal; Based on the processed first feature of the dark matter candidate signal, the second feature of the dark matter candidate signal is obtained through the bottleneck layer and the output layer of the autoencoder, wherein the second feature includes identifying the abnormal pattern of the light curve, the color change and the spatial distribution characteristics through the nonlinear transformation and feature selection mechanism of the autoencoder; According to the second feature, large-scale structure simulation software is used to simulate the distribution of dark matter in the universe and its gravitational lens effect on light, and a generative adversarial network is used to synthesize simulated data of light affected by the distribution of dark matter. The simulated data is compared with the actual observation data to identify the prediction results of the potential location of dark matter, including: Based on the second characteristic, cosmological principles and historical observational data, after N-body simulation processing, a simulated distribution map of dark matter in the universe is obtained, where the simulated distribution map includes the growth, evolution and aggregation of dark matter halos described by the positions and velocities of particles; The gravitational lens effect is simulated on the simulated distribution map through simulation software to obtain the deflection effect of the light path. The calculation formula is as follows: In the formula, δθ is the deflection angle of the light, G is the gravitational constant, c is the speed of light, ρ (x′) is the dark matter density at position x′, is the distance between the light source position x and the dark matter position x′, is a volume element in three-dimensional space; The light path deflection data obtained by simulating the gravitational lens effect is used as the training target of the generative adversarial network to generate simulated data similar to the real astronomical data, where the training goal of the generative adversarial network is to minimize the difference between the generated data and the real data; The structural similarity index is used to compare the simulated data synthesized by the generative adversarial network with the actual observed data to identify the prediction results of the potential location of dark matter; According to the second estimated value, a Monte Carlo simulation method is used to simulate the interaction between the dark matter particles and the detector material, wherein the simulation process includes the collision probability of the dark matter particles and the atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles generated, including: According to the second estimate, the physical characteristics of dark matter particles and the properties of the detector material, the collision probability of dark matter particles and atomic nuclei in the detector is obtained through Monte Carlo simulation. The collision probability of dark matter particles and atomic nuclei in the detector includes the calculation of the flux of dark matter particles and the effective area of ​​the detector and the consideration of the interaction cross section between dark matter particles and atomic nuclei. The calculation formula is as follows: In the formula, σ is the interaction cross section between dark matter particles and atomic nuclei, N is the number of nuclei in the detector, v is the speed of dark matter particles, P collision is the collision probability; Based on the collision probability, after calculating the energy transfer efficiency, the efficiency of energy transfer after the dark matter particle collides with the atomic nucleus in the detector is obtained. The calculation formula is as follows: In the formula, is the energy deposited in the detector, is the initial energy of the dark matter particle, is the energy transfer efficiency; Through the simulation of energy transfer efficiency produced by secondary particles, the types and quantities of secondary particles produced after the collision between dark matter particles and atomic nuclei in the detector are obtained; The dark matter signal extracted by combining identification and background noise analysis results is calculated to obtain the dark matter detection analysis results, including: According to the characteristics of dark matter signals and the distribution of background noise, the signal intensity and noise level are quantified to obtain the signal-noise distinction metric, which is calculated as follows: In the formula, S is the average signal strength, N is the average intensity of noise, SNR is the signal-to-noise ratio; Based on the signal-to-noise ratio, the confidence level of dark matter detection is calculated and processed by confidence level. The calculation formula is as follows: In the formula, is the probability of observing the current data under the null hypothesis, which is that there is no dark matter signal, and P is the probability. The hypothesis of no effect or no difference; The confidence of dark matter detection is analyzed. The analysis process is to determine whether the probability of the current data is less than a preset threshold. If so, the detected signal is considered to be a candidate event for dark matter. If not, the detected signal is not considered to be a candidate event for dark matter, and the dark matter detection analysis results are obtained.

2. The dark matter detection and analysis method based on an astronomical telescope according to claim 1, characterized in that: The compressed representation is processed by using the decoding layer of the autoencoder, and the data dimension is changed by using the transposed convolution layer and the upsampling layer to obtain the first feature of the processed dark matter candidate signal, which includes: Compressed representation of the output of the encoding layer based on the autoencoder, processed by the transposed convolution layer, using the transposed convolution kernel W Perform a convolution operation to obtain the upsampled feature map, where the calculation formula for the transposed convolution is as follows: In the formula, Y is the output feature map, X is the input feature map, is the transposed convolution kernel, b is the bias term; The upsampled feature map is processed by bilinear interpolation to obtain the feature map after the upsampling layer processing, wherein the feature map after the upsampling layer processing includes the recognition spatial resolution, wherein the calculation formula of the upsampling layer is as follows: In the formula, Y ′ is the feature map after processing by the upsampling layer, Y is the output feature map, Upsample is an operation in deep learning and image processing, used to enlarge the size of the data; The feature map processed by the upsampling layer is input to the decoding layer to reconstruct the feature representation of the original input data and obtain the first feature of the processed dark matter candidate signal, where the original input data is the initial data processed by the encoder.

3. The dark matter detection and analysis method based on an astronomical telescope according to claim 1, characterized in that: The prediction results of the potential position of dark matter are processed by using the spatiotemporal correlation algorithm, and the candidate dark matter signals at different time and space scales are analyzed to obtain a candidate event set of dark matter particle annihilation or decay events, including: Based on the prediction results, a spatiotemporal correlation graph is constructed, where the spatiotemporal correlation graph includes multiple nodes and edges. The nodes are observation data points at different time points and spatial positions, and the edges are the spatiotemporal correlations between the observation data points. After the graph construction process, the dark matter candidate signals at different time and space scales are analyzed to obtain the correlation strength. The clustering algorithm is used to identify the correlation strength, and the nodes belonging to the same dark matter event are grouped together to obtain the collective result, which is recorded as the candidate event set of dark matter particle annihilation or decay events.

4. A dark matter detection and analysis system based on an astronomical telescope, based on the dark matter detection and analysis method based on an astronomical telescope according to claim 1, characterized in that: include: Collection module: used to collect multi-band astronomical observation data acquired by the Hubble Space Telescope, Chandra X-ray Observatory and Fermi Gamma-ray Space Telescope, where the multi-band astronomical observation data includes optical, infrared, X-ray and gamma-ray bands, and pre-process the multi-band astronomical observation data to obtain a calibrated multi-band astronomical data set; Feature extraction module: used to perform deep feature extraction on the calibrated multi-band astronomical data set using a convolutional neural network to obtain the first feature of the dark matter candidate signal, and to perform dimensionality reduction and feature enhancement on the first feature of the dark matter candidate signal through an autoencoder to obtain the second feature; Identification module: used to simulate the distribution of dark matter in the universe and its gravitational lensing effect on light using large-scale structure simulation software based on the second feature, and use generative adversarial networks to synthesize simulated data of light affected by the distribution of dark matter, compare the simulated data with the actual observation data, and identify the prediction results of the potential location of dark matter; Processing module: used to process the prediction results of the potential position of dark matter using the spatiotemporal correlation algorithm, analyze the candidate dark matter signals at different time and space scales, and obtain the candidate event set of dark matter particle annihilation or decay events; apply the Bayesian inference method to process the candidate event set to obtain the first estimated value of the dark matter model parameters; use the Markov chain Monte Carlo method to sample the parameter space of the first estimated value and verify the model to obtain the second estimated value of the dark matter model parameters including density distribution, velocity distribution and energy spectrum; Simulation module: used to simulate the interaction between dark matter particles and detector materials using the Monte Carlo simulation method based on the second estimated value. The simulation process includes the collision probability of dark matter particles and atomic nuclei in the detector, the energy transfer efficiency, and the secondary particles generated. Based on the simulation results, a signal response model is constructed to characterize its energy, time, and spatial distribution characteristics. The signal response model is used to identify and extract dark matter signals from the detector readings through the peak detection method. Computing module: used to analyze the background noise in the detector, including cosmic rays, environmental radioactivity and other interference sources, determine the baseline noise level of the detector, and obtain the background noise analysis results; The extracted dark matter signal and background noise analysis results are combined to calculate the confidence of dark matter detection and obtain the dark matter detection analysis results.

5. The dark matter detection and analysis system based on an astronomical telescope according to claim 4, characterized in that: The feature extraction module includes: The first processing unit is used to obtain the first feature of the dark matter candidate signal by performing convolution operation using multiple convolution kernels according to the calibrated multi-band astronomical data set after processing by the convolution layer in the convolutional neural network, wherein the first feature includes the shape parameters, color change and spatial distribution characteristics of the light curve extracted by the convolution layer and activation function in the CNN; The second processing unit is used for processing the first feature of the dark matter candidate signal through the encoding layer of the autoencoder, which includes multiple convolutional layers and pooling layers, and reducing the spatial dimension of the feature map by downsampling to obtain a compressed representation; The third processing unit is used to process the compressed representation using the decoding layer of the autoencoder, and change the data dimension using the transposed convolution layer and the upsampling layer to obtain the first feature of the processed dark matter candidate signal; The fourth processing unit is used to obtain the second feature of the dark matter candidate signal based on the processed first feature of the dark matter candidate signal through the bottleneck layer and the output layer of the autoencoder, wherein the second feature includes identifying abnormal patterns of light curves, color changes, and spatial distribution characteristics through the nonlinear transformation and feature selection mechanism of the autoencoder.

6. The dark matter detection and analysis system based on an astronomical telescope according to claim 5, characterized in that: The third processing unit includes: Operation unit: used for the compressed representation of the output of the encoding layer based on the autoencoder, processed by the transposed convolution layer, using the transposed convolution kernel W Perform a convolution operation to obtain the upsampled feature map, where the calculation formula for the transposed convolution is as follows: In the formula, Y is the output feature map, X is the input feature map, is the transposed convolution kernel, b is the bias term; The fifth processing unit is used to process the upsampled feature map using a bilinear interpolation method to obtain a feature map processed by an upsampling layer, wherein the feature map processed by the upsampling layer includes an identification spatial resolution, wherein the calculation formula of the upsampling layer is as follows: In the formula, Y ′ is the feature map after processing by the upsampling layer, Y is the output feature map, Upsample is an operation in deep learning and image processing, used to enlarge the size of the data; Reconstruction unit: used to input the feature map processed by the upsampling layer to the decoding layer to reconstruct the feature representation of the original input data and obtain the first feature of the processed dark matter candidate signal, wherein the original input data is the initial data processed by the encoder.

Images